Research on Event Extraction Method Based on A Lite BERT And Conditional Random Field Model

摘要

Information extraction technology is used to extract hot information with high attention from unstructured text data. Event extraction technology is a challenging research direction in the field of information extraction. The purpose of event extraction is to extract key elements describing events from unstructured text data and present them in a structured way. Event extraction can be divided into two important stages: event trigger word extraction and event element extraction. Regarding event extraction as a sequence labeling task, the ALBERT pretraining model was used to learn the features, and the CRF feature function of conditional random field was introduced to learn the sequence features. By limiting the label relationship within a specific window, the performance of sequence labeling was improved. Event extraction experiments are carried out on ACE2005 standard corpus. the experimental results show that the ALBERT-CRF model has improved recall rate and F-score in the task of trigger word recognition and classification compared with the existing models.