kernel_id
int64
24.2k
23.3M
prompt
stringlengths
8
1.85M
completetion
stringlengths
1
182k
comp_name
stringlengths
5
57
21,301,029
def remove_emoji(text): emoji_pattern = re.compile("[" u"\U0001F600-\U0001F64F" u"\U0001F300-\U0001F5FF" u"\U0001F680-\U0001F6FF" u"\U0001F1E0-\U0001F1FF" u"\U00002702-\U000027B0" u"\U000024C2-\U0001F251" "]+", flags=re.UNICODE) return emoji_pattern.sub(r'', text )<feature_engineering>
train_cleaned_df = train.copy()
Natural Language Processing with Disaster Tweets
21,301,029
train['text'] = train['text'].apply(lambda s : remove_emoji(s)) test ['text'] = test ['text'].apply(lambda s : remove_emoji(s)) <string_transform>
tokenizer = AutoTokenizer.from_pretrained('bert-large-uncased') bert = TFBertModel.from_pretrained('bert-large-uncased' )
Natural Language Processing with Disaster Tweets
21,301,029
def create_vocab(df): vocab = Counter() for i in range(df.shape[0]): vocab.update(df.text[i].split()) return(vocab) <categorify>
tokenizer('Shine on you crazy diamond.' )
Natural Language Processing with Disaster Tweets
21,301,029
master=pd.concat(( train,test)).reset_index(drop=True) vocab = create_vocab(master) len(vocab )<set_options>
print("max len of tweets",max([len(x.split())for x in train.text]))
Natural Language Processing with Disaster Tweets
21,301,029
vocab.most_common(50) <define_variables>
x_train = tokenizer( text=train.text.tolist() , add_special_tokens=True, max_length=73, truncation=True, padding=True, return_tensors='tf', return_token_type_ids = False, return_attention_mask = True, verbose = True)
Natural Language Processing with Disaster Tweets
21,301,029
final_vocab = [] min_occur = 2 for k,v in vocab.items() : if v >= min_occur: final_vocab.append(k )<string_transform>
train.target.value_counts()
Natural Language Processing with Disaster Tweets
21,301,029
def filter(tweet): sentence = "" for word in tweet.split() : if word in final_vocab: sentence = sentence + word + ' ' return(sentence )<feature_engineering>
max_len = 73 input_ids = Input(shape=(max_len,), dtype=tf.int32, name="input_ids") input_mask = Input(shape=(max_len,), dtype=tf.int32, name="attention_mask") embeddings = bert(input_ids,attention_mask = input_mask)[1] out = tf.keras.layers.Dropout(0.1 )(embeddings) out = Dense(128, activation='relu' )(out) out = t...
Natural Language Processing with Disaster Tweets
21,301,029
train['text'] = train['text'].apply(lambda s : filter(s)) test ['text'] = test ['text'].apply(lambda s : filter(s))<define_variables>
optimizer = Adam( learning_rate=5e-05, epsilon=1e-08, decay=0.01, clipnorm=1.0) loss = BinaryCrossentropy(from_logits = True) metric = BinaryAccuracy('accuracy'), model.compile( optimizer = optimizer, loss = loss, metrics = metric )
Natural Language Processing with Disaster Tweets
21,301,029
real = train[train.target==1].reset_index() fake = train[train.target==0].reset_index()<statistical_test>
train_history = model.fit( x ={'input_ids':x_train['input_ids'],'attention_mask':x_train['attention_mask']} , y = y_train, epochs=12, batch_size=32 )
Natural Language Processing with Disaster Tweets
21,301,029
def get_ngrams(data,n): all_words = [] for i in range(len(data)) : temp = data["text"][i].split() for word in temp: all_words.append(word) tokenized = all_words esBigrams = ngrams(tokenized, n) esBigram_wordlist = nltk.FreqDist(esBigrams) top100 = esBigram_wordlist.most_common(100) top100 = dict(top100) df_ngrams ...
x_test = tokenizer( text=test.text.tolist() , add_special_tokens=True, max_length=73, truncation=True, padding=True, return_tensors='tf', return_token_type_ids = False, return_attention_mask = True, verbose = True)
Natural Language Processing with Disaster Tweets
21,301,029
real_unigrams = get_ngrams(real,1) fake_unigrams = get_ngrams(fake,1 )<categorify>
predicted = model.predict({'input_ids':x_test['input_ids'],'attention_mask':x_test['attention_mask']} )
Natural Language Processing with Disaster Tweets
21,301,029
real_bigrams = get_ngrams(real,2) fake_bigrams = get_ngrams(fake,2 )<categorify>
y_predicted = np.where(predicted>0.5,1,0 )
Natural Language Processing with Disaster Tweets
21,301,029
real_trigrams = get_ngrams(real,3) fake_trigrams = get_ngrams(fake,3 )<string_transform>
y_predictedd = y_predicted.reshape(( 1,3263)) [0]
Natural Language Processing with Disaster Tweets
21,301,029
def word_cloud(df): comment_words = '' stopwords = set(STOPWORDS) for val in df.text: val = str(val) tokens = val.split() for i in range(len(tokens)) : tokens[i] = tokens[i].lower() comment_words += " ".join(tokens)+" " wordcloud = WordCloud(width = 800, height = 800, background_color ='white', stopwords = stopwords,...
