Quick vertical target recognition and classification method, classification system and classification device

Abstract

The invention provides an identifier classification method, and belongs to the technical field of internet content recognition and classification. The method comprises the steps of obtaining a labeledsample set, the labeled sample set having vectors of known class digital content of known class objects and vectors of known class identifiers of different hierarchies of the known class digital content; and obtaining a target class sample set and a target set with identifiers to be classified, and identifying the identifiers to be classified in the target set as target classes the same as the target class sample set according to a transfer learning method in combination with the labeled sample set.

Description

technical field [0001] The present invention relates to the technical field of Internet content identification and classification, in particular to an identifier classification method, an identifier application method, an identifier classification system, a classification device and a computer-readable storage medium. Background technique [0002] With the popularization and rapid development of the Internet, users are constantly looking for practical, practical and affordable information on the Internet. Users have more and more extensive demands for Internet information and hope that the information will become more and more abundant. Traditional web search is being upgraded to content search. The first generation of large and comprehensive horizontal websites (also known as comprehensive websites) can no longer fully meet the needs of users, and focus on specific areas or vertical resources that provide specific services. As a new bright spot of the Internet, it is attrac...

AI Technical Summary

Problems solved by technology

[0009] According to the above, the main disadvantages of the prior art solutions are the high labor cost, slow speed and narrow coverage of vertical resource identification:

[0010] (1) High labor cost: first screen the sites, and then label the links by the labeling team

Especially for the labeling process of a single link, a lot of manpower is required for multi-day labeling; in the process of delineating the site and expanding, it still needs a lot of manpower consumption; making a recognition model from scratch also requires a lot of large labeling samples, the effect A better deep learning model requires about 100,000 samples;

[0011] (2) Slow speed: In the labeling process, there are generally about a thousand links that can be marked at the day level, and it takes weeks or even months to realize the labeling only to identify vertical resources in a small number of delineated sites; every time the site needs to be expanded, it takes It takes a long period of time to complete the identification of vertical resources in the newly added site; training the model from scratch also takes a long time for labeling and training because of the large amount of data and many model parameters;

[0012] (3) Narrow coverage: firstly, manual delineation of sites can only obtain a small number of well-known large sites. Faced with such a wealth of Internet resources, manual coverage cannot be achieved in all directions; secondly, only part of the resources under many sites belong to the target vertical category , while manual screening and sampling tend to ignore these resources, training the model from scratch can theoretically improve coverage, but due to the large number of samples required and the long cycle, the accuracy of the initial model is low, and the coverage of vertical resources The situation is not ideal, and the time cost directly leads to the inability of the deep learning solution to classify and collect different types of vertical resources

Images