Literature retrieval method based on semantic small-word model

Abstract

The invention discloses a document retrieval method based on a semantic small-world model. Firstly, a latent semantic index is used to extract document feature vectors, and the dimensionality thereof is reduced on the premise of retaining document features as much as possible, so as to reduce the amount of document information storage; and then a support vector machine is used. Classify all the shared documents of the node to form classification information, which marks the interest ratio of the node to the literature category; finally, using the small-world phenomenon in the social network, all the nodes in the peer-to-peer network have direct connections. There are few short link nodes with similar interests, and there are very few long links with a very high interest ratio in a certain literature category, forming a network topology with semantic small world characteristics. The invention is characterized in that the query message is routed to the node most likely to answer the request, improving query efficiency; making full use of long links, the query statement can be quickly routed to other parts of the network, improving the recall rate and reducing network load.

Description

technical field [0001] The invention belongs to distributed computing and information retrieval in the computer field, and specifically relates to a document retrieval method based on a semantic small world model, which mainly uses the semantic small world model to solve efficient information storage and retrieval in a peer-to-peer network for document information sharing question. Background technique [0002] Peer-to-peer network system has attracted more and more attention in the field of large-scale information retrieval due to its characteristics of scalability, fault tolerance, autonomy and self-organization. However, in the peer-to-peer network for document information sharing, how to store and retrieve information effectively is still a very challenging problem. [0003] The small world phenomenon widely exists in social networks, that is, everyone in the world can be connected through a very short social relationship chain, and the length of the social relationship...

AI Technical Summary

Problems solved by technology

However, the above methods all require precise metadata matching (such as file names or keywords) to complete the search requirements. Since the semantic information of other nodes in the network cannot be obtained, it is necessary to blindly search a large number of nodes to ensure the recall rate of information retrieval. causing severe network load

Guiding query messages through improved neighbor node index information (such as local indexes) can improve query performance, but updating index information requires very large additional overhead

A structured peer-to-peer network (such as CAN, Chord) based on a distributed hash table can provide good scalability and effective search performance, but it can only support the keyword / value lookup method. For the information retrieval field Full-text search is not suitable, and the overhead of maintaining a structured peer-to-peer network structure is very high

Images