sample['id'] = test.id sample['target'] = y_predictedd
Natural Language Processing with Disaster Tweets
21,301,029
def get_f1(y_true, y_pred): true_positives = K.sum(K.round(K.clip(y_true * y_pred, 0, 1))) possible_positives = K.sum(K.round(K.clip(y_true, 0, 1))) predicted_positives = K.sum(K.round(K.clip(y_pred, 0, 1))) precision = true_positives /(predicted_positives + K.epsilon()) recall = true_positives /(possible_positives...
sample.to_csv('submission_a.csv',index = False )
Natural Language Processing with Disaster Tweets
21,018,653
def create_tokenizer(lines): tokenizer = Tokenizer() tokenizer.fit_on_texts(lines) return tokenizer<prepare_x_and_y>
df = pd.read_csv("/kaggle/input/nlp-getting-started/train.csv", na_filter=False) df.head()
Natural Language Processing with Disaster Tweets
21,018,653
X = train.text y = train.target test_id = test.id test.drop(["id","location","keyword"],1,inplace = True )<split>
nlp = spacy.load("en_core_web_sm") def preprocess(text): doc = nlp(text) token_semstop = [word for word in doc if not word.is_stop if not word.text == ' text = ' '.join(token.lower_ for token in token_semstop) text = re.sub(r'(@\w+|https?:\S+)', '', text) text = text.replace(r'&amp;?', r'and') text = re.sub(r'(&gt...
Natural Language Processing with Disaster Tweets
21,018,653
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.2, random_state = 42 )<string_transform>
train_df['text'] = train_df['text'].apply(preprocess) train_df.head()
Natural Language Processing with Disaster Tweets
21,018,653
tokenizer = create_tokenizer(X_train) X_train_set = tokenizer.texts_to_matrix(X_train, mode = 'freq') <choose_model_class>
second_df = pd.read_csv("/kaggle/input/nlp-getting-started/test.csv", na_filter=False) second_df.head() test_df = second_df[['text']].copy() test_df['text'] = test_df['text'].apply(preprocess)
Natural Language Processing with Disaster Tweets
21,018,653
def define_model(n_words): model = Sequential() model.add(Dense(128, input_shape=(n_words,), activation='relu')) model.add(Dense(1, activation='sigmoid')) model.compile(loss='binary_crossentropy', optimizer='adam', metrics = [get_f1]) model.summary() plot_model(model, to_file='model.png', show_shapes=True) return mod...
vectorizer = TfidfVectorizer(use_idf=True, ngram_range=(1,2), preprocessor=preprocess) tfidf_data = vectorizer.fit_transform(train_df['text'])
Natural Language Processing with Disaster Tweets
21,018,653
model.fit(X_train_set,y_train,epochs=10,verbose=2 )<predict_on_test>
labels = train_df['target'].values mnb = MultinomialNB() mnb.fit(tfidf_data, labels) X_train, X_test, y_train, y_test = train_test_split(tfidf_data, labels, random_state=42, test_size=0.2) print(classification_report(y_test, mnb.predict(X_test), digits=4))
Natural Language Processing with Disaster Tweets
21,018,653
X_test_set = tokenizer.texts_to_matrix(X_test, mode = 'freq') y_pred = model.predict_classes(X_test_set )<compute_test_metric>
submission = pd.read_csv("/kaggle/input/nlp-getting-started/sample_submission.csv") tfidf_test = vectorizer.transform(test_df['text']) test_df['target'] = mnb.predict(tfidf_test) test_df.head()
Natural Language Processing with Disaster Tweets
21,018,653
<prepare_x_and_y><EOS>
submission['target'] = test_df['target'] submission.to_csv("sample_submission.csv", index=False) submission.head()
Natural Language Processing with Disaster Tweets
17,733,566
<SOS> metric: meanfscore Kaggle data source: natural-language-processing-with-disaster-tweets<predict_on_test>
for dirname, _, filenames in os.walk('/kaggle/input'): for filename in filenames: print(os.path.join(dirname, filename))
Natural Language Processing with Disaster Tweets
17,733,566
y_test_pred = model.predict_classes(test_set )<save_to_csv>
df_train = pd.read_csv("/kaggle/input/nlp-getting-started/train.csv") df_test = pd.read_csv("/kaggle/input/nlp-getting-started/test.csv" )
Natural Language Processing with Disaster Tweets
17,733,566
sub = pd.DataFrame() sub['Id'] = test_id sub['target'] = y_test_pred sub.to_csv('submission_1.csv',index=False )<train_model>
def format_keyword(df): df["keyword"] = df["keyword"].fillna(".") df["keyword"] = df.keyword.str.replace("%20"," " )
Natural Language Processing with Disaster Tweets
17,733,566
t = Tokenizer() t.fit_on_texts(X_train.tolist() )<define_variables>
df_train.loc[df_train.target==0]["keyword"].value_counts()
Natural Language Processing with Disaster Tweets
17,733,566
vocab_size = len(t.word_index)+ 1<load_from_csv>
df_count = df_train.text.str.split().str.len() max(df_count )
Natural Language Processing with Disaster Tweets
17,733,566
embeddings_index = dict() f = open('.. /input/glove6b100dtxt/glove.6B.100d.txt', mode='rt', encoding='utf-8') for line in f: values = line.split() word = values[0] coefs = asarray(values[1:], dtype='float32') embeddings_index[word] = coefs f.close() print('Loaded %s word vectors.' % len(embeddings_index))<categorify>
def process_text(text): text=text.replace(" ","") text = re.sub(r'@\S+','',text) text = re.sub(r' text = re.sub(r'https?://\S+|www\.\S+|http?://\S+','',text) text = re.sub('[%s]' % re.escape (
Natural Language Processing with Disaster Tweets
17,733,566
encoded_docs = t.texts_to_sequences(X_train.tolist()) max_length = 100 padded_docs = pad_sequences(encoded_docs, maxlen=max_length, padding='post') print(padded_docs )<define_variables>
df_train["text"] = df_train.text.transform(lambda x: process_text(x)) df_test["text"] = df_test.text.transform(lambda x: process_text(x))
Natural Language Processing with Disaster Tweets
17,733,566
mis_spelled = [] embedding_matrix = zeros(( vocab_size, 100)) for word, i in t.word_index.items() : embedding_vector = embeddings_index.get(word) if embedding_vector is not None: embedding_matrix[i] = embedding_vector else: mis_spelled.append(word )<train_on_grid>
df_train["appears"]=df_train.groupby("text" ).text.transform("count" )
Natural Language Processing with Disaster Tweets
17,733,566
model = Sequential() e = Embedding(vocab_size, 100, weights=[embedding_matrix], input_length=100, trainable=False) model.add(e) model.add(Flatten()) model.add(Dense(1, activation='sigmoid')) model.compile(optimizer='adam', loss='binary_crossentropy', metrics=[get_f1]) model.summary() model.fit(padded_docs, y_train,...
