Multi-domain non-cooperative distributed retrieval result fusion system and fusion method thereof

Abstract

The invention discloses a multi-domain non-cooperative distributed retrieval result fusion system and a fusion method thereof. The system comprises a data processing module, a feature extraction module, a model training module, a central processing module and a correlation scoring module; the data processing module is used for storing corpus data and cached data needed by the system and providing interfaces for the outside to interact with other modules; the feature extraction module is used for extracting multi-domain features from training corpus or retrieval results; the model training module is used for training a model, wherein the training of the model includes model initialization and model learning; the central processing module is responsible for flow processing of distributed retrieval result fusion and interacting with the other modules and is the implementation part of the core logic of the system; and the correlation scoring module is sued for performing global correlation scoring on query results. With the multi-domain non-cooperative distributed retrieval result fusion system and the fusion method thereof of the invention adopted, the problem of incapability of directly comparing different retrieval results due to the difference of retrieval models under a non-cooperative environment can be solved, and the problem of complex parameter adjustment of a heuristic method can be also solved.

Description

technical field [0001] The invention relates to the technical field of distributed retrieval, in particular to a multi-field non-cooperative distributed retrieval result fusion system and a fusion method thereof. Background technique [0002] In distributed retrieval, the quality of retrieval results presented to users is closely related to the method of fusing query results from different resource bases, especially when the results come from non-cooperative search engines. The so-called non-cooperative search engines mean that different retrieval models are used between search engines. In this case, search results from different sources cannot be directly compared based on ranking or score. [0003] The current distributed retrieval result fusion system in a non-cooperative environment mostly uses a simple linear fusion method, and the model parameters are determined by a heuristic method. Although this type of method has a fast calculation speed and is conducive to the o...

AI Technical Summary

Problems solved by technology

Although this type of method has a fast calculation speed and is conducive to the online operation of the system, its accuracy is low, and the heuristic method is troublesome to adjust parameters.

Images