A sequence marking system and method

Abstract

The invention relates to a sequence marking system, which comprises a model marking module, an adjustment module, and a strategy base; an output end of the model marking module and an input end of theadjustment module. The model annotation module is used to annotate the input text data serially. One or more policies are stored in the policy library, and the adjustment module is used for retrieving the policies from the policy library, and adjusting the annotation results output by the model annotation module according to the policies and the input text data. As that system or the method of the invention label the text sequence, the accuracy and the applicability of the original model labeling can be enhance.

Description

technical field [0001] The invention relates to the technical field of natural language processing, in particular to a sequence labeling system and method. Background technique [0002] Most of the knowledge and information of human society are recorded in the form of text. These knowledge and information are described in the form of human language, which cannot be directly recognized by machines. Natural language processing is an algorithmic technology for processing human natural language texts, among which Words Segmentation, POS Tagging and Named Entity Recognition are the basic tasks. Word segmentation is to divide a sentence from a sequence of words into a sequence of words; part-of-speech tagging is to assign a part of speech to each word, such as nouns, verbs, adjectives, etc.; named entity recognition is to extract specific types of nouns in the text, such as " Xiao Ming" (type: person name), "this morning" (type: time). Word segmentation, part-of-speech tagging,...

AI Technical Summary

Problems solved by technology

However, due to the large differences between the new data and the training data, such as the appearance of some proper nouns (such as the name "Abraham", which does not exist in the training data), or the incomplete coverage and uneven distribution of the training data, it will lead to training A good model cannot process some texts correctly, and re-labeling data training is time-consuming and laborious

Images