df_train["target_std"]=df_train.groupby("text" ).target.transform(np.std) df_train["target_mean"]=df_train.groupby("text" ).target.transform(np.mean )
Natural Language Processing with Disaster Tweets
17,733,566
loss, accuracy = model.evaluate(padded_docs, y_train, verbose=0 )<categorify>
duplicate_ids = df_train.loc[df_train.target_std>0].sort_values(by=["appears","text"],ascending=False ).index
Natural Language Processing with Disaster Tweets
17,733,566
encoded_docs = t.texts_to_sequences(X_test.tolist()) padded_docs = pad_sequences(encoded_docs, maxlen=max_length, padding='post' )<predict_on_test>
df_train = df_train.drop(index = duplicate_ids )
Natural Language Processing with Disaster Tweets
17,733,566
y_pred = model.predict_classes(padded_docs )<categorify>
df_train = df_train.drop_duplicates(subset=["text"] )
Natural Language Processing with Disaster Tweets
17,733,566
encoded_docs = t.texts_to_sequences(test.text.tolist()) padded_docs = pad_sequences(encoded_docs, maxlen=max_length, padding='post' )<predict_on_test>
df_train.reset_index(drop=True,inplace=True) df_train
Natural Language Processing with Disaster Tweets
17,733,566
y_test_pred = model.predict_classes(padded_docs )<save_to_csv>
nlp = spacy.load("en_core_web_lg") keyword_train = np.array([nlp(text ).vector for text in df_train.keyword]) keyword_test = np.array([nlp(text ).vector for text in df_test.keyword] )
Natural Language Processing with Disaster Tweets
17,733,566
sub = pd.DataFrame() sub['Id'] = test_id sub['target'] = y_test_pred sub.to_csv('submission_2.csv',index=False )<train_model>
def nlp_vectors(text): res = [] doc = nlp(text) for token in doc: if not token.is_space: res.append(token.vector) return res def build_nlp_vectors(df_text): spacy_vectors =([nlp_vectors(text)for text in df_text]) max_length = 0; for vector in spacy_vectors: max_length = max(max_length, len(vector)) print(f"Maximum L...
Natural Language Processing with Disaster Tweets
17,733,566
<categorify>
nlp_train = build_nlp_vectors(df_train.text )
Natural Language Processing with Disaster Tweets
17,733,566
encoded_docs = t.texts_to_sequences(X_train.tolist()) padded_docs = pad_sequences(encoded_docs, maxlen=max_length, padding='post') print(padded_docs )<define_variables>
tokenizer = ppb.DistilBertTokenizer.from_pretrained("distilbert-base-uncased") bert_model = ppb.DistilBertModel.from_pretrained("distilbert-base-uncased" )
Natural Language Processing with Disaster Tweets
17,733,566
vocab_size<choose_model_class>
def process_data(df_text): tokens = df_text.apply(lambda text: tokenizer.encode(text,add_special_tokens=True)) max_len = 0; i = 0; for token in tokens.values: max_len = max(max_len,len(token)) print(f"Max Length: {max_len}") padded = np.array([i+[0]*(max_len-len(i)) for i in tokens.values]) attention_mask = np.where(...
Natural Language Processing with Disaster Tweets
17,733,566
def define_model(vocab_size, max_length): model = Sequential() model.add(Embedding(vocab_size, 100, input_length=max_length)) model.add(Conv1D(filters=32, kernel_size=8, activation='relu')) model.add(MaxPooling1D(pool_size=2)) model.add(Flatten()) model.add(Dense(10, activation='relu')) model.add(Dense(1, activation='...
