The Research And Implementation Of Query Expansion Based On Mongolian Semantics

With the continued growth of the Internet user’s number and website scale, the amount of information on the Internet increasing rapidly, so information retrieval technology appeals more and more people’s attention. Mongolian, as the main National Characters and the official language in Inner Mongolia Autonomous Region, has been widely used in various fields. Nowadays, Mongolian information retrieval becomes one of pressing issues which is needed to solve immediately, with the times of digital information arrives.The purpose of information retrieval is finding out the user’s required documents from various documents. The semantics associate has been ignored in the traditional approaches, which cannot express the user’s intention from the essence. In this thesis, we try to expand query start with Mongolian semantics.The basic thought of this paper is that, utilizes the Associate-Dictionary to expand the original query, and uses the expansion terms retrieve with different weights. In other words, we can analysis professional corpus by taking advantage of the IT. That we can find the expansion of terms which closely related to the query, and can rank&distribute according to related degree. The central algorithm of the method adopted in this thesis is to rearrange associated terms (Expected Rank) and re-weight expansion terms. The Rearrangement (Expected Rank) algorithm is by assigning different factors for rearrangement algorithm based on four traditional algorithms (Dice’s Coefficient, Mutual Information, Expected Mutual Information Measure, and Person’s Chisquaered (χ2) Measure), and finally get the associated terms’expected rank. The expansion terms re-weight algorithm is assigned different weights level to expansion terms with different rankings by utilizing proper mathematical functions. Then finish the retrieve with expansion terms.We will discuss the semantic space model window-size parameters, rearrangement(ER) algorithm parameters and the number of expansion terms in detail. After experimental trainings, we build a better retrieval performance Associated-Dictionary. The experimental results indicate that we can build Associate-Dictionary for a professional corpus to improve the retrieval performance. In short, experiments show that the method used in this article is effective.