A method for commonsense causal reasoning on short texts

Abstract

The invention provides a method for carrying out commonsense causal reasoning on a short text. The method comprises the following steps that 1, a causal knowledge network is constructed; 2, the short text is subjected to causal analysis based on the causal reasoning algorithm, the sorting mechanism of commonsense causal relation strength included in the short text is obtained, the technical method can be applied to an existing question-answering system, and the answering result to a causal question (why-question) is improved. According to the method, the causal relation of the short text is analyzed through a causal mode, the coverage face is guaranteed by setting the size of a sliding window, and compared with a traditional Internet text corpus causal relation extracting method, the method is more accurate; the mode of combining sufficient causes and effects and necessary causes and effects is utilized for achieving the aim of effectively measuring the causal relation strength between lexical items.

Description

technical field [0001] The invention relates to the technical field of natural language processing in the field of computer technology, in particular to a method for performing commonsense causal reasoning on short texts. Background technique [0002] Commonsense causal reasoning embodies the most essential challenge in the realization of artificial intelligence, but related work has progressed slowly in recent years. The causal knowledge in the existing knowledge base is all manually compiled, and the quantity scale is extremely limited. There is an urgent need for a large-scale commonsense causal knowledge base to help achieve research breakthroughs in commonsense causal inference problems. [0003] The identification and extraction of causal relations can be seen as a preprocessing step for causal inference. Some human codification of commonsense causal knowledge exists (see Liu, H., and Singh, P. Commonsense reasoning in and overnatural language, Springer 2004). Howev...

AI Technical Summary

Problems solved by technology

First of all, in natural language, causality does not have a relatively simple and clear binary relationship text representation like the entity relationship "Located-in", and people describe causality in various ways; secondly, causality does not exist between entities but rather It exists between events, so we cannot extract causal knowledge by identifying entities like extracting entity relationships; this makes the extraction of causal knowledge more difficult

[0007] Due to the complexity of the above-mentioned causal relationship, there is currently no effective and reliable causal knowledge base to help question answering systems answer causal questions

Images