def __init__(self,
batch_size=32,
hidden_layer=0,
build_model=False,
comparison=compare.SimMul(),
positions=False,
regression=False,
trainable=False,
siamese=False):
self.batch_size = batch_size
self.trainable = trainable
self.sentence_length = 250
self.hidden_layer = hidden_layer
self.build_model = build_model
# self.tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
self.tokenizer = self.load_tokenizer()
self.cleantext = lambda t: t
self.comparison = comparison
self.positions = positions
self.regression_model = regression
self.siamese = siamese
def __init__(self,
sentence_length=SENTENCE_LENGTH,
batch_size=128,
embeddings=True,
hidden_layer=0,
build_model=False,
comparison=compare.SimMul(),
positions=False,
regression=False,
siamese=False):
if build_model:
BasicNN.__init__(self, sentence_length, batch_size, embeddings,
hidden_layer, build_model)
else:
# Do not load embeddings if we do not need to build the model
BasicNN.__init__(self, sentence_length, batch_size, False,
hidden_layer, build_model)
self.embeddings = embeddings
self.build_model = build_model
self.comparison = comparison
self.positions = positions
self.regression_model = regression
self.siamese = siamese
def __init__(self,
corpusFile,
rougeFile,
metric=['SU4'],
rouge_labels=True,
labels="topn",
labelsthreshold=5,
nb_epoch=3,
verbose=2,
classification_type="Bi-LSTM",
embeddings=True,
hidden_layer=0,
dropout=0.5,
regression=False,
batch_size=128):
"""Initialise the classification system."""
BaseClassification.__init__(self,
corpusFile,
rougeFile,
metric=metric,
rouge_labels=rouge_labels,
labels=labels,
labelsthreshold=labelsthreshold)
self.nb_epoch = nb_epoch
self.verbose = verbose
self.dropout = dropout
self.nnc = None
if classification_type == "BasicNN":
self.nnc = nnc.BasicNN(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size)
elif classification_type == "BERT":
self.nnc = nnc.BasicBERT(hidden_layer=hidden_layer,
batch_size=batch_size)
elif classification_type == "SimilaritiesBERT":
self.nnc = nnc.SimilaritiesBERT(hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimMul(),
regression=regression,
positions=True)
elif classification_type == "LSTMSimilaritiesBERT":
self.nnc = nnc.LSTMSimilaritiesBERT(hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimMul(),
regression=regression,
positions=True)
elif classification_type == "SimilaritiesBioBERT":
self.nnc = nnc.SimilaritiesBioBERT(hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimMul(),
regression=regression,
positions=True)
elif classification_type == "LSTMSimilaritiesBioBERT":
self.nnc = nnc.LSTMSimilaritiesBioBERT(hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimMul(),
regression=regression,
positions=True)
elif classification_type == "SimilaritiesBERTTrainable":
self.nnc = nnc.SimilaritiesBERT(hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimMul(),
regression=regression,
positions=True,
trainable=True)
elif classification_type == "Similarities":
self.nnc = nnc.Similarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimMul(),
regression=regression,
positions=True)
elif classification_type == "SimilaritiesYu":
self.nnc = nnc.Similarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimYu(),
regression=regression,
positions=True)
elif classification_type == "SimilaritiesEuc":
self.nnc = nnc.Similarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimEuc(),
regression=regression,
positions=True)
elif classification_type == "CNNSimilarities":
self.nnc = nnc.CNNSimilarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimMul())
elif classification_type == "CNNSimilaritiesYu":
self.nnc = nnc.CNNSimilarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimYu())
elif classification_type == "CNNSimilaritiesEuc":
self.nnc = nnc.CNNSimilarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimEuc())
elif classification_type == "LSTMSimilarities":
self.nnc = nnc.LSTMSimilarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimMul(),
regression=regression,
positions=True)
elif classification_type == "LSTMSiameseSimilarities":
self.nnc = nnc.LSTMSimilarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimMul(),
regression=regression,
positions=True,
siamese=True)
elif classification_type == "LSTMSimilaritiesYu":
self.nnc = nnc.LSTMSimilarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimMul(),
regression=regression,
positions=True)
elif classification_type == "LSTMSimilaritiesEuc":
self.nnc = nnc.LSTMSimilarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimEuc(),
regression=regression,
positions=True)
def __init__(self, layer1=50, batch_size=128, comparison=compare.SimMul()):
self.layer1 = layer1
self.batch_size = batch_size
self.comparison = comparison
self.name = "NNR-%s-relu(%i)" % (self.comparison.name, self.layer1)
def __init__(self,
corpusFile,
rougeFile,
metric=['SU4'],
nb_epoch=3,
verbose=2,
regression_type="Bi-LSTM",
embeddings=True,
use_peepholes=False,
hidden_layer=0,
dropout=0.5,
batch_size=128):
"""Initialise the regression system."""
BaseRegression.__init__(self, corpusFile, rougeFile, metric)
self.nb_epoch = nb_epoch
self.use_peepholes = use_peepholes
self.verbose = verbose
self.dropout = dropout
self.lstm = None
if regression_type == "BasicNN":
self.lstm = lstm.BasicNN(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size)
elif regression_type == "CNN":
self.lstm = lstm.CNN(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size)
elif regression_type == "LSTM":
self.lstm = lstm.LSTM(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size)
elif regression_type == "Bi-LSTM":
self.lstm = lstm.LSTMBidirectional(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size)
elif regression_type == "Similarities":
self.lstm = lstm.Similarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimMul(),
positions=True)
elif regression_type == "SimilaritiesYu":
self.lstm = lstm.Similarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimYu(),
positions=True)
elif regression_type == "SimilaritiesEuc":
self.lstm = lstm.Similarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimEuc(),
positions=True)
elif regression_type == "CNNSimilarities":
self.lstm = lstm.CNNSimilarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimMul())
elif regression_type == "CNNSimilaritiesYu":
self.lstm = lstm.CNNSimilarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimYu())
elif regression_type == "CNNSimilaritiesEuc":
self.lstm = lstm.CNNSimilarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimEuc())
elif regression_type == "LSTMSimilarities":
self.lstm = lstm.LSTMSimilarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimMul(),
positions=True)
elif regression_type == "LSTMSimilaritiesYu":
self.lstm = lstm.LSTMSimilarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimMul(),
positions=True)
elif regression_type == "LSTMSimilaritiesEuc":
self.lstm = lstm.LSTMSimilarities(embeddings=embeddings,
hidden_layer=hidden_layer,
batch_size=batch_size,
comparison=compare.SimEuc(),
positions=True)
elif args.regression_type == "TfidfNNR":
self.lstm = regressionbaselines.TfidfNNR(
batch_size=args.batch_size, n_components=args.svd_components)
elif args.regression_type == "TfidfSimNNR":
self.lstm = regressionbaselines.TfidfSimNNR(
batch_size=args.batch_size, n_components=args.svd_components)
elif args.regression_type == "TfidfSim2NNR":
self.lstm = regressionbaselines.TfidfSim2NNR(
batch_size=args.batch_size,
n_components=args.svd_components,
comparison=compare.SimMul())
elif args.regression_type == "TfidfSimYuNNR":
self.lstm = regressionbaselines.TfidfSim2NNR(
batch_size=args.batch_size,
n_components=args.svd_components,
comparison=compare.SimYu())
elif args.regression_type == "TfidfSimEucNNR":
self.lstm = regressionbaselines.TfidfSim2NNR(
batch_size=args.batch_size,
n_components=args.svd_components,
comparison=compare.SimEuc())