Publications

Humans tend to follow the Uniform Information Density (UID) principle by distributing information evenly in utterances. We study if decoding algorithms implicitly follow this UID principle‚ and under what conditions adherence to UID might be desirable for dialogue generation. We generate responses using different decoding algorithms with GPT-2 on the Persona-Chat dataset and collect human judgments on their quality using Amazon Mechanical Turk. We find that (i) surprisingly‚ model-generated responses follow the UID principle to a greater extent than human responses‚ and (ii) decoding algorithms that promote UID do not generate higher-quality responses. Instead‚ when we control for surprisal‚ non-uniformity of information density correlates with the quality of responses with very low/high surprisal. Our findings indicate that encouraging non-uniform responses is a potential solution to the “likelihood trap” problem (quality degradation in very high-likelihood text). Our dataset containing multiple candidate responses per dialog history along with human-annotated quality ratings is available at: https://huggingface.co/datasets/saranya132/dialog_uid_gpt2.

Knowledge-grounded dialogue systems powered by large language models often generate responses that‚ while fluent‚ are not attributable to a relevant source of information. Progress towards models that do not exhibit this issue requires evaluation metrics that can quantify its prevalence. To this end‚ we introduce the Benchmark for Evaluation of Grounded INteraction (Begin)‚ comprising 12k dialogue turns generated by neural dialogue systems trained on three knowledge-grounded dialogue corpora. We collect human annotations assessing the extent to which the models’ responses can be attributed to the given background information. We then use Begin to analyze eight evaluation metrics. We find that these metrics rely on spurious correlations‚ do not reliably distinguish attributable abstractive responses from unattributable ones‚ and perform substantially worse when the knowledge source is longer. Our findings underscore the need for more sophisticated and robust evaluation metrics for knowledge-grounded dialogue. We make BEGIN publicly available at https://github.com/google/BEGIN-dataset.

For open-domain conversational question answering (CQA)‚ it is important to retrieve the most relevant passages to answer a question‚ but this is challenging compared with standard passage retrieval because it requires understanding the full dialogue context rather than a single query. Moreover‚ it can be expensive to re-train well-established retrievers such as search engines that are originally developed for non-conversational queries. To facilitate their use‚ we develop a query rewriting model CONQRR that rewrites a conversational question in context into a standalone question. It is trained with a novel reward function to directly optimize towards retrieval and can be adapted to any fixed blackbox retriever using reinforcement learning. We show that CONQRR achieves state-of-the-art results on a recent open-domain CQA dataset‚ a combination of conversations from three different sources. We also conduct extensive experiments to show the effectiveness of CONQRR for any given fixed retriever.

AI researchers have posited Dungeons and Dragons (D&D) as a challenge problem to test systems on various language-related capabilities. In this paper‚ we frame D&D specifically as a dialogue system challenge‚ where the tasks are to both generate the next conversational turn in the game and predict the state of the game given the dialogue history. We create a gameplay dataset consisting of nearly 900 games‚ with a total of 7‚000 players‚ 800‚000 dialogue turns‚ 500‚000 dice rolls‚ and 58 million words. We automatically annotate the data with partial state information about the game play. We train a large language model (LM) to generate the next game turn‚ conditioning it on different information. The LM can respond as a particular character or as the player who runs the game–i.e.‚ the Dungeon Master (DM). It is trained to produce dialogue that is either in-character (roleplaying in the fictional world) or out-of-character (discussing rules or strategy). We perform a human evaluation to determine what factors make the generated output plausible and interesting. We further perform an automatic evaluation to determine how well the model can predict the game state given the history and examine how well tracking the game state improves its ability to produce plausible conversational output.

With recent improvements in natural language generation (NLG) models for various applications‚ it has become imperative to have the means to identify and evaluate whether NLG output is only sharing verifiable information about the external world. In this work‚ we present a new evaluation framework entitled Attributable to Identified Sources (AIS) for assessing the output of natural language generation models‚ when such output pertains to the external world. We first define AIS and introduce a two-stage annotation pipeline for allowing annotators to appropriately evaluate model output according to AIS guidelines. We empirically validate this approach on generation datasets spanning three tasks (two conversational QA datasets‚ a summarization dataset‚ and a table-to-text dataset) via human evaluation studies that suggest that AIS could serve as a common framework for measuring whether model-generated statements are supported by underlying sources. We release guidelines for the human evaluation studies.

Why do bilinguals switch languages within a sentence? The present observational study asks whether word surprisal and word entropy predict code-switching in bilingual written conversation. We describe and model a new dataset of Chinese-English text with 1476 clean code-switched sentences‚ translated back into Chinese. The model includes known control variables together with word surprisal and word entropy. We found that word surprisal‚ but not entropy‚ is a significant predictor that explains code-switching above and beyond other well-known predictors. We also found sentence length to be a significant predictor‚ which has been related to sentence complexity. We propose high cognitive effort as a reason for code-switching‚ as it leaves fewer resources for inhibition of the alternative language. We also corroborate previous findings‚ but this time using a computational model of surprisal‚ a new language pair‚ and doing so for written language.

Computational models of distributional semantics (a.k.a. word embeddings) represent a word’s meaning in terms of its relationships with all other words. We examine what grammatical information is encoded in distributional models and investigate the role of indirect associations. Distributional models are sensitive to associations between words at one degree of separation‚ such as ‘tiger’ and ‘stripes’‚ or two degrees of separation‚ such as ‘soar’ and ‘fly’. By recursively adding higher levels of representations to a computational‚ holographic model of semantic memory‚ we construct a distributional model sensitive to associations between words at arbitrary degrees of separation. We find that word associations at four degrees of separation increase the similarity assigned by the model to English words that share part-of-speech or syntactic type. Word associations at four degrees of separation also improve the ability of the model to construct grammatical English sentences. Our model proposes that human memory uses indirect associations to learn part-of-speech and that the basic associative mechanisms of memory and learning support knowledge of both semantics and grammatical structure.

We face delays in a variety of situations. They are either inevitable‚ e.g.‚ due to system limits‚ or are intentionally added‚ e.g.‚ advertisements. In many situations‚ a visual feedback is provided during the delay to manage expectations. This feedback is usually provided through progress bars‚ percentages‚ or countdowns‚ depending on design limitations such as screen size. In this article‚ we use 15-second delays and examine (a) how delays affect users’ decision-making and task satisfaction‚ and (b) how to manipulate time perception to reduce the negative consequences of delays. Experiment 1 ( N = 421 ) shows that faster countdowns increase task satisfaction and lead to more rational decisions in the subsequent task. In Experiment 2‚ we investigate the effect of countdown speed on delay perception and recall ( N = 531 ). We show that faster countdowns lead to shorter perceived delays‚ while the delay will be recalled as longer after the task. The opposite is obtained for slower countdowns. We also increased the countdown rate and found a limit for the effect of increased speed. Thus‚ designers have to trade-off between how delays are perceived at the moment of experience and how they are recalled. We discuss the implications of these findings for user interface design.

We demonstrate that the key components of cognitive architectures—declarative and procedural memory—and their key capabilities—learning‚ memory retrieval‚ probability judgement‚ and utility estimation—can be implemented as algebraic operations on vectors and tensors in a high-dimensional space using a distributional semantics model. High-dimensional vector spaces underlie the success of modern machine learning techniques based on deep learning. However‚ while neural networks have an impressive ability to process data to find patterns‚ they do not typically model high-level cognition‚ and it is often unclear how they work. Symbolic cognitive architectures can capture the complexities of high-level cognition and provide human-readable‚ explainable models‚ but scale poorly to naturalistic‚ non-symbolic‚ or big data. Vector-symbolic architectures‚ where symbols are represented as vectors‚ bridge the gap between the two approaches. We posit that cognitive architectures‚ if implemented in a vector-space model‚ represent a useful‚ explanatory model of the internal representations of otherwise opaque neural architectures. Our proposed model‚ Holographic Declarative Memory (HDM)‚ is a vector-space model based on distributional semantics. HDM accounts for primacy and recency effects in free recall‚ the fan effect in recognition‚ probability judgements‚ and human performance on an iterated decision task. HDM provides a flexible‚ scalable alternative to symbolic cognitive architectures at a level of description that bridges symbolic‚ quantum‚ and neural models of cognition.

This paper examines the effects of working memory size in incremental grammatical encoding during language production. Our experiment tests different variants of a computational-cognitive model that combines an empirically validated framework of general cognition‚ ACT-R‚ with a linguistic theory‚ Combinatory Categorial Grammar. The model is induced from a corpus of spoken dialogue. This methodology facilitates comparison of different strategies and working memory capacities according to the similarity of the model’s produced sentences to the corpus sentences. The experiment presented shows that while having more working memory available improves performance‚ using less working memory during realization does as well‚ even after controlling sentence length. Sentences realized with a more incremental strategy also appear to more closely track the naturalistic data. As high incrementality is correlated with low working memory usage‚ this study offers a possible mechanism by which syntactic incrementality can be explained. Finally‚ this paper proposes a multi-disciplinary modeling and simulation-based approach to empirical psycholinguistic inquiry.

In this paper‚ we examine the origination and dispersion of neologisms from the perspective of both communities and cognitive modeling. We use the Reddit corpus to identify words that were first used by Reddit communities between 2013-2014. We induce a hierarchy on Reddit based on the specificity of the topic. Generally‚ less specific communities have more users‚ while more specific communities likely feature closer social ties. We ask whether larger numbers of people or closer social ties are better environments to foster the adoption of new words. We found that the majority of new words are first adopted/created in more general communities; though this account is relativized by the size of the communities. We also examined the pace of dispersion of such words in new communities‚ and discuss how this relates to models of memory by using an ACT-R cognitive model. We discuss some parameterizations of such a model of memory and its implications in the word adoption paradigm. Lastly‚ we show that there is an increasing trend in the number of new words being adopted/created in Reddit communities‚ even during a relatively short time period.

