Cambridge Language Sciences Symposium for Early-Career Researchers June 2020

I really enjoyed the diversity of the topics covered in the oral and poster presentations - it's made me realise the very rich and varied inputs Cambridge researchers are contributing to the field of Language Sciences.

The presentations and poster session were recorded. Please visit the Early-Career Researchers Symposium June 2020 page (Raven log-in required) to access recordings of the live sessions, slides & extended abstracts from the plenary presenters and a link to the virtual poster session blog which will remain live until 31 July 2020.

We would like to thank everyone who contributed to make this such a successful event, in particular the speakers, the poster presenters, and Dr Avin Mirawdeli, the Programme Organiser who also did a fantastic job of chairing both of the live sessions.

Cambridge Language Sciences currently runs two Symposia a year: a Symposium for Early-Career Researchers in the summer and an Annual Symposium in November for all members of our research community. Details of the next Symposium in November 2020 will be confirmed in due course.

Programme

Adaptivity in Automated Language Teaching & Assessment

Dr Andrew Caines, Computer Science & Technology / ALTA Institute

RESEARCHER PROFILE (500-word article)

I will give an overview of research by the ALTA Institute into adaptive educational technology for the purposes of automated language teaching and assessment. The talk gives an introduction to key concepts in this area, explains our research questions and presents a sample of our work. This includes randomised controlled trials measuring the effectiveness of adaptive teaching interventions for English grammatical knowledge, training a machine to select homework tasks based on past selections by human teachers on a physics platform for school-aged children, learning to identify the knowledge components required by educational tasks by mining user data, and predicting the forgetting curves of language learners with English vocabulary of varying complexity. This body of work represents a group of faculty, postdocs and students at the Computer Laboratory, and collaboration with Cambridge Assessment. I will talk about our plans to apply these research techniques in real-world settings, and future objectives and research directions of the group.

Empowering Speech and Language Therapists to provide support for children with autism in linguistically and culturally diverse contexts: The case of India and the UK

Mélanie Gréaux, Faculty of Education

VIDEO: watch speaker introduction by Mélanie Gréaux

RESEARCHER PROFILE (short article)

Autism is a neurodevelopmental condition often associated with speech, language and communication difficulties. Whilst experiencing social communication difficulties is one of the key diagnostic criteria, autistic individuals may also present with additional language difficulties (e.g. one in four autistic people speak few or no words). Early Speech and Language Therapy (SLT) interventions are evidenced to support the development of functional communication skills, some of the best predictors for enhanced quality of life in autistic adults. Yet, whilst SLT services represent a promising avenue, most clinical studies fail to include and represent Linguistically and Culturally Diverse (LCD) individuals. This calls for more inclusive and evidence-based communication support for the autistic community.

Across the globe and despite varied clinical responsibilities and resources, SLTs seem to face a number of fundamental and somehow ‘universal’ challenges when providing services to LCD populations. For example, there are conceptual difficulties to distinguish disorder from diversity (e.g. is a ‘language error’ representative of an underlying language disorder, the effect of limited linguistic exposure, or the expression of a culturally appropriate linguistic pattern?). There are also overwhelming limitations to services (e.g. lack of linguistically valid resources in multiple languages) as well as systemic failures (e.g. unequal access to services as reflected by under- or over-referrals in LCD populations).

In this presentation, the author will discuss the conceptualisation of a cross-cultural knowledge exchange programme for SLTs in the UK and India to address these challenges together, learn from one another, share experiences and resources for the support of LCD autistic children. A brief introduction to the key theoretical and methodological underpinnings of this project will be discussed, and the findings of a survey collecting SLTs’ demographic information and professional learning priorities in both countries will be shared.

