Python nltk.corpus.wordnet.all_synsets() Examples

def process_definitions(self):
        self.definition_map = {}
        self.lemmakey_to_synset = {}
        n_empty_definitions = 0
        print ("Processing definitions")
        all_synsets = wn.all_synsets()
        for s in tqdm(all_synsets):
            definition = s.definition().strip()
            if len(definition) == 0:
                n_empty_definitions = n_empty_definitions + 1

            self.definition_map[s.name()] = definition

            lemmas = s.lemmas()
            for lemma in lemmas:
                key = lemma.key()
                self.lemmakey_to_synset[key] = s.name()

        print ("#Empty definitions {}/{}".format(n_empty_definitions, len(self.definition_map)))

        synsets = sorted(self.definition_map.keys())
        #self.synset_to_idx = {v:i for i,v in enumerate(self.synset_to_definition.keys())}
        self.synset_to_idx = {v:i for i,v in enumerate(synsets)}
        self.idx_to_synset = {v:i for i,v in self.synset_to_idx.items()}
        self.definitions = [self.definition_map[k] for k in synsets] 

def buildNouns():
    """
        Returns the set of all nouns of NLTK
    """
    return {x.name().split('.', 1)[0] for x in wn.all_synsets('n')} 

def buildVerbs():
    """
        Returns the set of all verbs of NLTK
    """
    return {x.name().split(".", 1)[0] for x in wn.all_synsets("v")} 

def extract_wordnet_from_nltk(entity_output_file, relation_output_file):
    from nltk.corpus import wordnet as wn
    import json

    # each node is a synset or synset+lemma
    # synsets have POS
    # synsets have several lemmas associated with them
    #       each lemma is keyed by something like able%3:00:00::
    #       where string = lemma, first number is POS, then sense id
    #
    # in addition to the synset-synset and lemma-lemma relationships,
    # we will also add synset_lemma relationship for lemmas contained
    # in each synset
    with open(entity_output_file, 'w') as fent, \
         open(relation_output_file, 'w') as frel:

        for synset in wn.all_synsets():
            node = {
                'id': synset.name(),
                'pos': synset.pos(),
                'lemmas': [lem.key() for lem in synset.lemmas()],
                'examples': synset.examples(),
                'definition': synset.definition(),
                'type': 'synset',
            }
            fent.write(json.dumps(node) + '\n')
    
            # synset-synset relationships
            for relation in SYNSET_RELATIONSHIP_TYPES:
                entity2 = [rel_synset.name()
                           for rel_synset in getattr(synset, relation)()]
                for e2 in entity2:
                    frel.write('{}\t{}\t{}\n'.format(synset.name(), 'synset_' + relation, e2))

            # now get synset-lemma and lemma-lemma relationships
            for lemma in synset.lemmas():
                node = {
                    'id': lemma.key(),
                    'pos': synset.pos(),
                    'synset': synset.name(),
                    'type': 'lemma',
                    'count': lemma.count(),
                }
                fent.write(json.dumps(node) + '\n')

                frel.write('{}\t{}\t{}\n'.format(synset.name(), 'synset_lemma', lemma.key()))

                # lemma-lemma
                for relation in LEMMA_RELATIONSHIP_TYPES:
                    entity2 = [rel_lemma.key()
                           for rel_lemma in getattr(lemma, relation)()]
                    for e2 in entity2:
                        frel.write('{}\t{}\t{}\n'.format(synset.name(), 'lemma_' + relation, e2))