Our two other themes are methods and positions.
Frans Gregersen (University of Copenhagen): Changes across linguistic levels and across life stages: In search of a pattern
This paper aims at providing a corpus linguistic counterpart to the current hypotheses about the effects of ageing on the various subsystem of language. The LANCHART Corpus, a corpus composed of more than 1000 hours of audio material, where we have studied language change in real time by recording informants at two different points in time almost twenty years apart and observed micro-diachronic changes within the same generation. The data as well as related studies are used to exemplify changes in the phonological, morphological subsystems as well as in pragmatics, interaction, and discourse.
The approach is a socio-historical one, analysing the interaction between how ageing is perceived in various communities and evidence from conducted linguistic studies.
Sylvie Ratté (École de Technologie Supérieure, Montreal): How can Machine Learning and Natural Language Processing help researchers to study communication behaviors in later age?
The Cécilia research program aims to automatically characterize and monitor changes in Alzheimer’s disease (AD). As a follow-up to CLARe 2017 (Ratté et al. 2017), the main objective of this talk is two-fold. First, we will present the results obtained by the Cécilia team during the past two years on automatic annotations and features learning in video/sound-recorded conversations in various settings. Second, we want to propose an interdisciplinary collaboration program that will diminish the cost of manual annotations and instigate a back-and-forth dialog between computer sciences (and in particular the ML and NLP communities) and researchers in the CLARe network.
The Cécilia research program focuses on proposing a computer-based analysis of verbal and nonverbal behaviors in multiple modalities: speech, discourse, and facial and corporal expressions. More specifically, we are pursuing three objectives: 1. Distinguish between AD patients and ailing patients using properties automatically extracted from transcriptions and audio sources of conversations with elderly subjects (analysis of verbal behaviors). 2. Distinguish between AD patients and ailing patients using properties extracted from video recordings of conversations with elderly subjects (analysis of nonverbal behaviors). 3. Construct a ML/NLP system that will use every property obtained in (1) and (2) to identify various stages of AD.
To accomplish these tasks, we are using (among others) three recognized datasets: the DementiaBank Pitt Corpus (DB, Becker et al. 1994), the Carolina Conversations Collection (CCC, Pope & Davis, 2011), and the CorpAGEst corpus (Bolly & Boutet, 2018). The first one contains conversations made during the application of cognitive tests to patients. The last two contain spontaneous conversations. The languages covered include Spanish, English, and French.
In the first experiment (Hernández-Domínguez et al. 2018), we analyzed transcripts of the DB corpus. We automatically combined and extracted linguistic metrics (rhythm, part of speech, vocabulary richness, idea density, syntactic complexity) and phonetic metrics which previous literature correlated to early stage AD with a new measure based on asymmetry of information coverage. This new measure implies the construction of a referent by merging the conversations of a subset of healthy patients. Once the referent is constructed, the rest of the population (healthy or not) is compared to it. We trained two learners to distinguish Healthy Control (HC) from cognitively impaired individuals. Our measures significantly (P <.001) correlated with the severity of the cognitive impairment and the Mini-Mental State Examination score. The classification sensitivity was 81% and 85% between HCs and AD and between HCs and AD and mild cognitively impaired, respectively.
The second experiment takes as a source videos (CorpAGEst and CCC). Our goal was two-fold: first, to illustrate how automatized techniques could be used to annotate videos according to a standard; second, to let the algorithm identify features. The results suggest two avenues for further explorations. First, it is possible to automatically annotate nonverbal behaviors in videos; second, ML algorithms have the capacity to identify features that are not totally in sync with what humans are used to. This characteristic clearly opens a new dialog between artificial intelligence and researchers in the CLARe network in terms of interpretation of results and features to consider in videos.
The impacts of the Cécilia program are twofold. First, it contributes to building stronger facial and gesture recognition techniques adapted to the aging population. Second, it contributes to developing automated tools to monitor AD and other types of dementia (e.g., the effects of medications on language abilities), giving tools to the community of researchers to explore specificities in an automatic fashion. The results of this proposal are not meant to diagnose or to replace specialists (e.g., clinicians, doctors, geriatricians); they are aimed at assisting these professionals in innovative and non-invasive ways, helping them to understand the disease and its progression in a more efficient way.