X_train = process_data(df_train.text )
Natural Language Processing with Disaster Tweets
17,733,566
model = define_model(vocab_size, max_length) model.fit(padded_docs, y_train, epochs=10, verbose=2 )<compute_test_metric>
y_train = df_train.target
Natural Language Processing with Disaster Tweets
17,733,566
loss, accuracy = model.evaluate(padded_docs, y_train, verbose=0 )<categorify>
X_tr, X_val, nlp_tr, nlp_val, kw_tr, kw_val, y_tr, y_val = train_test_split(X_train,nlp_train, keyword_train, y_train, test_size=0.25, train_size=0.75,shuffle=True )
Natural Language Processing with Disaster Tweets
17,733,566
encoded_docs = t.texts_to_sequences(X_test.tolist()) padded_docs = pad_sequences(encoded_docs, maxlen=max_length, padding='post' )<predict_on_test>
def build_nn() : model = tf.keras.Sequential() model.add(layers.Input(shape=(768,))) model.add(layers.Dense(128,activation='tanh')) model.add(layers.Dropout(0.6)) model.add(layers.Dense(32,activation='tanh')) model.add(layers.Dropout(0.6)) model.add(layers.Dense(8,activation='tanh')) model.add(layers.Dense(1,activatio...
Natural Language Processing with Disaster Tweets
17,733,566
y_pred = model.predict_classes(padded_docs )<categorify>
kfold = KFold(n_splits=4, shuffle=True, random_state=1 )
Natural Language Processing with Disaster Tweets
17,733,566
encoded_docs = t.texts_to_sequences(test.text.tolist()) padded_docs = pad_sequences(encoded_docs, maxlen=max_length, padding='post' )<predict_on_test>
def eval_f1_score(X_val, y_val, model): pred_val =(model.predict(X_val)>0.5) f1 = f1_score(y_val,pred_val) return f1
Natural Language Processing with Disaster Tweets
17,733,566
y_test_pred = model.predict_classes(padded_docs )<save_to_csv>
EPOCHS = 100 BATCH_SIZE = 64
Natural Language Processing with Disaster Tweets
17,733,566
sub = pd.DataFrame() sub['Id'] = test_id sub['target'] = y_test_pred sub.to_csv('submission_cnn.csv',index=False )<save_model>
fold = 0 history_by_fold = [] cv_results = [] for train,val in kfold.split(X_train,y_train): nn_model = build_nn() history = nn_model.fit(X_train[train],y_train[train], validation_data=(X_train[val],y_train[val]), epochs=EPOCHS, batch_size=BATCH_SIZE, verbose=0) scores = nn_model.evaluate(X_train[val],y_train[val],ver...
Natural Language Processing with Disaster Tweets
17,733,566
model.save('model.h5' )<choose_model_class>
nn_model = build_nn() history = nn_model.fit(X_tr,y_tr, validation_data=(X_val,y_val), epochs=EPOCHS, batch_size=BATCH_SIZE,verbose=0) scores= nn_model.evaluate(X_val,y_val,verbose=0) print(f"Accuracy: {scores[1]}") print(f"F1 Score: {eval_f1_score(X_val,y_val,nn_model)}" )
Natural Language Processing with Disaster Tweets
17,733,566
def define_model(length, vocab_size): inputs1 = Input(shape=(length,)) embedding1 = Embedding(vocab_size, 100 )(inputs1) conv1 = Conv1D(32, 4, activation='relu' )(embedding1) drop1 = Dropout(0.5 )(conv1) pool1 = MaxPooling1D()(drop1) flat1 = Flatten()(pool1) inputs2 = Input(shape=(length,)) embedding2 = Embedding(...
def build_LSTM() : lstm_model = tf.keras.Sequential() lstm_model.add(layers.Input(shape=(None,300))) lstm_model.add(layers.LSTM(16)) lstm_model.add(layers.Dense(8, activation="tanh")) lstm_model.add(layers.Dense(8, activation="tanh")) lstm_model.add(layers.Dense(1,activation="sigmoid")) lstm_model.compile(loss=tf.kera...
Natural Language Processing with Disaster Tweets
17,733,566
encoded_docs = t.texts_to_sequences(X_train.tolist()) padded_docs = pad_sequences(encoded_docs, maxlen=max_length, padding='post') <train_model>
EPOCHS = 30; BATCH_SIZE = 64;
Natural Language Processing with Disaster Tweets
17,733,566
model = define_model(max_length,vocab_size) model.fit([padded_docs,padded_docs,padded_docs], array(y_train), epochs=7, batch_size=16 )<compute_test_metric>
kfold = KFold(n_splits=4, shuffle=True, random_state=1 )
Natural Language Processing with Disaster Tweets
17,733,566
loss, accuracy = model.evaluate([padded_docs,padded_docs,padded_docs], y_train, verbose=0 )<categorify>
fold = 0 history_by_fold = [] cv_results = [] for train, val in kfold.split(nlp_train,y_train): lstm_model = build_LSTM() history = lstm_model.fit(nlp_train[train],y_train[train], validation_data=(nlp_train[val],y_train[val]), epochs=EPOCHS,batch_size=BATCH_SIZE,verbose=0) scores = lstm_model.evaluate(nlp_train[val],y...