To advance models of multimodal context‚ we introduce a simple yet powerful neural architecture for data that combines vision and natural language. The ‘‘Bounding Boxes in Text Transformer’’ (B2T2) also leverages referential information binding words to portions of the image in a single unified architecture. B2T2 is highly effective on the Visual Commonsense Reasoning benchmark (visualcommonsense.org)‚ achieving a new state-of-the-art with a 25% relative reduction in error rate compared to published baselines and obtaining the best performance to date on the public leaderboard. A detailed ablation analysis shows that the early integration of the visual features into the text analysis is key to the effectiveness of the new architecture. A reference implementation of our models is provided as supplementary material.

Language is inseparable from its physical context‚ according to the theory of situated cognition. That is‚ children learn lexicon and grammar in the context of the objects or concepts to which words and phrases refer. Therefore‚ we examine the generalization ability of neural language models trained with a visual context. A multi-modal neural architecture is introduced that outperform its equivalent trained on language alone with a 2% decrease in perplexity‚ even when no visual context is available at test. Fine-tuning the embeddings of a pre-trained state-of-the-art bidirectional language model (BERT) in the language modeling framework yields a 3.5% improvement. The advantage for training with visual context when testing without is robust across different languages (English‚ German and Spanish) and different models (GRU‚ LSTM‚ Delta-RNN‚ as well as those that use BERT embeddings). These results provide evidence that language models perform better when they learn like a baby: in a rich environment that is not restricted to language alone.

Does the social power status of people influence how others speak towards them? Previous studies said yes‚ but they overlooked the effect low-level linguistic features. We find that after controlling for low-level features (lexical surprisal and temporal clustering)‚ the effect of power on alignment vanishes or is reversed.

This paper addresses the role of personality characteristics in decisions on the timing of an action‚ such as in the context of security and safety choices. Examples of such decisions include when to check log files for intruders and when to monitor financial accounts for fraud or errors. Two behavioral studies (n = 461) are conducted. Individual risk propensity and need for cognition are obtained via scales. The task is a game against an opaque computer opponent in which participants make decisions about the timing of actions in response to an unknown external risk factor. The task is not payoff-neutral w.r.t. risk. Difficulty is varied through the availability of explicitly given or decision-critical information‚ which is observable visually (Study 1) or in temporal memory (Study 2). Across this problem space‚ we find that risk propensity is not generally a hindrance in timing tasks. Participants of average risk propensity generally benefit from a high need for cognition‚ particularly when externalized memory is available‚ as in Study 1. In the more difficult temporal-estimation task‚ need for cognition was associated with increased payoffs from task experience. In both tasks‚ higher risk propensity in participants was associated with increased improvements in payoffs from task experience.

Linguistic alignment between dialogue partners has been claimed to be affected by their relative social power. A common finding has been that interlocutors of higher power tend to receive more alignment than those of lower power. However‚ these studies overlook some low-level linguistic features that can also affect alignment‚ which casts doubts on these findings. This work characterizes the effect of power on alignment with logistic regression models in two datasets‚ finding that the effect vanishes or is reversed after controlling for low-level features such as utterance length. Thus‚ linguistic alignment is explained better by low-level features than by social power. We argue that a wider range of factors‚ especially cognitive factors‚ need to be taken into account for future studies on observational data when social factors of language use are in question.

This paper explores the time course of lexical memory retrieval by modeling fluent language production. The duration of retrievals is predicted using the ACT-R cognitive architecture. In a large-scale observational study of a spoken corpus‚ we find that language production at a time point preceding a word is sped up or slowed down depending on activation of that word. This computational analysis has consequences for the theoretical model of language production. The results point to interference between lexical and phonological stages as well as a quantifiable buffer for lexical information that opens up the possibility of non-sequential retrievals.

On the surface‚ bi- and multilingualism would seem to be an ideal context for exploring questions of typological proximity. The obvious intuition is that the more closely related two languages are‚ the easier it should be to implement the two languages in one mind. This is the starting point adopted here‚ but we immediately run into the difficulty that the overwhelming majority of cognitive‚ computational‚ and linguistic research on bi- and multilingualism exhibits a monolingual bias (i.e.‚ where monolingual grammars are used as the standard or comparison for outputs from bilingual grammars). The primary questions so far have focused on how bilinguals balance and switch between their two languages‚ but our perspective on typology leads us to consider the nature of bi- and multi-lingual systems as a whole. Following an initial proposal from Hsin (2014)‚ we conjecture that bilingual grammars are neither isolated‚ nor (completely) conjoined with one another in the bilingual mind‚ but rather exist as integrated source grammars that are further mitigated by a common‚ combined grammar (Cook‚ 2016; Goldrick et al.‚ 2016a‚b; Putnam & Klosinski‚ to appear). Here we conceive such a combined grammar in a parallel‚ distributed‚ and gradient architecture implemented in a shared vector-space model that employs compression through routinization and dimensionality reduction. We discuss the emergence of such representations and functioning in the minds of bilinguals. This architecture aims to be consistent with empirical results on bilingual cognition and memory representations in computational cognitive architectures.

The principle of entropy rate constancy (ERC) states that language users distribute information such that words tend to be equally predictable given previous contexts. We examine the applicability of this principle to spoken dialogue‚ as previous findings primarily rest on written text. The study takes into account the joint-activity nature of dialogue and the topic shift mechanisms that are different from monologue. It examines how the information contributions from the two dialogue partners interactively evolve as the discourse develops. The increase of local sentence-level information density (predicted by ERC) is shown to apply to dialogue overall. However‚ when the different roles of interlocutors in introducing new topics are identified‚ their contribution in information content displays a new converging pattern. We draw explanations to this pattern from multiple perspectives: Casting dialogue as an information exchange system would mean that the pattern is the result of two interlocutors maintaining their own context rather than sharing one. Second‚ we present some empirical evidence that a model of Interactive Alignment may include information density to explain the effect. Third‚ we argue that building common ground is a process analogous to information convergence. Thus‚ we put forward an information-theoretic view of dialogue‚ under which some existing theories of human dialogue may eventually be unified.

Decisions on when to act are critical in many health care‚ safety and security situations‚ where acting too early or too late can both lead to huge costs or losses. In this paper‚ impatience is investigated as a bias affecting timing decisions‚ and is successfully manipulated and moderated. Experiment 1 (N = 123) shows that in different tasks with the same duration‚ participants perform better when acting early is advantageous‚ as compared to when acting late is. Experiment 2 (N = 701) manipulates impatience and shows that impatience induced by delays (a) affects timing decisions in the subsequent tasks‚ (b) increases a tendency to receive information faster‚ only for a few seconds‚ with cost and no gain‚ and (c) reduces satisfaction in the subsequent task. Furthermore‚ impatience is significantly moderated by showing fast countdowns during the delays. Experiment 3 (N = 304) shows that the mechanism behind this impatience moderation is altered time perception and presents trade-offs between duration perception and duration recall.

Blockchains are inherently secure‚ encrypted technologies‚ enabling electronic transaction processing‚ communications‚ authentication‚ and recordkeeping. Blockchain circumvents traditional (banking) systems‚ promising transaction cost savings and speculative profits. Blockchain is an open‚ encrypted and distributed ledger system underlying crypto-currencies (BitCoin) and initial coin offerings (ICO). Both masquerade as media of exchange. Future blockchain applications include supply chains‚ identity management and healthcare information exchange. Blockchains’ technical complexity obscure comprehension by many legislators‚ regulators‚ lawyers and judges. Libertarian technologists cheer these barriers to regulation. This paper explores blockchain operations‚ reviews evolving cryptocurrency policies‚ and proposes regulatory approaches‚ urging expanded jurisdiction over blockchain bubbles.

This paper investigates how cognition facilitates the adoption of new words through a study of the large-scale Reddit corpus‚ which contains written‚ threaded conversations conducted over the internet. Parameters for the cognitive architecture are estimated. Using ACT-R’s account of declarative memory‚ the activation of memory chunks representing words is traced and compared to usage statistics sampled from a year of data. Potential values for decay and retrieval threshold are identified according to model fit and growth rates of word adoption. The resulting estimate for the decay parameter‚ d‚ is 0.22‚ and the estimate for the retrieval threshold parameter‚ rt‚ lies between 3.4 and 4.5.

Learning useful information across long time lags is a critical and difficult problem for temporal neural models in tasks such as language modeling. Existing architectures that address the issue are often complex and costly to train. The Differential State Framework (DSF) is a simple and high-performing design that unifies previously introduced gated neural models. DSF models maintain longer-term memory by learning to interpolate between a fast-changing data-driven representation and a slowly changing‚ implicitly stable state. This requires hardly any more parameters than a classical‚ simple recurrent network. Within the DSF framework‚ a new architecture is presented‚ the Delta-RNN. In language modeling at the word and character levels‚ the Delta-RNN outperforms popular complex architectures‚ such as the Long Short Term Memory (LSTM) and the Gated Recurrent Unit (GRU)‚ and‚ when regularized‚ performs comparably to several state-of-the-art baselines. At the subword level‚ the Delta-RNN’s performance is comparable to that of complex gated architectures.

