Online retrieval method and system

Abstract

The invention discloses an online retrieval method and system. The method comprises the steps of constructing query points based on query operation; searching a neighbor point of which the direction is closest to the query point in a direction navigation map; and determining at least one return point closest to the query point based on the operation of searching the neighbor point closest to the query point in the direction. According to the method, the concept of the direction navigation graph is introduced on the basis of the direction vector, and the attribute that the direction navigation graph can be used for performing direction pruning on neighbors is utilized, so that the graph retrieval efficiency is greatly improved, and the method can adapt to online retrieval application scenes with larger data scales and higher data dimensions.

Description

Technical field [0001] The present disclosure relates to the fields of retrieving and recommendations, in particular to an online search method and system. Background technique [0002] In recent years, in search and recommendation related areas, vector retrieval applications are more extensive, and an important research direction in vector retrieval are the neighbor search of Approximate K-neighbor graph (knn graph), namely Approximate Nearest Neighbor Search (ANNS) problem. Especially under high dimensional and large-scale data scenarios, how to improve the search efficiency of the graph has always been a research hotspot in the field of graph retrieval. [0003] Optimization of existing neighbor retrieval is often concentrated on the original KNN construct. However, in conventional KNN, the higher the vertex in the figure, the worse the retrieval performance, and it is impossible to meet the needs of large-scale efficient retrieval, especially online retrieval. [0004] To thi...

AI Technical Summary

Problems solved by technology

However, in conventional kNN, the higher the degree of the vertices in the graph, the worse the retrieval performance, so it cannot meet the needs of large-scale and efficient retrieval, especially online retrieval.

Images