Entity Relationship Management

A work-in-progress project aimed at building a entity-relationship builder system using NLP

Goals of the project (For reference)

• Identify the subject and object entities from various data sources like news, blogs etc
• Identify the relationship and categorical scores and sentiments
• Build a historical knowledge graph of entities and various scores
• Analyse how entities are related, how the sentiment varies over a period of time
• Predict/forecast the futuristic sentiment and categorical scores

easticsearch Container Setup

The elasticsearch docker container version 7.6.1 is used. More details: https://hub.docker.com/_/elasticsearch/

Start the elastic search container using

docker run -d --name elasticsearch -p 9200:9200 -p 9300:9300 -e "discovery.type=single-node" elasticsearch:7.6.1

Components

indexer.py

Index the newly available news articles into elasticsearch

entity_extractor.py

Tokenize and parse the sentences, identify the different entities and relationships

TODO:

• USE: https://www.johnsnowlabs.com/finance-nlp/
• Clean the sentences
• Detect active and passive voices before finding the relationships
• Improvise the subject and object detection
• Infer the pronouns in the sequence of sentences and replace pronouns with the actual subjects/objects

References

• https://realpython.com/natural-language-processing-spacy-python/#dependency-parsing-using-spacy