nai-php/naipostagger

A part of speech tagger written in PHP.

v0.2 2021-08-22 08:39 UTC

Requires

• php: >=7.2.0
• aura/sql: ^3.0
• monolog/monolog: ^2.3

MIT

• Giorgio G. Rey

• tag
• semantic
• pos
• keywords
• natural language processing
• ai
• linguistics
• chatbot
• part of speech
• pos tagging

README

A lightweight framework-agnostic library in pure PHP for part-of-speech tagging. Can be used for chatbots, personal assistants, keywords extraction etc. Being written in PHP, it can be easily integrated in pre existent or new applications, giving the real ability to understand what users write.

It is based on vocabularies and predefined grammatical rules, without wrappers to third part systems, neural networks, machine learning or models that requires huge resources.

This is the english version. Documentation and TODO are coming, more info and demo on n-ai.cloud

Precision

In this table I'll put results of differents type of sentences corpus.

Corpus	Total tokens	Correctly tagged	Not correctly tagged	% of total correct
"Just Shoot Me" movie subtitles	3403	3381	22	99,35

Installation

1. in your project folder e.g. "myproject" install the package via composer;

2. create folder "dictionaries";

3. inside folder "dictionaries" clone or download the english dictionary repository;

4. run this example script:

use NaiPosTagger\Pipelines\PipelinePosTagging;
use NaiPosTagger\Models\NaiPosArr;


include('vendor/autoload.php');

include(__DIR__ . '/vendor/nai-php/naipostagger/src/Utilities/common_functions_helper.php');

define('DICTIONARIES_PATH', __DIR__ . '/./dictionaries/dictionaries-');

define('TRAITS_PATH', __DIR__ . '/./vendor/nai-php/naipostagger/src/');

$sentence = 'my name is Fred';

$PipelinePosTagging = new PipelinePosTagging();

$PipelinePosTagging->language = 'en';

$pos_arr = $PipelinePosTagging->transform($sentence);

// for a clear output, better hide metadata
$pos_arr = NaiPosArr::clearMetadata($pos_arr);

// and further simplify the output
$pos_arr = NaiPosArr::flatPosArr($pos_arr);

diex($pos_arr);

And the output will be:

Array
(
    [0] => Array
        (
            [form] => .
            [lemma] => .
            [features] => SENT
            [sh-feat] => SENT
            [label] => 
            [rule] => 
            [pos_score] => 0
        )

    [1] => Array
        (
            [form] => my
            [lemma] => my
            [features] => ADJ:pos+m+s
            [sh-feat] => ADJ
            [label] => 
            [rule] => 
            [pos_score] => 0
        )

    [2] => Array
        (
            [form] => name
            [lemma] => name
            [features] => NOUN-m:s
            [sh-feat] => NOUN
            [label] => 
            [rule] => 
            [pos_score] => 0
        )

    [3] => Array
        (
            [form] => is
            [lemma] => is
            [features] => VER:ind+pres+3+s
            [sh-feat] => VER
            [label] => 
            [rule] => 
            [pos_score] => 0
        )

    [4] => Array
        (
            [form] => Fred
            [lemma] => Fred
            [features] => NPR
            [sh-feat] => NPR
            [label] => 
            [rule] => 
            [pos_score] => 0
        )

    [5] => Array
        (
            [form] => .
            [lemma] => .
            [features] => SENT
            [sh-feat] => SENT
            [label] => 
            [rule] => 
            [pos_score] => 0
        )

)

To do list

• Find contributors
• Clean, check, fix and tag term in dictionaries
• Clean, check, fix brill rules
• Add more ngrams
• Add more tests, expecially for filters
• Collect and load frill words
• Better Oop for some classes?
• In module for logical analysis (yet not published) collect synonyms and temporal expressions