🤗 NLP Toolkit

A multi-feature NLP application built with Streamlit and powered by modern Transformer models. Supports both English and Portuguese (PT-BR).

---

Features

Feature	Description	Models / Libraries
📝 Text Summarization	Abstractive summarization via Hugging Face Inference API	BART-large-CNN (EN), PTT5-base-summ-xlsum (PT-BR)
🏷️ Named Entity Recognition	Entity extraction with displaCy visualization and CSV export	spaCy en_core_web_sm (18 entity types), pt_core_news_sm (4 types)
✂️ Text Chunking	Fixed-size and semantic chunking with side-by-side comparison	LangChain RecursiveCharacterTextSplitter, HF embeddings
🔗 Semantic Similarity	Cosine similarity between texts with language-aware model selection	BERTimbau (PT-BR), MiniLM / MPNet (multilingual)

Architecture

app.py                  # Entry point — st.navigation multi-page routing
pages/
  summarization.py      # HF Inference API summarization
  ner.py                # spaCy NER + displaCy rendering
  chunking.py           # Fixed-size & semantic chunking
  similarity.py         # Sentence-transformer similarity
utils/
  config.py             # Centralized models, colors, thresholds, example texts
  hf_client.py          # Shared InferenceClient with retry logic
  spacy_models.py       # Cached spaCy model loading

Tech Stack

• Streamlit — Multi-page UI with st.navigation
• Hugging Face Inference API — Serverless model inference (free tier)
• spaCy 3.7 — NER pipelines + displaCy visualization
• LangChain — Text splitters for chunking strategies
• NumPy / Pandas — Similarity metrics and data export

Getting Started

Prerequisites

• Python 3.11+
• A Hugging Face API token (free)

Installation

git clone https://github.com/marinaramalhete/NLP-Text-Summary-App.git
cd NLP-Text-Summary-App

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Configuration

Create a .streamlit/secrets.toml file with your HF token:

HF_TOKEN = "hf_..."

Running

streamlit run app.py

Deployment

This app is designed for Streamlit Community Cloud. To deploy your own instance:

1. Push the repo to GitHub
2. Go to share.streamlit.io and connect your repo
3. Add HF_TOKEN in the app's Secrets settings
4. Deploy

Author

Marina Ramalhete Masid — GitHub · LinkedIn