Python nltk.word_tokenize() Examples

def predict(self, query):
        usr = query
        print('Mem2Seq usr:', usr)
        #example input: 'please find a restaurant called nusha .'
        self.t += 1
        print('Mem2Seq turn:', self.t)
        usr = ' '.join(word_tokenize(usr.lower()))
        self.memory += generate_memory(usr, '$u', self.t)
        src_plain = (self.memory+[['$$$$']*MEM_TOKEN_SIZE],)
        src_seqs = plain2tensor(self.lang.word2index, src_plain[0])
        words = self.model.evaluate_batch(1, src_seqs, [len(src_plain[0])], None, None, None, None, src_plain)
        row = np.transpose(words)[0].tolist()
        if '<EOS>' in row:
            row = row[:row.index('<EOS>')]
        sys = ' '.join(row)
        sys = denormalize(sys)
        print('Mem2Seq sys:', sys)
        self.memory += generate_memory(sys, '$s', self.t)
        return sys 

def analyze_en():
    translation_path = os.path.join(train_translation_folder, train_translation_en_filename)

    with open(translation_path, 'r') as f:
        sentences = f.readlines()

    sent_lengths = []

    for sentence in tqdm(sentences):
        sentence_en = sentence.strip().lower()
        tokens = [normalizeString(s) for s in nltk.word_tokenize(sentence_en)]
        seg_list = list(jieba.cut(sentence.strip()))
        # Update word frequency
        sent_lengths.append(len(seg_list))

    num_bins = 100
    n, bins, patches = plt.hist(sent_lengths, num_bins, facecolor='blue', alpha=0.5)
    title = 'English Sentence Lengths Distribution'
    plt.title(title)
    plt.show() 

def extract_features(corpus):
    '''Extract TF-IDF features from corpus'''

    stop_words = nltk.corpus.stopwords.words("english")

    # vectorize means we turn non-numerical data into an array of numbers
    count_vectorizer = feature_extraction.text.CountVectorizer(
        lowercase=True,  # for demonstration, True by default
        tokenizer=nltk.word_tokenize,  # use the NLTK tokenizer
        min_df=2,  # minimum document frequency, i.e. the word must appear more than once.
        ngram_range=(1, 2),
        stop_words=stop_words
    )
    processed_corpus = count_vectorizer.fit_transform(corpus)
    processed_corpus = feature_extraction.text.TfidfTransformer().fit_transform(
        processed_corpus)

    return processed_corpus 

def extract_features(corpus):
    '''Extract TF-IDF features from corpus'''

    stop_words = nltk.corpus.stopwords.words("english")

    # vectorize means we turn non-numerical data into an array of numbers
    count_vectorizer = feature_extraction.text.CountVectorizer(
        lowercase=True,  # for demonstration, True by default
        tokenizer=nltk.word_tokenize,  # use the NLTK tokenizer
        min_df=2,  # minimum document frequency, i.e. the word must appear more than once.
        ngram_range=(1, 2),
        stop_words=stop_words
    )
    processed_corpus = count_vectorizer.fit_transform(corpus)
    processed_corpus = feature_extraction.text.TfidfTransformer().fit_transform(
        processed_corpus)

    return processed_corpus 

def __init__(self, sentences_file, stopwords):
        self.dictionary = None
        self.corpus = None
        f_sentences = codecs.open(sentences_file, encoding='utf-8')
        documents = list()
        count = 0
        print("Gathering sentences and removing stopwords")
        for line in f_sentences:
            line = re.sub('<[A-Z]+>[^<]+</[A-Z]+>', '', line)

            # remove stop words and tokenize
            document = [word for word in nltk.word_tokenize(line.lower()) if word not in stopwords]
            documents.append(document)
            count += 1
            if count % 10000 == 0:
                sys.stdout.write(".")

