Financial text entity relationship extraction method and system

An entity relationship and text technology, which is applied in the field of financial text entity relationship extraction method and system, can solve problems such as inflexibility, discarding, and inability to dynamically capture multi-hop relationships, and achieve the effect of enhancing modeling capabilities and improving performance

Pending Publication Date: 2021-11-19
CHONGQING UNIV
View PDF0 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The previous technology simply simulated the relationship between entity sequences. The mainstream uses left-to-right and right-to-left RNN to model text. However, text often contains multiple entities, and the relationship between entities may present complex situations such as cross-references. , t

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Financial text entity relationship extraction method and system
  • Financial text entity relationship extraction method and system
  • Financial text entity relationship extraction method and system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0076] Such as figure 1 As shown, a financial text entity relationship extraction method includes the following steps:

[0077] Step S1: Pre-train word embeddings using BERT on the financial dataset. The pre-training of BERT is divided into two stages, one is random masked word prediction, and the other is masked word prediction of financial entities;

[0078] Step S2: Establish an entity-relationship diagram. The entities in the financial text are regarded as nodes, and the texts where the entities are located are regarded as the relationship edges between the nodes. Entity nodes in the graph are coded in three categories: entities involved in relation extraction, start and end entities of text sequences related to entity relations, and irrelevant entities. The text where the entity is located is encoded by BERT stage 1 to obtain word embedding, and after the splicing position is embedded, it is sent to BiGRU encoding to obtain the relationship transition matrix between en...

Embodiment 2

[0082] Such as figure 2 As shown, a financial text entity relationship extraction system includes the following modules:

[0083] Word embedding pre-training module: From the financial text corpus, word embedding suitable for the financial field is trained, including the general word embedding of stage 1 and the financial entity word embedding of stage 2, for the construction of graph neural network;

[0084] Entity-relationship graph construction module: extract entities from financial texts as vertices, encode the text where the entities are located as edges, and construct entity-relationship graphs;

[0085] Entity-relationship diagram update module: disseminate information between entities, so that entity nodes in the entity-relationship diagram can obtain information about adjacent points;

[0086] Relationship prediction module: Embed entity pairs that need to be predicted from entity pairs and fusion representations between different layers, and predict the affiliatio...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses a financial text entity relationship extraction method, and relates to the technical field of artificial intelligence, and the specific scheme is as follows: S1: pre-training word embedding on a financial data set by using BERT; S2, establishing an entity relation graph: taking entities in the financial text as nodes, and taking texts where the entities are located as relation edges among the nodes; S3, updating the entity relation graph: updating node embedding by using an aggregation function through a transfer matrix of a relation edge, and repeatedly updating the entity relation graph; and S4, predicting a relationship category: obtaining the embedded output of each layer of target entity pair for updating the entity relationship graph for a plurality of times from the process of S3, performing transformation splicing, then sending the output into a multi-layer perceptron for classification, and selecting the category with the maximum probability as the relationship output. The method and system are established on the basis of financial data, so that the method and system have field advantages on entity relation extraction of financial field texts.

Description

technical field [0001] The present invention relates to the technical field of artificial intelligence, more specifically, it relates to a financial text entity relationship extraction method and system. Background technique [0002] Financial text entity relationship extraction is an important technology for information intelligence. The current method roughly follows a 4-step framework: 1. Map words to corresponding embedding features; 2. Use RNN or CNN to learn a sentence representation from word embeddings; 3. Use attention mechanism to fuse word-level features and sentence-level features, Obtain a final representation; 4. Classify using a neural network based on the obtained representation. [0003] However, in the financial entity relationship recognition task, the existing technical methods show three obvious defects: [0004] 1. The scope of the text content is large, and the adaptability to the financial field is poor. Most of the existing technologies are traine...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
IPC IPC(8): G06F16/35G06F16/36G06F16/901G06N3/08
CPCG06F16/35G06F16/367G06F16/9024G06N3/08
Inventor 杨智翔邹东升陈香张晓彤宋心仪杨钰铭席康
Owner CHONGQING UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap