A Case Study on Chinese Text Information Filtering Method Based on User Ontology Model

With the dramatic increase of information on Web, text filtering is a key technology in content processing. Now more and more researchers know that filter must be based on the meaning of the word but not on a specific sequence of signs, so they pay more and more attention on filtering method based on semantics recently. However, the usability of these methods is inconvenient up to now, because they must be supported by plentiful rules and domain knowledge. To improve filtering precision and recall, this paper presents a novel method to text information filtering based on user ontology model (UOM), so that the machines can understand user´s requirements and text content to some extent, and the filtering results are more suitable for the users´ requirements. The method includes user model building, text structure analyzing, text conception extracting, semantic correlation computing and so on. The filtering method based on UOM can not only express the complex requirements but also avoid building domain ontology, so its efficiency and usability have a great promotion. In order to enlarge the filtering result set and to improve the recall rate, the virtual relationship, a kind of fuzzy semantic relationship, is introduced in the semantic similarity estimation effectively. This method is applied to the system of intelligent on-demand services for teaching resource on Internet. The results show that it can provide filtering services effectively. Although the experiments are only tested in Chinese text documents, these methods can be employed in other languages.