Design And Implementation Of A General Query Language For Academic Search Based On Automatic Question Answering

Academic search engine optimization has always been a key research direction for scholars.Among them,there are two sub-hot spots,one is the behavior analysis of user search,and the other is the innovation of the search method based on the improvement of functions.The research on query intent belongs to the analysis of user search behavior.There are many mature and widely used large academic search engines.Their functions and implementation methods are different,but users' query intents are homogeneous.How can we not only give a unified description of academic search intentions in theory,but also unify the functional requirements of the academic search engine during development,so as to form a reusable framework output? This research proposes a common query language to describe the academic search intention in a unified and standardized way,and applies it to the prototype of the academic search engine developed in this research for verification.At the same time,this research also explores the academic search method 's innovation-conversational search oriented to academic search tasks,applied to the prototype system instead of traditional text input.The general query language defined in this article——SSL is a domain-specific language in the field of academic search.It is a semantic representation of the user 's search intent in academic search.It includes the category of the purpose information,query method and query refinement.It makes different academic searches understand query intent in a unified way.When defining SSL,this article refers to the traditional design process of domain-specific languages.In terms of semantic definition,based on query log analysis and academic search engine functional analysis,the classification and composition of academic search query intentions are defined;in terms of grammatical definition,based on JSON format,the corresponding grammatical rules are described through ABNF.In the design and implementation of conversational academic search,the functional modules of the traditional task-oriented dialogue system used in this study include question understanding,dialogue management and search modules.For the intention recognition task in the question understanding module,this article has constructed a variety of classifiers and conducted intent recognition experiments based on the defined query intent categories.Among them,the model based on RNN has the best effect;not only that,this article also defines question templates to to describe the search method;in the dialog management module,a strategy based on finite state is adopted for state tracking,and a dialog is generated based on the knowledge base and template;before retrieval,the SSL mapping and parsing module is integrated into the system,with the help of that,user intent can be completely replaced by SSL expressions.Our method can be applied to different search engine implementations,providing a reusable reference for academic search from user input to general query language mapping.