Research And Implementation Of Intelligent Question-and-answer System For Real Estate

The quality of customer service will directly affect customers' satisfaction with the company's products.With the explosive growth of Internet users,traditional manual customer service cannot cope with a large number of users,and cannot solve the problem all day.The efficiency of manual customer service cannot be guaranteed,and traditional manual customer service requires a lot of manpower and financial resources.Therefore,designing and implementing an intelligent Q&A system to replace or assist manual customer service personnel can greatly reduce the costs and improve the efficiency.Intelligent Q&A system can be divided into open domain Q&A system and closed domain Q&A system in the application domain.According to technology of Q&A system,it can be divided into Q&A system based on retrieval(answer selection sort),Q&A system based on ontology database or knowledge map,and generative Q&A system based on deep learning models.This thesis ueses two Q&A system solutions,namely FAQ Q&A solution and ontology Q&A solution,to realize the intelligent customer service Q&A system for the real estate field.The main research contents of this thesis are as follows:(1)This thesis proposes a problem classification method that integrates multivariate feature fusion and named entity recognition.When choosing the two schemes of FAQ and ontology,this thesis uses the method of question classification as the important basis for scheduling the two schemes.Named entity recognition adopts LSTM+CRF method.Text classification adopts the method of multi-feature fusion model,and the prediction probability of text classification is weighted and corrected by the result of named entity recognition.It is proved by experiment that the problem classification method that integrates multivariate feature and named entity recognition is better than other single text classification models based on deep learning.(2)In this thesis,an improved deep learning model is proposed on the basis of Matchpyramid model and BIMPM model.In FAQ Q&A,semantic similarity calculation is the core content.With the rapid development of deep learning in recent years,the performance of semantic similarity calculation method based on deep learning model has almost crushed the traditional NLP methods.In this thesis,the improved deep learning model is used to implement the semantic similarity computing task,and the experimental results show that the performance of the model in the test set is better than Matchpyramid model and BIMPM model.(3)This thesis adopts both FAQ and ontology methods.Because FAQ method,that is,the solution based on retrieval,can only extract answers from existing answers and cannot completely cover some questions about real estate information,so the ontology solution is introduced.The ontology Q&A method is a supplement to the FAQ Q&A method.In the ontology Q&A scheme,the ontology database is constructed manually according to the corpus after word segmentation.Keyword matching and named entity recognition are carried out on the sentences of the question,and the results are used to construct SPARQL expressions to complete the retrieval of the ontology library,so as to obtain the user's answers.And based on FAQ Q&A and ontology Q&A two methods,design a complete Q&A system for real estate.