Research On Visualization Platform Of Agricultural Text Information Retrieval

With the development of the agricultural informatization, more and more agricultural users want to find the agricultural information which they need quickly and efficiently. General search engine can not meet the requirements of agricultural users in information retrieval precision, the current exists’ agricultural search engine provide users with the retrieval results according to the input search keyword. However, due to the words are often exist ambiguity in natural language, the search results are relatively scattered and users would take more time to find their real subjects of interest.In order to solve these problems has mentioned above, this paper has do some research on agricultural information vertical search engine model, takes advantage of the Journal of agricultural technology in the Wanfang Data Knowledge Service Platform as its information sources and improves the retrieval results.This paper completed the following tasks:(1) The research on methods to dynamically obtain agricultural domain concepts. Web data extraction, Chinese word segmentation and cleaning technology are used to obtain the candidate domain concepts, which based on all kinds of agricultural web literatures. Experimental results are given to show the extracted precision is above 95%, and the value of F-index is around 85%. The success ratio is improved by more than 9% after the using of failures retry mechanism. At the same time, using improved forward algorithm for maximum matching word segmentation improved the agricultural words recognition correct rate to 87.03%.(2) The research on the agriculture text information visualization model. Firstly, based on the information visualization model, this paper is to discuss a construction of agriculture text information visualization model, which included the information entity, the association with the information entity and the network structure. Secondly, discovering the relationship between domain concepts with the pretreatment of visual data and web data mining technology. Lastly, the system is used to implement visual effect for the retrieval results. Experiments showed that, compared with the primary hierarchical cluster algorithm, this method improves the effect of relation clustering between the domain concepts and reduces total consuming time. The value of F-Measure was increased from 0.675 to 0.751, and the execute time reduced from 52.893 s to 16.342 s.(3) Designed and implemented the visualization platform of agriculture information retrieval. The object-oriented programming is used to design and realize the final system, which provides users with lots of convenience such as dynamically obtaining agricultural domain concepts, information visualization and optimizing the retrieval process. Software tests show the system has achieved the prospected requirements.