App updated

README.md

@@ -1,8 +1,8 @@
 ---
-title: News
-emoji: 📉
-colorFrom: blue
-colorTo: indigo
+title: Tweet Evaluator
+emoji: 🐦
+colorFrom: DeepSkyBlue
+colorTo: LightSkyBlue
 sdk: streamlit
 app_file: app.py
 pinned: false

app.py

@@ -1,4 +1,30 @@
 import streamlit as st
 
-x = st.slider("Select a value")
-st.write(x, "squared is", x * x)
+from tweet_pipeline import TweetPipeline
+
+EMOTION_EMOJIS = {"Anger": "😡", "Joy": "😂", "Optimism": "😉", "Sadness": "😢"}
+OFFENSIVE_EMOJIS = {"Offensive": "😈", "Non-Offensive": "😇"}
+SENTIMENT_EMOJIS = {"Negative": "❌", "Neutral": "🤷‍♂️", "Positive": "✅"}
+
+tweet_eval = TweetPipeline()
+
+st.title("🐦 Tweet Evaluator")
+input_text = st.text_input("")
+
+button = st.button("Evaluate!")
+
+if button and input_text != "":
+    with st.spinner("Evaluating tweet..."):
+        prediction = tweet_eval(input_text)
+    st.success("Tweet successfully evaluated!")
+    st.markdown(
+        f"{EMOTION_EMOJIS[prediction['emotion']]} **Emotion:** {prediction['emotion']}"
+    )
+    st.markdown(
+        f"{OFFENSIVE_EMOJIS[prediction['offensive']]} **Offensive:** {'Yes' if prediction['offensive'] == 'Offensive' else 'No'}"
+    )
+    st.markdown(
+        f"{SENTIMENT_EMOJIS[prediction['sentiment']]} **Sentiment:** {prediction['sentiment']}"
+    )
+elif button and not input_text:
+    st.warning("Please, introduce a tweet to eval.")

tweet_pipeline.py

@@ -0,0 +1,37 @@
+import os
+from typing import Dict
+
+from transformers import (
+    AutoModelForSequenceClassification,
+    AutoTokenizer,
+    TextClassificationPipeline,
+)
+
+
+class TweetPipeline:
+    def __init__(self):
+        self.emotion_pipeline = TextClassificationPipeline(
+            model=AutoModelForSequenceClassification.from_pretrained(
+                "elozano/tweet_emotion_eval"
+            ),
+            tokenizer=AutoTokenizer.from_pretrained("elozano/tweet_emotion_eval"),
+        )
+        self.offensive_pipeline = TextClassificationPipeline(
+            model=AutoModelForSequenceClassification.from_pretrained(
+                "elozano/tweet_offensive_eval"
+            ),
+            tokenizer=AutoTokenizer.from_pretrained("elozano/tweet_offensive_eval"),
+        )
+        self.sentiment_pipeline = TextClassificationPipeline(
+            model=AutoModelForSequenceClassification.from_pretrained(
+                "elozano/tweet_sentiment_eval"
+            ),
+            tokenizer=AutoTokenizer.from_pretrained("elozano/tweet_sentiment_eval"),
+        )
+
+    def __call__(self, text: str) -> Dict[str, str]:
+        return {
+            "emotion": self.emotion_pipeline(text)[0]["label"],
+            "offensive": self.offensive_pipeline(text)[0]["label"],
+            "sentiment": self.sentiment_pipeline(text)[0]["label"],
+        }