Discourse-level multi-event extraction method based on argument subgraph prompt generation and guidance

Abstract

The invention discloses a chapter-level multi-event extraction method based on argument subgraph prompt generation and guidance. According to the method, a chapter-level long text encoder is used to obtain complete text features, and chapter-level information and sentence-level information can be utilized at the same time. Anaphora and positioning of multiple events are realized by extracting the generated event sketch through the multi-element argument relationship, and the argument classification is realized by performing event slot filling by using a pre-training model method based on a prompt normal form, so that the multi-event extraction accuracy is improved. The method does not need to use a trigger word, and reduces the annotation burden of the data set.

Description

technical field [0001] The invention relates to the technical field of natural language processing, in particular to a chapter-level multi-event extraction method based on argument subgraph prompt generation and guidance. Background technique [0002] With the rapid development of Internet technology, massive amounts of data are pouring into people's lives. In order to quickly process large-scale data and mine potentially valuable information in the data, there is an increasing demand for information extraction technology. Event extraction is an important task in the field of information extraction, which aims to detect the occurrence of events from unstructured natural language texts, judge the type of events, extract important elements involved in events, and present the results in a structured presented in a manner. Event extraction has a wide range of application values. On the one hand, event extraction can assist in providing structured multi-relational information, ...

AI Technical Summary

Problems solved by technology

[0005] 2. The accuracy of multi-event extraction is not high

There are often multiple events in a document in the actual field. Existing methods rely on trigger words to refer to events for multi-event extraction. However, trigger words in real scenes are often difficult to judge. There is an event containing multiple trigger words, a Trigger words correspond to multiple event types, no obvious trigger words, etc.

Therefore, the method of relying on trigger words will cause redundancy or omission of extraction results, resulting in poor final multi-event extraction effect.

[0006] 3. Too much reliance on trigger words brings burden to data labeling

Existing methods often use trigger words as the medium, but trigger words are only intermediate results of event extraction, not necessary, and labeling is very difficult, which increases the burden of manually constructing data sets

[0007] In summary, the existing technical solutions have shortcomings such as ignoring chapter-level information, low accuracy of multi-event extraction, and excessive reliance on trigger words.

Images