These are template analyzers that appear in the New Analyzer selection list of the NLP++ Language Extension for VSCode. When a user creates a new analyzer from the extension, they can pick one of these templates as a starting point and then extend it for their own domain.
For more on the NLP++ language and tooling, see the VisualText / NLP++ project site.
Each folder in this repository is a self-contained analyzer template with its own README.md, sample input/, knowledge base (kb/), grammar/passes (spec/), and working directory (tmp/).
A parser for American (USPS-format) postal addresses. Extracts house number, street name and suffix/type, city, state, ZIP code, and address type. Also handles Rural Route and Highway Contract addresses, exposing the extracted fields as JSON.
A bare-minimum parser bundled with the full English dictionary. A good starting point when you need dictionary lookup but don't want any domain-specific grammar pre-wired.
The smallest possible analyzer skeleton — no dictionary, no rules. Use this when you want to build an analyzer from scratch with nothing in the way.
A date and time analyzer that recognizes common date and time expressions in text.
Identifies email addresses across a variety of formats. For each match it extracts the local name, domain name, and top-level domain (TLD), and emits the result as JSON.
A starter analyzer that builds, saves, and exports a knowledge base as JSON. Shows the minimal plumbing: an init pass that creates a kb global concept and populates it, and an output pass that persists it with SaveKB and JsonKB from the shared KB function library.
A two-way data interface between the NLPPlus packages (Python and NPM) and an analyzer: imports JSON handed in by a script into the knowledge base (via the json2kbb pass) and exports the knowledge base back out as output.json.
Lays out the basics of paragraph and sentence segmentation with dictionary lookup. Intended as a starting point — each domain has its own quirks (abbreviations, lists, etc.) that can fool sentence enders and will need domain-specific handling.
Recognizes telephone numbers in many international and US formats (e.g. 212.456.7890, +1 212-456-7890, 1-212-456-7890). For each number it extracts the area code, prefix, station, and phone type (mobile/landline).
Identifies hyperlinks of many shapes and extracts the scheme (http/https/ftp/…), sub-domain, domain name, page path, and top-level domain, emitting the result as JSON.
A utility analyzer that masks all alphanumeric text as x. Useful for sanitizing or redacting sample text while preserving structure.
- Install the NLP++ Language Extension in VSCode.
- Create a new analyzer from the extension and choose one of the templates above.
- Drop the text you want to parse into the analyzer's
input/folder and run the analyzer. - Inspect the generated output (e.g.
input/<file>.txt_log/output.json) and extend the rules inspec/and the knowledge base inkb/for your domain.
These are the rules for submitting a block analyzer to this repository:
- The folder name should be a two-to-three-word descriptive name for the analyzer.
- The first line in the README should be the same title with a
#header. - The next non-empty line should be a short description of the analyzer — this text appears in the NLP++ VSCode Extension's template selection list.
- The rest of the README should explain how to use and extend the analyzer.
Example:
# Moose Analyzer
Allows for analyzing text for articles on moose (appears in the description of the selection block analyzer selection list)
## Using the Moose Analyzer
The way in which to best utilize this analyzer is...
This repo participates in the VisualText cross-repo release "percolation"
system: submodule bumps flow downstream automatically via repository_dispatch.
See nlp-engine/docs/PERCOLATION.md for the full map.