Automatic extraction of subordinate clauses and its application in second language acquisition research

Area of Science:

• Linguistics
• Psycholinguistics
• Computational Linguistics
• Cognitive and Behavioral Sciences

Background:

• Clause subordination is a key linguistic feature studied across multiple disciplines.
• Existing research often treats subordination as a unitary phenomenon, potentially overlooking complexities.

Purpose of the Study:

• To introduce AutoSubClause, a novel tool for extracting subordinate clause (SC) information from natural language.
• To investigate the influence of first language (L1) typology on second language (L2) English subordinate clause acquisition.
• To analyze subordinate clauses as a multi-componential construct.

Main Methods:

• Developed AutoSubClause, a tool utilizing dependency parsing (Stanford CoreNLP) for SC extraction.
• Extracted information on SC types (complement, adverbial, relative), internal structure, and clause relations.
• Analyzed a large-scale learner corpus to examine L1 effects on L2 subordination.

Main Results:

• AutoSubClause demonstrated robust performance in extracting SC information.
• Learners with typologically different L1s exhibited distinct developmental trajectories in L2 English subordination compared to those with similar L1s.
• Developmental patterns varied across different types of SCs, supporting a multi-componential view.

Conclusions:

• Subordinate clause acquisition is influenced by L1 typology and should be viewed as a multi-componential construct.
• NLP tools like AutoSubClause can facilitate large-scale linguistic analysis and generate new insights across disciplines.
• Findings highlight the need for nuanced approaches to studying language acquisition and cross-linguistic influence.