Study And Implementation Of Text Recommendation System Based On Similarity Of Word Sense

With the development of information technology, “information overload” phenomenon caused by massive text messages becomes more serious. How to efficiently get the interested texts from massive alternative text sets becomes a new research hotspot in the field of information retrieval. The content-based text recommendation system is a kind of personalized information retrieval system. It first extracts the user interest from the texts which have been read by the user, and then recommends some texts to the user based on the similarity between the user interest and target texts. Compared with the traditional search engines, recommend system can efficiently meet user demand for personalized information, but common text recommendation system only considers the strict matching of morphology between feature words, and does not consider the effect of context on morphology and synonyms. To solve the problem, this thesis studies and implements the text recommendation system based on the similarity of word sense. The main contributions of the thesis include:1. When using the synonymous relationship in the text recommendation, the weak synonyms and polysemy relationship may introduce errors in the calculation of similarity between user interest and document feature, degrade the accuracy of recommendation. The thesis takes the ideological of link prediction, optimizes synonym network from the perspective of network, and presents a method of acquiring strong synonyms network based on link prediction.2. Based on the strong synonyms network, combined with the shortest path optimal matching algorithm and the application of lemmatization, the thesis presents a text recommendation algorithm based on strong synonyms network.3. Based on the recommendation algorithm mentioned above, IOCP communication model, multi-client multi-server C/S architecture, load balancing and other technologies, the thesis designs and implements a text recommendation system based on similarity of word sense.The thesis firstly studies the information retrieval model and methods to structure represent the text; implements a basic content-based recommendation method and system; after studying the way to use synonym relationship in text recommendation, the thesis puts forward a method to obtain strong synonym network from the perspective of network to avoid the adverse effect of weak synonyms and polysemy relationship; then presents a text recommendation algorithm based on strong synonyms network, the simulation results verify its better performance in text recommendation; finally design and implement the recommendation system based on similarity of word sense.