Research On Chinese Frequent Question-Answering System Based On Semantic Comprehension

With the deeply development of internet, the research on Question-Answering(QA) becomes the most active aspect in domain of Natural Language Processing (NLP). The application of QA system based on the Frequently Asked Questions(FAQ) is becoming more and more extensive simultaneously. Some technology is related to it such as knowledge representation and information retrieval. It is applied to in some regions such as education on internet and client's service of company. The arithmetic of similarity degree is the key of Chinese information processing. It directly decides developmental status of some fields named automatic question answering, machine translation, information retrieval and automatic text summarization. The method of similarity degree is an important aspect of natural language research for long time.In the research of traditional Chinese QA based on FAQ, some technology of similarity degree which is based on the key words or based on the statistics is often applied, but the effect is not perfect. The calculation method of the similarity degree is the most important factor. In this paper, HowNet is the basis of the similarity degree calculation and the semantic structure is analyzed detailedly. Then the module of similarity degree calculation based on HowNet is actualized and a Chinese QA system based on FAQ is also realized.The phenomena of multivocal word are prevalent in the natural language. It restricts the development of similarity degree method and affects the performance of QA based on FAQ. To settle this problem, the technique of Word Sense Disambiguation is applied in this paper. The module of word sense disambiguation based on HowNet is realized. Moreover semantic files in HowNet are amended to improve the performance of the module of word sense disambiguation. The author combines the advanced technique of word sense disambiguation and the arithmetic of semantic similarity degree based on HowNet. Then a practical Chinese QA system based on FAQ is actualized grounded upon the combined arithmetic.In this paper, the test set of the module of similarity degree and word sense disambiguation is all collected and compiled by the author. The test set of QA base on FAQ is offered by the Network Education Institute of Northeast Finance and Economics University. The results of the experiments show that the precision of the similarity degree calculation is increased. It is said that the advanced method of semantic similarity degree is superiorer. At the same time the total performenc of FAQ system also can be applied to the practical...