diff --git a/README.md b/README.md index 4b3f0f1..2d4ad7c 100644 --- a/README.md +++ b/README.md @@ -6,6 +6,10 @@ Use Wordnet to answer dictionary like questions Uses [Wordnet](https://wordnet.princeton.edu/) to provide information. +checks language specific wordnets for `'en', 'als', 'arb', 'bg', 'cmn', 'da', 'el', 'fi', 'fr', 'he', 'hr', 'is', 'it', 'it-iwn', 'ja', 'ca', 'eu', 'gl', 'es', 'id', 'zsm', 'nl', 'nn', 'nb', 'pl', 'pt', 'ro', 'lt', 'sk', 'sl', 'sv', 'th'` + +definitions are sometimes missing, when this happens the english text definition for the word will be machine translated + ## Examples * "what is the definition of ..." diff --git a/__init__.py b/__init__.py index 37be1e5..f6bdee5 100644 --- a/__init__.py +++ b/__init__.py @@ -11,32 +11,250 @@ # limitations under the License. # import random +from typing import Optional, Dict + import nltk -from ovos_classifiers.opm.nltk import WordnetSolverPlugin +from nltk.corpus import wordnet as wn +from nltk.data import find +from ovos_plugin_manager.templates.language import LanguageTranslator +from ovos_utils.lang import standardize_lang_tag from ovos_workshop.decorators import intent_handler -from ovos_workshop.skills.common_query_skill import CommonQuerySkill, CQSMatchLevel +from ovos_workshop.skills.ovos import OVOSSkill + + +def download_nltk_resource(res: str, res_type: str = "taggers"): + """ + Download necessary NLTK resource if not already downloaded. + """ + + resource_name = f'{res_type}/{res}.zip' + try: + find(resource_name) + except LookupError: + # Download resource if not already present + nltk.download(res) + + +class Wordnet: + LANGMAP = {'en': 'eng', 'als': 'als', 'arb': 'arb', 'bg': 'bul', 'cmn': 'cmn', + 'da': 'dan', 'el': 'ell', 'fi': 'fin', 'fr': 'fra', 'he': 'heb', + 'hr': 'hrv', 'is': 'isl', 'it': 'ita', 'it-iwn': 'ita_iwn', 'ja': 'jpn', + 'ca': 'cat', 'eu': 'eus', 'gl': 'glg', 'es': 'spa', 'id': 'ind', 'zsm': 'zsm', + 'nl': 'nld', 'nn': 'nno', 'nb': 'nob', 'pl': 'pol', 'pt': 'por', 'ro': 'ron', + 'lt': 'lit', 'sk': 'slk', 'sl': 'slv', 'sv': 'swe', 'th': 'tha'} + translator: Optional[LanguageTranslator] = None + download_nltk_resource("wordnet", "corpora") + download_nltk_resource("omw-1.4", "corpora") + + @staticmethod + def get_synsets(word, pos=wn.NOUN, lang: Optional[str] = "en"): + lang = Wordnet.LANGMAP[standardize_lang_tag(lang.split("-")[0])] + synsets = wn.synsets(word, pos=pos, lang=lang) + if not len(synsets): + return [] + return synsets + + @staticmethod + def get_definition(word, pos=wn.NOUN, synset=None, lang: Optional[str] = "en"): + lang = Wordnet.LANGMAP[standardize_lang_tag(lang.split("-")[0])] + if synset is None: + synsets = wn.synsets(word, pos=pos, lang=lang) + if not len(synsets): + return [] + synset = synsets[0] + defi = synset.definition(lang=lang) + if not defi: + # translate if possible + if Wordnet.translator is not None: + return Wordnet.translator.translate(text=synset.definition(lang="eng"), + target=standardize_lang_tag(lang), + source="en") + return defi + + @staticmethod + def get_examples(word, pos=wn.NOUN, synset=None, lang: Optional[str] = "en"): + lang = Wordnet.LANGMAP[standardize_lang_tag(lang.split("-")[0])] + if synset is None: + synsets = wn.synsets(word, pos=pos, lang=lang) + if not len(synsets): + return [] + synset = synsets[0] + return synset.examples(lang=lang) + + @staticmethod + def get_lemmas(word, pos=wn.NOUN, synset=None, lang: Optional[str] = "en"): + lang = Wordnet.LANGMAP[standardize_lang_tag(lang.split("-")[0])] + if synset is None: + synsets = wn.synsets(word, pos=pos, lang=lang) + if not len(synsets): + return [] + synset = synsets[0] + return [l.name().replace("_", " ") for l in synset.lemmas(lang=lang)] + + @staticmethod + def get_hypernyms(word, pos=wn.NOUN, synset=None, lang: Optional[str] = "en"): + lang = Wordnet.LANGMAP[standardize_lang_tag(lang.split("-")[0])] + if synset is None: + synsets = wn.synsets(word, pos=pos, lang=lang) + if not len(synsets): + return [] + synset = synsets[0] + + # Translate hypernyms to lang + lang_h = [] + for hypernym in synset.hypernyms(): + lang_h += [lemma.name().split(".")[0].replace("_", " ") + for lemma in hypernym.lemmas(lang=lang)] + return lang_h + + @staticmethod + def get_hyponyms(word, pos=wn.NOUN, synset=None, lang: Optional[str] = "en"): + lang = Wordnet.LANGMAP[standardize_lang_tag(lang.split("-")[0])] + if synset is None: + synsets = wn.synsets(word, pos=pos, lang=lang) + if not len(synsets): + return [] + synset = synsets[0] + # Translate hyponyms to lang + lang_h = [] + for hyponym in synset.hyponyms(): + lang_h += [lemma.name().split(".")[0].replace("_", " ") + for lemma in hyponym.lemmas(lang=lang)] + return lang_h + + @staticmethod + def get_holonyms(word, pos=wn.NOUN, synset=None, lang: Optional[str] = "en"): + lang = Wordnet.LANGMAP[standardize_lang_tag(lang.split("-")[0])] + if synset is None: + synsets = wn.synsets(word, pos=pos, lang=lang) + if not len(synsets): + return [] + synset = synsets[0] + # Translate holonyms to lang + lang_h = [] + for holonym in synset.member_holonyms(): + lang_h += [lemma.name().split(".")[0].replace("_", " ") + for lemma in holonym.lemmas(lang=lang)] + return lang_h + + @staticmethod + def get_root_hypernyms(word, pos=wn.NOUN, synset=None, lang: Optional[str] = "en"): + lang = Wordnet.LANGMAP[standardize_lang_tag(lang.split("-")[0])] + if synset is None: + synsets = wn.synsets(word, pos=pos, lang=lang) + if not len(synsets): + return [] + synset = synsets[0] + # Translate hypernyms to lang + lang_h = [] + for hypernym in synset.root_hypernyms(): + lang_h += [lemma.name().split(".")[0].replace("_", " ") + for lemma in hypernym.lemmas(lang=lang)] + return lang_h + + @staticmethod + def common_hypernyms(word, word2, pos=wn.NOUN, lang: Optional[str] = "en"): + lang = Wordnet.LANGMAP[standardize_lang_tag(lang.split("-")[0])] + synsets = wn.synsets(word, pos=pos, lang=lang) + if not len(synsets): + return [] + synset = synsets[0] + synsets = wn.synsets(word2, pos=pos, lang=lang) + if not len(synsets): + return [] + synset2 = synsets[0] + return [l.name().split(".")[0].replace("_", " ") for l in + synset.lowest_common_hypernyms(synset2, lang=lang)] + @staticmethod + def get_antonyms(word, pos=wn.NOUN, synset=None, lang: Optional[str] = "en"): + lang = Wordnet.LANGMAP[standardize_lang_tag(lang.split("-")[0])] + if synset is None: + synsets = wn.synsets(word, pos=pos, lang=lang) + if not len(synsets): + return [] + synset = synsets[0] + lemmas = synset.lemmas(lang=lang) + if not len(lemmas): + return [] + lemma = lemmas[0] + antonyms = lemma.antonyms() + + return [l.name().split(".")