Get our free extension to see links to code for papers anywhere online!

Chrome logo  Add to Chrome

Firefox logo Add to Firefox

Studying Alignment in Spontaneous Speech via Automatic Methods: How Do Children Use Task-specific Referents to Succeed in a Collaborative Learning Activity?

Apr 09, 2021
Utku Norman, Tanvi Dinkar, Barbara Bruno, Chloé Clavel

Share this with someone who'll enjoy it:

A dialogue is successful when there is alignment between the speakers, at different linguistic levels. In this work, we consider the dialogue occurring between interlocutors engaged in a collaborative learning task, and explore how performance and learning (i.e. task success) relate to dialogue alignment processes. The main contribution of this work is to propose new measures to automatically study alignment, to consider completely spontaneous spoken dialogues among children in the context of a collaborative learning activity. Our measures of alignment consider the children's use of expressions that are related to the task at hand, their follow-up actions of these expressions, and how it links to task success. Focusing on expressions related to the task gives us insight into the way children use (potentially unfamiliar) terminology related to the task. A first finding of this work is the discovery that the measures we propose can capture elements of lexical alignment in such a context. Through these measures, we find that teams with bad performance often aligned too late in the dialogue to achieve task success, and that they were late to follow up each other's instructions with actions. We also found that while interlocutors do not exhibit hesitation phenomena (which we measure by looking at fillers) in introducing expressions pertaining to the task, they do exhibit hesitation before accepting the expression, in the role of clarification. Lastly, we show that information management markers (measured by the discourse marker 'oh') occur in the general vicinity of the follow up actions from (automatically) inferred instructions. However, good performers tend to have this marker closer to these actions. Our measures still reflect some fine-grained aspects of learning in the dialogue, even if we cannot conclude that overall they are linked to the final measure of learning.

* * The authors contributed equally to this work. This article a preprint under review 

   Access Paper Source

Share this with someone who'll enjoy it: