Loading

Company related event extraction for risk assessment

hero-image
  • Region

    Europe

  • Industry

    NLP

  • Type

    Web Application

  • Duration

    3 Years

Challenge

An EU-based client-targeted the development of knowledge-discovery software for real-time processing of

  • English
  • French
  • Spanish

The news streams and, therefore, engaged the Envion Software NLP team to implement a relation extraction solution for the three languages in question.

Solution

The purpose of the solution was to extract mentions of certain companies in the context of events denoting potential risks of the companies:

  • operational
  • social
  • fraudulent

Envion Software’s NLP team met the client’s requirements by developing a rule-based application using the GATE NLP framework and the Java Annotation Patterns Engine.

In order to get the highest accuracy of the company-event relation extraction, the relations were established based on knowledge of the syntactic structure of the text.

For this reason, SyntaxNet – an open-source deep-learning language model – was utilized within the application.

Technical details

  • Gate
  • JAPE
  • Treetagger
  • Java 8
  • Hunspeller
  • SyntaxNet

Result

As a result, the client’s system achieved a significant increase in information extraction accuracy, as well as an easily-extendable relation extraction module.

Similar projects

The multilingual named – entity recognition for Journal Article Summarization

The semantic reasoning tool for revealing patterns across versatile datastores

    Contact us

    Interested in working together? Ok, don't be shy. Just fill out the form below.

    Don't like forms? That's ok, just email us.