Natural Language Processing with Disaster Tweets
17,733,566
encoded_docs = t.texts_to_sequences(X_test.tolist()) padded_docs = pad_sequences(encoded_docs, maxlen=max_length, padding='post' )<predict_on_test>
lstm_model = build_LSTM() history = lstm_model.fit(nlp_tr,y_tr,validation_data=(nlp_val,y_val), epochs=EPOCHS, batch_size=BATCH_SIZE )
Natural Language Processing with Disaster Tweets
17,733,566
_, acc = model.evaluate([padded_docs,padded_docs,padded_docs], array(y_test), verbose=0) print('Train Accuracy: %.2f' %(acc*100))<categorify>
valid_predict =(lstm_model.predict(nlp_val)> 0.5) f1 = f1_score(y_val, valid_predict) print(f" F1 Score: {f1}" )
Natural Language Processing with Disaster Tweets
17,733,566
encoded_docs = t.texts_to_sequences(test.text.tolist()) padded_docs = pad_sequences(encoded_docs, maxlen=max_length, padding='post' )<predict_on_test>
lr_keywords = LogisticRegression(max_iter=500) lr_keywords.fit(kw_tr,y_tr) val_pred = lr_keywords.predict(kw_val) print(f"Accurcay: {accuracy_score(y_val, val_pred)}") print(f"F1 score: {f1_score(y_val,val_pred)}" )
Natural Language Processing with Disaster Tweets
17,733,566
y_test_pred = model.predict([padded_docs,padded_docs,padded_docs] )<save_to_csv>
nn_tr_predict = nn_model.predict(X_tr) kw_tr_predict = lr_keywords.predict_proba(kw_tr)[:,1] lstm_tr_predict = lstm_model.predict(nlp_tr) nn_val_predict = nn_model.predict(X_val) kw_val_predict = lr_keywords.predict_proba(kw_val)[:,1] lstm_val_predict = lstm_model.predict(nlp_val) kw_tr_predict = kw_tr_predict.resh...
Natural Language Processing with Disaster Tweets
17,733,566
sub = pd.DataFrame() sub['Id'] = test_id sub['target'] = y_test_pred sub.to_csv('submission_multi-cnn.csv',index=False )<load_from_url>
lr = LogisticRegression() lr.fit(concat_tr,y_tr) val_pred = lr.predict(concat_val) print(f"Accurcay: {accuracy_score(y_val, val_pred)}") print(f"F1 score: {f1_score(y_val,val_pred)}" )
Natural Language Processing with Disaster Tweets
17,733,566
!wget --quiet https://raw.githubusercontent.com/tensorflow/models/master/official/nlp/bert/tokenization.py<import_modules>
X_test = process_data(df_test.text )
Natural Language Processing with Disaster Tweets
17,733,566
from tensorflow.keras.layers import Dense, Input from tensorflow.keras.optimizers import Adam from tensorflow.keras.models import Model from tensorflow.keras.callbacks import ModelCheckpoint import tensorflow_hub as hub import tokenization<load_from_csv>
nlp_test = build_nlp_vectors(df_test.text )
Natural Language Processing with Disaster Tweets
17,733,566
train= pd.read_csv('.. /input/nlp-getting-started/train.csv') test=pd.read_csv('.. /input/nlp-getting-started/test.csv' )<categorify>
df_test["nn_predict"]= nn_model.predict(X_test) df_test["lstm_predict"]= lstm_model.predict(nlp_test) df_test["keyword_predict"] = lr_keywords.predict_proba(keyword_test)[:,1] features = ["nn_predict","keyword_predict","lstm_predict"] test_features = df_test[features] predict = lr.predict(test_features )
Natural Language Processing with Disaster Tweets
17,733,566
<choose_model_class><EOS>
output = pd.DataFrame({"id":df_test.id, "target":predict}) output.to_csv("submission.csv",index=False) output
Natural Language Processing with Disaster Tweets
17,691,511
<SOS> metric: meanfscore Kaggle data source: natural-language-processing-with-disaster-tweets<choose_model_class>
%matplotlib inline InteractiveShell.ast_node_interactivity = 'all' !pip install chart_studio plotly.offline.init_notebook_mode(connected=True) cufflinks.go_offline() cufflinks.set_config_file(world_readable=True, theme='pearl') warnings.filterwarnings('ignore' )
Natural Language Processing with Disaster Tweets
17,691,511
%%time module_url = "https://tfhub.dev/tensorflow/bert_en_uncased_L-24_H-1024_A-16/1" bert_layer = hub.KerasLayer(module_url, trainable=True )<data_type_conversions>
data = pd.read_csv('.. /input/nlp-getting-started/train.csv' )
Natural Language Processing with Disaster Tweets
17,691,511
vocab_file = bert_layer.resolved_object.vocab_file.asset_path.numpy() do_lower_case = bert_layer.resolved_object.do_lower_case.numpy()<choose_model_class>
def create_corpus(target): corpus = [] for i in data[data['target']==target]['text'].str.split() : for x in i: corpus.append(x) return corpus
Natural Language Processing with Disaster Tweets
17,691,511
tokenizer = tokenization.FullTokenizer(vocab_file, do_lower_case) <categorify>
lemmatizer = WordNetLemmatizer() def preprocess_data(data): text = re.sub(r'https?://\S+|www\.\S+|http?://\S+',' ',data) text = re.sub(r"won't", " will not", text) text = re.sub(r"won't've", " will not have", text) text = re.sub(r"can't", " can not", text) text = re.sub(r"don't", " do not", text) text = re.sub(r"c...
Natural Language Processing with Disaster Tweets
17,691,511
train_input = bert_encode(train.text.values, tokenizer, max_len=160) test_input = bert_encode(test.text.values, tokenizer, max_len=160) train_labels = train.target.values<train_model>
common_words = ['via','like','build','get','would','one','two','feel', 'lol','fuck','take','way','may','first','latest','want', 'make','back','see','know','let','look','come','got', 'still','say','think','great','pleas','amp'] def text_cleaning(data): return ' '.join(i for i in data.split() if i not in common_words) d...
