The product
VisualText & NLP++
VisualText is an integrated development environment for building text analyzers. Think of it as Visual C++ for natural language processing. Analyzers are written in NLP++, a C++-like language for grammars, patterns, heuristics and knowledge.
This page describes the original VisualText. The current, open-source versions, including NLP++ for Visual Studio Code, are at visualtext.org.
What it does
VisualText turns unstructured text into structured data. For example, you can extract business events from news articles and load them into a searchable database. It has been used to build resume analyzers, web-page categorizers, a monitor for financial-transaction chat, email analyzers and focused web crawlers.
Analyzers are “glass box”: every rule, grammar and piece of knowledge is visible and editable. Non-programmers can extend an analyzer by highlighting new examples in a text, and VisualText generates the rules automatically. Changes can be tested right away, without waiting to compile, so prototypes come together in days or even hours.
How it is built
- NLP++, a programming language for building analyzers quickly, pass by pass.
- Conceptual Grammar, a knowledge base for linguistic, conceptual and domain knowledge.
- Rule generation that learns from highlighted samples.
- A runtime engine that runs analyzers interpreted, or compiles them to C++ libraries for speed and embedding on any platform that runs C++, including Linux.
What people built with it
Information extraction, named-entity extraction, indexing, filtering, categorization, text mining, summarization, test grading, automatic coding of resumes and medical or police reports, natural-language query, and routing documents to the right people. Application areas included the web, business, law, medicine and government.
Explore further: the quick tour, the in-depth tour, or the full help documentation.
What developers said
“I think this is one of the most interesting tools I've ever seen. It's beautifully composed, and a pleasure to use (more accurately — addicting!)”
“Being able to move through the passes in an analysis and have source text highlighted appropriately — and being able to generate the different versions of the parse tree — absolutely wonderful! My compliments to the chefs!”
“After working through all tutorials I was amazed by this program… it is I think the most innovative application I have ever seen.”
“A whole lot easier than writing a bunch of ad hoc C or Perl parsers.”