A text implication relation recognition method based on multi-granularity information fusion

Abstract

The present invention provides a text implication relation recognition method which fuses multi-granularity information, and proposes a modeling method which fuses multi-granularity information fusionand interaction between words and words, words and words, words and sentences. The invention firstly uses convolution neural network and Highway network layer in character vector layer to establish word vector model based on character level, and splices with word vector pre-trained by GloVe; Then the sentence modeling layer uses two-way long-short time memory network to model the word vector of fused word granularity, and then interacts and matches the text pairs through the sentence matching layer to fuse the attention mechanism, finally obtains the category through the integration classification layer; After the model is established, the model is trained and tested to obtain the text implication recognition and classification results of the test samples. This hierarchical structure method which combines the multi-granularity information of words, words and sentences combines the advantages of shallow feature location and deep feature learning in the model, so as to further improve the accuracy of text implication relationship recognition.

Description

technical field [0001] The present invention relates to the field of natural language processing, and more specifically, relates to a method for identifying text implication relations by fusing multi-granularity information. Background technique [0002] Textual entailment recognition research refers to judging the implied relationship (implicative, contradictory, or neutral) between two given texts (premise text and hypothetical text). This is an important task in the field of natural language processing. Traditional research methods mainly rely on the support of feature engineering, external semantic resources and tools, combined with machine learning methods to complete the text entailment classification. This method based on feature classification usually requires more time and computational cost. With the excellent performance of deep neural networks in various tasks of artificial intelligence, the mainstream research work of text entailment recognition is to use deep ...

AI Technical Summary

Problems solved by technology

This method based on feature classification usually requires more time and computational cost

With the excellent performance of deep neural networks in various tasks of artificial intelligence, the mainstream research work of text entailment recognition is to use deep neural networks to model text sequences to complete the representation and matching of sentences. There are some shortcomings: First, the word2vec, GloVe or Str2Matrix used in the sentence representation process all rely on the existing corpus, and more and more new words, low-frequency words or compound words do not appear or rarely appear in the training corpus , and these words are not all included in the pre-training word vector, the large proportion of new word vector missing will affect the training effect of the model

The second is that the vector representation of two sentences in the mainstream method calculates the distance relationship matrix or the method based on text similarity does not have the ability of semantic reasoning

The third is that there is little discussion on the interactive information between the premise text and the hypothetical text. The existing technology is to encode or map the two texts separately, then simply aggregate the vectors, and then go through the deep sequence model and predict the classification. This process ignores the interaction information between sentences, etc.

Images