With the rapid development of the Internet, people enjoy the convenience of rapid circulation of information, but they also have to face explosive growth in the amount of data. In the era of popularization of the Internet and explosion of knowledge, books borrowing, online shopping, information searching, and so on can all be completed over the Internet. In such an era, what methods should be used to locate useful information and what users should face in the search for information, that is how to filter out hundreds of useful sources information for users. The user must judge the information that he wants in his own right. In the judgment process, he will face the problem of information management and filtering. This study based on a multi-dimensional information retrieval framework and aims to propose a universal adaptability multi-dimensional information retrieval framework and designed a retrieval system using catering information services as an example. In the aspect of knowledge representation, this study used the three dimensions of time, geography, and intent to conduct data search for information retrieval frameworks, supplemented by Memory-based Reasoning (MBR), an analysis method in Data Mining, was used to learn userʼs habitual intent from past search cases and predict possible future search results that are timely, appropriate and adaptive. In practical design, this study designed a set of catering information recommendation platform to demonstrate the feasibility of the adaptive multidimensional information retrieval framework.