A longitudinal multi-modal dataset for dementia monitoring and diagnosis.

Dementia affects cognitive functions of adults, including memory, language, and behaviour. Standard diagnostic biomarkers such as MRI are costly, whilst neuropsychological tests suffer from sensitivity issues in detecting dementia onset. The analysis of speech and language has emerged as a promising and non-intrusive technology to diagnose and monitor dementia. Currently, most work in this direction ignores the multi-modal nature of human communication and interactive aspects of everyday conversational interaction. Moreover, most studies ignore changes in cognitive status over time due to the lack of consistent longitudinal data. Here we introduce a novel fine-grained longitudinal multi-modal corpus collected in a natural setting from healthy controls and people with dementia over two phases, each spanning 28 sessions. The corpus consists of spoken conversations, a subset of which are transcribed, as well as typed and written thoughts and associated extra-linguistic information such as pen strokes and keystrokes. We present the data collection process and describe the corpus in detail. Furthermore, we establish baselines for capturing longitudinal changes in language across different modalities for two cohorts, healthy controls and people with dementia, outlining future research directions enabled by the corpus.

Comparing neural sentence encoders for topic segmentation across domains: not your typical text similarity task.

Neural sentence encoders (NSE) are effective in many NLP tasks, including topic segmentation. However, no systematic comparison of their performance in topic segmentation has been performed. Here, we present such a comparison, using supervised and unsupervised segmentation models based on NSEs. We first compare results with baselines, showing that the use of NSEs does often provide improvements, except for specific domains such as news shows. We then compare over three different datasets a range of existing NSEs and a new NSE based on ad hoc pre-training strategy. We show that existing literature documenting general performance gains of NSEs does not always conform to the results obtained by the same NSEs in topic segmentation. If Transformers-based encoders do improve over previous approaches, fine-tuning in sentence similarity tasks or even on the same topic segmentation task we aim to solve does not always equate to better performance, as results vary across method being used and domains of application. We aim to explain this phenomenon and the relative poor performance of NSEs in news shows by considering how well different NSEs encode the underlying lexical cohesion of same-topic segments; to do so, we introduce a new metric, ARP. The results from this study suggest that good topic segmentation results do not always rely on good cohesion modelling on behalf of the segmenter and that is dependent upon what kind of text we are trying to segment. Also, it appears evident that traditional sentence encoders fail to create topically cohesive clusters of segments when used on conversational data. Overall, this work advances our understanding of the use of NSEs in topic segmentation and of the general factors determining the success (or failure) of a topic segmentation system. The new proposed metric can quantify the lexical cohesion of a multi-topic document under different sentence encoders and, as such, might have many different uses in future research, some of which we suggest in our conclusions.

"Who's there?": Depicting identity in interaction.

Social robots have limited social competences. This leads us to view them as depictions of social agents rather than actual social agents. However, people also have limited social competences. We argue that all social interaction involves the depiction of social roles and that they originate in, and are defined by, their function in accounting for failures of social competence.

Power and vulnerability: managing sensitive language in organizational communication.

Organizational responsibilities can give people power but also expose them to scrutiny. This tension leads to divergent predictions about the use of potentially sensitive language: power might license it, while exposure might inhibit it. Analysis of peoples' language use in a large corpus of organizational emails using standardized Linguistic Inquiry and Word Count (LIWC) measures shows a systematic difference in the use of words with potentially sensitive (ethnic, religious, or political) connotations. People in positions of relative power are ~3 times less likely to use sensitive words than people more junior to them. The tendency to avoid potentially sensitive language appears to be independent of whether other people are using sensitive language in the same email exchanges, and also independent of whether these words are used in a sensitive context. These results challenge a stereotype about language use and the exercise of power. They suggest that, in at least some circumstances, the exposure and accountability associated with organizational responsibilities are a more significant influence on how people communicate than social power.

Investigating cross-lingual training for offensive language detection.

Platforms that feature user-generated content (social media, online forums, newspaper comment sections etc.) have to detect and filter offensive speech within large, fast-changing datasets. While many automatic methods have been proposed and achieve good accuracies, most of these focus on the English language, and are hard to apply directly to languages in which few labeled datasets exist. Recent work has therefore investigated the use of cross-lingual transfer learning to solve this problem, training a model in a well-resourced language and transferring to a less-resourced target language; but performance has so far been significantly less impressive. In this paper, we investigate the reasons for this performance drop, via a systematic comparison of pre-trained models and intermediate training regimes on five different languages. We show that using a better pre-trained language model results in a large gain in overall performance and in zero-shot transfer, and that intermediate training on other languages is effective when little target-language data is available. We then use multiple analyses of classifier confidence and language model vocabulary to shed light on exactly where these gains come from and gain insight into the sources of the most typical mistakes.

Re-Representing Metaphor: Modeling Metaphor Perception Using Dynamically Contextual Distributional Semantics.

In this paper, we present a novel context-dependent approach to modeling word meaning, and apply it to the modeling of metaphor. In distributional semantic approaches, words are represented as points in a high dimensional space generated from co-occurrence statistics; the distances between points may then be used to quantifying semantic relationships. Contrary to other approaches which use static, global representations, our approach discovers contextualized representations by dynamically projecting low-dimensional subspaces; in these ad hoc spaces, words can be re-represented in an open-ended assortment of geometrical and conceptual configurations as appropriate for particular contexts. We hypothesize that this context-specific re-representation enables a more effective model of the semantics of metaphor than standard static approaches. We test this hypothesis on a dataset of English word dyads rated for degrees of metaphoricity, meaningfulness, and familiarity by human participants. We demonstrate that our model captures these ratings more effectively than a state-of-the-art static model, and does so via the amount of contextualizing work inherent in the re-representational process.

