Training method and device for word sense disambiguation model

Abstract

The embodiment of the invention provides a training method and device for a word sense disambiguation model, and the method comprises the steps: obtaining a word co-occurrence graph and a semantic association graph; and selecting a first word from the training text; obtaining a positive example sample and a negative example sample corresponding to the first word; calculating the similarity betweeneach word in the training text and the word represented by each node in each semantic association graph, and selecting a target association graph based on the similarity;based on the target association graph, determining a semantic vector of the first word, and based on the word co-occurrence graph, determining word vectors of other words; based on the determined semantic vector and word vector,encoding by utilizing an encoder; based on the word co-occurrence graph, determining a word vector of each word in the two samples; carrying out encoding by using an encoder according to the determined word vecto; based on the coding result, calculating a first text distance between the training text and the positive example sample, and calculating a second text distance between the training textand the negative example sample; and training an encoder by taking the condition that the first text distance is smaller than the second text distance as a target.

Description

technical field [0001] One or more embodiments of this specification relate to the field of computer technology, and in particular to a method and device for training a word sense disambiguation model. Background technique [0002] Word sense disambiguation refers to automatically judging the current meaning of a polysemous word based on the context of the word. In traditional techniques, word sense disambiguation is usually based on supervised learning methods. For example, based on the context C, the posterior probability P(s_i|C) of each sense s_i of the word to be disambiguated is obtained through a supervised learning method. The meaning s_k=argmaxP(s_i|C) of the maximum posterior probability is taken as the meaning determined after disambiguation. [0003] However, when word sense disambiguation is performed by this method, the disambiguation results are usually not accurate enough. Therefore, a more accurate word sense disambiguation method needs to be provided. ...

AI Technical Summary

Problems solved by technology

[0003]However, when using this method for word sense disambiguation, the disambiguation results are usually not accurate enough

Images