We propose a perspective on dialogue that focuses on relative information contributions of conversation partners as a key to successful communication. We predict the success of collaborative task in English and Danish corpora of task-oriented dialogue. Two features are extracted from the frequency domain representations of the lexical entropy series of each interlocutor‚ power spectrum overlap (PSO) and relative phase (RP). We find that PSO is a negative predictor of task success‚ while RP is a positive one. An SVM with these features significantly improved on previous task success prediction models. Our findings suggest that the strategic distribution of information density between interlocutors is relevant to task success.

This paper improves on several aspects of a sieve-based event ordering architecture‚ CAEVO (Chambers et al.‚ 2014)‚ which creates globally consistent temporal relations between events and time expressions. First‚ we examine the usage of word embeddings and semantic role features. With the incorporation of these new features‚ we demonstrate a 5% relative F1 gain over our replicated version of CAEVO. Second‚ we reformulate the architecture’s sieve-based inference algorithm as a prediction reranking method that approximately optimizes a scoring function computed using classifier precisions. Within this prediction reranking framework‚ we propose an alternative scoring function‚ showing an 8.8% relative gain over the original CAEVO. We further include an in-depth analysis of one of the main datasets that is used to evaluate temporal classifiers‚ and we show that in spite of the density of this corpus‚ there is still a danger of overfitting. While this paper focuses on temporal ordering‚ its results are applicable to other areas that use sievebased architectures.

Language processing is not an isolated capacity‚ but is embedded in other aspects of our cognition. However‚ it is still largely unexplored to what extent and how language processing interacts with general cognitive resources. This question can be investigated with cognitively constrained computational models‚ which simulate the cognitive processes involved in language processing. The theoretical claims implemented in cognitive models interact with general architectural constraints such as memory limitations. This way‚ it generates new predictions that can be tested in experiments‚ thus generating new data that can give rise to new theoretical insights. This theory-model-experiment cycle is a promising method for investigating aspects of language processing that are difficult to investigate with more traditional experimental techniques. This review specifically examines the language processing models of Lewis and Vasishth (2005)‚ Reitter‚ Keller‚ and Moore (2011)‚ and Van Rij‚ Van Rijn‚ and Hendriks (2010)‚ all implemented in the cognitive architecture ACT-R (Anderson‚ Bothell‚ Byrne‚ Douglass‚ Lebiere‚ & Qin‚ 2004). These models are all limited by the assumptions about cognitive capacities provided by the cognitive architecture‚ but use different linguistic approaches. Because of this‚ their comparison provides insight into the extent to which assumptions about general cognitive resources influence concretely implemented models of linguistic competence. For example‚ the sheer speed and accuracy of human language processing is a current challenge in the field of cognitive modeling‚ as it does not seem to adhere to the same memory and processing capacities that have been found in other cognitive processes. Architecture-based cognitive models of language processing may be able to make explicit which language-specific resources are needed to acquire and process natural language. The review sheds light on cognitively constrained models of language processing from two angles: we discuss 1) whether currently adopted cognitive assumptions meet the requirements for language processing‚ and 2) how validated cognitive architectures can constrain linguistically motivated models‚ which‚ all other things being equal‚ will increase the cognitive plausibility of these models. Overall‚ the evaluation of cognitively constrained models of language processing will allow for a better understanding of the relation between data‚ linguistic theory‚ cognitive assumptions‚ and explanation.

Linguistic alignment in text-based communication means that people tend to adjust their language use to one another both in terms of word choice and sentence structure. Previous studies about linguistic alignment suggested that these two forms of adaptation are correlated with each other‚ and that they build up to alignment at a higher representational level‚ such as pragmatic alignment for support functions. Two types of social support have been identified as important for online health communities (OHCs): emotional and informational support between support seekers and support providers. Do the two lower-level alignment measures (lexical and syntactic) relate to these two types of social support in the same way or‚ are they different? Our hypothesis was that they are similar‚ due to their correlation relationship. However‚ we found that‚ based on an analysis of a 10-year online forum for cancer survivors‚ the lower-level alignment measures have distinct relationships to the two higher-level support functions. In this paper‚ we describe this finding and its implications regarding potential refinement of the Inter- active Alignment Model.

This study examines how novel words originate in and disperse through online communities. It asks whether larger numbers of people or closer social ties are better environments to foster the adoption of new words. The data stem from Reddit‚ a large sample of web-mediated‚ asynchronous conversations. Reddit communities are divided by size in this study: larger communities are based on discussing general topics and have weak social ties‚ whereas small communities are based on discussing specific topics and have strong social ties. The analysis shows that the majority of new words are created / first adopted in larger communities.

Using corpus data of spoken dialogue‚ we examine the convergence of syntactic complexity levels between interlocutors in natural conversations‚ as it occurs within spans of topic episodes. The findings of general convergence in the Switchboard and BNC corpora are compatible with an information-theoretic model of dialogue and with Interactive Alignment Theory.

The applicability of entropy rate constancy to dialogue is examined on two spoken dialogue corpora. The principle is found to hold; however‚ new entropy change patterns within the topic episodes of dialogue are described‚ which are different from written text. Speaker’s dynamic roles as topic initiators and topic responders are associated with decreasing and increasing entropy‚ respectively‚ which results in local convergence between these speakers in each topic episode. This implies that the sentence entropy in dialogue is conditioned on different contexts determined by the speaker’s roles. Explanations from the perspectives of grounding theory and interactive alignment are discussed‚ resulting in a novel‚ unified information-theoretic approach of dialogue.

Countdowns and progress bars provide computer users with estimates of remaining wait times. These types of feedback are intended to manage their expectations and allow users to direct attention elsewhere. We suggest that they also moderate user’s impatience‚ which affects decision-making in the subsequent task. In an experiment with 421 participants‚ impatience in a timing decision task was effectively and systematically manip- ulated through a countdown‚ as it affected timing and performance of the user’s actions in the task. The effect persisted even after users gained task experience. More rapid countdowns reduced impatience. Post-hoc analysis also showed increased task satisfaction with rising countdown speed and suggested greater task satisfaction with a rapid countdown than with no waiting period at all.

Much of the data used to measure conflict is extracted from news reports. This is typically accomplished using either expert coders to quantify the relevant information or machine coders to automatically extract data from documents. Although expert coding is costly‚ it produces quality data. Machine coding is fast and inexpensive‚ but the data are noisy. To diminish the severity of this tradeoff‚ we introduce a method for analyzing news documents that uses crowdsourcing‚ supplemented with computational approaches. The new method is tested on documents about Militarized Interstate Disputes‚ and its accuracy ranges between about 68 and 76 percent. This is shown to be a considerable improvement over automated coding‚ and to cost less and be much faster than expert coding.

Armed conflicts around the world produce displacement‚ in- jury‚ and death. This study examines how anonymous and pseudony- mous Internet commenters discuss such conflicts. Specifically‚ we ask how permissible it is to express positive or negative sentiments about these conflicts as a function of variables including region‚ conflict nature‚ and severity. Data from the Armed Conflicts Database is aggregated to identify a number of potential factors that may influence views on ac- ceptable sentiments. A large-scale sample of the Reddit comment forum was coded for positive and negative sentiments using sentiment analy- sis techniques. Permissibility is judged using the Reddit voting features. Regressions reveal that positive sentiments are found not permissible for higher numbers of fatalities‚ and that negative sentiments are found permissible for certain regions‚ more permissible for older conflicts‚ and less permissible for territorial conflicts. A number of alternative‚ non- correlated predictors are presented.

Most literature on symmetry perception has focused on bilateral reflection symmetry‚ with some suggesting that it is the only type of symmetry humans can perceive (Wilson&Wilkinson‚ Vis. Res. 42(5)‚ 2002). Using image-stimuli generated from mathematically well-defined 17 wallpaper groups (Kohler et al.‚ J. Neurosci.‚ in press) and timed trials‚ we seek to demonstrate that humans can discriminate various symmetries found in 2D wallpaper patterns (Liu‚ et al Found.&Trends in Comp.Graph.&Vis.‚ 2010). Furthermore‚ we examine which features play an essential role in wallpaper pattern perception. The features include: reflection‚ maximum order of rotation‚ glide reflection‚ tile shape and subgroup distance‚ referring to the shortest path between the two wallpaper groups in the group hierarchy. We recruited 106 individuals (Amazon Mechanical Turks) to compare three images (target‚ probe‚ probe) and to choose the probe most similar to the target within five seconds. Each participant performed 272 (=17x16) trials. Participants compare among all possible wallpaper group pairs. Every group comparison except one is distinguishable (p< 0.05); all are likely distinguishable (p< 0.10). After modeling every defining feature of symmetry groups using linear mixed-effects regression (GLMM)‚ we used AIC selection to compare models. Our selected model includs subgroup distance‚ 3-fold rotation‚ 4-fold rotation‚ and the T1 and D1 axes. Even in a model with all other features‚ subgroup distance is a significant additional predictor (p< 0.0001). This distance measure has shown to be a better proxy describing human perceptual processes than individual features. This suggests that the mathematically motivated symmetry group hierarchy may be a valid model of pattern perception.

We present a novel fine-tuning algorithm in a deep hybrid architecture for semi-supervised text classification. During each increment of the online learning process‚ the fine-tuning algorithm serves as a top-down mechanism for pseudo-jointly modifying model parameters following a bottom-up generative learning pass. The resulting model‚ trained under what we call the Bottom-Up-Top-Down learning algorithm‚ is shown to outperform a variety of competitive models and baselines trained across a wide range of splits between supervised and unsupervised training data.

