📊 Excel Semantic Search

A Streamlit-based web app that lets you search inside Excel or CSV files semantically — not just by exact keywords. It uses SentenceTransformer embeddings to understand the meaning of your query and find similar rows, even if the words are different.

---

🧠 What this app does

This tool helps you explore tabular data in a more human-like way. Instead of typing exact terms, you can describe what you’re looking for in natural language, and the app will retrieve conceptually related entries.

For example:

Searching for “solar energy materials” might return rows describing “photovoltaic systems for desert environments” or “nanostructured catalysts for hydrogen production.”

---

🚀 Key Features

• Upload .xlsx or .csv files directly.
• Choose the column to index and search semantically.
• Add contextual columns to show in the results.
• Adjust similarity threshold and number of results.
• Uses multilingual model paraphrase-multilingual-MiniLM-L12-v2.
• Download the results as a CSV file.
• Works fully offline once the model is downloaded.

---

⚙️ Installation (recommended with Conda)

You can easily run this project on Python 3.11.

1. Clone the repository

git clone https://github.com/<your-username>/excel-semantic-search.git
cd excel-semantic-search

2. Create and activate a conda environment

conda create -n excelsearch python=3.11
conda activate excelsearch

---

🧩 Dependencies

You can install the dependencies using pip or conda.

📦 Option 1 — requirements.txt

Install using pip:

pip install -r requirements.txt

requirements.txt:

streamlit>=1.37.0
pandas>=2.2.0
numpy>=1.26.0
sentence-transformers>=2.7.0
scikit-learn>=1.5.0
openpyxl>=3.1.2

Explanation:

• streamlit → runs the web interface
• pandas, numpy → handle tabular data
• sentence-transformers → generate semantic embeddings
• scikit-learn → compute cosine similarity
• openpyxl → required to read .xlsx files

---

🧬 Option 2 — environment.yml

You can also create the full conda environment in one command:

name: excelsearch
channels:
  - conda-forge
dependencies:
  - python=3.11
  - pip
  - pip:
      - streamlit>=1.37.0
      - pandas>=2.2.0
      - numpy>=1.26.0
      - sentence-transformers>=2.7.0
      - scikit-learn>=1.5.0
      - openpyxl>=3.1.2

Create and activate it:

conda env create -f environment.yml
conda activate excelsearch

---

▶️ How to Run

Once the environment is ready, launch the Streamlit app:

streamlit run app.py

Then open the link that appears in your terminal (usually http://localhost:8501).

---

🧩 How It Works

1. The app loads your Excel/CSV file into a DataFrame.
2. You select which column contains the text you want to search.
3. Each entry in that column is converted into an embedding vector using a pretrained SentenceTransformer model.
4. When you type a query, it is embedded as well.
5. The app computes the cosine similarity between your query and all rows.
6. Results are sorted by similarity and filtered by a threshold you can adjust interactively.

---

💡 Example Use Cases

• Searching project descriptions or grant proposals.
• Exploring survey responses semantically.
• Finding related abstracts, titles, or keywords.
• Comparing textual fields across large datasets.

---

🧰 Tech Stack

• Streamlit: interactive UI
• SentenceTransformers: text embeddings
• scikit-learn: cosine similarity
• Pandas: data handling
• Python 3.11 (recommended)

---

⚡ Quick Start

For the fastest setup:

git clone https://github.com/<your-username>/excel-semantic-search.git
cd excel-semantic-search
conda env create -f environment.yml
conda activate excelsearch
streamlit run app.py

---

📊 Example Data

An example dataset (example_data.csv) is included in this repository for testing and demonstration purposes. All data — including researcher names and project descriptions — are fictional and were synthetically generated to illustrate semantic search behavior.

🧪 The only real researcher included is Antonio Alvarez, whose authentic work can be reviewed here: 🔗 GitHub – antadlp/rms-aot-methanol

This dataset is safe to use for teaching, experimentation, and public demos.

---

📄 License

This project is released under the MIT License — free to use, modify, and share with attribution.

---

Would you like me to integrate this paragraph directly into the full README (so you have the final consolidated version in one file, ready to upload)?

---

📦 Future Improvements

• Add support for FAISS-based search for faster performance on large datasets.
• Enable multi-column semantic fusion (search across several text fields).
• Add GPU detection for automatic acceleration.

---

📄 License

This project is released under the MIT License — free to use, modify, and share with attribution.

---