ITACA_Insurance_Core_v4 / sentiment_analysis_v2.py
dperales's picture
Upload 2 files
65ffad7
raw
history blame
3.96 kB
from transformers import AutoTokenizer, AutoModelForSequenceClassification
from transformers_interpret import SequenceClassificationExplainer
import torch
import pandas as pd
class SentimentAnalysis:
"""
Sentiment on text data.
Attributes:
tokenizer: An instance of Hugging Face Tokenizer
model: An instance of Hugging Face Model
explainer: An instance of SequenceClassificationExplainer from Transformers interpret
"""
def __init__(self):
# Load Tokenizer & Model
hub_location = 'cardiffnlp/twitter-roberta-base-sentiment'
self.tokenizer = AutoTokenizer.from_pretrained(hub_location)
self.model = AutoModelForSequenceClassification.from_pretrained(hub_location)
hub_location_sp = 'finiteautomata/beto-sentiment-analysis'
self.tokenizer_sp = AutoTokenizer.from_pretrained(hub_location_sp)
self.model_sp = AutoModelForSequenceClassification.from_pretrained(hub_location_sp)
# Change model labels in config
self.model.config.id2label[0] = "Negative"
self.model.config.id2label[1] = "Neutral"
self.model.config.id2label[2] = "Positive"
self.model.config.label2id["Negative"] = self.model.config.label2id.pop("LABEL_0")
self.model.config.label2id["Neutral"] = self.model.config.label2id.pop("LABEL_1")
self.model.config.label2id["Positive"] = self.model.config.label2id.pop("LABEL_2")
# Instantiate explainer
self.explainer = SequenceClassificationExplainer(self.model, self.tokenizer)
self.explainer_sp = SequenceClassificationExplainer(self.model_sp, self.tokenizer_sp)
def justify(self, text, lang):
"""
Get html annotation for displaying sentiment justification over text.
Parameters:
text (str): The user input string to sentiment justification
Returns:
html (hmtl): html object for plotting sentiment prediction justification
"""
if lang == 'es':
word_attributions = self.explainer_sp(text)
html = self.explainer_sp.visualize("example.html")
else:
word_attributions = self.explainer(text)
html = self.explainer.visualize("example.html")
return html
def classify(self, text, lang):
"""
Recognize Sentiment in text.
Parameters:
text (str): The user input string to perform sentiment classification on
Returns:
predictions (str): The predicted probabilities for sentiment classes
"""
if lang == 'es':
tokens = self.tokenizer_sp.encode_plus(text, add_special_tokens=False, return_tensors='pt')
outputs = self.model_sp(**tokens)
probs = torch.nn.functional.softmax(outputs[0], dim=-1)
probs = probs.mean(dim=0).detach().numpy()
predictions = pd.Series(probs, index=["Negative", "Neutral", "Positive"], name='Predicted Probability')
else:
tokens = self.tokenizer.encode_plus(text, add_special_tokens=False, return_tensors='pt')
outputs = self.model(**tokens)
probs = torch.nn.functional.softmax(outputs[0], dim=-1)
probs = probs.mean(dim=0).detach().numpy()
predictions = pd.Series(probs, index=["Negative", "Neutral", "Positive"], name='Predicted Probability')
return predictions
def run(self, text, lang):
"""
Classify and Justify Sentiment in text.
Parameters:
text (str): The user input string to perform sentiment classification on
Returns:
predictions (str): The predicted probabilities for sentiment classes
html (hmtl): html object for plotting sentiment prediction justification
"""
predictions = self.classify(text, lang)
html = self.justify(text, lang)
return predictions, html