Information Retrieval Framework Based on Filtering-HMM Algorithm

Today, with the rapid development of the Internet, text information is increasing rapidly. Therefore, there is a need to mine text information so that the relevant information can be detected and organized in the text. Especially, the multimodal information retrieval is mainstream, to solve the “heterogeneous gap” problem in cross-modal retrieval, most of the existing main methods rely on the idea of representation learning. Specifically, cross-modal retrieval algorithms generate isomorphic unified high-level feature representations for different modal data by setting up a model framework that can establish the semantic relationships between heterogeneous data. Hence, in this study, the filtering algorithm is firstly applied to clean the data sets, then, the modified HMM model is applied to achieve the information retrieval framework. In the experiment, from the aspects of accuracy and time usage, the proposed model is tested compared with the SVM and traditional HMM, the results show the proposed algorithm performs much better.