[0].replace("_", " ") for l in antonyms] + + @classmethod + def query(cls, query, pos=wn.NOUN, synset=None, lang: Optional[str] = "en"): + lang = Wordnet.LANGMAP[standardize_lang_tag(lang.split("-")[0])] + if synset is None: + synsets = wn.synsets(query, pos=pos, lang=lang) + if not len(synsets): + return {} + synset = synsets[0] + res = {"lemmas": cls.get_lemmas(query, pos=pos, synset=synset, lang=lang), + "antonyms": cls.get_antonyms(query, pos=pos, synset=synset, lang=lang), + "holonyms": cls.get_holonyms(query, pos=pos, synset=synset, lang=lang), + "hyponyms": cls.get_hyponyms(query, pos=pos, synset=synset, lang=lang), + "hypernyms": cls.get_hypernyms(query, pos=pos, synset=synset, lang=lang), + "root_hypernyms": cls.get_root_hypernyms(query, pos=pos, synset=synset, lang=lang), + "definition": cls.get_definition(query, pos=pos, synset=synset, lang=lang)} + return res + + +class WordnetSkill(OVOSSkill): -class WordnetSkill(CommonQuerySkill): def initialize(self): - nltk.download('punkt_tab') - nltk.download('averaged_perceptron_tagger_eng') - self.wordnet = WordnetSolverPlugin() + Wordnet.translator = self.translator + + @staticmethod + def get_data(query: str, lang: Optional[str] = "en", pos="auto") -> Dict[str, str]: + """ + Retrieves WordNet data for the given query. + + Args: + query (str): The query string. + lang (Optional[str]): The language of the query. Defaults to None. + + Returns: + Dict[str, str]: A dictionary containing WordNet data such as lemmas, antonyms, definitions, etc. + """ + p = wn.NOUN if pos not in [wn.NOUN, wn.ADJ, wn.VERB] else pos + + lang = Wordnet.LANGMAP[standardize_lang_tag(lang.split("-")[0])] + synsets = wn.synsets(query, pos=p, lang=lang) + if not synsets and pos == "auto" and p == wn.NOUN: + # try looking for an adjective + p = wn.ADJ + synsets = wn.synsets(query, pos=p, lang=lang) + if not synsets and pos == "auto" and p == wn.ADJ: + # try looking for a verb + p = wn.VERB + synsets = wn.synsets(query, pos=p, lang=lang) + + if not synsets: + return {} + + synset = synsets[0] + res = { + "postag": "ADJ" if p == wn.ADJ else "VERB" if p == wn.VERB else "NOUN", + "lemmas": Wordnet.get_lemmas(query, pos=pos, synset=synset, lang=lang), + "antonyms": Wordnet.get_antonyms(query, pos=pos, synset=synset, lang=lang), + "holonyms": Wordnet.get_holonyms(query, pos=pos, synset=synset, lang=lang), + "hyponyms": Wordnet.get_hyponyms(query, pos=pos, synset=synset, lang=lang), + "hypernyms": Wordnet.get_hypernyms(query, pos=pos, synset=synset, lang=lang), + "root_hypernyms": Wordnet.get_root_hypernyms(query, pos=pos, synset=synset, lang=lang), + "definition": Wordnet.get_definition(query, pos=pos, synset=synset, lang=lang) + } + return res # intents @intent_handler("search_wordnet.intent") def handle_search(self, message): - query = message.data["query"] - summary = self.wordnet.spoken_answer(query, lang=self.lang) - if summary: - self.speak(summary) - else: - self.speak_dialog("no_answer") + self.handle_definition(message) @intent_handler("definition.intent") def handle_definition(self, message): query = message.data["query"] - res = self.wordnet.search(query, lang=self.lang).get("definition") + res = self.get_data(query, lang=self.lang).get("definition") if res: self.speak(res) else: @@ -47,7 +265,7 @@ def handle_definition(self, message): @intent_handler("lemma.intent") def handle_lemma(self, message): query = message.data["query"] - res = self.wordnet.search(query, lang=self.lang).get("lemmas") + res = self.get_data(query, lang=self.lang).get("lemmas") if res: self.speak(random.choice(res)) else: @@ -56,7 +274,7 @@ def handle_lemma(self, message): @intent_handler("antonym.intent") def handle_antonym(self, message): query = message.data["query"] - res = self.wordnet.search(query, lang=self.lang).get("antonyms") + res = self.get_data(query, lang=self.lang).get("antonyms") if res: self.speak(random.choice(res)) else: @@ -65,7 +283,7 @@ def handle_antonym(self, message): @intent_handler("holonym.intent") def handle_holonym(self, message): query = message.data["query"] - res = self.wordnet.search(query, lang=self.lang).get("holonyms") + res = self.get_data(query, lang=self.lang).get("holonyms") if res: self.speak(random.choice(res)) else: @@ -74,7 +292,7 @@ def handle_holonym(self, message): @intent_handler("hyponym.intent") def handle_hyponym(self, message): query = message.data["query"] - res = self.wordnet.search(query, lang=self.lang).get("hyponyms") + res = self.get_data(query, lang=self.lang).get("hyponyms") if res: self.speak(random.choice(res)) else: @@ -83,48 +301,33 @@ def handle_hyponym(self, message): @intent_handler("hypernym.intent") def handle_hypernym(self, message): query = message.data["query"] - res = self.wordnet.search(query, lang=self.lang).get("hypernyms") + res = self.get_data(query, lang=self.lang).get("hypernyms") if res: self.speak(random.choice(res)) else: self.speak_dialog("no_answer") - # common query - def CQS_match_query_phrase(self, phrase): - summary = self.wordnet.spoken_answer(phrase, lang=self.lang) - if summary: - self.log.info(f"Wordnet answer: {summary}") - return (phrase, CQSMatchLevel.CATEGORY, summary, - {'query': phrase, - 'answer': summary}) - - def CQS_action(self, phrase, data): - pass - if __name__ == "__main__": + print(list(Wordnet.LANGMAP)) from ovos_utils.fakebus import FakeBus d = WordnetSkill(skill_id="wordnet.ovos", bus=FakeBus()) query = "what is the definition of computer" - ans = d.wordnet.search("computer", context={"lang": "es-es"}) - print(ans) - # {'lemmas': ['computer', 'computing machine', 'computing device', 'data processor', 'electronic computer', 'information processing system'], + ans = d.get_data("computador", lang="pt") + print("pt", ans) + # {'postag': 'NOUN', + # 'lemmas': ['Calculadoras', 'calculador', 'calculadora', 'calculista', 'computador'], # 'antonyms': [], # 'holonyms': [], - # 'hyponyms': ['analog computer', 'digital computer', 'home computer', 'node', 'number cruncher', 'pari-mutuel machine', 'predictor', 'server', 'turing machine', 'web site'], - # 'hypernyms': ['machine'], - # 'root_hypernyms': ['entity'], - # 'definition': 'a machine for performing calculations automatically'} + # 'hyponyms': ['quipo', 'máquina de somar', 'Ossos de Napier', 'Ossos de napier', 'Abaco', 'ábaco'], + # 'hypernyms': ['maquinaria', 'máquina'], + # 'root_hypernyms': ['ente', 'entidade', 'ser'], + # 'definition': "Uma máquina pequena utilizada para cálculos matemáticos"} # full answer - ans = d.wordnet.spoken_answer(query) - print(ans) + ans = d.get_data("computer")["definition"] + print("en", ans) # a machine for performing calculations automatically - - # bidirectional auto translate by passing lang - sentence = d.wordnet.spoken_answer("qual é a definição de computador", lang="pt-pt") - print(sentence) - # uma máquina para realizar cálculos automaticamente diff --git a/requirements.txt b/requirements.txt index ec67788..8596d18 100644 --- a/requirements.txt +++ b/requirements.txt @@ -1,6 +1,5 @@ ovos-utils>=0.0.35,<1.0.0 ovos_workshop>=0.0.11,<4.0.0 -ovos-classifiers>=0.0.0a57 ovos-translate-server-plugin ovos-config>=0.0.11,<1.0.0 ovos-plugin-manager>=0.0.26,<1.0.0