NLP++ analyzer template

A starting point for an NLP++ text analyzer, with continuous integration that runs it on every push.

NLP++ is a programming language for natural language processing. An analyzer is a sequence of passes; each pass matches pattern rules against a parse tree and runs code when a rule fires, and a built-in knowledge base holds what the analyzer learns as it reads.

Start a project

1. Click Use this template above, then Create a new repository.
2. Add the topic nlp-plus-plus to your repository (the gear next to About) so other NLP++ projects can find it.
3. Clone it and open the folder in VS Code with the VisualText extension installed. The extension installs the NLP engine and its help.

What is in it

spec/analyzer.seq   the pass sequence
spec/*.nlp          the passes
input/              text files to analyze
kb/user/            knowledge base sources: .dict dictionaries and .kbb knowledge bases

The analyzer builds a knowledge base and exports it as JSON:

• kbinit creates a kb concept off the knowledge-base root, keeps it in the global variable G("kb"), and fills in a small example (animals with dog and cat). Replace the example with concepts your grammar passes find in the text.
• output saves the knowledge base with SaveKB and exports it with JsonKB.
• KBFuncs is VisualText's shared knowledge-base function library, where SaveKB and JsonKB come from. Leave it as it is and put your own functions in a separate @DECL pass after it.

Add your grammar passes between kbinit and output. Results appear in input/<file>_log/.

Run it

In VS Code, select a file in input/ and run the analyzer from the VisualText panel.

From the command line:

nlp -ANA <this folder> -IN <this folder>/input/text.txt -WORK <engine folder>

-WORK is the engine's own folder (the one containing the nlp executable and its data/ folder), not the analyzer's. With the VS Code extension installed it is ~/.vscode/extensions/dehilster.nlp-<version>/nlp-engine.

Continuous integration

.github/workflows/analyze.yml installs the latest NLP engine release on every push and pull request, runs the analyzer over every input/*.txt, and fails if the engine reports an error. Each file's output log is kept as a downloadable artifact named output-logs.

Learn more

• VisualText and the NLP++ textbook
• NLP engine
• More templates (dates, addresses, emails, telephone numbers, URLs, English parsing) ship with the VS Code extension.

License

MIT.