aalto1 untyped-item.component.html
Automatic assessment of L2 interactional competency
Loading...
URL
Journal Title
Journal ISSN
Volume Title
School of Electrical Engineering |
Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.
Authors
Date
Department
Major/Subject
Mcode
Language
en
Pages
51
Series
Abstract
Despite interaction being a major component of speech, it is underrepresented in language assessment, including automatic language assessment. While non-verbal behavior has also been noted for its importance in speech assessment literature, it is likewise missing from most speech assessment, both manual and automatic. These problems go hand in hand.
This thesis suggests the application of traditional techniques for speech assessment. Modern language learning is focused on large neural network and text embeddings, which makes it less applicable to multi-modal and multi-speaker problems. Natural methods for comparison of speech channels are provided in the form of segmentation and reactive features. Features are subsequently analyzed and put to a test in a speech ability estimator. The methods used are agnostic towards text, circumventing the problems of low-resource languages. The channels and features are analyzed in a bottom up approach, based in theory surrounding interactional competence.
The results are exploratory and not exhaustive but promising for future potential of interactional competence assessment. Examples of estimation are present as well as a list of statistically significant features for interactional competence assessment. The importance of non-verbal behavior in interaction is apparent.