A natural language reasoning method for hierarchical semantic representation of image enhancement

Abstract

The invention discloses a natural language reasoning method of image enhanced hierarchical semantic representation, which comprises the following steps: acquiring natural language sentence pairs withheterogeneous data structure and corresponding image information; obtaining natural language sentence pairs with heterogeneous data structure and corresponding image information. The natural languagesentence pairs are semantically modeled at different granularities of word level, phrase level and sentence level respectively, and corresponding semantic representations are obtained; the semantic representations at word level, phrase level and sentence level are enhanced by using the corresponding image information. According to the semantic representation of word level, phrase level and sentence level after enhanced processing, the representation vector of natural language sentence pair is processed by the matching method in natural language reasoning, so as to judge the inference relationship between two sentences in natural sentence pair. The method realizes the comprehensive and accurate understanding and representation of sentence semantics, and then efficiently models the semanticinteraction between two sentences, and finally accurately judges the semantic inference relationship between the two sentences.

Description

technical field [0001] The invention relates to the technical fields of deep learning and natural language understanding, in particular to a natural language reasoning method for image-enhanced hierarchical semantic representation. Background technique [0002] Natural Language Inference (NLI) is an important part of the field of natural language understanding. The main problem to be solved is to judge the semantic inference relationship between the Premise Sentence and the Hypothesis Sentence. The relationship is mainly divided into three categories: 1) Entailment: the semantics of the assumption sentence can be inferred from the semantics of the premise sentence, 2) Contradiction: the semantics of the assumption sentence cannot be inferred from the semantics of the premise sentence; 3) Neutral: it cannot be judged The semantic relationship between the hypothesis sentence and the premise sentence. Therefore, a primary problem to be solved in this task is the semantic repre...

AI Technical Summary

Problems solved by technology

In real life, the semantic expression of a natural language sentence is highly dependent on its external context. The same sentence can express different meanings depending on the external environment. Therefore, the semantics of a natural language sentence is complex. Ambiguity and ambiguity, and these methods do not take the external information of the sentence into account when modeling the sentence semantics, so the resulting sentence semantic representation is more of a fusion representation of the possible multiple semantics of the sentence, and does not To solve the complexity, ambiguity and ambiguity of the sentence, it is impossible to accurately express the semantics of the sentence

At the same time, natural language sentences can achieve different expressions of semantics by changing a word, but the above-mentioned natural language reasoning work does not consider the semantic expressions of natural language sentences at different granularities. Therefore, when the repetition of words in two sentences is high, The above method will not be able to distinguish the different semantic expressions of the two sentences

Images