EmotionDetectionLSTM: Text-Based Emotion Recognition with LSTM

EmotionDetectionLSTM is an NLP model designed to classify emotions from text using a Long Short-Term Memory (LSTM) deep learning architecture.
It analyzes the linguistic and semantic patterns of input text to predict one of six emotions:
joy, sadness, anger, fear, love, or surprise.

---

⚠️ Disclaimer:
This model is created for educational, research, and demonstration purposes only.
Please ensure responsible and ethical use of AI systems.

---

🧩 Model Details

Key Features:

• Includes preprocessing: tokenization, stopword removal, and lemmatization
• Uses LSTM layers for sequential text understanding
• Outputs emotion predictions with probabilities and expressive emojis
• Deployed on Hugging Face Spaces with a Streamlit-based interface

Skills & Technologies Used:

• TensorFlow / Keras for deep learning
• NLTK for text preprocessing
• Streamlit for deployment
• Matplotlib and Seaborn for data visualization

---

• Developed by: Rawan Alwadeya
• Model Type: Text Classification (LSTM)
• Language: English
• License: MIT

---

💡 Uses

This model can be applied in:

• Emotion and sentiment analysis in text data
• Chatbots and virtual assistants for emotion-aware responses
• Human-computer interaction and affective computing research
• Customer feedback or social media emotion analysis

---

📊 Performance

The LSTM model achieved strong accuracy on the test set:

• Accuracy: 92%

This demonstrates reliable emotion recognition performance across multiple emotion categories.

---

🚀 Deployment

• Hugging Face Repo: EmotionDetectionLSTM
• Demo Video: Watch on LinkedIn

Users can input text and instantly receive a predicted emotion label accompanied by an expressive emoji, making the interaction both informative and user-friendly.

---

👩‍💻 Author

Rawan Alwadeya
AI Engineer | Generative AI & Deep Learning | Data Scientist

• 📧 Email: [email protected]
• 🌐 LinkedIn Profile

---

🧠 Example Usage

from huggingface_hub import hf_hub_download
from tensorflow.keras.models import load_model
from tensorflow.keras.preprocessing.sequence import pad_sequences
import pickle
import numpy as np

# Download model and tokenizer files from Hugging Face Hub
model_path = hf_hub_download(repo_id="RawanAlwadeya/EmotionDetectionLSTM", filename="lstm.h5")
tokenizer_path = hf_hub_download(repo_id="RawanAlwadeya/EmotionDetectionLSTM", filename="tokenizer.pkl")
label_path = hf_hub_download(repo_id="RawanAlwadeya/EmotionDetectionLSTM", filename="label_encoder.pkl")

# Load model and preprocessing objects
model = load_model(model_path)
with open(tokenizer_path, "rb") as f:
    tokenizer = pickle.load(f)
with open(label_path, "rb") as f:
    label_encoder = pickle.load(f)

# Example text
text = "Being a mom has made me feel more vulnerable than I have ever felt before."

# Preprocess
seq = tokenizer.texts_to_sequences([text.lower()])
padded = pad_sequences(seq, maxlen=35, padding='post')

# Predict
pred = model.predict(padded)
emotion = label_encoder.inverse_transform([np.argmax(pred)])[0]
print("Predicted emotion:", emotion)