A Method Of Automatic Grammar Learning Oriented To Inquiry Sentences

Grammar has the outstanding characteristics which can deal with complex nested structure, a complete grammar library is a basic understanding of security issues automatically the quality and accuracy. If the statistical method can be introduced to the various stages of the design grammar, that will be able to greatly improve the overall efficiency and coverage grammar system. Grammar learning of inquiry sentences refers to learn the grammar after repeated experiments by studying methods of knowledge acquisition from such a document. With the development of information technology, the inquiry sentence sexists in the form of electronic documents is an important source for knowledge acquisition, the design by hand with great limitations grammar.Due to huge data of consulting statement documentation; direct grammar learning will waste a lot of time, so we propose two preprocessing methods. First, we proposed the concept of language-independent and language-independent recognition, and with the artificial collection of the language-independent as our seed language-independent, and then we propose a method based on seed language-independent recognition. Similarity calculating inquiry sentences would be more accurate after recognizing language-independent in consulting statements. Second, this paper presents hierarchical clustering method based on similarity function of2gram between sentences; the result is a collection, which composes by the inquiry sentences being roughly the same meaning.Pattern must appear frequently in the inquiry sentences, it is more characteristic of grammar. In order to find a special the internal relationship and laws in the natural language inquiry sentences between word-class, especially structural features and semantic relationships between words pattern, we use k-frequent sequences with an average spacing limit (k=2) coverage as a standard to judge the quality of the word pattern. First, we obtain k-frequent sequences through frequent sequences algorithm and calculates the coverage of each word pattern. In the vast corpus, we can’t manually read and extract the semantic association between the entries, but you can extract semantic relations probability and spacing between entries, experiments show that the method is effective. And on this basis, we last add process of ambiguous alternative Items.Grammar learning method proposed in this paper has good universality.it can be applied in the many field, such as telecommunication.