A hybrid architecture is presented capable of online learning from both labeled and unlabeled samples. It combines both generative and discriminative objectives to derive a new variant of the Deep Belief Network‚ i.e.‚ the Stacked Boltzmann Experts Network model. The model’s training algorithm is built on principles developed from hybrid discriminative Boltzmann machines and composes deep architectures in a greedy fashion. It makes use of its inherent ‘‘layer-wise ensemble" nature to perform useful classification work. We (1) compare this architecture against a hybrid Denoising Auto-encoder version of itself as well as several other models and (2) investigate training in the context of an incremental learning procedure. The best-performing hybrid model‚ the Stacked Boltzmann Experts Network‚ consistently outperforms all others.

A discriminative model is presented for crowd-sourcing the annotation of news stories to produce a structured dataset about incidents involving militarized disputes between nation-states. We used a question tree to gather partially redundant data from each crowd worker. A lattice of Bayesian Networks was then applied to error correct the individual worker annotations‚ the results of which were then aggregated via majority voting. The resulting hybrid model outperformed comparable‚ state-of-the-art aggregation models in both accuracy and computational scalability.

We report a way to build a series of GOMS-like cognitive user models representing a range of performance at different stages of learning. We use a spreadsheet task across multiple sessions as an example task; it takes about 20–30 min. to perform. The models were created in ACT-R using a compiler. The novice model has 29 rules and 1‚152 declarative memory task elements (chunks)—it learns to create procedural knowledge to perform the task. The expert model has 617 rules and 614 task chunks (that it does not use) and 538 command string chunks—it gets slightly faster through limited declarative learning of the command strings and some further production compilation; there are a range of intermediate models. These models were tested against aggregate and individual human learning data‚ confirming the models’ predictions. This work suggests that user models can be created that learn like users while doing the task.

Task-solving in dialogue depends on the convergence of the situation models held by the dialogue partners. The Interactive Alignment Model (Pickering and Garrod‚ 2004) suggests that this convergence is the result of an interactive alignment process‚ which is based on mechanistic repetition at a number of linguistic levels. In this paper‚ we develop two predictions arising from the theory‚ along with two methods to quantify the known structural priming effects in the full inventory of syntactic choices found in text and speech corpora. (a) Under a rational perspective‚ we expect increased repetition in task-oriented dialogue compared to spontaneous conversation. We find within- and between-speaker priming in a corpus of spontaneous conversations‚ but stronger priming in task-oriented dialogue. (b) The Interactive Alignment Model predicts linguistic adaptation to be correlated with task success. We show this effect in a corpus of task-oriented dialogue‚ where we find a positive correlation of long-term adaptation and a quantifiable task success measure. We argue that the repetition tendency relevant for the high-level alignment of situation models is based on slow adaptation rather than short-term priming. We demonstrate that lexical and syntactic repetition are reliable and computationally exploitable predictors of task success.

This paper addresses security and safety choices that involve a decision on the timing of an action. Examples of such decisions include when to check log files for intruders and when to monitor financial accounts for fraud or errors. To better understand how performance in timing-related security situations is shaped by individuals’ cognitive predispositions‚ we effectively combine survey measures with economic experiments. Two behavioral experiments are presented in which the timing of online security actions is the critical decision-making factor. The feedback modality in the decision-environment is varied between visual feedback with history (Experiment 1)‚ and temporal feedback without history (Experiment 2). Using psychometric scales‚ we study the role of individual difference variables‚ specifically risk propensity and need for cognition. The analysis is based on the data from over 450 participants. We find that risk propensity is not a hindrance in timing tasks. Participants of average risk propensity generally benefit from a reflective disposition (high need for cognition)‚ particularly when visual feedback is given. Overall‚ participants benefit from need for cognition; however‚ in the more difficult‚ temporal-estimation task‚ this requires familiarity with the task.

Two information decay methods are examined that help multi-agent systems cope with dynamic environments. The agents in this simulation have human-like memory and a mechanism to moderate their communications: they forget internally stored information via temporal decay‚ and they forget distributed information by filtering it as it passes through a communication network. The agents play a foraging game‚ in which performance depends on communicating facts and requests and on storing facts in internal memory. Parameters of the game and agent models are tuned to human data. Agent groups with moderated communication in small-world networks achieve optimal performance for typical human memory decay values‚ while non-adaptive agents benefit from stronger memory decay. The decay and filtering strategies interact with the properties of the network graph in ways suggestive of an evolutionary co-optimization between the human cognitive system and an external social structure.

The psycholinguistic literature has identified two syntactic adaptation effects in language production: rapidly decaying short-term priming and long-lasting adaptation. To explain both effects‚ we present an ACT-R model of syntactic priming based on a wide-coverage‚ lexicalized syntactic theory that explains priming as facilitation of lexical access. In this model‚ two well-established ACT-R mechanisms‚ base-level learning and spreading activation‚ account for long-term adaptation and short-term priming‚ respectively. Our model simulates incremental language production and in a series of modeling studies we show that it accounts for (a) the inverse frequency interaction; (b) the absence of a decay in long-term priming; and (c) the cumulativity of long-term adaptation. The model also explains the lexical boost effect and the fact that it only applies to short-term priming. We also present corpus data that verifies a prediction of the model‚ i.e.‚ that the lexical boost affects all lexical material‚ rather than just heads.

We simulate the evolution of a domain vocabulary in small communities. Empirical data show that human communicators can evolve graphical languages quickly in a constrained task (Pictionary)‚ and that communities converge towards a common language. We propose that simulations of such cultural evolution incorporate properties of human memory (cue-based retrieval‚ learning‚ decay). A cognitive model is described that encodes abstract concepts with small sets of concrete‚ related concepts (directing)‚ and that also decodes such signs (matching). Learning captures conventionalized signs. Relatedness of concepts is characterized by a mixture of shared and individual knowledge‚ which we sample from a text corpus. Simulations show vocabulary convergence of agent communities of varied structure‚ but idiosyncrasy in vocabularies of each dyad of models. Convergence is weakened when agents do not alternate between encoding and decoding‚ predicting the necessity of bi-directional communication. Convergence is improved by explicit feedback about communicative success. We hypothesize that humans seek out subtle clues to gauge success in order to guide their vocabulary acquisition.

We present a cognitive model performing the Dynamic Stocks&Flows control task‚ in which subjects control a system by counteracting a systematically changing external variable. The model uses a metacognitive layer that chooses a task strategy drawn from of two classes of strategies: precise calculation and imprecise estimation. The model‚ formulated within the ACT-R theory‚ monitors the success of each strategy continuously using instance-based learning and blended retrieval from declarative memory. The model underspecifies other portions of the task strategies‚ whose timing was determined as unbiased estimate from empirical data. The model’s predictions were evaluated on data collected from novel experimental conditions‚ which did not inform the model’s development and included discontinuous and noisy environmental change functions and a control delay. The model as well as the data show sudden changes in subject error and general learning of control; the model also correctly predicted oscillations of plausible magnitude. With its predictions‚ the model ranked first among the entries to the 2009 Dynamic Stocks&Flows modeling challenge.

The cognitive modeling and artificial general intelligence research communities may reap greater scientific return on research investments – may achieve an improved understanding of architectures and models – if there is more emphasis on systematic sensitivity and necessity analyses during model development‚ evaluation‚ and comparison. We demonstrate this methodological prescription with two of the models submitted for the Dynamic Stocks and Flows (DSF) Model Comparison Challenge‚ exploring the complex interactions among architectural mechanisms‚ knowledge-level strategy variants‚ and task conditions. To cope with the computational demands of these analyses we use a predictive analytics approach similar to regression trees‚ combined with parallelization on high performance computing clusters‚ to enable large scale‚ simultaneous search and exploration.

Planning a path to a destination‚ given a number of options and obstacles‚ is a common task. We suggest a two-component cognitive model that combines retrieval of knowledge about the environment with search guided by visual perception. In the first component‚ subsymbolic information‚ acquired during navigation‚ aids in the retrieval of declarative information representing possible paths to take. In the second component‚ visual information directs the search‚ which in turn creates knowledge for the first component. The model is implemented using the ACT-R cognitive architecture and makes realistic assumptions about memory access and shifts in visual attention. We present simulation results for memory-based high-level navigation in grid and tree structures‚ and visual navigation in mazes‚ varying relevant cognitive (retrieval noise and visual finsts) and environmental (maze and path size) parameters. The visual component is evaluated with data from a multi-robot control experiment‚ where subjects planned paths for robots to explore a building. We describe a method to compare trajectories without referring to aligned points in the itinerary. The evaluation shows that the model provides a good fit‚ but also that planning strategies may differ between task loads.

Task-solving in dialogue depends on the linguistic alignment of the interlocutors‚ which Pickering & Garrod (2004) have suggested to be based on mechanistic repetition effects. In this paper‚ we seek confirmation of this hypothesis by looking at repetition in corpora‚ and whether repetition is correlated with task success. We show that the relevant repetition tendency is based on slow adaptation rather than short-term priming and demonstrate that lexical and syntactic repetition is a reliable predictor of task success given the first five minutes of a task-oriented dialogue.

This paper presents a corpus-based account of structural priming in human sentence processing‚ focusing on the role that syntactic representations play in such an account. We estimate the strength of structural priming effects from a corpus of spontaneous spoken dialogue‚ annotated syntactically with Combinatory Categorial Grammar (CCG) derivations. This methodology allows us to test a range of predictions that CCG makes about priming. In particular‚ we present evidence for priming between lexical and syntactic categories encoding partially satisfied subcategorization frames‚ and we show that priming effects exist both for incremental and normal-form CCG derivations.

Syntactic priming effects‚ modeled as increase in repetition probability shortly after a use of a syntactic rule‚ have the potential to improve language processing components. We model priming of syntactic rules in annotated corpora of spoken dialogue‚ extending previous work that was confined to selected constructions. We find that speakers are more receptive to priming from their interlocutor in task-oriented dialogue than in sponaneous conversation. Low-frequency rules are more likely to show priming.

