Question Generation Algorithm Based On Legal Debate Document

As data continues to grow at an exponential rate,more and more data-driven computing artificial intelligence models have been applied to various industries,such as medical,legal and other fields.Huge amount of knowledge,rules and patterns have been derived from those data,providing meaningful assistance to users in different industries.The artificial intelligence models are playing important roles in today’s world.During a court trial,the plaintiff and the defendant will first make their claims and arguments.Based on focus of the arguments,the judge will screen out the controversy focus of the dispute,clarify the cause of the case,and form basis for the final verdicts.This paper takes legal cases on private loans as research topic,and studies the methods of applying artificial intelligence algorithms to assist judges in case verdicts,especially the tasks to automatically generate leading questions.Leading question refers to the first question raised by the judge in a legal court trial against the controversy focus between the plaintiff and the defendant.In the private loan cases,the plaintiffs usually present comprehensive and complete evidence to the court.The judge can verdict a case based solely on the information provided by the plaintiff,without defendant’s claims.Therefore,the information presented by the plaintiff provides valuable "fuel" for the artificial intelligence algorithm to generate the leading question,thereby assisting the judge in final verdict.In practice,most of the defendants in private loan cases did not even appear in court to defend or respond.In conclusion,the focus of this research is to generate the leading question based on the content of the plaintiff’s claim,while ignoring defendant’s argument.In other words,the content of this article is to identify the controversy focus among plaintiffs,defendants and trial text,help judges to generate inquiries,and assist judges in case verdicts,using materials provided by plaintiffs.The main work of this paper is as follows:(1)Present a leading question generation method combining a seq2 seq attentional model and a pointer generator network.The advantage of this method is that it not only utilizes the generation ability of the sequence attention model,but also uses the copy function of the pointer network.Therefore,to some extent,the generated questions are consistent with the content of the plaintiff’s material,and the novelty of the generated questions is improved.Hereafter,the pointer generator network is referred to as PGN,and the pointer network is referred as PN.(2)Research the methods of mining controversy focus from trial text.The controversy focus of the trial dialogue is the gist of the debate between the plaintiff and the defendant in which the judge is concerned.This paper introduces the seq2 seq attentional model and PGN to mine the controversy focus.(3)We take the controversy focus as an auxiliary task and combine it with the leading question generation task to design a multi-task learning model.In this method,generation of the leading question and mining of the controversy focus are used as two tasks for collaborative learning,which effectively improves the effect of the leading question generation.