Archived — legacy sample analyzer
This sample analyzer dates from the Windows-only VisualText IDE era and needs that retired IDE to build and run. It is kept here as a historical description only.
Current sample analyzers live at github.com/VisualText and install straight from the NLP++ Language Extension for VS Code.
See the VisualText archive for the full list of retired material.
TAIParse was Text Analysis International’s general-purpose English analyzer, built to need as little knowledge as possible (“just-in-time” knowledge) while still doing part-of-speech tagging, entity extraction and shallow parsing. It tagged parts of speech at 94% accuracy on a blind test over a business-article corpus, and shipped with a full lexicon with part-of-speech information in its knowledge base.
It was meant as a starting point for customising your own text analysis, and it demonstrated these capabilities:
- Zoning and “parsing-per-line” to characterise regions and formats in text
- Dynamic, context-dependent part-of-speech assignment and parsing
- Successive segmentation of text in a divide-and-conquer strategy
- Treatment of unknown words
- Noun phrase extraction
- A semantic and discourse processing framework tied into an ontology and a dynamic representation of the analysis in the knowledge base
- Analysis independent of capitalisation, so that all-uppercase regions could still be analysed
- Robust analysis in the face of errors, misspellings and ungrammatical text
For a current English parser, see the analyzers available through the NLP++ Language Extension for VS Code and the repositories at github.com/VisualText.
