mirror of
https://github.com/gsi-upm/senpy
synced 2024-11-22 00:02:28 +00:00
182 lines
6.7 KiB
Python
182 lines
6.7 KiB
Python
#!/usr/local/bin/python
|
|
# coding: utf-8
|
|
|
|
from future import standard_library
|
|
standard_library.install_aliases()
|
|
|
|
import os
|
|
import re
|
|
import sys
|
|
import string
|
|
import numpy as np
|
|
from six.moves import urllib
|
|
from nltk.corpus import stopwords
|
|
|
|
from senpy import EmotionBox, models
|
|
|
|
|
|
def ignore(dchars):
|
|
deletechars = "".join(dchars)
|
|
tbl = str.maketrans("", "", deletechars)
|
|
ignore = lambda s: s.translate(tbl)
|
|
return ignore
|
|
|
|
|
|
class DepecheMood(EmotionBox):
|
|
'''
|
|
Plugin that uses the DepecheMood emotion lexicon.
|
|
|
|
DepecheMood is an emotion lexicon automatically generated from news articles where users expressed their associated emotions. It contains two languages (English and Italian), as well as three types of word representations (token, lemma and lemma#PoS). For English, the lexicon contains 165k tokens, while the Italian version contains 116k. Unsupervised techniques can be applied to generate simple but effective baselines. To learn more, please visit https://github.com/marcoguerini/DepecheMood and http://www.depechemood.eu/
|
|
'''
|
|
|
|
author = 'Oscar Araque'
|
|
name = 'emotion-depechemood'
|
|
version = '0.1'
|
|
requirements = ['pandas']
|
|
nltk_resources = ["stopwords"]
|
|
|
|
onyx__usesEmotionModel = 'wna:WNAModel'
|
|
|
|
EMOTIONS = ['wna:negative-fear',
|
|
'wna:amusement',
|
|
'wna:anger',
|
|
'wna:annoyance',
|
|
'wna:indifference',
|
|
'wna:joy',
|
|
'wna:awe',
|
|
'wna:sadness']
|
|
|
|
DM_EMOTIONS = ['AFRAID', 'AMUSED', 'ANGRY', 'ANNOYED', 'DONT_CARE', 'HAPPY', 'INSPIRED', 'SAD',]
|
|
|
|
def __init__(self, *args, **kwargs):
|
|
super(DepecheMood, self).__init__(*args, **kwargs)
|
|
self.LEXICON_URL = "https://github.com/marcoguerini/DepecheMood/raw/master/DepecheMood%2B%2B/DepecheMood_english_token_full.tsv"
|
|
self._denoise = ignore(set(string.punctuation)|set('«»'))
|
|
self._stop_words = []
|
|
self._lex_vocab = None
|
|
self._lex = None
|
|
|
|
def activate(self):
|
|
self._lex = self.download_lex()
|
|
self._lex_vocab = set(list(self._lex.keys()))
|
|
self._stop_words = stopwords.words('english') + ['']
|
|
|
|
def clean_str(self, string):
|
|
string = re.sub(r"[^A-Za-z0-9().,!?\'\`]", " ", string)
|
|
string = re.sub(r"[0-9]+", " num ", string)
|
|
string = re.sub(r"\'s", " \'s", string)
|
|
string = re.sub(r"\'ve", " \'ve", string)
|
|
string = re.sub(r"n\'t", " n\'t", string)
|
|
string = re.sub(r"\'re", " \'re", string)
|
|
string = re.sub(r"\'d", " \'d", string)
|
|
string = re.sub(r"\'ll", " \'ll", string)
|
|
string = re.sub(r"\.", " . ", string)
|
|
string = re.sub(r",", " , ", string)
|
|
string = re.sub(r"!", " ! ", string)
|
|
string = re.sub(r"\(", " ( ", string)
|
|
