Algorithms Study On Answer Selection Task Using Deep Learning Technology

Answer selection task is an important foundation in the field of automatic question answering.It is a set of candidate answers for a given question and corresponding to the question.It can find the correct answer from the set,so it can be regarded as the problem of matching between the questions and answers.This paper mainly studies the way of introducing the context information of question-answer data set into the model and constructing knowledge network through in deep learning neural network,representing the question-answer sentences,and designing comparative experiments to prove the effectiveness of the model.The main work of this paper is as follows:1.Improve the current way of manual preprocessing of answer selection task,distinguish all the questions in the data set through the model,make full use of the existing data set while reducing labor costs,and enhance the practicability of the model.2.By introducing the context semantic information of the same candidate answer set into the encoding of answer sentences through the structure of neural network,each candidate answer is representated,which helps the model to model the semantic relevance between the question and the answer in a better way,thus enhancing the effect of the model.3.The knowledge memory network is constructed by using cyclic neural network,and the knowledge weight vector corresponding to the question is introduced to complete the re-representation of the question and answer vector.Experiments prove that the model can better learn the important semantic information between question and answer pairs.4.Combining context information with knowledge network,and using knowledge network and context information to re-represent the questions and candidate answers respectively,the experimental data reflect the effectiveness of context and knowledge network in improving task indicators.