Loading

The semantic reasoning tool for revealing patterns across versatile datastores

hero-image
  • Country

    USA

  • Industry

    NLP

  • Type

    Web Application

  • Duration

    5 Years

Challenge

A US-located client was eager to obtain a knowledge-management solution to enrich user insights into data discovery, allowing them to perform comprehensive data analysis based on a broad range of 3rd-party data resources.

Solution

Envion Software’s NLP, Semantic Web, and ETL experts contributed to the successful implementation of the solution.

Our joint team developed a solution consisting of the following core modules:

  • NLP module for processing natural-language requests of a user: it was built using Python and Spacy so that the system was able to capture key concepts of interest to the data analyst while searching through the multiple datastores.
  • ETL module allowed to unify datasets of versatile format (csv, xls, rdf, owl, ttl) into the canonical one for further semantic reasoning (Virtuoso Server, Pentaho Kettle, Google Refine, AlignAPI, Java 8).
  • Semantic Web module included numerous industry-standard ontologies and per-business-case semantic reasoning logic over triple stores (Protégé Framework, OWL, RDF, SPARQL, Python 3.5).

Result

As a result, the entire solution made it possible to identify the most relevant datasets for a given analytical problem, empowering a data scientist with valuable insights regarding non-obvious patterns hidden in raw data.  

Furthermore, the technology developed was flexible enough to customize it and apply it to multiple data analysis requests within numerous domains.

Technical details

  • Python 3.5
  • Java 8
  • Protégé Framework
  • SPARQL
  • Google Refine
  • OWL
  • RDF
  • Pentaho Kettle

Similar projects

Company related event extraction for risk assessment

The multilingual named – entity recognition for Journal Article Summarization

    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.