maana-nlp-numbers v1.0.5

Maana-NLP-Numbers

This package implements a simple, deterministic, numeric entity extractor suitable for finding natural language number mentions in text. It recognizes whole numbers between zero and 999,999,999,999, as well customary numeric quantities (e.g. dozen, score, ...). The parser is built using monadic parser combinators and can easily be extended to support additional numeric expressions.

Installation

You can install the latest version of this package from npm:

npm i maana-nlp-numbers

Parsing numeric entities

Given a string containing a surface form for a numeric entity, the parse function returns the normal form ( a.k.a. the numeric value ) of the entity. If the function is not a valid surface form for a numeric entity, or contains extra characters before or after the numeric entity mention, the parser will fail. Consider the following examples:

const { parse } = require('maana-nlp-numbers')

console.log(parse("nineteen nintey nine")) // 1999
console.log(parse("sixty five thousand, five hundred and thiry five"))// 65535
console.log(parse("four score and seven")) // 87
console.log(parse("one day at a time")) // error

Extracting numeric entities

Given a string containing zero or more surfaces forms for numeric entities, the extract function returns a list of entity mentions. For each numeric entity, it returns a structured object which contains the starting end ending position (start and end fields) of the entity mention, and the normal form (value field). Consider the following example:

const { extract } = require('maana-nlp-numbers')

console.log(extract(`Nine out of ten respondents would buy the product for 
fifty four dollars or less per dozen.   The optimum price point is thirty seven 
dollars.
`)).map( x => x.value )

// [9,10,54,12,37]

The parser

The numeric parser is built using modadic parser combinators from the parsimmon npm package. We export the numeric parser as parser to facilitate construction of more sophisticated parsers in combination with other monadic parsers. For example:

const { parser } = require('maana-nlp-numbers')
const P = require('parsimmon')

/** a parser for a comma separated list of numbers */
const manyNumbers = P.sepBy( parser, P.string(","))

console.log(manyNumbers.tryParse("one,two,three"))
// [1,2,3]