Design And Implementation Of Intelligent Question Answering System In Automobile Field

In the field of automobile customer service,the number of online users has grown rapidly,resulting in a surge in the workload of online customer service.The intelligent question answering system can greatly reduce the workload of online customer service and provide users with car question answering services.However,the following problems still exist:（1）Under the condition of high concurrent requests,the system response time cannot meet the requirements of low online delay of the system.（2）The data fusion between the system and the automotive field is not close enough,resulting in inaccurate answers.In order to improve the performance and accuracy of the intelligent question answering system in the automotive field,a multi-level parallel cluster system structure is designed and implemented in combination with natural language processing and information retrieval technology to improve the system performance.The system consists of three clusters,including question answering system cluster,inverted index table cluster and reading comprehension model cluster.At the same time,a Reading Comprehension（RCAEE）model based on the embedded names of entities in the automotive field is designed and implemented to improve the accuracy of answers.In order to improve the performance and accuracy of the intelligent question answering system in the automotive field,a multi-level parallel cluster system structure is designed and implemented in combination with natural language processing and information retrieval technology to improve the system performance.The system consists of three clusters,including question answering system cluster,inverted index table cluster and reading comprehension model cluster.At the same time,a Reading Comprehension（RCAEE）model based on the embedded names of entities in the automotive field is designed and implemented to improve the accuracy of answers.First of all,a question input and answer display module is designed to provide users with visual interactive pages to complete the sending of questions and the display of answers.At the same time,the questions will be sent to the question answering service on the question answering system cluster,and the question answering service will realize all subsequent functions of the system.Secondly,the question semantic analysis module is designed to preprocess the question and correct the text errors in the question.At the same time,the car entities in the question are extracted to provide the basis for the answer generation.Thirdly,the answer retrieval algorithm is designed to retrieve the candidate answers from the inverted index table cluster and narrow the answer range.With the help of text similarity algorithm,the candidate answers are refined to further narrow the answer range.Then,use the designed reading comprehension model to obtain accurate answers,deploy the reading comprehension model on the reading comprehension model cluster,and call the model on the cluster to reduce the system response time.Finally,a question answer generation module is designed to generate the final answer according to the exact answer obtained,combined with the matching question template and the extracted car entity.The actual system test results show that the system obtained 86.21 F₁ scores on the marked 6182 automobile question and answer test data sets.Under 50 concurrent requests,test the interface for 1000 times,and the average response time is 192 ms.The system meets the application requirements in terms of function and performance.In addition,on the car question and answer test data,the RCAEE model is 0.65 points higher than the SOTA model in reading comprehension,F₁ is 1.05 points higher than EM.