We present the Embra system‚ a first-time entry to DUC for 2005 which performed at or above median for the manual assessment of responsiveness and on 4 out of 5 linguistic quality questions. The system takes a novel approach to relevance and redundancy‚ modeling sentence similarity using a latent semantic space constructed over a very large corpus. We present a simple approach to modeling specificity based on named entities which shows a small improvement over baseline. Finally‚ we discuss coherence and present a sentence reordering algorithm with a component-level evaluation demonstrating a positive effect.

UI on the Fly is a system that dynamically presents coordinated multimodal content through natural language and a small-screen graphical user interface. It adapts to the user’s preferences and situation. Multimodal Functional Unification Grammar (MUG) is a unification-based formalism that uses rules to generate content that is coordinated across several communication modes. Faithful variants are scored with a heuristic function.

Most text displays an internal coherence structure‚ which can be analyzed as a tree structure of relations that hold between short segments of text. We present a machine-learning governed approach to such an analysis in the framework of Rhetorical Structure Theory. Our rhetorical analyzer observes a variety of textual properties‚ such as cue phrases‚ part-of-speech information‚ rhetorical context and lexical chaining. A two-stage parsing algorithm uses local and global optimization to find an analysis. Decisions during parsing are driven by an ensemble of support vector classifiers. This training method allows for a non-linear separation of samples with many relevant features. We define a chain of annotation tools that profits from a new underspecified representation of rhetorical structure. Classifiers are trained on a newly introduced German language corpus‚ as well as on a large English one. We present evaluation data for the recognition of rhetorical relations.

We demonstrate that the key components of cognitive architectures—declarative and procedural memory—and their key capabilities—learning‚ memory retrieval‚ probability judgement‚ and utility estimation—can be implemented as algebraic operations on vectors and tensors in a high-dimensional space using a distributional semantics model. High-dimensional vector spaces underlie the success of modern machine learning techniques based on deep learning. However‚ while neural networks have an impressive ability to process data to find patterns‚ they do not typically model high-level cognition‚ and it is often unclear how they work. Symbolic cognitive architectures can capture the complexities of high-level cognition and provide human-readable‚ explainable models‚ but scale poorly to naturalistic‚ non-symbolic‚ or big data. Vector-symbolic architectures‚ where symbols are represented as vectors‚ bridge the gap between the two approaches. We posit that cognitive architectures‚ if implemented in a vector-space model‚ represent a useful‚ explanatory model of the internal representations of otherwise opaque neural architectures. Our proposed model‚ Holographic Declarative Memory (HDM)‚ is a vector-space model based on distributional semantics. HDM accounts for primacy and recency effects in free recall‚ the fan effect in recognition‚ probability judgements‚ and human performance on an iterated decision task. HDM provides a flexible‚ scalable alternative to symbolic cognitive architectures at a level of description that bridges symbolic‚ quantum‚ and neural models of cognition.

AI researchers have posited Dungeons and Dragons (D&D) as a challenge problem to test systems on various language-related capabilities. In this paper‚ we frame D&D specifically as a dialogue system challenge‚ where the tasks are to both generate the next conversational turn in the game and predict the state of the game given the dialogue history. We create a gameplay dataset consisting of nearly 900 games‚ with a total of 7‚000 players‚ 800‚000 dialogue turns‚ 500‚000 dice rolls‚ and 58 million words. We automatically annotate the data with partial state information about the game play. We train a large language model to generate the next game turn‚ conditioning it on different information. The LM can respond as a particular character or as the player who runs the game—i.e.‚ the Dungeon Master (DM). It is trained to produce dialogue that is either in-character (roleplaying in the fictional world) or out-of-character (discussing rules or strategy). We perform a human evaluation to determine what factors make the generated output plausible and interesting. We further perform an automatic evaluation to determine how well the model can predict the game state given the history and examine how well tracking the game state improves its ability to produce plausible conversational output.

We investigated how the language background (L1) of bilinguals influences the representation and use of the second language (L2) through computational models. With the essays part from The International Corpus Network of Asian Learners of English (ICNALE)‚ we compared variables indicating syntactic complexity in their L2 production to predict L1. We then trained neural language models based on BERT to predict the L1 of these English learners. Results showed the systematic influence of L1 syntax properties on English learners’ L2 production‚ which further confirmed integrations of syntactic knowledge across languages in bilingual speakers. Results also found neural models can learn to represent and detect such L1 impacts‚ while multilingually trained models have no advantage in doing so.

With the essays part from The International Corpus Network of Asian Learners of English (ICNALE) and the TOEFL11 corpus‚ we finetuned neural language models based on BERT to predict English learners’ native languages. Results showed neural models can learn to represent and detect such native language impacts‚ but multilingually trained models have no advantage in doing so.

Native speakers can judge whether a sentence is an acceptable instance of their language. Acceptability provides a means of evaluating whether computational language models are processing language in a human-like manner. We test the ability of language models‚ simple language features‚ and word embeddings to predict native speakers’ judgments of acceptability on English essays written by non-native speakers. We find that much sentence acceptability variance can be captured by a combination of misspellings‚ word order‚ and word similarity (r = 0.494). While predictive neural models fit acceptability judgments well (r = 0.527)‚ we find that a 4-gram model is just as good (r = 0.528). Thanks to incorporating misspellings‚ our 4-gram model surpasses both the previous unsupervised state-of-the art (r = 0.472)‚ and the average native speaker (r = 0.46)‚ demonstrating that acceptability is well captured by n-gram statistics and simple language features.

Recent models of language have eliminated syntactic-semantic dividing lines. We explore the psycholinguistic implications of this development by comparing different types of sentence embeddings in their ability to encode syntactic constructions. Our study uses contrasting sentence structures known to cause syntactic priming effects‚ that is‚ the tendency in humans to repeat sentence structures after recent exposure. We compare how syntactic alternatives are captured by sentence embeddings produced by a neural language model (BERT) or by the composition of word embeddings (BEAGLE‚ HHM‚ GloVe). Dative double object vs. prepositional object and active vs. passive sentences are separable in the high-dimensional space of the sentence embeddings and can be classified with a high degree of accuracy. The results lend empirical support to the modern‚ computational‚ integrated accounts of semantics and syntax‚ and they shed light on the information stored at different layers in deep language models such as BERT.

We use a variant of word embedding model that incorporates subword information to characterize the degree of compositionality in lexical semantics. Our models reveal some interesting yet contrastive patterns of long-term change in multiple languages: Indo-European languages put more weight on subword units in newer words‚ while conversely Chinese puts less weights on the subwords‚ but more weight on the word as a whole. Our method provides novel evidence and methodology that enriches existing theories in evolutionary linguistics. The resulting word vectors also has decent performance in NLP-related tasks.

What role does the study of natural language play in the task of developing a unified theory and common model of cognition? Language is perhaps the most complex behaviour that humans exhibit‚ and‚ as such‚ is one of the most difficult problems for understanding human cognition. Linguistic theory can both inform and be informed by unified models of cognition. We discuss (1) how computational models of human cognition can provide insight into how humans produce and comprehend language and (2) how the problem of modelling language processing raises questions and creates challenges for widely used computational models of cognition. Evidence from the literature suggests that behavioural phenomena‚ such as recency and priming effects‚ and cognitive constraints‚ such as working memory limits‚ affect how language is produced by humans in ways that can be predicted by computational cognitive models. But just as computational models can provide new insights into language‚ language can serve as a test for these models. For example‚ simulating language learning requires the use of more powerful machine learning techniques‚ such as deep learning and vector symbolic architectures‚ and language comprehension requires a capacity for on-the-fly situational model construction. In sum‚ language plays an important role in both shaping the development of a common model of the mind‚ and‚ in turn‚ the theoretical understanding of language stands to benefit greatly from the development of a common model.

Computational models of distributional semantics can analyze a corpus to derive representations of word meanings in terms of each word’s relationship to all other words in the corpus. While these models are sensitive to topic (e.g.‚ tiger and stripes) and synonymy (e.g.‚ soar and fly)‚ the models have limited sensitivity to part of speech (e.g.‚ book and shirt are both nouns). By augmenting a holographic model of semantic memory with additional levels of representations‚ we present evidence that sensitivity to syntax is supported by exploiting associations between words at varying degrees of separation. We find that sensitivity to associations at three degrees of separation reinforces the relationships between words that share part-of-speech and improves the ability of the model to construct grammatical sentences. Our model provides evidence that semantics and syntax exist on a continuum and emerge from a unitary cognitive system.

We examine working memory use and incrementality using a cognitive model of grammatical encoding. Our model com- bines an empirically validated framework‚ ACT-R‚ with a lin- guistic theory‚ Combinatory Categorial Grammar‚ to target that phase of language production. By building the model with the Switchboard corpus‚ it can attempt to realize a larger set of sentences. With this methodology‚ different strategies may be compared according to the similarity of the model’s sentences to the test sentences. In this way‚ the model can still be evalu- ated by its fit to human data‚ without overfitting to individual experiments. The results show that while having more work- ing memory available improves performance‚ using less work- ing memory during realization is correlated with a closer fit‚ even after controlling for sentence complexity. Further‚ sen- tences realized with a more incremental strategy are also more similar to the corpus sentences as measured by edit distance. As high incrementality is correlated with low working mem- ory usage‚ this study offers a possible mechanism by which incrementality can be explained.

