Research On Campus Q&A Based On Interactive Attention Network

In the life of campus,there are a lot of problems every day.The way to obtain information mainly uses search engines and information services.The traditional search engines are mainly based on keyword search and return a lot of irrelevant information,which leads to poor user experience.At the same time,the traditional information services have no unified information query entrance,which can not meet the needs of users to obtain information efficiently.Q & A is a novel way to obtain information,which can answer users' questions quickly and accurately.Applying Q & A to the field of campus,we can design a campus Q & A system,provide a unified entrance for consultation,return accurate information,and provide convenient information services for users.The core part of campus Q & A is to match the question with the database to get the answer information,while the secondary part is to analyze the types of questions to reduce the amount of matching tasks.This paper mainly studies campus Q & A from text matching and question analysis,the main work is as follows:(1)In this paper,the semantic matching algorithm for campus text matching is studied.At present,only from the perspective of word or character to analyze the Chinese text in the field of campus,can not get accurate semantic representation.For feature expressing and learning,this paper presents a semantic matching model based on interactive attention network,which is a hybrid representation of text from the perspective of word,character and position,and interactive learning of the relationship between features combined with the advantages of attention network.Firstly,the text is represented from the perspective of word and character.Secondly,the representation of word and character is encoded and learned through position embedding and long short term memory network to obtain the mixed representation of the text.Then,the interactive attention network is constructed to learn the association information of text features.Finally,the classifier is constructed to get the matching result.(2)In this paper,a fast short text classification algorithm for campus question analysis is studied.The essence of campus Q & A task is to get the most matching similar questions from a question set according to the questions.The number of question sets greatly affects the efficiency of Q & A.In the process of campus Q & A,if we classify the questions first,we can get the same kind of questions from the database through the categories of questions for semantic matching,which can reduce the number of questions to be matched.Therefore,this paper proposes a classification model based on self-attention network.First,the word order information of features is learned by N-gram,then a self-attention network is constructed to focus on important features to reduce the impact of redundant features on the classification task,and finally a classifier is constructed to get the classification results.(3)In this paper,the data set of campus field is constructed.At present,there is no public data set for the campus field,so this paper grabs the relevant data of forum and post bar through crawler technology,improves and labels the data with reference to other high-quality and open source data sets,obtains the campus common question pair data set,and applies the data set to the research of campus text matching.For the research of campus text classification,this paper annotates the captured data and obtains the campus common question data set.The construction of two campus domain datasets provides data support for this study.(4)In this paper,the campus Q & A system is designed and developed.This paper combines the IANSM model and SANCM model to complete the campus Q & A task.Firstly,the SANCM model is used to classify the questions,and then the information of the same kind is obtained from the database according to the categories.Finally,the questions and information are input into the IANSM model for matching,and the answers are returned to the user.At the same time,the system also provides the function of viewing personal information such as historical records and collection records,reading hot issues and expanding knowledge.