Semi-supervised Extractive Question Summarization Using Question-Answer Pairs

Abstract

Neural extractive summarization methods often require much labeled training data, for which headlines or lead summaries of news articles can sometimes be used. Such directly useful summaries are not always available, however, especially for user-generated content, such as questions posted on community question answering services. In this paper, we address an extractive summarization (i.e., headline extraction) task for such questions as a case study and consider how to alleviate the problem by using question-answer pairs, instead of missing-headline pairs. To this end, we propose a framework to examine how to use such unlabeled paired data from the viewpoint of training methods. Experimental results show that multi-task training performs well with undersampling and distant supervision.

Semi-Supervised Extractive Question Summarizer Using Question-Answer Pairs and its Learning Methods

“…Gao (Liu, Li, Cao, Lin, Han, and Yu 2008;Morita, Sakai, and Okumura 2011;Chan, Zhou, Wang, and Chua 2012;Ishigaki, Huang, Takamura, Chen, and Okumura 2020) CQA (Surdeanu, Ciaramita, and Zaragoza 2008;Celikyilmaz, Thint, and Huang 2009;Bhaskar 2013) (Lei, Joshi, Barzilay, Jaakkola, Tymoshenko, Moschitti, and Màrquez 2016;Romeo, Da San Martino, Barrón-Cedeño, Moschitti, Belinkov, Hsu, Zhang, Mohtarami, and Glass 2016;Nakov, Hoogeveen, Màrquez, Moschitti, Mubarak, Baldwin, and Verspoor 2017) Ishigaki, T., Machida, K., Kobayashi, H., Takamura, H., and Okumura, M. (2020). "Distant Supervision for Question Summarization."…”