Natural Language Processing with Disaster Tweets
17,691,511
train_history = model.fit( train_input, train_labels, validation_split=0.2, epochs=3, batch_size=16 )<predict_on_test>
def top_ngrams(data,n,grams): count_vec = CountVectorizer(ngram_range=(grams,grams)).fit(data) bow = count_vec.transform(data) add_words = bow.sum(axis=0) word_freq = [(word, add_words[0, idx])for word, idx in count_vec.vocabulary_.items() ] word_freq = sorted(word_freq, key = lambda x: x[1], reverse=True) return w...
Natural Language Processing with Disaster Tweets
17,691,511
test_pred = model.predict(test_input )<save_to_csv>
common_uni = top_ngrams(data["Cleaned_text"],10,1) common_bi = top_ngrams(data["Cleaned_text"],10,2) common_tri = top_ngrams(data["Cleaned_text"],10,3) common_uni_df = pd.DataFrame(common_uni,columns=['word','freq']) common_bi_df = pd.DataFrame(common_bi,columns=['word','freq']) common_tri_df = pd.DataFrame(common...
Natural Language Processing with Disaster Tweets
17,691,511
submission=pd.DataFrame() submission['Id']=test_id submission['target'] = test_pred.round().astype(int) submission.to_csv('submission_3.csv', index=False) <set_options>
X_inp_clean = data['Cleaned_text'] X_inp_original = data['text'] y_inp = data['target']
Natural Language Processing with Disaster Tweets
17,691,511
np.random.seed(1) nltk.download('stopwords') tf.random.set_seed(1) pd.set_option('display.max_colwidth', 500) warnings.filterwarnings('ignore' )<load_from_csv>
word_tokenizer = Tokenizer() word_tokenizer.fit_on_texts(X_inp_clean.values) vocab_length = len(word_tokenizer.word_index)+ 1
Natural Language Processing with Disaster Tweets
17,691,511
train = pd.read_csv(".. /input/nlp-getting-started/train.csv") test = pd.read_csv(".. /input/nlp-getting-started/test.csv") print("Train Shape :", train.shape) print("Test Shape :", test.shape )<count_missing_values>
def embed(corpus): return word_tokenizer.texts_to_sequences(corpus) longest_train = max(X_inp_clean.values, key=lambda sentence: len(word_tokenize(sentence))) length_long_sentence = len(word_tokenize(longest_train)) padded_sentences = pad_sequences(embed(X_inp_clean.values), length_long_sentence, padding='post' )
Natural Language Processing with Disaster Tweets
17,691,511
train.isnull().sum()<count_missing_values>
embeddings_dictionary = dict() embedding_dim = 100 glove_file = open('.. /input/glove6b100dtxt/glove.6B.100d.txt') for line in glove_file: records = line.split() word = records[0] vector_dimensions = np.asarray(records[1:], dtype='float32') embeddings_dictionary [word] = vector_dimensions glove_file.close()
Natural Language Processing with Disaster Tweets
17,691,511
train.isnull().sum()<count_values>
embedding_matrix = np.zeros(( vocab_length, embedding_dim)) for word, index in word_tokenizer.word_index.items() : embedding_vector = embeddings_dictionary.get(word) if embedding_vector is not None: embedding_matrix[index] = embedding_vector
Natural Language Processing with Disaster Tweets
17,691,511
train['target'].value_counts(normalize = True )<count_values>
X_train, X_val, y_train, y_val = train_test_split(padded_sentences, y_inp.values,test_size=0.2,random_state=1 )
Natural Language Processing with Disaster Tweets
17,691,511
train.keyword.value_counts()<count_values>
def CNN(hp): model = keras.Sequential() hp_learning_rate = hp.Choice('learning_rate', values=[3e-2, 3e-3, 3e-4, 3e-5]) model.add(Embedding(vocab_length, 100, weights=[embedding_matrix], input_length=length_long_sentence,trainable=False)) model.add(Conv1D(filters=hp.Int('conv_1_filter',min_value=21,max_value=200,step=1...
Natural Language Processing with Disaster Tweets
17,691,511
train.location.value_counts()<filter>
tuner_CNN = kt.Hyperband(CNN,objective='val_accuracy', max_epochs=15,factor=5, directory='my_dir', project_name='DisasterTweets_kt', overwrite=True )
Natural Language Processing with Disaster Tweets
17,691,511
real_tweets = train[train['target']==1]['text'] real_tweets.values[0:5]<define_variables>
stop_early = EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=10) tuner_CNN.search(X_train, y_train, epochs=15, validation_data=(X_val,y_val),callbacks=[stop_early]) best_hps_CNN=tuner_CNN.get_best_hyperparameters(num_trials=1)[0]
Natural Language Processing with Disaster Tweets
17,691,511
fake_tweets = train[train['target']==0]['text'] fake_tweets.values[0:5]<string_transform>
model_CNN = tuner_CNN.hypermodel.build(best_hps_CNN) checkpoint = ModelCheckpoint( 'model_CNN.h5', monitor = 'val_loss', verbose = 1, save_best_only = True ) history_CNN = model_CNN.fit(X_train, y_train,epochs=50, validation_data=(X_val,y_val), callbacks=[checkpoint,stop_early] )
Natural Language Processing with Disaster Tweets
17,691,511
def clean_text(text): text = text.lower() text = re.sub('\[.*?\]', '', text) text = re.sub('https?://\S+|www\.\S+', '', text) text = re.sub('<.*?>+', '', text) text = re.sub('[%s]' % re.escape(string.punctuation), '', text) text = re.sub(' ', '', text) text = re.sub('\w*\d\w*', '', text) return text<feature_engin...
def MultichannelCNN(hp): inputs1 = Input(shape=(length_long_sentence,)) embedding1 = Embedding(vocab_length, 100, weights=[embedding_matrix], input_length=length_long_sentence, trainable=False )(inputs1) conv1 = Conv1D(filters=hp.Int('conv_1_filter',min_value=21,max_value=150,step=14), kernel_size=hp.Choice('conv_1_ke...
