Tool · Digital Humanities & Social Science

Semantic Term Extractor
& Lexicon Explorer

Seed-guided vocabulary expansion, dense semantic matching, and 2D cluster manifold visualization in Arabic & English. Insert target seed terms (e.g. political concepts, jargon, or slang) to extract and rank all conceptually similar expressions from your text, accompanied by real corpus sentence context and interactive cluster maps.

🌱 Seed term expansion 🌍 Arabic & English 📄 PDF & CSV upload 📊 Frequency & context snippets 🗺️ 2D UMAP/PCA manifold 🏷️ K-Means clustering 📥 CSV & JSON export
Step 01
📤
Upload your corpus
Paste plain text, upload a multi-page PDF, or load a CSV with your document column. Arabic and English supported.
Step 02
🌱
Define seed concepts
Provide target seed terms (e.g. corruption, protests, medical terms, or slang). The tool extracts candidate words and phrases.
Step 03
⚙️
Dense semantic matching
Candidate terms and seeds are embedded via multilingual transformers to calculate cosine similarity and nearest seed targets.
Step 04
🗺️
Cluster map & export
Explore an interactive 2D manifold with seed landmarks and candidate clusters. Inspect sentence context and export to CSV.
🤗
paraphrase-multilingual-MiniLM-L12-v2
sentence-transformers · Hosted on Hugging Face Spaces
Languages
50+ incl. Arabic & English
Embedding dim
384
Architecture
MiniLM-L12
Max input
128 tokens / chunk
Similarity metric
Cosine similarity
Dimensionality reduction
UMAP → PCA fallback
Live Interactive Tool
Open Space in New Tab ↗ 🤗 Powered by Hugging Face Spaces

Texts are processed in the sandboxed inference environment and are never stored or logged. Direct Space URL: huggingface.co/spaces/Muhsabrys/semantic-embeddings.