        f_sentences.close()

        self.dictionary = corpora.Dictionary(documents)
        self.corpus = [self.dictionary.doc2bow(text) for text in documents]
        self.tf_idf_model = TfidfModel(self.corpus)

        print(len(documents), "documents red")
        print(len(self.dictionary), " unique tokens") 

def preprocess(text):
	min_length = 3
	text = re.sub('\d+','#',text)
	text = re.sub('\.',' eos ',text)
	# Tokenize
	words = map(lambda word: word.lower(), word_tokenize(text))
	tokens = words
	# Remove non characters
	p = re.compile('[a-zA-Z#]+')
	# Filter tokens (we do not remove stopwords)
	filtered_tokens = list(filter(lambda token: p.match(token) and len(token)>=min_length and (token not in english_stopwords), tokens))
	# Encode to ascii
	filtered_tokens = [token.encode('ascii','ignore') for token in filtered_tokens]

	return filtered_tokens


# Modify this path 

def text_to_num(text):
    tokenized = nltk.word_tokenize(text);
    tags = nltk.pos_tag(tokenized)
    print(tags)
    chunkPattern = r""" Chunk0: {((<NN|CD.?|RB>)<CD.?|VBD.?|VBP.?|VBN.?|NN.?|RB.?|JJ>*)<NN|CD.?>} """
    chunkParser = nltk.RegexpParser(chunkPattern)
    chunkedData = chunkParser.parse(tags)
    print(chunkedData)

    for subtree in chunkedData.subtrees(filter=lambda t: t.label() in "Chunk0"):
        exp = ""
        for l in subtree.leaves():
            exp += str(l[0]) + " "
        exp = exp[:-1]
        print(exp)
        try:
            text = text.replace(exp, str(t2n.text2num(exp)))
        except Exception as e:
            print("error text2num ->", e.args)
        print("text2num -> ", text)
    return text 

def __init__(self):
		import nltk
		from nltk.tag import PerceptronTagger
		from nltk.tokenize import TreebankWordTokenizer
		#return pkgutil.get_data('scattertext',
		#                        'data/viz/semiotic_new.html').decode('utf-8')
		path = os.path.dirname(sys.modules['scattertext'].__file__)+'/data/'
		tokenizer_fn = path + 'punkt.english.pickle'
		tagger_fn = path + 'averaged_perceptron_tagger.pickle'
		#tokenizer_fn = os.path.abspath(resource_filename('scattertext.data', 'punkt.english.pickle'))
		#tagger_fn = os.path.abspath(resource_filename('scattertext.data', 'averaged_perceptron_tagger.pickle'))
		# Load the tagger
		self.tagger = PerceptronTagger(load=False)
		self.tagger.load(tagger_fn)

		# note: nltk.word_tokenize calls the TreebankWordTokenizer, but uses the downloader.
		#       Calling the TreebankWordTokenizer like this allows skipping the downloader.
		#       It seems the TreebankWordTokenizer uses PTB tokenization = regexes. i.e. no downloads
		#       https://github.com/nltk/nltk/blob/develop/nltk/tokenize/treebank.py#L25
		self.tokenize = TreebankWordTokenizer().tokenize
		self.sent_detector = nltk.data.load(tokenizer_fn)

	# http://www.nltk.org/book/ch05.html 

def get_story_question_answer_triples(sqa_file):
    sqatriples = []
    fsqa = open(sqa_file, "rb")
    for line in fsqa:
        line = line.strip().decode("utf8").encode("ascii", "ignore")
        if line.startswith("#"):
            continue
        story, question, answer, correct = line.split("\t")
        swords = []
        story_sents = nltk.sent_tokenize(story)
        for story_sent in story_sents:
            swords.extend(nltk.word_tokenize(story_sent))
        qwords = nltk.word_tokenize(question)
        awords = nltk.word_tokenize(answer)
        is_correct = int(correct) == 1
        sqatriples.append((swords, qwords, awords, is_correct))
    fsqa.close()
    return sqatriples 

def convert_string(text):
    """Convert a str of text into tokenized and selected list of words.

    Parameters
    ----------
    text : str
        Text as one long string.

    Returns
    -------
    words_cleaned : list of str
        List of tokenized words, after processing.

    Notes
    -----
    This function sets text to lower case, and removes stopwords and punctuation.
    """

    words = word_tokenize(text)
    words_cleaned = [word.lower() for word in words if (
        (not word.lower() in stopwords.words('english')) & word.isalnum())]

    return words_cleaned 

def evaluate_sentiment(text):
    pos_score = 0
    neg_score = 0
    tokened = nltk.word_tokenize(text.decode('utf8', 'ignore').replace('<br />',' '))
    pos_pairs = nltk.pos_tag(tokened)
    for tuple in pos_pairs:
        pos = ''
        if tuple[1] == "NN":
            pos = 'n/'
        if tuple[1] == "JJ":
            pos = 'a/'
        if tuple[1] == "VB":
            pos = 'v/'
        if tuple[1] == "RB":
            pos = 'r/'
        try:
            pos_score += sentiwordnet[pos+tuple[0].lower()][0]
            neg_score += sentiwordnet[pos+tuple[0].lower()][1]
        except:
            pass
    return pos_score, neg_score 

