The Application Of Sentence Similarity In The Intelligent Question Answering System

As an essential activity of instruction, question answering become is receiving more and more attention. How to develop an automatic question answering system running on the Internet becomes a hotspot in the field of intelligent tutoring. Some representative algorithms, such as, the keywords matching algorithm and the sentence similarity algorithm are implied to improve the intelligence of the system, which is the core of the current research in and out of the country. Although those algorithms mentioned above possess characteristic of intelligent, they inherently have defects in low accuracy rate of answer and unfriendly interface. To improve this situation, a novel closed-domain question answering system is constructed, which combines with features of distance education based on the principle of the intelligent systems.There are two main improvements in this paper.1. By analyzing approaches of sentence similarity computation mentioned above and its shortages, a new sentence similarity computation is proposed, which ensure the similarity between original sentence and its changed one whose clause or phrase moves much. The algorithm has advantage over others algorithm in respect of intelligence, preciseness, and practicability.2. To improve the efficiency of searching for the most similar sentence by traversal, a search algorithm based on inverted index and sentence length index was proposed. And the average search time of this algorithm is unremarkably affected by the size of the corpora over that of other algorithms.