gnoss sherlock nlp-nlu

1. What is NERD-NLU? Named Entity Recognition and Disambiguation (NERD) is a compilation of APIs that enables textual analysis through natural language processing and interpretation (NLU).

NERD-NLU locates, contextualises, classifies and disambiguates the entities designated within a text, offering:

• An engine in the form of an entity recognition and disambiguation service that understands natural language and is able to analyse texts and extract metadata, such as concepts, entities or key words, based on their content.
• Learning and training services aimed at building personalised language processing models that achieve results in specific and private knowledge domains.

2. How does it work?

• NERD identifies, disambiguates and extracts entities in a text based on the recognised context in which they appear, where confidence is expressed by a percentage.
• The greater and richer the context, the more confident NERD will be (providing higher percentages) when relating the suggested and extracted entities.
• NERD employs disambiguation enhancement engines, a system of tools that boosts its confidence and success rate.

3. Performance and Scope of Service

• Speed. The NERD-NLU solution was devised for processing large volumes of information in very competitive time frames:
  • The New Testament - 1,000,000 characters and 10,000 lines - analysed in less than one minute.
  • Able to produce a set of metadata to represent any text with a volume of less than 5,000 words in less than one second.
• Flexibility:
  • The NERD-NLU solution is capable of identifying and extracting entities from any published text, whether produced by a media outlet, a renowned cultural institution, educational publishing houses or any other organisation.
  • NERD-NLU offers the same performance for private knowledge sources or areas on the basis of initial training that incorporates the necessary contexts for interpretation into the service.  NERD-NLU uses Machine Learning technology to learn from and accompany knowledge evolution in every area.

What is Sherlock-Discovery?

Discovery is an API for the construction of an advanced Graph that interlinks and consolidates the entities in a repository, enriching them with third-party content.

• Discovery extends Connect’s capacity, making it possible to enrich the resource-specific dynamic Knowledge Graph with third-party content (from DBpedia-Wikipedia or Geonames for example) or other, more specialised knowledge bases, including ones belonging to the client.
• Discovery enables the individual creation, reading, updating and deletion of resources.

Sherlock Connect and Discovery Administration

Users can configure the following options on the Connect and Discovery administration panel:

• Number of levels and nodes in the Graph.
• Properties used to generate the Graph.
• Types of entities to be displayed in the Graph.
• Private client datasheets to be integrated into the Graph (only available with Discovery).