def evaluate_sentiment(text):
    pos_score = 0
    neg_score = 0
    tokened = nltk.word_tokenize(text)
    pos_pairs = nltk.pos_tag(tokened)
    for tuple in pos_pairs:
        pos = ''
        if tuple[1] == "NN":
            pos = 'n/'
        if tuple[1] == "JJ":
            pos = 'a/'
        if tuple[1] == "VB":
            pos = 'v/'
        if tuple[1] == "RB":
            pos = 'r/'
        try:
            pos_score += sentiwordnet[pos+tuple[0].lower()][0]
            neg_score += sentiwordnet[pos+tuple[0].lower()][1]
        except:
            pass
    return pos_score, neg_score 

def _nltk_process_sents(self, sents):
        for sentence in sents:
            if isinstance(sentence, STRING_TYPES):
                if self._tokenizer_lang is None:
                    raise ValueError(
                        "No word tokenizer available for this language. "
                        "Please tokenize before calling the parser."
                        )
                sentence = nltk.word_tokenize(sentence, self._tokenizer_lang)

            if IS_PY2:
                sentence = [
                    word.decode('utf-8', 'ignore') if isinstance(word, str) else word
                    for word in sentence
                    ]

            if not self._provides_tags:
                sentence = nltk.pos_tag(sentence)
                yield [word for word, tag in sentence], sentence
            else:
                yield sentence, sentence 

def encode_text(text, word_embeddings, max_sen_len):
    '''
    Encode a sequence of words into corresponding vector representation
    Input:
        text (string) : text (space separated words, etc..)
        word_embeddings (dict) : dictionary mapping from words to their representation
        max_sen_len (int) : maximum sentence length (in words)
    Returns:
        X (np.matrix) : matrix of shape (max_sen_len, embedding_size) after zero padding
    '''
    
    default_embed = np.zeros(300)
    words = word_tokenize(text)[:max_sen_len]
    embeds = [word_embeddings.get(x, default_embed) for x in words]
    embeds += [default_embed] * (max_sen_len - len(embeds))
    return np.array(embeds, dtype=np.float32) 

def encode_text(text, word_embeddings):
    '''
    Encode a sequence of words into corresponding vector representation
    Input:
        text (string) : text (space separated words, etc..)
        word_embeddings (dict) : dictionary mapping from words to their representation
        max_sent_len (int) : maximum sentence length (in words)
    Returns:
        X (np.array) : array of shape (embedding_size,) averaging all word vectors of text
    '''
    
    embed = np.zeros(300)
    count = 0
    words = word_tokenize(text)
    for word in words:
        if word in word_embeddings:
            embed += word_embeddings[word]
            count += 1
    return embed / count 

def mixed_segmentation(in_str, rm_punc=False):
	in_str = str(in_str).decode('utf-8').lower().strip()
	segs_out = []
	temp_str = ""
	sp_char = ['-',':','_','*','^','/','\\','~','`','+','=',
			   '，','。','：','？','！','“','”','；','’','《','》','……','·','、',
			   '「','」','（','）','－','～','『','』']
	for char in in_str:
		if rm_punc and char in sp_char:
			continue
		if re.search(ur'[\u4e00-\u9fa5]', char) or char in sp_char:
			if temp_str != "":
				ss = nltk.word_tokenize(temp_str)
				segs_out.extend(ss)
				temp_str = ""
			segs_out.append(char)
		else:
			temp_str += char

	#handling last part 

def predict(self,Sentence):
        Sentence = words =  word_tokenize(Sentence)
        Sentence = self.addCharInformation(Sentence)
        Sentence = self.padding(self.createTensor(Sentence,self.word2Idx,self.case2Idx,self.char2Idx))
        tokens, casing,char = Sentence
        tokens = np.asarray([tokens])     
        casing = np.asarray([casing])
        char = np.asarray([char])
        pred = self.model.predict([tokens, casing,char], verbose=False)[0]   
        pred = pred.argmax(axis=-1)
        pred = [self.idx2Label[x].strip() for x in pred]
        return list(zip(words,pred)) 