Natural Language Processing with Disaster Tweets
17,691,511
train['cleaned_text'] = train['text'].apply(lambda x: clean_text(x)) test['cleaned_text'] = test['text'].apply(lambda x: clean_text(x)) train['cleaned_text'].head()<feature_engineering>
tuner_MCNN = kt.Hyperband(MultichannelCNN,objective='val_accuracy', max_epochs=15,factor=5, directory='my_dir', project_name='DisasterTweetsMCNN_kt', overwrite=True) stop_early = EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=10) tuner_MCNN.search([X_train,X_train], y_train, epochs=15, validation_d...
Natural Language Processing with Disaster Tweets
17,691,511
!pip install nlppreprocess nlp = NLP() train['stopwords_cleaned'] = train['cleaned_text'].apply(nlp.process) test['stopwords_cleaned'] = test['cleaned_text'].apply(nlp.process )<categorify>
model_MCNN = tuner_MCNN.hypermodel.build(best_hps_MCNN) checkpoint = ModelCheckpoint( 'model_MCNN.h5', monitor = 'val_loss', verbose = 1, save_best_only = True ) history_MCNN = model_MCNN.fit([X_train,X_train], y_train,epochs=50, validation_data=([X_val,X_val], y_val), callbacks=[checkpoint,stop_early] )
Natural Language Processing with Disaster Tweets
17,691,511
en_model = spacy.load('en', disable=['parser', 'ner']) def lemmatization(texts): output = [] for i in texts: s = [token.lemma_ for token in en_model(i)] output.append(' '.join(s)) return output<categorify>
def BiLSTM(hp): model = Sequential() model.add(Embedding(input_dim=embedding_matrix.shape[0], output_dim=embedding_matrix.shape[1], weights = [embedding_matrix], input_length=length_long_sentence,trainable = False)) model.add(Bidirectional(CuDNNLSTM(units = hp.Int('dense_1', min_value=21,max_value=120,step=14) ,return...
Natural Language Processing with Disaster Tweets
17,691,511
train['lemmatized_text'] = lemmatization(train['stopwords_cleaned']) test['lemmatized_text'] = lemmatization(test['stopwords_cleaned'] )<split>
tuner_BiLSTM = kt.Hyperband(BiLSTM,objective='val_accuracy', max_epochs=15,factor=5, directory='my_dir', project_name='DisasterTweetsBiLSTM_kt', overwrite=True) stop_early = EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=12) tuner_BiLSTM.search(X_train, y_train, epochs=15, validation_data=(X_val, y...
Natural Language Processing with Disaster Tweets
17,691,511
x_train, x_test, y_train, y_test = train_test_split(train['stopwords_cleaned'], train['target'], test_size = 0.2, random_state = 1 )<choose_model_class>
model_BiLSTM = tuner_BiLSTM.hypermodel.build(best_hps_BiLSTM) checkpoint = ModelCheckpoint( 'model_BiLSTM.h5', monitor = 'val_loss', verbose = 1, save_best_only = True ) history_BiLSTM = model_BiLSTM.fit(X_train, y_train, epochs=50, validation_data=(X_val, y_val), callbacks=[checkpoint,stop_early] )
Natural Language Processing with Disaster Tweets
17,691,511
hub_layer = hub.KerasLayer('https://tfhub.dev/google/universal-sentence-encoder/4', input_shape = [], output_shape = [512], dtype = tf.string, trainable = True) model = tf.keras.models.Sequential() model.add(hub_layer) model.add(tf.keras.layers.Dense(128, activation = 'relu')) model.add(tf.keras.layers.Dense(32, acti...
onehot_encoder = OneHotEncoder(sparse=False) y =(np.asarray(y_inp)).reshape(-1,1) Y = onehot_encoder.fit_transform(y) X_train, X_val, y_train, y_val = train_test_split(X_inp_clean,Y, test_size=0.2, random_state=1 )
Natural Language Processing with Disaster Tweets
17,691,511
model.compile(optimizer = 'adam', loss = 'binary_crossentropy', metrics = ['accuracy'] )<train_model>
model_checkpoint = "distilbert-base-uncased-finetuned-sst-2-english" tokenizer = AutoTokenizer.from_pretrained(model_checkpoint, use_fast=True )
Natural Language Processing with Disaster Tweets
17,691,511
model.fit(x_train, y_train, epochs = 1, validation_data =(x_test, y_test))<predict_on_test>
tokenizer("Hello, this one sentence!", "And this sentence goes with it." )
Natural Language Processing with Disaster Tweets
17,691,511
pred = model.predict_classes(test['stopwords_cleaned'] )<save_to_csv>
def regular_encode(texts, tokenizer, maxlen=512): enc_di = tokenizer.batch_encode_plus( texts, return_token_type_ids=False, pad_to_max_length=True, max_length=maxlen, add_special_tokens = True, truncation=True ) return np.array(enc_di['input_ids']) X_train_t = regular_encode(list(X_train), tokenizer, maxlen=512) X...
