Research on the Identification of Organizational Entities in Fraudulent Texts Via Bi-GRU-Flat

The analysis of fraudulent texts requires the extraction of entities in the text. Organization is an important entity type in fraudulent texts. Recognizing and extracting the entity of organization is of great significance to text analysis. This paper proposes a deep learning network model based on Bi- GRU-Flat for identifying entities of organization in fraudulent texts. The model uses a modified ELMo module to perform deeper contextualized word representations on the input sequence, and then uses Flat's special position coding and lattice structure transformation to help extract more lexical information, and finally makes further constrained corrections to the output through the CRF layer. The experiments show that the model outperforms other related models in terms of precision, recall and F1 value, and the recognition effect is better.