Gasevic, Dragan

Lucas, Christopher

Lopez, Adam

Sinclair, Arabella Jane

Engineering and Physical Sciences Research Council (EPSRC)

2020-04-29T14:29:39Z

2020-04-29T14:29:39Z

2020-06-25

Understanding how tutors and students adapt to one another within Second Language (L2) learning is an important step in the development of better automated tutoring tools for L2 conversational practice. Such an understanding can not only inform conversational agent design, but can be useful for other pedagogic applications such as formative assessment, self reflection on tutoring practice, learning analytics, and conversation modelling for personalisation and adaptation. Dialogue is a challenging domain for natural language processing, understanding, and generation. It is necessary to understand how participants adapt to their interlocutor, changing what they express and how they express it as they update their beliefs about the knowledge, preferences, and goals of the other person. While this adaptation is natural to humans, it is an open problem for dialogue systems, where managing coherence across utterances is an active area of research, even without adaptation. This thesis extends our understanding of adaptation in human dialogue, to better implement this in agent-based conversational dialogue. This is achieved through comparison to fluent conversational dialogues and across student ability levels. Specifically, we are interested in how adaptation takes place in terms of the linguistic complexity, lexical alignment and the dialogue act usage demonstrated by the speakers within the dialogue. Finally, with the end goal of an automated tutor in mind, the student alignment levels are used to compare dialogues between student and human tutor with those where the tutor is an agent. We argue that the lexical complexity, alignment and dialogue style adaptation we model in L2 human dialogue are signs of tutoring strategies in action, and hypothesise that creating agents which adapt to these aspects of dialogue will result in better environments for learning. We hypothesise that with a more adaptive agent, student alignment may increase, potentially resulting in improved engagement and learning. We find that In L2 practice dialogues, both student and tutor adapt to each other, and this adaptation depends on student ability. Tutors adapt to push students of higher ability, and to encourage students of lower ability. Complexity, dialogue act usage and alignment are used differently by speakers in L2 dialogue than within other types of conversational dialogue, and changes depending on the learner proficiency. We also find different types of learner behaviours within automated L2 tutoring dialogues to those present in human ones, using alignment to measure this. This thesis contributes new findings on interlocutor adaptation within second language practice dialogue, with an emphasis on how these can be used to improve tutoring dialogue agents.

https://hdl.handle.net/1842/37009

http://dx.doi.org/10.7488/era/310

en

The University of Edinburgh

Sinclair, A., Ferreira, R., Lopez, A., Lucas, C.& Gasevic, D.(2019), I wanna talk like you: Speaker adaptation to dialogue style in l2 practice conversation, in ‘Proceedings of Artificial Intelligence in Education - 20th International Conference

Sinclair, A., Lopez, A., Lucas, C. & Gasevic, D. (2018), Does ability affect alignment in second language tutorial dialogue?, in ‘Proceedings of the 19th Annual SIGdial Meeting on Discourse and Dialogue’, pp.41–50.

Sinclair, A., McCurdy, K., Lopez, A., Lucas, C. & Gasevic, D. (2019), Tutorbot corpus: Evidence of human-agent verbal alignment in second language learner dialogues, in ‘Proceedings of Educational Data Mining - 12th International Conference’.

Sinclair, A., Oberlander, J.& Gasevic, D. (2017), Finding the zone of proximal development: Student-tutor second language dialogue interactions, in ‘Proc. SEMDIAL 2017 (SaarDial) Workshop on the Semantics and Pragmatics of Dialogue’, pp.107– 115.

natural language processing

dialogue

conversation analysis

second language

machine learning

AI

linguistic alignment

linguistic complexity

alignment

adaptation

computational linguistics

dialogue agent

language learning

Modelling speaker adaptation in second language learner dialogue

Thesis or Dissertation

Doctoral

PhD Doctor of Philosophy