Analyzes tweet content sentiment using named model, providing sentiment label and score

hf_twitter_sentiment_zapier_code.py

import requests
import time
# Add the model ID and your Hugging Face Token
HF_TOKEN = "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"
# MODEL = "cardiffnlp/twitter-roberta-base-sentiment-latest"
SENTIMENT_MODEL = "cardiffnlp/twitter-xlm-roberta-base-sentiment"

# Create the API request
API_BASE = "https://api-inference.huggingface.co/models/"
SENTIMENT_API_URL = API_BASE + SENTIMENT_MODEL
headers = {"Authorization": "Bearer %s" % (HF_TOKEN)}

# Preprocess text (username and link placeholders)
def preprocess(text):
    new_text = []
    for t in text.split(" "):
        t = '@user' if t.startswith('@') and len(t) > 1 else t
        t = 'http' if t.startswith('http') else t
        new_text.append(t)
    return " ".join(new_text)

# Returns sentiment analysis on the given payload
def analysis(payload):
  response = requests.post(SENTIMENT_API_URL, headers=headers, json=payload)
  while "is currently loading" in str(response.json()): # retry request while model loads
    print(response.json())
    time.sleep(30)
    response = requests.post(SENTIMENT_API_URL, headers=headers, json=payload)
  return response.json()

# Do the request to the Inference API and process the resutls
input = preprocess(str(input_data['input_data']))
result = analysis({"inputs": input})
prelim_scores = result[0]
# Get the scores
neutral = prelim_scores[0]['score']
positive = prelim_scores[1]['score']
negative = prelim_scores[2]['score']

scores = [{'label': 'negative', 'score': negative}, {'label': 'neutral', 'score': neutral}, {'label': 'positive', 'score': positive}]
top_score = max(item['score'] for item in scores)
top_sentiment = next(item['label'] for item in scores if item["score"] == top_score)

# Define the output for Zapier
output = [{'sentiment_label': top_sentiment, 'sentiment_score': top_score}]