string = re.sub(r"\)", " ) ", string)
|
|
string = re.sub(r"\?", " ? ", string)
|
|
string = re.sub(r"\s{2,}", " ", string)
|
|
return string.strip().lower()
|
|
|
|
def preprocess(self, text):
|
|
if text is None:
|
|
return None
|
|
tokens = self._denoise(self.clean_str(text)).split(' ')
|
|
tokens = [tok for tok in tokens if tok not in self._stop_words]
|
|
return tokens
|
|
|
|
def estimate_emotion(self, tokens, emotion):
|
|
s = []
|
|
for tok in tokens:
|
|
s.append(self._lex[tok][emotion])
|
|
dividend = np.sum(s) if np.sum(s) > 0 else 0
|
|
divisor = len(s) if len(s) > 0 else 1
|
|
S = np.sum(s) / divisor
|
|
return S
|
|
|
|
def estimate_all_emotions(self, tokens):
|
|
S = []
|
|
intersection = set(tokens) & self._lex_vocab
|
|
for emotion in self.DM_EMOTIONS:
|
|
s = self.estimate_emotion(intersection, emotion)
|
|
S.append(s)
|
|
return S
|
|
|
|
def download_lex(self, file_path='DepecheMood_english_token_full.tsv', freq_threshold=10):
|
|
|
|
import pandas as pd
|
|
|
|
try:
|
|
file_path = self.find_file(file_path)
|
|
except IOError:
|
|
file_path = self.path(file_path)
|
|
filename, _ = urllib.request.urlretrieve(self.LEXICON_URL, file_path)
|
|
|
|
lexicon = pd.read_csv(file_path, sep='\t', index_col=0)
|
|
lexicon = lexicon[lexicon['freq'] >= freq_threshold]
|
|
lexicon.drop('freq', axis=1, inplace=True)
|
|
lexicon = lexicon.T.to_dict()
|
|
return lexicon
|
|
|
|
def predict_one(self, features, **kwargs):
|
|
tokens = self.preprocess(features[0])
|
|
estimation = self.estimate_all_emotions(tokens)
|
|
return estimation
|
|
|
|
test_cases = [
|
|
{
|
|
'entry': {
|
|
'nif:isString': 'My cat is very happy',
|
|
},
|
|
'expected': {
|
|
'onyx:hasEmotionSet': [
|
|
{
|
|
'onyx:hasEmotion': [
|
|
{
|
|
'onyx:hasEmotionCategory': 'wna:negative-fear',
|
|
'onyx:hasEmotionIntensity': 0.05278117640010922
|
|
},
|
|
{
|
|
'onyx:hasEmotionCategory': 'wna:amusement',
|
|
'onyx:hasEmotionIntensity': 0.2114806151413433,
|
|
},
|
|
{
|
|
'onyx:hasEmotionCategory': 'wna:anger',
|
|
'onyx:hasEmotionIntensity': 0.05726119426520887
|
|
},
|
|
{
|
|
'onyx:hasEmotionCategory': 'wna:annoyance',
|
|
'onyx:hasEmotionIntensity': 0.12295990731053638,
|
|
},
|
|
{
|
|
'onyx:hasEmotionCategory': 'wna:indifference',
|
|
'onyx:hasEmotionIntensity': 0.1860159893608025,
|
|
},
|
|
{
|
|
'onyx:hasEmotionCategory': 'wna:joy',
|
|
'onyx:hasEmotionIntensity': 0.12904050973724163,
|
|
},
|
|
{
|
|
'onyx:hasEmotionCategory': 'wna:awe',
|
|
'onyx:hasEmotionIntensity': 0.17973650399862967,
|
|
},
|
|
{
|
|
'onyx:hasEmotionCategory': 'wna:sadness',
|
|
'onyx:hasEmotionIntensity': 0.060724103786128455,
|
|
},
|
|
]
|
|
}
|
|
]
|
|
}
|
|
}
|
|
]
|
|
|
|
|
|
if __name__ == '__main__':
|
|
from senpy.utils import easy, easy_load, easy_test
|
|
# sp, app = easy_load()
|
|
# for plug in sp.analysis_plugins:
|
|
# plug.test()
|
|
easy_test(debug=False)
|