We explore replacing the declarative memory system of the ACT-R cognitive architecture with a distributional semantics model. ACT-R is a widely used cognitive architecture‚ but scales poorly to big data applications and lacks a robust model for learning association strengths between stimuli. Distribu- tional semantics models can process millions of data points to infer semantic similarities from language data or to infer product recommendations from patterns of user preferences. We demonstrate that a distributional semantics model can account for the primacy and recency effects in free recall‚ the fan effect in recognition‚ and human performance on iterated decisions with initially unknown payoffs. The model we propose provides a flexible‚ scalable alternative to ACT-R’s declarative memory at a level of description that bridges symbolic‚ quantum‚ and neural models of cognition. Our intent is to advance toward a cognitive architecture capable of modeling human performance at all scales of learning.

Decisions over the timing of actions are critical in several safety‚ security and healthcare scenarios. These decisions‚ sim- ilar to discrete decisions‚ can be influenced by biases and individual traits. In this paper‚ a bias of impatience is studied in an experiment with 626 participants‚ with a focus on gender differences. Impatience was moderated with a manipula- tion of a variable-speed countdown. Men and women differed in how they expressed impatience. While men systematically and irrationally act earlier when become impatient following the slower countdowns‚ women react by irrationally requesting earlier information about the outcome of each trial‚ and impulsively pressing an inactive key.

Linguistic alignment has emerged as an important property of conversational language and a driver of mutual understanding in dialogue. While various computational measures of linguistic alignment in corpus and experimental data have been devised‚ a systematic evaluation of them is missing. In this study‚ we first evaluate the sensitivity and distributional properties of three measures‚ indiscriminate local linguistic alignment (LLA)‚ Spearman’s correlation coefficient (SCC)‚ and repetition decay (RepDecay). Then we apply them in a study of interactive alignment and individual differences to see how well they conform to the Interactive Alignment Model (IAM)‚ and how well they can reveal the individual differences in alignment propensity. Our results suggest that LLA has the overall best performance.

Linguistic alignment‚ such as lexical and syntactic alignment‚ is a universal phenomenon influencing dialogue participants in online conversations. Besides linguistic alignment‚ social support‚ such as emotional support and informational support‚ also provides important benefits to their members. More recently‚ identifying and analyzing social support facilitated insights in the construction of support oriented communities. In this paper‚ we study whether members show pragmatic alignment on social support. While adaptation can occur at lexical‚ syntactic and pragmatic levels‚ relationships between alignments at multiple levels are neither theoretically nor empirically well understood. Our results indicate pragmatic alignment of forum members along the axis of support type. We also find that lexical alignment is correlated with emotional support‚ and that the amount of lexical alignment is also correlated with the amount of pragmatic alignment. This finding can contribute to improving our understanding about the linguistic signature of different types of support‚ and enhancing theoretical model of the Interactive Alignment Model in a multi-party peer support conversation context.

Previous studies of alignment have focused on two-party conversations. We study multi-party conversation in online health communities‚ which have shown benefits for their members from forum conversations. So far‚ our understanding of the relationship between alignment in such multi-party conversations and its possible connection to member benefits has been limited. This paper quantifies linguistic alignment in the oldest and the largest cancer online forum. Alignment at lexical and syntactic levels‚ as well as decay of alignment was observed in forum threads‚ although the decay was slower than commonly found in psycholinguistic studies. The different pattern of adaptation to the initial post on a thread suggests that specific roles in the online forum (e.g.‚ seeking support from the community) can potentially be revealed through alignment theory and its extensions.

Games of timing reflect dynamic decision-making under uncertainty‚ as it takes place in many real-world situations‚ including health care‚ safety and security. Rather than making discrete decisions‚ participants choose one or more points in time that determine the outcome. We study individual’s biases and characteristics in such games of timing. We examine risk propensity as a personal preference affecting timing decisions and document a bias‚ impatience. Experiment 1 analyzes people’s strategy in timing games in relation to a rational model. Contrasting two cognitive models suggests that individuals apply risk propensity to the probability distributions underlying short games and when unfamiliar with the situation‚ but that‚ over time‚ impatience takes over as a linear adjustment. In Experiment 2‚ impatient participants risk their incentive payment in order to play early‚ even if they receive no advantage from doing so.

Continuous games of timing represent an ecologically relevant case of decision-making under uncertainty with many applications to tactical security scenarios. We present two experiments collecting data using the two-player‚ 20-second FlipIt game‚ in which a player has to react to an estimated‚ externally controlled quantity in light of a continuous payoff distribution. The quantity is represented once as time interval‚ and once as distance in time and space. Risk-taking propensity is an individual parameter obtained via a standardized survey instrument. We find that risk propensity affects participants’ decisions in early and late rounds of the game and show that risk-seeking individuals fail to play well initially in the difficult‚ time-interval estimation task‚ but that they can be trained to make rational choices. Participants’ biases are gaussian-distributed. They do not reflect the distribution of payoffs or expected utility. We present an ACT-R model using established error distributions for temporal and visual perception and instance-based learning.

In a congestion game‚ individuals exhaust a common resource out of selfish behavior. In this scenario‚ drivers create traffic jams by choosing the shortest route according to their individual knowledge. They can avoid them by communicating their belief states about the traffic situation in real-time through a peer-to-peer network‚ assuming unlimited bandwidth. We study two potential‚ cognitively inspired models of human behavior: 1) categorization (quantized memorization and communication)‚ which dampens communication and belief adoption‚ but leads to undesired oscillations and lower performance. 2) Instance-based blending with memory decay‚ which achieves good dynamics and near-optimal performance without the same bandwidth needs. We argue that this supports our hypothesis of co-adaptation of cognitive function and communicating communities.

The development and refinement of natural-language communication systems among networked individuals is not well understood and difficult to study. This paper uses a task providing a controlled environment for the goal-oriented‚ collaborative exchange of short‚ natural-language messages between experimental participants (20 per group) in order to demonstrate lexical convergence. A technique for illustrating convergence based on graph layout by multidimensional scaling is described. While reliable convergence is shown‚ it is limited to the collaborative or communicative situation: participants did not adopt group terms in a separately administered post-test questionnaire.

In multimodal human computer interaction users can often select between specific input modalities. Modality choice is influenced by various factors including user attributes‚ system attributes‚ the task and the environment (e.g. Lemmelä et al.‚ 2008). Here‚ we describe on-going research into cognitive models of input modality selection.

In an experiment involving teams of humans playing a cooperative game‚ we study the effect of local communication policies on the efficiency and the performance of teams and of individuals in different positions within a network. This design provides an experimental model of human communities‚ where information may spread from peer to peer by word of mouth. With this model‚ we explain the realistic tradeoff between liberal information dumping and targeted information sharing by human peers. Human subjects exchanged natural-language messages with relevance to a task‚ thereby sharing knowledge across a community. Communication took place along the edges of a small-world graph. Cooperation and individual efforts were incentivized. In one condition‚ participants were asked to request specific information and only supply information that they knew was needed. In another condition‚ they were asked to supply and forward as much information as possible. We found that a targeted communication policy was successfully implemented by the participants‚ increased task success‚ shortened the time it took to get answers to questions‚ increased efficiency (task success per communication bandwidth)‚ and may have done so selectively for nodes with fewer connections.

In social tasks‚ communication and task execution are often co-dependent‚ yet they represent a tradeoff. We present a cognitive model of an experimental task consisting of a collaborative and competitive game played by groups of human participants organized in a smallworld graph. In an experiment involving teams of humans playing a cooperative game‚ the effect of local communication policies on the efficiency and the performance of networked participants was observed. The model follows the ACT-R theory and provides a formalization of the decision-making processes and tradeoffs involved. A simulation of the hypothetical case of unnatural memory decay shows decreased performance and supports a prediction of the thesis that memory limitations have co-evolved with social structure.

The Lemonade Game is a three-player game in which players have to pick locations on a circular board‚ which are as far away as possible from those chosen independently by other players. Players may observe other player’s moves and infer their strategies. The game was studied using a competition of cognitively motivated agents‚ which inherit properties of adaptivity and stochasticity from human memory and decision-making‚ and simplistic‚ yet effective agents implementing fixed strategies. We argue that metacognition is the unique attribute that allows sophisticated agents to adapt to unforeseen conditions‚ cooperators and competitors.

Natural language as well as other commu- nication forms are constrained by cognitive function and evolved through a social process. Here‚ we examine whether hu- man memory may be uniquely adapted to the social structures prevalent in groups‚ specifically small-world networks. The emergence of domain languages is simulated using an empirically evaluated ACT-R-based cognitive model of agents in a naming game played within communities. Several community structures are examined (grids‚ trees‚ random graphs and small-world networks). We present pre- liminary results from small-scale simulations‚ showing relative robustness of cognitive models to network structure.

ACT-UP is a toolbox implementation of the ACT-R cognitive architecture‚ aimed at allowing rapid prototyping of complex models. With ACT-UP‚ we propose Accountable Modeling‚ where the only model components that are specified are those supported by empirical evidence and part of the model’s theoretical claims. ACT-UP is a library providing a programmatic interface to the classical ACT-R functionality. Implemented in a functional programming paradigm‚ models are reusable in other contexts. The toolbox is demonstrated using five implemented and evaluated cognitive models.

We simulate the evolution of a domain language in small speaker communities. Data from experiments (Garrod et al.‚ 2007; Fay et al.‚ 2008) show that human communicators can evolve graphical languages quickly in a constrained task (Pictionary)‚ and that communities converge towards a common language even in the absence of feedback about the success of each communication. We postulate that simulations of such horizontal evolution have to take into account properties of human memory (cue-based retrieval‚ learning‚ decay). We implement a model that can draw abstract concepts through sets of non-abstract‚ related concepts‚ and recognize such drawings. The knowledge base is a network with association strengths randomly sampled from a natural distribution found in a text corpus; it is a mixture of knowledge shared between agents and individual knowledge. In three experiments‚ we show that the agent communities converge‚ but that initial convergence is stronger when communities are structured so that the same pairs of agents interact throughout. Convergence is weaker in communities when agents do not swap roles (between recognizing and drawing)‚ predicting the necessity of bi-directional communication in domain language evolution. Average and ultimate recognition performance depends on how much of the knowledge agents share initially.