The costs of faking it: exploring the role of accent in human social signalling

Jonathan R. Goodman, The Leverhulme Centre for Human Evolutionary Studies

VIDEO: watch speaker introduction by Jonathan R. Goodman

RESEARCHER PROFILE (500-word article)

Accents are known to affect listener preferences and probably have historic evolutionary importance. Here we propose that accents evolved as ‘Greenbeard’ signals, or mechanisms of group-member recognition, with a possible origin in kin selection. The preferential treatment given to fellow group members because of these signals led to a selective pressure for mimicry, which enabled non-group members to benefit from this mechanism through dishonest signalling. In turn, honest signallers would have developed both more complex signalling systems and a sensitive apparatus to mimicry, creating a co-evolutionary relationship between honest signallers and mimics, with a trajectory towards complexity, which may help to explain the complexity of accents in modern languages. Based on these speculations we conducted a trial of 49 native speakers of English, in which we scored participants on their abilities to mimic accents, and to detect mimicry, and compared these scores with responses to an accompanying questionnaire. We found that mimicry does not improve with short-term practice, and further that age may negatively correlate with ability to mimic effectively. Speakers of culturally dominant accents were comparatively worse at mimicry than non-dominant speakers, which may reflect both experiential need to mimic in an individual, and longer term reduced demands on dominant language speakers. Mimicry detection varied by listener geographic origin and native accent, though was strong across groups. We note how our findings accord with an evolutionary account of accent complexity, and suggest further avenues of inquiry.

The Contextual Linguistic Profile Questionnaire: Demonstrating the need for Contextual Linguistic Diversity in a Holistic Measure of Language Experience

Mandy Wigdorowitz, Theoretical & Applied Linguistics

VIDEO: watch speaker introduction by Mandy Wigdorowitz

RESEARCHER PROFILE (500-word article)

Individual reports of language history, use, and proficiency are generally considered sufficient for language profiling; however, these variables alone neglect contextual linguistic experience as a factor that contributes to one’s overall language repertoire. Including a measure of contextual linguistic experience as a feature of one’s sociolinguistic context can indicate where the speaker is linguistically immersed, including how certain languages may be privileged over others, and the fluidity and intermingling of language use. A language profiling measure that captures variables pertinent to language knowledge and experience overall should be a goal towards best practice.

In this talk, I present such a measure, the Contextual Linguistic Profile Questionnaire (CLiP-Q), which captures holistic linguistic information, including contextual linguistic experience and socio-economic status. I then show that lingualism status classifications (such as mono-, bi-, and multilingualism) are often compared based on self-reports or scores without taking contextual linguistic diversity into account. I highlight a shortcoming to this approach by evaluating whether self-reported mono-, bi-, and multilingual groups differ in their linguistic experience across two linguistically different countries where English is the lingua franca. A linear mixed model analysis was conducted to compare whether 119 participants from England age matched to 119 South African participants (Mage = 25.84, SD = 7.95) differ in terms of their contextual linguistic diversity as a feature of their lingualism status and socio-economic standing. Results indicate an effect of nationality and lingualism status on contextual linguistic diversity, where South Africans, across mono-, bi-, and multilingual groups scored higher on multilingual exposure, engagement, and linguistic diversity promotion than that of participants from England. Using the CLiP-Q, I illustrate the importance of contextual linguistic diversity when evaluating linguistically diverse groups and show that lingualism status descriptors alone yield different outcomes on linguistic experience when the contexts of language use differ.

Posters

Reducing Gender Bias in Neural Machine Translation as a Domain Adaptation Problem

Danielle Saunders, Prof. Bill Byrne, Department of Engineering 

The Efficacy of a Guided Play Intervention on the Language Skills of 4- and 5-Year-Olds

Tanya Paes, Dr Michelle Ellefson, Faculty of Education 

“Trouble, Trouble, Trouble”: Resolving the Problematic influence of Diacritics on Lexical Distance

Itamar Shatz, Theoretical and Applied Linguistics

Context is Key: Grammatical Error Detection with Contextual Word Representations

Samuel Bell, Dr Helen Yannakoudakis, Dr Marek Rei, Department of Psychology  

Autoencoding Pixies: Amortised Variational Inference with Graph Convolutions for Functional Distributional Semantics               

Dr Guy Emerson, Department of Computer Science and Technology 

Impact of bilingualism and literacy on language production in bilingual children from underprivileged contexts in India

Dr Anusha Balasubramanian, Prof. Ianthi Tsimpli, Dr Suvarna Alladi, Abhigna Reddy, Theoretical and Applied Linguistics    

Fact-Checking Fake News: Improving Classification in FEVER Systems

Jonty Page, Dr Marcus Tomalin, Department of Engineering    

The processing of subject pronouns in native Spanish: a review of studies testing the Position of Antecedent Strategy

Fernando Martín-Villena, Theoretical and Applied Linguistics