Recognition method of event temporal relationship combining dependency relationship and discourse rhetorical relationship

Abstract

The invention discloses a method for identifying event time-sequence relations which integrates dependency relations and discourse rhetorical relations. The present invention is an event time sequence relationship recognition method that integrates dependency relationship and discourse rhetorical relationship, which is characterized in that it includes: a given event pair e 1 ,e 2 , in terms of extracting the shortest dependency path; considering the complex syntactic structure of the dependency syntax tree, an additional heuristic rule is added to enrich event-related information; child nodes are added to the dependency path, and finally the extracted words are sorted according to the order in which the event sentences appear ; Afterwards, the obtained event-to-dependent path will be processed in the same way as the benchmark model, and the vector representation Y' will be obtained through the embedding layer, the bidirectional LSTM layer and the hidden layer. The beneficial effect of the present invention is to add the information coding of the chapter tree into the model. This method divides the relationship between events into two parts to represent, one is the dependency path information of the sentence where the event is located.

Description

technical field [0001] The invention relates to the field of language processing, in particular to a method for identifying event time-series relationships that integrates dependency relationships and discourse rhetorical relationships. Background technique [0002] The temporal relationship of events generally exists among events, and it refers to the time sequence of events described in the text, which is a natural connection of event information units. The recognition of temporal relationships between events is actually a basic research at the semantic level, which is of great significance to tasks such as story timeline construction, automatic question answering, text summarization, and information extraction. For example, in the sentence "Finally today, we learned that the space agency has finally taken giant leap forward.", the event "learned" occurred after the event "taken", so the temporal relationship between them is "AFTER". [0003] In terms of Chinese temporal ...

AI Technical Summary

Problems solved by technology

This does not bring direct information to the inter-sentence event to the dependency path, but violently superimposes the two dependency paths, and there is a problem of missing information

Images