Planning a path to a destination‚ given a number of options and obstacles‚ is a common task. We suggest a two-component cognitive model that combines retrieval of knowledge about the environment with search guided by visual perception. In the first component‚ subsymbolic information‚ acquired during navigation‚ aids in the retrieval of declarative information representing possible paths to take. In the second component‚ visual information directs the search‚ which in turn creates knowledge for the first component. The model is implemented using the ACT-R cognitive architecture and makes realistic assumptions about memory access and shifts in visual attention. We present simulation results for memory-based high-level navigation in grid and tree structures‚ and visual navigation in mazes‚ varying relevant cognitive (retrieval noise‚ visual finsts) and environmental (maze and path size) parameters.

We discuss an experiment involving visual path planning for multiple‚ remote robots in a partially visible building‚ with a partial 2D map available. Participants in the experiment defined waypoints for each robot to circumnavigate obstacles and explore the building. A cognitively plausible model of visual planning is evaluated using a normalized metric of the fit between model and subject itineraries. We discuss variation in the data and model fit‚ indicating individual differences in strategies to cope with task demands.

Structural priming‚ i.e.‚ the tendency to repeat linguistic material‚ can be explained by two alternative representational assumptions: either as the repetition of hierarchical representations generated by syntactic rules‚ or as the repetition of lexical sequences. We present two studies that test these explanations by investigating priming effects in a dialogue corpus. We compare syntactic constituents with distituents‚ i.e.‚ part-of-speech pairs that cross constituent boundaries. We find a reliable short-term priming effect for constituents‚ but no priming for distituents. This result supports the rule-based view of priming‚ which does not predict priming of distituents. The data are incompatible with a sequence priming analysis‚ which cannot distinguish between constituents and distituents. In a second corpus study‚ we study long-term priming and find priming effects for both constituents and distituents. This indicates that the mechanism underlying long-term adaptation differs substantially from short-term priming.

We describe results from analyzing a corpus of human-human tutorial dialogue‚ which are aimed at identifying‚ formalizing and automating natural tutoring in the domain of calculus‚ in particular symbolic differentiation. We analyzed a corpus of human-human tutoring dialogues‚ focusing on three areas important to system development: interleaved symbolic and natural language‚ domain modeling‚ and tutorial dialogue strategies. We provide empirical verification of previous results on interleaving natural and symbolic language‚ and show that the properties of interleaving are highly dependent on the input modality. We describe a task model for our domain‚ and provide corpus data to show that the model must cover basic algebra skills as well‚ which are involved in differentiation. We verify the applicability of an existing annotation scheme for tutoring algebra in our domain‚ and propose that it be extended to cover student initiative.

A key task in an extraction system for query-oriented multi-document summarisation‚ necessary for computing relevance and redundancy‚ is modelling text semantics. In the Embra system‚ we use a representation derived from the singular value decomposition of a term co-occurrence matrix. We present methods to show the reliability of performance improvements. We find that Embra performs better with dimensionality reduction.

Previous work provided corpus evidence for structural priming for specific syntactic constructions. The present paper extends these results by investigating priming effects involving arbitrary syntactic rules in spoken dialogue corpora. We demonstrate the existence of within- and between-speaker priming in both spontaneous conversation (the Switchboard corpus) and task-oriented dialogue (the Map Task corpus). We also find that between-speaker priming is stronger in the Map Task corpus. This supports the hypothesis that in task-oriented dialog‚ low-level priming is linked to higher-level alignment of situation models.

In the context of the FASiL project‚ we have studied natural language interactions in a unimodal (speech only) and multimodal (speech and graphics) interface to a personal information management database. We collected multilingual corpora to investigate these interactions in three languages: Portuguese‚ English and Swedish. The corpora are used to train language models‚ to update acoustic models‚ to study semantic concepts‚ multimodal interactions‚ and dialogue management strategies. The corpora are annotated in a uniform way‚ with timings‚ transcriptions‚ and semantics. In this paper‚ we report on the structure and design of the corpora which are now available via ELRA.

Adaptable multimodal systems are difficult to test. We present a methodology for evaluating parallel multimodal output which is generated in response to a specific set of user‚ device and situation constraints. Our method involves ranking of many potential output variants using a fitness function‚ and selecting well-differentiated variants for user testing. We focus on the generation of multiple variants of graphical user interfaces for small-screen devices and natural language voice output‚ within a system we term UI on the Fly.

When grammar-based techniques for natural language generation (and analysis alike) find their way into collaborative projects or actual application‚ big grammars tend to become hard to extend and debug. The MUG system represents a new tool set with a graphical debugging environment for functional unification grammars‚ which is designed to help grammar developers inspect the results of their work. The particular formalism supported is Multimodal Functional Unification Grammar (MUG‚ [4])‚ which is similar to Functional Unification Grammars (FUG: [2]‚ [1])‚ but supports several coordinated modes‚ such as voice prompts or structural and/or language-based screen displays. For each input description‚ the grammar can generate a range of coherent realization variants‚ which are ranked by a scoring function in order to optimize the output towards situational and device-related factors.

While quite a few linguistic corpora with syntactic annotations are available today‚ resources are scarce on the level of discourse annotation. A flexible‚ extendible annotation format speeds up development. We therefore propose an XML format for annotating rhetorical structure trees. In human and automatic analysis‚ rhetorical structure is often difficult and assigned incrementally. Thus‚ the format allows for underspecification. The paper discusses the various design decisions involved‚ illustrates the format with an example‚ and sketches some applications.

I present a statistical-based approach to the part-of-speech guessing problem. I see assigning a part-of-speech‚ such as Adjective or Noun ‚ as a classification problem. My guessing framework‚ which relies on automated learning of a language model‚ is described in detail. The rich feature analysis presented is suitable for linguistic data‚ such as the ones observed in German. I use a large margin classifier learning algorithm to select relevant features and learn appropriate labelling. The system is evaluated using a German corpus.

We describe our ongoing work on an application of XML/XSL technology to a dictionary‚ from whose source representation various views for the human reader as well as for automatic text generation and understanding are derived. Our case study is a dictionary of discourse markers‚ the words (often‚ but not always‚ conjunctions) that signal the presence of a disocurse relation between adjacent spans of text.

CyMON-NLU can inform‚ chat and gather user information using an advanced natural language understanding engine. It combines statistical morphosyntactic disambiguation methods (trigram tagging)‚ a stemming algorithm and a robust parser for a large semantic grammar implemented in an XML formalism. The scalable CyMON-NLU engine is implemented in C++ and provides interfaces to the agent-based CRM platform CyMON. Further features include automatic language detection and dialog tracking using a semantic network interface. A development kit enables language engineers to easily create semantic grammars for the specific domain.

Studies on linguistic alignment explain the emergence of mutual understanding in dialogue. They connect to psycholinguistic models of language processing. Recently‚ more computational cognitive models of language processing in the dialogue setting have been published‚ and they increasingly are based on observational‚ large datasets rather than hypothesis-driven experimentation. I review literature in data-intensive computational psycholinguistics and demonstrate the approach in a study that elucidates the mechanisms of alignment. I describe consistent results on data from large online forums‚ Reddit‚ and the Cancer Survivors Network‚ which suggest that linguistic alignment is a consequence of the architecture of the human memory system and not an explicit‚ consciously controlled process tailored to the recipient of a speaker’s contributions to the dialogue.

The Wolves and Sheep game is an educational‚ multi-agent computer simulation for students of programming classes. Players move about in a two- dimensional landscape‚ aiming to reach targets or catch other players. Students write programs that control individual players in the game. The programs then participate in a tournament. Students practice basic programming and algorithmic thinking‚ object-orientation and the role of interfaces‚ and are even exposed to basic Artificial Intelligence. Gamification helps motivate students and creates positive‚ but challenging learning experiences.

Computational models of memory presented in this issue reflect varied empirical data and levels of representation. From mathematical models to neural and cognitive architectures‚ all aim to converge on a unified theory of the mind.

