BAS: An Answer Selection Method Using BERT Language Model

  • 2019-11-04 23:16:47
  • Jamshid Mozafari, Afsaneh Fatemi, Mohammad Ali Nematbakhsh
  • 7

Abstract

In recent years, Question Answering systems have become more popular and usedmore by users. Despite the increasing popularity of these systems, theperformance of them is not even sufficient for textual data and requiresfurther research. These systems consist of several parts that one of them isthe Answer Selection component. This component detects the most relevant answerfrom a list of candidate answers. The methods presented in previous researcheshave attempted to provide an independent model to undertake theanswer-selection task. An independent model cannot comprehend the syntactic andsemantic features of questions and answers with a small training dataset. Tofill this gap, language models can be employed in implementing the answerselection part. This action enables the model to have a better understanding ofthe language in order to understand questions and answers better than previousworks. In this research, we will present the "BAS" which stands for "BERTAnswer Selection", which uses the BERT language model to comprehend language.The empirical results of applying the model on the TrecQA Raw, TrecQA Clean,and WikiQA datasets demonstrate that using a robust language model such as BERTcan enhance the performance. Using a more robust classifier also enhances theeffect of the language model on the answer selection component. The resultsdemonstrate that language comprehension is an essential requirement in naturallanguage processing tasks such as answer-selection.

 

Quick Read (beta)

loading the full paper ...