---
tags:
- sentence-transformers
- feature-extraction
- embeddings
- mplads
- lok-sabha
- indian-government
- multilingual
license: apache-2.0
language:
- en
- hi
- ta
- te
- bn
- mr
- gu
- kn
- ml
- pa
- ur
- or
- as
- sd
- ne
- si
datasets:
- ZeroiJ/mplads-works
---

# MPLADS Lok Sabha Embedding Model

Fine-tuned [paraphrase-multilingual-MiniLM-L12-v2](https://huggingface.co/sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2) for Lok Sabha MPLADS work descriptions.

## Use with sentence-transformers

```python
from sentence_transformers import SentenceTransformer
model = SentenceTransformer("ZeroiJ/mplads-ls-encoder")
emb = model.encode("construction of pucca road in Gram Panchayat")
```

## Use with the Hugging Face Inference API

```bash
curl https://api-inference.huggingface.co/models/ZeroiJ/mplads-ls-encoder \
  -X POST \
  -H "Authorization: Bearer ${HF_TOKEN}" \
  -H "Content-Type: application/json" \
  -d '{"inputs": "construction of pucca road in Gram Panchayat"}'
```

## Model details

- Base: `sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2`
- Fine-tuning objective: cosine-similarity contrastive loss on curated MPLADS work pairs
- Output: 384-dimensional L2-normalized embeddings
- Validation cosine accuracy: ~96.3%
- Built for the MPLADS (SIH26102) fraud-detection demo.