Extract recurring noun phrases from English source strings to build or enhance your bilingual glossary (EN > IT), with real context examples.
This tool is designed to support Localization Quality Assurance (LQA) and Translation Memory (TM) management by programmatically identifying candidate terms for glossary inclusion.
It leverages spaCy for NLP-based noun chunking and pandas for filtering, with progress tracking via tqdm. Output terms include real examples from source and target strings, and can be checked against an existing glossary for overlap.
Large projects often suffer from inconsistent terminology or missing glossary entries. This tool helps:
- Detect high-frequency candidate terms in EN source text.
- Review them with real EN/IT sentence pairs.
- Identify gaps or mismatches in your current glossary.
Ideal for PMs, linguists, and LQA specialists working on games, apps, or any high-volume localization project.
- 🔍 Noun phrase extraction from EN source strings
- 📈 Frequency filtering (customizable)
- 🔤 Support for multi-word expressions (n-grams)
- ✅ Check against existing glossary (EN > IT)
- 🧹 Rejects noisy/invalid terms and logs them
- 💬 Outputs context-rich examples for human validation
- 📦 Export to CSV for review or integration into glossaries
You will need an Excel file with at least these two columns:
| EN | Italian |
|---|---|
| The final boss is here! | Il boss finale è qui! |
Optionally, provide a second glossary file with:
| en Term1 | it Term1 |
|---|---|
| final boss | boss finale |
- Clone this repo
- Install dependencies
pip install -r requirements.txt
-
python cli#2.py
- Follow the prompts to input:
- Your Translation Excel File
- Your Glossary file
- View your results:
- terms_filtered.xlsx -> Cleaned candidates
- rejected_terms.xlsx -> Filtered-out noisy terms
📌 Notes
- Language model: spaCy en_core_web_sm
- Current language pair: EN > IT
- Extendable to other pairs with simple tweaks
📬 Contributors Leoth — Localization QA Specialist @ Moonton