Research On Ontology Based Mobile Question Answering System

With the wireless technology developing dramatically and the wireless devices becoming popular gradually, the wide wireless applications with increasing demand promote mobile information systems to become a research hotspot at present, especially the mobile network based information search. Answering questions automatically is an important form of information search. An ontology based text classification model and an improved sentence similarity calculating model based on ontology and How-net are proposed to help answers retriving with high efficiency and precision. An experimental mobile question answering system with restricted domain knowledge is designed and implemented, to which the two models are applied. Several experiments are designed and carried out for verification. IR (Information Retrieval) technology is utilized in this system to find the most similar question from the FAQ (Frequently Ask Question) corpus with the user's question. This prototype system function modules mainly include three parts as follows:Ontology construction, management and storage;Question-answer sets classified according to topics;Automatic question answering.Firstly, an overview of the researches on ontology and IR technologies up to now both for domestic and overseas is given. The general steps of building ontology knowledge are introduced.Secondly, the key technologies of ontology based mobile question answering system are introduced as follows:Using RFSA(Reversed Finite State Automata) to parse words.Integrating VSM (Vector Space Model) and ontology, the hierarchy classification method is applied to complete text topic classification.Extending queries based on the concepts' relationship of ontology.Analyzing and improving the concept similarity calculating based on geometrical distance, a sentence similarity calculating model based on ontology and How-net is proposed.Thirdly, a prototype system is discussed which is in C/S (Client/Server) architecture. The OWL (Web Ontology Language) ontology file is mapped to database using Jena API. The application system developed by J2ME is running in the mobile phone. This client applications access the Web Server in which a Servlet process is running. The Servletanalyzes users' questions, extends queries, matches the most similarity question, orders the results and then returns the final results to the users.Finally, several experiments are designed to calculate the precision rate and recall rate of some samples to verify the topic classification model. Comparing with the traditional concept similarity calculating model and calculating the precision rate and response time of the question answering system, the results show that the models are feasible.