Computational Models of Miscommunication Phenomena.

Miscommunication phenomena such as repair in dialogue are important indicators of the quality of communication. Automatic detection is therefore a key step toward tools that can characterize communication quality and thus help in applications from call center management to mental health monitoring. However, most existing computational linguistic approaches to these phenomena are unsuitable for general use in this way, and particularly for analyzing human-human dialogue: Although models of other-repair are common in human-computer dialogue systems, they tend to focus on specific phenomena (e.g., repair initiation by systems), missing the range of repair and repair initiation forms used by humans; and while self-repair models for speech recognition and understanding are advanced, they tend to focus on removal of "disfluent" material important for full understanding of the discourse contribution, and/or rely on domain-specific knowledge. We explain the requirements for more satisfactory models, including incrementality of processing and robustness to sparsity. We then describe models for self- and other-repair detection that meet these requirements (for the former, an adaptation of an existing repair model; for the latter, an adaptation of standard techniques) and investigate how they perform on datasets from a range of dialogue genres and domains, with promising results.

Entraining IDyOT: Timing in the Information Dynamics of Thinking.

We present a novel hypothetical account of entrainment in music and language, in context of the Information Dynamics of Thinking model, IDyOT. The extended model affords an alternative view of entrainment, and its companion term, pulse, from earlier accounts. The model is based on hierarchical, statistical prediction, modeling expectations of both what an event will be and when it will happen. As such, it constitutes a kind of predictive coding, with a particular novel hypothetical implementation. Here, we focus on the model's mechanism for predicting when a perceptual event will happen, given an existing sequence of past events, which may be musical or linguistic. We propose a range of tests to validate or falsify the model, at various different levels of abstraction, and argue that computational modeling in general, and this model in particular, can offer a means of providing limited but useful evidence for evolutionary hypotheses.

Better late than Now-or-Never: The case of interactive repair phenomena.

Empirical evidence from dialogue, both corpus and experimental, highlights the importance of interaction in language use - and this raises some questions for Christiansen & Chater's (C&C's) proposals. We endorse C&C's call for an integrated framework but argue that their emphasis on local, individual production and comprehension makes it difficult to accommodate the ubiquitous, interactive, and defeasible processes of clarification and repair in conversation.

Twitter Language Use Reflects Psychological Differences between Democrats and Republicans.

Previous research has shown that political leanings correlate with various psychological factors. While surveys and experiments provide a rich source of information for political psychology, data from social networks can offer more naturalistic and robust material for analysis. This research investigates psychological differences between individuals of different political orientations on a social networking platform, Twitter. Based on previous findings, we hypothesized that the language used by liberals emphasizes their perception of uniqueness, contains more swear words, more anxiety-related words and more feeling-related words than conservatives' language. Conversely, we predicted that the language of conservatives emphasizes group membership and contains more references to achievement and religion than liberals' language. We analysed Twitter timelines of 5,373 followers of three Twitter accounts of the American Democratic and 5,386 followers of three accounts of the Republican parties' Congressional Organizations. The results support most of the predictions and previous findings, confirming that Twitter behaviour offers valid insights to offline behaviour.

Divergence in dialogue.

One of the best known claims about human communication is that people's behaviour and language use converge during conversation. It has been proposed that these patterns can be explained by automatic, cross-person priming. A key test case is structural priming: does exposure to one syntactic structure, in production or comprehension, make reuse of that structure (by the same or another speaker) more likely? It has been claimed that syntactic repetition caused by structural priming is ubiquitous in conversation. However, previous work has not tested for general syntactic repetition effects in ordinary conversation independently of lexical repetition. Here we analyse patterns of syntactic repetition in two large corpora of unscripted everyday conversations. Our results show that when lexical repetition is taken into account there is no general tendency for people to repeat their own syntactic constructions. More importantly, people repeat each other's syntactic constructions less than would be expected by chance; i.e., people systematically diverge from one another in their use of syntactic constructions. We conclude that in ordinary conversation the structural priming effects described in the literature are overwhelmed by the need to actively engage with our conversational partners and respond productively to what they say.

Using conversation topics for predicting therapy outcomes in schizophrenia.

Previous research shows that aspects of doctor-patient communication in therapy can predict patient symptoms, satisfaction and future adherence to treatment (a significant problem with conditions such as schizophrenia). However, automatic prediction has so far shown success only when based on low-level lexical features, and it is unclear how well these can generalize to new data, or whether their effectiveness is due to their capturing aspects of style, structure or content. Here, we examine the use of topic as a higher-level measure of content, more likely to generalize and to have more explanatory power. Investigations show that while topics predict some important factors such as patient satisfaction and ratings of therapy quality, they lack the full predictive power of lower-level features. For some factors, unsupervised methods produce models comparable to manual annotation.

Grammar resources for modelling dialogue dynamically.

This paper argues that by analysing language as a mechanism for growth of information (Cann et al. in The Dynamics of Language, Elsevier, Oxford, 2005; Kempson et al. in Dynamic Syntax, Blackwell, Oxford, 2001), not only does a unitary basis for ellipsis become possible, otherwise thought to be irredeemably heterogeneous, but also a whole range of sub-types of ellipsis, otherwise thought to be unique to dialogue, emerge as natural consequences of use of language in context. Dialogue fragment types modelled include reformulations, clarification requests, extensions, and acknowledgements. Buttressing this analysis, we show how incremental use of fragments serves to progressively narrow down the otherwise mushrooming interpretational alternatives in language use, and hence is central to fluent conversational interaction. We conclude that, by its ability to reflect dialogue dynamics as a core phenomenon of language use, a grammar with inbuilt parsing dynamics opens up the potential for analysing language as a mechanism for communicative interaction.