def spacy_sent_benchmark(datasets):
nlpS = load_spacy_model(textcat='sentiment', vectorError=True)
for dataset in datasets:
if dataset == 'euparlsent':
data = EuroparlSentiment1()
if dataset == 'lccsent':
data = LccSentiment()
df = data.load_with_pandas()
df['valence'] = df['valence'].map(to_label)
# predict with spacy sentiment
def predict(x):
doc = nlpS(x)
pred = max(doc.cats.items(), key=operator.itemgetter(1))[0]
#mathc the labels
labels = {
'positiv': 'positive',
'neutral': 'neutral',
'negativ': 'negative'
}
return labels[pred]
df['pred'] = df.text.map(lambda x: predict(x))
report(df['valence'], df['pred'], 'Spacy sentiment (polarity)',
dataset)
def spacy_sent_benchmark(datasets):
nlpS = load_spacy_model(textcat='sentiment', vectorError=True)
for dataset in datasets:
if dataset == 'euparlsent':
data = EuroparlSentiment1()
if dataset == 'lccsent':
data = LccSentiment()
df = data.load_with_pandas()
df['valence'] = df['valence'].map(sentiment_score_to_label)
# predict with spacy sentiment
def predict(x):
doc = nlpS(x)
pred = max(doc.cats.items(), key=operator.itemgetter(1))[0]
#match the labels
labels = {'positiv': 'positive', 'neutral': 'neutral', 'negativ': 'negative'}
return labels[pred]
spellings_map = {'subjective': 'subjektivt', 'objective': 'objektivt', 'positive': 'positiv', 'negative': 'negativ', 'neutral': 'neutral'}
start = time.time()
df['pred'] = df.text.map(lambda x: spellings_map[predict(x)])
print_speed_performance(start, len(df))
f1_report(df['valence'], df['pred'], 'Spacy sentiment (polarity)', dataset)
def afinn_benchmark(datasets):
afinn = Afinn(language='da', emoticons=True)
for dataset in datasets:
if dataset == 'euparlsent':
data = EuroparlSentiment1()
if dataset == 'lccsent':
data = LccSentiment()
df = data.load_with_pandas()
df['pred'] = df.text.map(afinn.score).map(to_label)
df['valence'] = df['valence'].map(to_label)
report(df['valence'], df['pred'], 'Afinn', dataset)
def sentida_benchmark(datasets):
"The scripts download from github from sentindaV2 and place it in cache folder"
DEFAULT_CACHE_DIR = os.path.join(str(Path.home()), '.danlp')
print(os.getcwd())
workdir = DEFAULT_CACHE_DIR + '/sentida'
print(workdir)
if not os.path.isdir(workdir):
os.mkdir(workdir)
url = "https://raw.githubusercontent.com/esbenkc/emma/master/SentidaV2/"
for file in ['SentidaV2.py', 'aarup.csv', 'intensifier.csv']:
urllib.request.urlretrieve(url + file, workdir + '/' + file)
sys.path.insert(1, workdir)
os.chdir(workdir + '/')
sys.stdout = open(os.devnull, 'w')
from SentidaV2 import sentidaV2
sys.stdout = sys.__stdout__
def sentida_score(sent):
return sentidaV2(sent, output='total')
for dataset in datasets:
if dataset == 'euparlsent':
data = EuroparlSentiment()
if dataset == 'lccsent':
data = LccSentiment()
df = data.load_with_pandas()
df['pred'] = df.text.map(sentida_score).map(to_label_sentida)
df['valence'] = df['valence'].map(to_label)
report(df['valence'], df['pred'], 'SentidaV2', dataset)
def bert_sent_benchmark(datasets):
model = load_bert_tone_model()
for dataset in datasets:
if dataset == 'euparlsent':
data = EuroparlSentiment1()
if dataset == 'lccsent':
data = LccSentiment()
df = data.load_with_pandas()
df['valence'] = df['valence'].map(to_label)
# predict with bert sentiment
df['pred'] = df.text.map(
lambda x: model.predict(x, analytic=False)['polarity'])
report(df['valence'], df['pred'], 'BERT_Tone (polarity)', dataset)
def afinn_benchmark(datasets):
afinn = Afinn(language='da', emoticons=True)
for dataset in datasets:
if dataset == 'euparlsent':
data = EuroparlSentiment1()
if dataset == 'lccsent':
data = LccSentiment()
df = data.load_with_pandas()
start = time.time()
df['pred'] = df.text.map(afinn.score).map(sentiment_score_to_label)
print_speed_performance(start, len(df))
df['valence'] = df['valence'].map(sentiment_score_to_label)
f1_report(df['valence'], df['pred'], 'Afinn', dataset)
def sentida_benchmark(datasets):
from sentida import Sentida
sentida = Sentida()
def sentida_score(sent):
return sentida.sentida(sent, output='total')
for dataset in datasets:
if dataset == 'euparlsent':
data = EuroparlSentiment1()
if dataset == 'lccsent':
data = LccSentiment()
df = data.load_with_pandas()
df['pred'] = df.text.map(sentida_score).map(to_label_sentida)
df['valence'] = df['valence'].map(to_label)
report(df['valence'], df['pred'], 'Sentida', dataset)
def bert_sent_benchmark(datasets):
model = load_bert_tone_model()
for dataset in datasets:
if dataset == 'euparlsent':
data = EuroparlSentiment1()
if dataset == 'lccsent':
data = LccSentiment()
df = data.load_with_pandas()
df['valence'] = df['valence'].map(sentiment_score_to_label)
# predict with bert sentiment
start = time.time()
df['pred'] = df.text.map(lambda x: model.predict(x, analytic=False)['polarity'])
print_speed_performance(start, len(df))
spellings_map = {'subjective': 'subjektivt', 'objective': 'objektivt', 'positive': 'positiv', 'negative': 'negativ', 'neutral': 'neutral'}
df['pred'] = df['pred'].map(lambda x: spellings_map[x])
f1_report(df['valence'], df['pred'], 'BERT_Tone (polarity)', dataset)
def test_lccsentiment(self):
sent = LccSentiment()
df = sent.load_with_pandas()
self.assertEqual(len(df), 499)