| Question answering system (QA) is one of the most important research subjects in nature language processing domain. It’s a research hotspot beyond the seas, but it’s at the initial stage in our country. In our country, many organizations who do research on QA are universities and research institutions, there’s a few of companies doing this. QA allows users ask their question with nature language, and it can give the user an accurate answer.There are two key technical points in QA:the first one is how to retrieve the answer according to user’s question? The existing QAs use vector space model as the sentence similarity algorithm while retrieving the answer, but the traditional vector space model does not consider the semantic meaning of words. To fix this issue, this thesis improves the traditional vector space model, and proposes a sentence similarity algorithm based on vector space mode with semantic meaning of word. This new algorithm adds word similarity into traditional Vector Space Model. The experiment proves that the new sentence similarity algorithm is better than traditional vector space model, it can get more accurate result. The second question is how to show the answer to the user properly? This thesis brings flow template engine into QA, and uses this engine to render the flow type answer. This engine makes QA more interactive.At last this thesis designs and implements a QA which focuses on an e-commerce site. This QA allow users ask their questions by using nature language, and then use the new sentence similarity algorithm to retrieve the answers corresponding to users’questions, at last show the answer to users. The implement of the QA proves the first two points are feasible in practice. |
