The Discussion Tracker Corpus of Collaborative Argumentation

• DOI: 10.48550/arxiv.2005.11344
• Publication Date: 2020-01-01
• Authors: Christopher Olshe...; Luca Lugini; Ravneet Singh; Diane Litman; Amanda Godley
• Keywords: FOS: Computer and information sciences; Computer Science - Computation and Language; 4. Education; 05 social sciences; 0503 education; Computation and Language (cs.CL)

Although Natural Language Processing (NLP) research on argument mining has advanced considerably in recent years, most studies draw corpora of asynchronous and written texts, often produced by individuals. Few published synchronous, multi-party argumentation are available. The Discussion Tracker corpus, collected American high school English classes, is an annotated dataset transcripts spoken, argumentation. corpus consists 29 discussions literature transcribed from 985 minutes audio. were for three dimensions collaborative argumentation: moves (claims, evidence, explanations), specificity (low, medium, high) collaboration (e.g., extensions disagreements about others' ideas). In addition to providing descriptive statistics the we provide performance benchmarks associated code predicting each dimension separately, illustrate use multiple annotations improve via multi-task learning, finally discuss other ways might be used further NLP research.