Becker, J. T., Boller, F., Lopez, O. L., Saxton, J., & McGonigle, K. L. (1994). The natural history of Alzheimer's disease: description of study cohort and accuracy of diagnosis. Archives of Neurology, 51(6), 585-594.
Bolly, C. T. & Boutet, D. (2018). The multimodal CorpAGEst corpus: Keeping an eye on pragmatic competence in later life. Corpora 13 (2).
Hernández-Domínguez, L., Ratté, S., Sierra-Martínez, G., & Roche-Bergua, A. (2018). Computer-based evaluation of Alzheimer’s disease and mild cognitive impairment patients during a picture description task. Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring, 10, 260-268.
Pope, C., & Davis, B. H. (2011). Finding a balance: The Carolinas Conversation Collection. Corpus Linguistics and Linguistic Theory, 7(1), 143-161.
Ratté, S., Hernández-Domínguez, L., Davis, B., Pope, C., Roche-Bergua, A., & Sierra-Martínez, G. (2017). Cécilia project: an international multidisciplinary collaboration on the study language in later life. Corpora for Language and Aging Research (CLARe 3 ) (Berlin, Allemagne, Mar. 06-08, 2017).
Christina Samuelsson (Linköping University): Communication involving people with dementia
Dementia comprises a number of symptoms that may be due to various diseases and injuries. While dementia is often primarily associated with memory deficits, communication and participation in social activities have been reported as areas where persons with dementia experience challenges (Johansson, Marcusson & Wressle 2015). As a dementia disease progresses, abilities to initiate and maintain interactions tend to gradually decline, resulting in diminishing social relations and increased social isolation (Örulv & Nikku 2007). Loneliness and social isolation are therefore significant problems for people living with dementia (Alzheimer’s Society, 2013). In addition, as the person’s communicative abilities decline, it becomes increasingly difficult to ensure that the person’s views are heard (Österholm & Hydén 2016).
In this presentation I will report from several projects on interaction involving persons with dementia demonstrating problematic areas, but also sequences where persons with dementia display their competence as interactants. We have shown that persons with dementia e.g. are able to initiate and perform repair work, albeit in a somewhat different pattern than the typical three turn format. I will also focus on how participants deal with instances of reality disjunctions in interaction involving persons with dementia. It is demonstrated that both the healthy participants as well as the person with dementia together skilfully avoid the face threats posed by reality disjunctive contributions by not pursuing argumentative lines that in the end might jeopardize both the collaborative and the personal relations.
When it comes to older people with dementia there exists very little rehabilitating intervention, and a common conception is that active intervention is not effective for people with progressive neurodegenerative diseases. Specific intervention for persons with dementia focusing on communication has been tried to some extent from an international perspective, and both the American Speech and Hearing Association (ASHA), and the Royal College of Speech Language Therapists (RCSLT) conclude that language intervention should be available for persons with dementia (RCSLT, 2014; ASHA, 2018). In a recent project, digital communication support in the form of two applications for tablet computers has been tried in interaction with persons with dementia. In this presentation, results on initiatives, topical management, pitfalls and possibilities when using digital communication support will be presented. It is shown that the conversations are longer with than without the digital communications support, that the communication support seems to engage the persons with dementia in conversation, and that it may be beneficial for eliciting personal stories. There are also positive results on using digital conversation support in group sessions for persons with dementia.
Johansson, M., Marcusson, J., & Wressle, E. (2015). Cognitive impairment and its consequences in everyday life: Experiences of people with mild cognitive impairment or mild dementia and their relatives. International Psychogeriatrics, 27(6), 949-958.
Örulv, L., & Nikku, N. (2007). Dignity work in dementia care: Sketching a microethical analysis. Dementia, 6(4), 507-525.
Österholm, J. H., & Hydén, L. C. (2018). Autobiographical occasions in assessment meetings involving persons with dementia. Qualitative Social Work, 17(1), 41-64.