English Short Text Similarity Research And Application Of Computer Book Recommendation

In recent years, accompanied by the rapid development of Internet, customers and the business have a strong requirement of recommender system. Most of the recommender systems are user-based nowadays. As the result of customers'less visit at the beginning of the system setting up, these systems will face the problem called "Cold Start". In this case, these systems can't play their role in recommendation. In contact, the recommender system based on content can solve the problem by focusing on commodity own properties not users.The content-based recommender system is based on text similarity study. Usually short text similarity built on the arithmetic for long text can't work well here. In this paper, short text similarity arithmetic with computer books as an example is proposed. Computer books corpus was snatched from Internet. With the process of splitting words, removing stop word, extracting words by statistics and rules, computer domain words were extracted. Then a level structure was built with the domain words just extracting. At last, a recommendation method is proposed combine with short text comparison arithmetic.Meanwhile, a recommender system is built on the theory we just discuss. The system takes users' demand into account and is designed properly followed by the practical operation circumstance. All the work we do provide a new thought for building e-business recommender system.