def _preprocess(text: str) -> Tuple[np.ndarray, np.ndarray]:
        tokens = word_tokenize(text)
        # split into lower-case word tokens, in numpy array with shape of (seq, 1)
        words = np.asarray([w.lower() for w in tokens]).reshape(-1, 1)
        # split words into chars, in numpy array with shape of (seq, 1, 1, 16)
        chars = [[c for c in t][:16] for t in tokens]
        chars = [cs + [""] * (16 - len(cs)) for cs in chars]
        chars = np.asarray(chars).reshape(-1, 1, 1, 16)
        return words, chars 

def profile(self, text):
        ''' Create FreqDist of trigrams within text '''
        from nltk import word_tokenize, FreqDist

        clean_text = self.remove_punctuation(text)
        tokens = word_tokenize(clean_text)
        
        fingerprint = FreqDist()
        for t in tokens:
            token_trigram_tuples = trigrams(self._START_CHAR + t + self._END_CHAR)
            token_trigrams = [''.join(tri) for tri in token_trigram_tuples]

            for cur_trigram in token_trigrams:
                if cur_trigram in fingerprint:
                    fingerprint[cur_trigram] += 1
                else:
                    fingerprint[cur_trigram] = 1

        return fingerprint 

def build_wordmap_en():
    translation_path = os.path.join(train_translation_folder, train_translation_en_filename)

    with open(translation_path, 'r') as f:
        sentences = f.readlines()

    word_freq = Counter()

    for sentence in tqdm(sentences):
        sentence_en = sentence.strip().lower()
        tokens = [normalizeString(s) for s in nltk.word_tokenize(sentence_en) if len(normalizeString(s)) > 0]
        # Update word frequency
        word_freq.update(tokens)

    # Create word map
    # words = [w for w in word_freq.keys() if word_freq[w] > min_word_freq]
    words = word_freq.most_common(input_lang_vocab_size - 4)
    word_map = {k[0]: v + 4 for v, k in enumerate(words)}
    word_map['<pad>'] = 0
    word_map['<start>'] = 1
    word_map['<end>'] = 2
    word_map['<unk>'] = 3
    print(len(word_map))
    print(words[:10])

    with open('data/WORDMAP_en.json', 'w') as file:
        json.dump(word_map, file, indent=4) 

def build_samples():
    word_map_zh = json.load(open('data/WORDMAP_zh.json', 'r'))
    word_map_en = json.load(open('data/WORDMAP_en.json', 'r'))

    for usage in ['train', 'valid']:
        if usage == 'train':
            translation_path_en = os.path.join(train_translation_folder, train_translation_en_filename)
            translation_path_zh = os.path.join(train_translation_folder, train_translation_zh_filename)
            filename = 'data/samples_train.json'
        else:
            translation_path_en = os.path.join(valid_translation_folder, valid_translation_en_filename)
            translation_path_zh = os.path.join(valid_translation_folder, valid_translation_zh_filename)
            filename = 'data/samples_valid.json'

        print('loading {} texts and vocab'.format(usage))
        with open(translation_path_en, 'r') as f:
            data_en = f.readlines()

        with open(translation_path_zh, 'r') as f:
            data_zh = f.readlines()

        print('building {} samples'.format(usage))
        samples = []
        for idx in tqdm(range(len(data_en))):
            sentence_zh = data_zh[idx].strip()
            seg_list = jieba.cut(sentence_zh)
            input_zh = encode_text(word_map_zh, list(seg_list))

            sentence_en = data_en[idx].strip().lower()
            tokens = [normalizeString(s) for s in nltk.word_tokenize(sentence_en) if len(normalizeString(s)) > 0]
            output_en = encode_text(word_map_en, tokens)

            if len(input_zh) <= max_len and len(
                    output_en) <= max_len and UNK_token not in input_zh and UNK_token not in output_en:
                samples.append({'input': list(input_zh), 'output': list(output_en)})

        with open(filename, 'w') as f:
            json.dump(samples, f, indent=4)

        print('{} {} samples created at: {}.'.format(len(samples), usage, filename)) 

def tokenize_words(targets):
    while True:
        sentence = (yield)
        words = nltk.word_tokenize(sentence)
        for target in targets:
            target.send(words) 

def extract_features(corpus):
    '''Extract TF-IDF features from corpus'''
    # vectorize means we turn non-numerical data into an array of numbers
    count_vectorizer = feature_extraction.text.CountVectorizer(
        lowercase=True,  # for demonstration, True by default
        tokenizer=nltk.word_tokenize,  # use the NLTK tokenizer
        stop_words='english',  # remove stop words
        min_df=1  # minimum document frequency, i.e. the word must appear more than once.
    )
    processed_corpus = count_vectorizer.fit_transform(corpus)
    processed_corpus = feature_extraction.text.TfidfTransformer().fit_transform(
        processed_corpus)

    return processed_corpus 

def extract_features(corpus):
    '''Extract TF-IDF features from corpus'''

    sa_stop_words = nltk.corpus.stopwords.words("english")