This thesis addresses the cognitive basis of syntactic adaptation‚ which biases speakers to repeat their own syntactic constructions and those of their conversational partners. I address two types of syntactic adaptation: short-term priming and long-term adaptation. I develop two metrics for syntactic adaptation within a speaker and between speakers in dialogue: one for short-term priming effects that decay quickly‚ and one for long-term adaptation over the course of a dialogue. Both methods estimate adaptation in large datasets consisting of transcribed human-human dialogue annotated with syntactic information. Two such corpora in English are used: Switchboard‚ a collection of spontaneous phone conversation‚ and HCRC Map Task‚ a set of task-oriented dialogues in which participants describe routes on a map to each other. I find both priming and long-term adaptation in both corpora‚ confirming well-known experimental results (Bock‚ 1986). I extend prior work by showing that syntactic priming effects not only apply to selected syntactic constructions that are alternative realizations of the same semantics‚ but still hold when a broad variety of syntactic phrase-structure rules are considered. Each rule represents a cognitive decision during syntactic processing. I show that the priming effect for a rule is inversely proportional to its frequency. With this methodology‚ I test predictions of the Interactive Alignment Model (Pickering & Garrod‚ 2004). The IAM claims that linguistic and situation-model agreement between interlocutors in dialogue is the result of a cascade of resource-free‚ mechanistic priming effects on various linguistic levels. I examine task-oriented dialogue in Map Task‚ which provides a measure of task success through the deviance of the communicated routes on the maps. I find that long-term syntactic adaptation predicts communicative success‚ and it does so earlier than lexical adaptation. The result is applied in a machine-learning based model that estimates task success based on the dialogue‚ capturing 14 percent of the variance in Map Task. Short-term syntactic priming differs qualitatively from long-term adaptation‚ as it does not predict task success‚ providing evidence against learning as a single cognitive basis of adaptation effects. I obtain further evidence for the correlation between semantic activity and syntactic priming through a comparison of the Map Task and Switchboard corpora‚ showing that short-term priming is stronger in task-oriented dialogue than in spontaneous conversation. This difference is evident for priming between and within speakers‚ which suggests that priming is a mechanistic rather than strategic effect. I turn to an investigation of the level at which syntactic priming influences language production. I establish that the effect applies to structural syntactic decisions as opposed to all surface sequences of lexical categories. To do so‚ I identify pairs of part-of-speech categories which consistently cross constituent boundaries defined by the phrase-structure analyses of the sentences. I show that such distituents are less sensitive to priming than pairs occurring within constituents. Thus‚ syntactic priming is sensitive to syntactic structure. The notion of constituent structure differs among syntactic models. Combinatory Categorial Grammar (CCG‚ Steedman‚ 2000) formalizes flexible constituent structure‚ accounting a varying degree of incrementality in syntactic sentence planning. I examine whether priming effects can support the predictions of CCG using the Switchboard corpus‚ which has been annotated with CCG syntax. I confirm the syntactic priming effect for lexical and non-lexical CCG categories‚ which encode partially satisfied subcategorization frames. I then show that both incremental and normal-form constituent structures exhibit priming‚ arguing for language production accounts that support flexible incrementality. The empirical results are reflected in a cognitive model of syntactic realization in language production. The model assumes that language production is subject to the same principles and constraints as any other form of cognition and follows the ACT-R framework (Anderson‚ 2004). Its syntactic process implements my empirical results on priming and is based on CCG. Syntactic planning can take place incrementally and non-incrementally. The model is able to generate simple sentences that vary syntactically‚ similar to the materials used in the experimental priming literature. Syntactic adaptation emerges due to a preferential and sped-up memory retrieval of syntactic categories describing linearization and subcategorization requirements. Long-term adaptation is explained as a form of learning‚ while short-term priming is the result of a combination of learning and spreading activation from semantic and lexical material. Simulations show that the model produces the adaptation effects and their inverse frequency interaction‚ as well as cumulativity of long-term adaptation.

The output of multimodal human-computer interfaces is what this thesis is concerned with. Rather than hard-coding graphical and spoken representations‚ methods are introduced that plan and realize coherent output‚ appropriate to the situation and the device. The generation system expects a mode- and language-independent representation‚ as it can be supplied by the dialogue management component of a dialogue system. The generator then assembles mode-specific rendering instructions simultaneously for each mode with the aid of a unification-based functional grammar. The approach proposed in this thesis abandons the canonical structure of pipelined planning and realization in natural language generation‚ in favor of hard constraints formulated in a grammar‚ and soft constraints that allow for the gradual adaptivity of the output. The grammar is constructed to ensure the coherence of output in different modalities‚ whose output is generated in a synchronized fashion rather than by separate‚ mode-specific generators. The soft constraints follow some of the Gricean maxims by incorporating two counteracting communicative goals: efficacy and efficiency. A fitness function encoding these goals takes into account situation- and user-specific factors‚ such as distractions in a single mode or the user’s sensory impairments. The function leads to the selection of an appropriate output from the variety of potential outputs generated by the grammar. It is evaluated in a study with human subjects. The thesis presents a unification based‚ hybrid grammar formalism which can combine pre-fabricated phrases and linguistically motivated grammar fragments‚ and an associated algorithm which integrates the formulation of grammars that lead to cross-modally coherent output. Methods are compared to efficiently implement a control strategy‚ combining hard and soft constraints as a constraint optimization problem. The cross-modal coherence implemented by the grammar formalism is motivated by known phenomena‚ such as cross-modal priming‚ or alignment between interlocutors. To optimize discourse coherence‚ central ideas of Centering Theory are implemented using the grammar formalism. Finally‚ novel methods and a ready-to-use implementation are introduced which allow user interface developers to inspect‚ maintain and extend grammars. The formalism and generation implementation is demonstrated with a grammar for a mobile‚ multimodal application‚ the Virtual Personal Assistant.

Most text displays an internal coherence structure‚ which can be analyzed as a tree structure of relations that hold between short segments of text. We present a machine-learning governed approach to such an analysis in the framework of Rhetorical Structure Theory. Our rhetorical analyzer observes a variety of textual properties‚ such as cue phrases‚ part-of-speech information‚ rhetorical context and lexical chaining. A two-stage parsing algorithm uses local and global optimization to find an analysis. Decisions during parsing are driven by an ensemble of support vector classifiers. This training method allows for a non-linear separation of samples with many relevant features. We define a chain of annotation tools that profits from a new underspecified representation of rhetorical structure. Classifiers are trained on a newly introduced German language corpus‚ as well as on a large English one. We present evaluation data for the recognition of rhetorical relations.

We present a novel lifelong neural architecture‚ the Temporal Neural Coding Network (TNCN)‚ and its learning algorithm‚ for uncovering multiple levels of distributed representations for language data streams. The TNCN model adapts its parameters iteratively as new samples are observed without resorting to the popular‚ but expensive back-propagation through time procedure needed to calculate gradients for recurrent neural networks‚ notably requiring no unrolling of internally defined recurrence relations. Furthermore‚ we integrate concepts from variational encoder-decoder models‚ such as those in [1]‚ to easily demonstrate how uncertainty can be characterized in the model’s next step predictions. We discuss how the proposed TNCN works specifically on language-based tasks formulated in the streaming setting‚ compared to various neural models‚ as well as how the TNCN may be easily adapted to other real-time tasks‚ such as video sequence modeling‚ yielding a natural framework for online multi-modal and online semi-supervised learning.

We study personal and general biases in timing decisions‚ which concern practically common continuous decisions about when to act when facing external uncertainty and dynamic risk. In a series of experiments using a two-player game called FlipIt‚ we studied the influence of individual difference variables of Risk Propensity (RP) and Need for Cognition (NfC) and of manipulations that induce impatience. Participants make several timing decisions per round‚ which aim to counteract the hidden actions made by an opponent. The opponent’s actions become visible only after the participant’s choices have been made. Earlier actions are riskier‚ but a rational time point‚ maximizing utility‚ can be defined. Two experiments were run with a total of 450 participants from a general population that were compensated to incentivize maximizing payoff. RP and NfC were obtained with standardized instruments. Participants with RP around the population average benefitted from higher NfC levels (p<0.01). We also found a general impatience bias that advanced timing independent of RP. With increasing task experience‚ models of linear impatience become better predictors of timing bias than risk-dependent ones. In a further experiment with 422 participants‚ impatience was manipulated by showing slow or fast-changing countdowns during a constant delay period. Timing (vs. rational strategy) was retarded or advanced‚ with fast countdowns reducing impatience (p<0.01) and accelerating countdowns yielding closer-to-rational performance even than the control (no delay at all). We conclude that timing decisions are affected by interacting risk and deep-processing preferences as well as situational impatience biases.

This is a multi-agent‚ bio-inspired simulation; it is a competitive game and a tournament written for 5-50 beginner programmers. It is designed as educational tool: Students of an intermediate programming class taught in Java write code that controls a single player in this simulation. Players are pitched against each other; typically 5 at a time. A tournament runs many combinations of players in many adjustable scenarios. The tournament is set up to run regularly on a server‚ taking .jar files provided by students. It uses standard visibility and sandboxing features of Java to provide a basic level of security.

This tool explores new ways to make bibliographies interactive‚ searchable and discoverable. Zot_bib_web generates HTML bibliographies based on one or more collections in a Zotero repository. These collections can be maintained by groups of people‚ using Zotero’s web interface or their desktop applications.

ACT-UP is a toolbox implementation of the ACT-R cognitive architecture‚ aimed at allowing rapid prototyping of complex models. With ACT-UP‚ we propose Accountable Modeling‚ where the only model components that are specified are those supported by empirical evidence and part of the model’s theoretical claims. ACT-UP is a library providing a programmatic interface to the classical ACT-R functionality. Implemented in a functional programming paradigm‚ models are reusable in other contexts.

Aquamacs is a user-friendly build of the powerful Emacs text editor. Aquamacs integrates with the Mac and offers the same comforts that any application on the Mac provides. Yet‚ it comes with all the ergonomy and extensibility you’ve come to expect from GNU Emacs. It has been authored by David Reitter‚ based on GNU Emacs by Richard Stallman and many others. Aquamacs has been available for about a decade and is used daily by more than twelve thousand academics‚ programmers‚ and authors. It is backed by two strong communities: Aquamacs users‚ and Emacs enthusiasts on all computing platforms (GNU/Linux/Unix‚ Windows‚ Mac).

This program compared files‚ typically text files or source code. It presented two versions of a file visually in a split screen and marked up differences‚ aided by an algorithm to detect changes.

Enables quick access to installed software.

Implementation of the popular Tetris game‚ with multi-player mode (via MIDI connections) and its own soundtrack.

X-Point was an integrated e-mail and bulletin board message client that connected to StarNet‚ a German network of bulletin board systems (BBS) via modems and the telephone network. These systems were an early computer network popular before the internet became widely available. X-Point included its own text editor‚ message viewer and implementation of the StarNet communication protocol.