phrase-matcher

JavaScript library to pattern-match English phrases

Installation

I'm not sure how well this installs! But for use in a browser, do:

git clone https://github.com/ianb/phrase-matcher.git
cd phrase-master
npm install
npm run rollup
cp dist/phraseMatcher.js ...my-project/public/

I haven't done much npm distribution, and would welcome patches to improve this part.

Usage

The bundled file exposes a phraseMatcher object with the following properties:

compile(rule, {entities, intentName})

Compiles a single rule into a FullPhrase() object. You can look at compile(...).toString() to confirm how the phrase was compiled.

Typically you will take several phrases and put them together.

convertEntities({entityType: ["word1", "word2"])})

Entities are when you have a slot like [n:number] (which matches 1, 2, etc under the slot n).

The actual entities the compiler expects is a specific matching objects, and this creates that object.

splitPhraseLines(bigString)

This takes a list of phrases and returns one string per line (but respecting things like {...} that goes over more than one line). It also removes empty lines and comment lines (starting with #).

new PhraseSet(phraseMatchers)

Given many phrases, this object can match against all those phrases and pick out the best match.

.match(utteranceString): this matches the utteranceString against the phrases, and returns null or the best match. The match object is shown below.

enumeratePhrases(fillerFunc): this enumerates all possible phrases, using fillerFunc(slotName) -> String to replace wildcard slots.

Match objects

This is returned by .match(...). It has these attributes:

.utterance: original matched string

.slots: an object containing the matched slots

.intentName: the intentName (as passed in to compile())