Data label mining method and data label mining system

Abstract

The invention provides a data label mining method, which comprises the following steps that: the number of occurrence times of open data labels obtained from Internet in advance and basic data labels in a pre-built basic data label system are counted; in addition, the number of co-occurrence times of the open data labels and the basic data labels are counted; the confidence degree is obtained by utilizing the counted number of occurrence times and the counted number of co-occurrence times; and when the number of co-occurrence times is greater than a preset first threshold or the confidence degree is greater than a preset second threshold, or the confidence degree is greater than a sum value of the average confidence degree and the confidence degree standard deviation, the open data labels are used as characteristic labels of the basic data labels to be added to a basic data label system. The invention also provides a data label mining system. According to the technical scheme provided by the invention, the hierarchical mining of the data labels can be realized, and the hyponymy accuracy of the mined data labels can be improved.

Description

【Technical field】 [0001] The invention relates to the field of Internet applications, in particular to a data tag mining method and system. 【Background technique】 [0002] At present, there are mainly the following two data label mining methods: [0003] The first labeling method is to manually construct a data label system, and achieve the effect of describing the data label system by formulating the upper and lower relationship of two data labels. For example, manually label the following data labels with the upper and lower relationship: [0004] Food [0005] food, restaurant [0006] food, restaurant, Chinese restaurant [0007] food, restaurant, western restaurant [0008] food, snacks [0009] food, bar [0010] Among them, each row of data labels describes the upper and lower relationships among a group of data labels. [0011] The second labeling method is to build a data labeling system on the Internet. Internet websites classify and index Internet pages. Fo...

AI Technical Summary

Problems solved by technology

[0014] Among them, the first method of manually constructing the data labeling system requires manual construction and the accuracy of the upper and lower relations, which brings a large workload and high labor costs. Therefore, the data in the manually constructed data labeling system The number of tags is small, and the upper and lower relationships between data tags are relatively simple, resulting in artificial construction results that cannot cover most of the data tags, so that the reasoning and verification of data tags cannot be realized

Although the second type of network construction data labeling system is to automatically construct a data labeling system, the accuracy of the upper and lower relations between data labels is low, and the construction results cannot be used for reasoning and verification

Images