Research And Implementation Of Chinese Question Answering System Based On Frequently Asked Questions

As a new type of question answering mode,Chinese question answering system relies on the technology of natural language processing and artificial intelligence,uses natural language to ask questions,fully understands user's intentions and returns answers in real time.It has the characteristics of high efficiency,accuracy and rapidity.Chinese question answering system based on FAQ is a kind of implementation mode with higher retrieval efficiency,which is suitable for restricted domain with strong professionalism.The concept of "Internet + education" promotes the further development of network education.Compared with the traditional education model,network education has incomparable advantages,and has its value and significance of existence.However,problems such as poor real-time performance and waste of educational resources exist in question and answer session of network education,which affect the learning initiative of students and the teaching quality of network education.In this paper,Chinese question answering system based on FAQ is applied to question and answer session of network education.The specific research work includes the following aspects:Firstly,the algorithm of category keyword extraction based on improved Text Rank is studied.In order to reduce the computational complexity of the algorithm,the candidate keyword determination algorithm based on improved ECE is used to filter common words and words with lower classification discrimination,and the candidate keyword set is obtained.Then the improved Text Rank is used to extract document keywords from the candidate keyword set,seeking a union of document keyword sets to obtain a category keyword set.The improvement of Text Rank includes four aspects: the construction of edges,the determination of edge weights,the initial weight assignment of vertices and the determination of random jump probability.Secondly,the method of question location based on category similarity and word sequence similarity is studied.According to the category keywords,the category semantics space is constructed,and the category similarity is obtained by calculating the similarity between the category vector and the question vector,and the candidate question sets are obtained by ocating the three categories with the higher similarity.Then the word sequence similarity is obtained by combining the similarity of word form and semantics on the candidate question sets.Lastly,according to the category similarity and the word sequence similarity,the question similarity is obtained,and the corresponding question-answer pairs of the three questions with the higher similarity are located in the FAQ.Finally,the effectiveness of the algorithm is validated by experiments,and the design and implementation of Chinese question answering system based on FAQ is completed,the practicability of the algorithm is verified in practical application scenarios.The experimental results show that extracting keyword on candidate keyword set can not only reduce the average time-consuming of the algorithm,but also improve the accuracy to some extent.When using improved Text Rank to extract keywords,the precision and recall rate of algorithm are significantly improved,however,due to the need to calculate word vector similarity,information entropy and expected cross entropy,the average time-consuming of improved algorithm increases.In addition,The p@k value of the category similarity calculation algorithm and mean reciprocal rank of the question location algorithm have good performance.According to the system test results,Chinese question answering system based on FAQ designed in this paper has good accuracy and response time.