Research On Question Answering Based Synchronous Collaborative Search System

Synchronous collaborative search system refers to a system that supports two or more users with similar search information need to interact simultaneously to perform collaborative search.Generally,synchronous collaborative search system provides a social engine to enable users within the same group to communicate.However,when the number of users in the social engine is insufficient,the social engine will encounter a cold start problem and cannot perform collaborative search well.In this paper,we propose a novel question answering based synchronous collaborative search system.We add a smart question answering engine Infobot into the social engine to support real-time interaction between users and the search engine,and address the cold start problem of the social engine.In the Infobot,we implement multiple rounds of interaction between users and the search engine through a reading comprehension model.We put the user's questions and related documents into the reading comprehension model,and the model gives the corresponding answers to the user.Our reading comprehension model is an improvement based on the Bert model.We propose two methods to improve the effectiveness of the model by jointly modeling the start and end positions and jointly modeling the answer content through conditional random field.Our synchronous collaborative search system provides a novel collaborative search pattern to support faster and smarter collaborative search.