Research On Domain Knowledge Based Chinese Short Message Understanding Model

With the wide application of SMS(Short Message Service) in China and rapid increase of amount of SMS users, an expectant information query in a database with SMS has already become one very convenient way of extracting information between users and application database. But the majorities of the interface of database query require the users to write several short messages in a fixed form, which brings many troubles and inconvenience to user's query. Natural language interface of database query doesn't need the user to write short messages in the fixed form. But the difficulties in understanding natural language enable the Chinese interface of querying database slowly to achieve the practical level.Therefore, the research on understanding Chinese short message of querying database has a positive function for the Chinese interface of database query to achieve practical level sooner. The work in the dissertation as followings:In the view of characteristic of the words in query sentence in the railroad ticket service domain, the words related to the application are divided into different classes to form several domain dictionaries which embody characteristics of the domain. Herein, a reversion maximum match algorithm is used to divide automatically the query sentence into some single words.In the syntax structure analysis, a grammar definition that has been able to cover the majorities of the query sentence patterns in the railroad ticket service domain is designed by analyzing carefully query goal and the query condition in the query sentence. It is proposed that a top-down parsing algorithm with a chart structure is integrated with a semantical database. The semantical database is obtained while the syntax ingredient is analyzed by this method.In the SQL command construction, a grammar analysis tree with the database semantics is constructed as a middle to transform a natural language sentence into a SQL command. By searching the syntax tree in priority of depth of the syntax tree, the corresponding informationis extracted from the node in the grammar tree, and then the SQL command is constructed.A answer template is designed to construct the answer sentence correspondingly according to the different query aims. At last, The answer sentence is created by combining the template and the query results.After summarizing the algorithms of understanding Chinese sentences query interface of database, a set of practical and feasible understanding models are proposed in a railroad ticket service domain. This model can provide a kind of solution for Chinese natural language based database query system in other domains except Chinese natural language based railroad ticket service database query.