    # words that might invert a sentence's meaning
    white_list = [
        'what', 'but', 'if', 'because', 'as', 'until', 'against',
        'up', 'down', 'in', 'out', 'on', 'off', 'over', 'under', 'again',
        'further', 'then', 'once', 'here', 'there', 'why', 'how', 'all', 'any',
        'most', 'other', 'some', 'such', 'no', 'nor', 'not', 'only', 'own',
        'same', 'so', 'than', 'too', 'can', 'will', 'just', 'don', 'should']

    # take these out of the standard NLTK stop word list
    sa_stop_words = [sw for sw in sa_stop_words if sw not in white_list]

    # vectorize means we turn non-numerical data into an array of numbers
    count_vectorizer = feature_extraction.text.CountVectorizer(
        lowercase=True,  # for demonstration, True by default
        tokenizer=nltk.word_tokenize,  # use the NLTK tokenizer
        min_df=2,  # minimum document frequency, i.e. the word must appear more than once.
        ngram_range=(1, 2),
        stop_words=sa_stop_words
    )
    processed_corpus = count_vectorizer.fit_transform(corpus)
    processed_corpus = feature_extraction.text.TfidfTransformer().fit_transform(
        processed_corpus)

    return processed_corpus 

def _get_metainfo(self, index):
        question = gdef.translate_question(self.questions[index])
        scene = gdef.translate_scene(self.scenes[question['image_index']])
        question['scene'] = scene

        question['image_index'] = question['image_index']
        question['image_filename'] = gdef.get_image_filename(scene)
        question['question_index'] = index
        question['question_tokenized'] = nltk.word_tokenize(question['question'])

        # program section
        has_program = False
        if 'program_nsclseq' in question:
            question['program_raw'] = question['program_nsclseq']
            question['program_seq'] = question['program_nsclseq']
            has_program = True
        elif 'program' in question:
            question['program_raw'] = question['program']
            question['program_seq'] = gdef.program_to_nsclseq(question['program'], question)
            has_program = True

        if has_program:
            question['program_tree'] = nsclseq_to_nscltree(question['program_seq'])
            question['program_qsseq'] = nsclseq_to_nsclqsseq(question['program_seq'])
            question['program_qstree'] = nscltree_to_nsclqstree(question['program_tree'])
            question['question_type'] = question['program_seq'][-1]['op']
        else:
            question['question_type'] = None

        return question 

def clean(s):
    tokens = nltk.word_tokenize(s.lower())
    tokens_clean = [token for token in tokens if token not in stopwords.words('english')]
    tokens_stemmed = [PorterStemmer().stem(token) for token in tokens_clean]
    return tokens_stemmed 

def _word_tokenize(text):
  # TODO(seominjoon): Consider using Stanford Tokenizer or othe tokenizers.
  return [
      word.replace('``', '"').replace("''", '"')
      for word in nltk.word_tokenize(text)
  ] 

def tokenize_question(text: str) -> List[str]:
    return word_tokenize(clean_question(text)) 

def create_qb_tokenizer(
        unigrams=True, bigrams=False, trigrams=False,
        zero_length_token='zerolengthunk', strip_qb_patterns=True):
    def tokenizer(text):
        if strip_qb_patterns:
            text = re.sub(
                '\s+', ' ',
                re.sub(regex_pattern, ' ', text, flags=re.IGNORECASE)
            ).strip().capitalize()
        import nltk
        tokens = nltk.word_tokenize(text)
        if len(tokens) == 0:
            return [zero_length_token]
        else:
            ngrams = []
            if unigrams:
                ngrams.extend(tokens)
            if bigrams:
                ngrams.extend([f'{w0}++{w1}' for w0, w1 in nltk.bigrams(tokens)])
            if trigrams:
                ngrams.extend([f'{w0}++{w1}++{w2}' for w0, w1, w2 in nltk.trigrams(tokens)])

            if len(ngrams) == 0:
                ngrams.append(zero_length_token)
            return ngrams

    return tokenizer 

def tokenize(sequence):
    tokens = [token.replace("``", '"').replace("''", '"').lower() for token in nltk.word_tokenize(sequence)]
    return tokens