Natural Language Processing with Disaster Tweets
17,691,511
submission = pd.read_csv(".. /input/nlp-getting-started/sample_submission.csv") submission['target'] = pred submission.to_csv('submission.csv', index=False )<save_to_csv>
AUTO = tf.data.experimental.AUTOTUNE batch_size = 16 train_dataset =( tf.data.Dataset .from_tensor_slices(( X_train_t, y_train)) .repeat() .shuffle(1995) .batch(batch_size) .prefetch(AUTO) ) valid_dataset =( tf.data.Dataset .from_tensor_slices(( X_val_t, y_val)) .batch(batch_size) .cache() .prefetch(AUTO) )
Natural Language Processing with Disaster Tweets
17,691,511
submission.to_csv("submission.csv", index = False )<set_options>
def build_model(transformer, max_len=512): input_word_ids = Input(shape=(max_len,), dtype=tf.int32, name="input_word_ids") sequence_output = transformer(input_word_ids)[0] cls_token = sequence_output[:, 0, :] out = Dense(2, activation='softmax' )(cls_token) model = Model(inputs=input_word_ids, outputs=out) model.com...
Natural Language Processing with Disaster Tweets
17,691,511
np.random.seed(0) plt.style.use('ggplot') stop=set(stopwords.words('english')) np.random.seed(1) for dirname, _, filenames in os.walk('/kaggle/input'): for filename in filenames: print(os.path.join(dirname, filename)) <load_from_csv>
transformer_layer = TFAutoModel.from_pretrained(model_checkpoint) model_DistilBert = build_model(transformer_layer )
Natural Language Processing with Disaster Tweets
17,691,511
train= pd.read_csv('.. /input/nlp-getting-started/train.csv') test=pd.read_csv('.. /input/nlp-getting-started/test.csv') train.head()<count_missing_values>
n_steps = X_train.shape[0] // batch_size history_DistilBert = model_DistilBert.fit(train_dataset, steps_per_epoch=n_steps, validation_data=valid_dataset, epochs=3 )
Natural Language Processing with Disaster Tweets
17,691,511
train.isnull().sum(axis=0 )<count_missing_values>
test = pd.read_csv('.. /input/nlp-getting-started/test.csv') test["Cleaned_text"] = test["text"].apply(preprocess_data) test["Cleaned_text"] = test["Cleaned_text"].apply(text_cleaning) test_sentences = pad_sequences(embed(test.Cleaned_text.values), length_long_sentence, padding='post' )
Natural Language Processing with Disaster Tweets
17,691,511
test.isnull().sum(axis=0 )<count_values>
predsCNN = model_CNN.predict_classes(test_sentences) predictions_test = pd.DataFrame(predsCNN) test_id = pd.DataFrame(test["id"]) submissionCNN = pd.concat([test_id,predictions_test],axis=1) submissionCNN.columns = ["id","target"] submissionCNN.to_csv("submissionCNN.csv",index=False )
Natural Language Processing with Disaster Tweets
17,691,511
keyword_cnt = train.keyword.value_counts() keyword_cnt<count_values>
predsMCNN = model_MCNN.predict([test_sentences,test_sentences]) predsMCNN =(predsMCNN[:,0] > 0.5 ).astype(np.int) predictions_test = pd.DataFrame(predsMCNN) submissionMCNN = pd.concat([test_id,predictions_test],axis=1) submissionMCNN.columns = ["id","target"] submissionMCNN.to_csv("submissionMCNN.csv",index=False )
Natural Language Processing with Disaster Tweets
17,691,511
train_fake = train[train['target'] == 1] keyword_cnt_fake = train_fake.keyword.value_counts() keyword_cnt_fake<string_transform>
predsBiLSTM = model_BiLSTM.predict(test_sentences) predsBiLSTM =(predsBiLSTM[:,0] > 0.5 ).astype(np.int) predictions_test = pd.DataFrame(predsBiLSTM) submissionBiLSTM = pd.concat([test_id,predictions_test],axis=1) submissionBiLSTM.columns = ["id","target"] submissionBiLSTM.to_csv("submissionBiLSTM.csv",index=False ...
Natural Language Processing with Disaster Tweets
17,691,511
n_corpus=[] for text in tqdm(train['text']): text = re.sub(r'https?://\S+|www\.\S+', '', text) text = re.sub(r'<.*?>', '', text) text = re.sub(r'[^a-zA-Z0-9]+', ' ', text) text = re.sub(r'[0-9]', '', text) text = text.lower() text = nltk.word_tokenize(text) ps = PorterStemmer() text = [ps.stem(word)for word in tex...
X_test = regular_encode(list(test.Cleaned_text), tokenizer, maxlen=512) test1 =(tf.data.Dataset.from_tensor_slices(X_test ).batch(batch_size)) pred = model_DistilBert.predict(test1,verbose = 0) pred = np.argmax(pred,axis=-1) pred = pred.astype('int32') res=pd.read_csv('.. /input/nlp-getting-started/sample_submissio...
Natural Language Processing with Disaster Tweets
10,225,902
train['text_n']=n_corpus train.drop('text',axis=1 )<string_transform>
import transformers import numpy as np import pandas as pd import torch import torch.nn as nn import torch.nn.functional as F from sklearn.model_selection import train_test_split
Natural Language Processing with Disaster Tweets