wrapperWSD 0.0.2

Word Sense Disambiguation wrapper

Word Sense Disambiguation wrapper

In natural language processing word sense disambiguation (WSD) is the problem of determining which "sense" (meaning) of a word is activated by the use of the word in a particular context, a process which appears to be largely unconscious in people.

This is a simple library that wrap two WSD methods: NLTK and Babelfy.

Requirements

You should run

pip3 install xmltodict
pip3 install nltk
pip3 install pywsd

The NLTK library requires more extra configurations, see this link to more details.

Methods

The wsdNLTK methods call the function pywsd.disambiguate which returns a mapping between words of the input text and their WornNet Synsets.

wsd = WrapperWSD()
wsd.wsdNLTK(u'My sister has a dog. She loves him.')
#output: [('sister', Synset('sister.n.02'), 3, 9), ('dog', Synset('pawl.n.01'), 16, 19), ('loves', Synset('sleep_together.v.01'), 25, 30)]

Instead of returning the WornNet Synsets, the method wsdNLTK_offset returns a mapping between words of the input text and their WornNet offset.

wsd.wsdNLTK_offset(u'My sister has a dog. She loves him.')
#output: [('president', 597265, 21, 30), ('USA', 8394922, 38, 41), ('best', 67379, 54, 58)]

A mapping between WordNet and Wikipedia was proposed in [Miller et al] available for download here. In the next example you can see some key-values of it.

wd2wiki = {
 1740: 'https://en.wikipedia.org/wiki/Madison_Square_Garden,_L.P.',
 2137: 'https://en.wikipedia.org/wiki/Abstraction',
 2452: 'https://en.wikipedia.org/wiki/Object_(philosophy)',
 2684: 'https://en.wikipedia.org/wiki/Computer_file',
 3553: 'https://en.wikipedia.org/wiki/Unit_of_alcohol',
 ...
 }

We used this mapping to link entities from Wikipedia for those cases where exists a correspondence.

wsd.wsdNLTK_links(u'My sister has a dog. She loves him.')
#output: [{'start': 38, 'end': 41, 'label': 'USA', 'link': 'United_States_Army'}]

On the other hand, we include Babelfy targetting BabelSynsets

wsd.wsdBabelfy(u'My sister has a dog. She loves him.')
#output: [('sister', 'bn:00071838n', 3, 9), ('dog', 'bn:00015267n', 16, 19), ('loves', 'bn:00090504v', 25, 30)]

Reference

[Miller et al] WordNet–Wikipedia–Wiktionary: Construction of a Three-way Alignment. Tristan Miller and Iryna Gurevych. 2014 https://pdfs.semanticscholar.org/90cd/22a9cd59dc1fc21f4ec36e9c7d95085f7fb6.pdf