Research Of The Intelligent Retrieval Technique Of Agricultural Knowledge Q&A System Based On Ontology

The rapid development of the network technology speeds up the pace of information in various fields, and information technology gradually penetrates to the agricultural field, which has greatly promoted the process of agricultural informatization. During the "Twelfth Five-Year Plan" period, agricultural informatization has clearly become an important task for China’s social development and the national economy forward. Knowledge of agricultural field has the characteristics of regional, timeliness, and complexity, but farmers as the service principal are generally with low level of scientific cognition, so how to get the information that people need from the massive variety of data has been become a research focus in the Agricultural Information Service. The emergence and development of the Q&A system has improved the shortcomings of traditional search, which allows users to ask questions in natural language, and with the precise answer feedback, instead of pages and information. The existing answering system mostly uses a keyword matching questions during the answer retrieving, and rarely involved in the semantic level of understanding of sentences or words.In response to the above problems, this paper introduces the concept of Ontology, applying the Agricultural Ontology into the organization and management of agricultural knowledge; provide a knowledge semantic network infrastructure for the retrieval of the Q&A system. Ontology is standardized description of the relationship between the concept, which builds a knowledge system of certain field, to let the knowledge possess better sharing and reusability. The Agricultural Ontology is to organize the relationship between the concepts in agricultural knowledge field, by using the formal description language recognized by computers. The introduction of the Ontology solves the weakness of the semantic understanding to a certain extent.Take the citrus as example, to build small citrus pest knowledge ontology to assist the accomplishment of the understanding, information retrieval and answers extraction of the answering system. On the basis of analyzing citrus pest domain knowledge, this paper defines the formal and the metadata semantic relations of the citrus pest knowledge ontology, determines the ontology construction framework and core body, and achieves the construction of the citrus pest knowledge ontology by using the ontology development tool of Protege.According to the characteristics of agricultural field, this paper has pretreated the questions raised by users, and used a particular sub-word for word segmentation. Question Semantic understanding based on domain ontology, extracts the key concepts and their extended concepts of the questions. Using Lucene to build inverted index for the problem-answer library, design retrieval program and determine the set of candidate issues.Combined with similarity algorithm of the existing sentence and on the basis of the ontology, this paper discusses the semantic similarity algorithm of concept based on domain ontology and the HowNet, and proposes a multi-information fusion model for sentence semantic similarity calculation, which considers the sentence’s surface similarity and semantic similarity. Similarity Calculation of users’ questions and candidate sets of questions, the answers to questions which reach the threshold return to the users in order. Finally, the paper introduces the system overall design, and proves the correctness and the validity of the model by experiments.