Question Answering over Electronic Devices: A New Benchmark Dataset and\n a Multi-Task Learning based QA Framework

Le résumé fourni par la source

Answering questions asked from instructional corpora such as E-manuals,\nrecipe books, etc., has been far less studied than open-domain factoid\ncontext-based question answering. This can be primarily attributed to the\nabsence of standard benchmark datasets. In this paper we meticulously create a\nlarge amount of data connected with E-manuals and develop suitable algorithm to\nexploit it. We collect E-Manual Corpus, a huge corpus of 307,957 E-manuals and\npretrain RoBERTa on this large corpus. We create various benchmark QA datasets\nwhich include question answer pairs curated by experts based upon two\nE-manuals, real user questions from Community Question Answering Forum\npertaining to E-manuals etc. We introduce EMQAP (E-Manual Question Answering\nPipeline) that answers questions pertaining to electronics devices. Built upon\nthe pretrained RoBERTa, it harbors a supervised multi-task learning framework\nwhich efficiently performs the dual tasks of identifying the section in the\nE-manual where the answer can be found and the exact answer span within that\nsection. For E-Manual annotated question-answer pairs, we show an improvement\nof about 40% in ROUGE-L F1 scores over the most competitive baseline. We\nperform a detailed ablation study and establish the versatility of EMQAP across\ndifferent circumstances. The code and datasets are shared at\nhttps://github.com/abhi1nandy2/EMNLP-2021-Findings, and the corresponding\nproject website is https://sites.google.com/view/emanualqa/home.\n

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.