COMI-LINGUA-NER

About Model

Hindi-English NER Tagger

A LoRA-adapted Transformer LLM fine-tuned for token-level Named Entity Recognition (NER) on Hindi–English (Hinglish) code-mixed text.

Supported Entity Tags

• PERSON - Names of individuals
• ORGANISATION - Institutions or companies
• LOCATION - Non-political physical locations
• DATE - Temporal expressions (dates)
• TIME - Temporal expressions (times)
• GPE - Geo-Political Entities
• HASHTAG - Words prefixed by ‘#’
• EMOJI - Emoticons conveying emotions
• MENTION - User mentions prefixed by ‘@’
• X / Other - Non-entity tokens (common words, punctuation, etc.)

Model Overview

• Model type: LoRA-adapted Transformer LLM
• Base model: CohereForAI/aya-expanse-8b
• Total parameters: 8B
• Trainable parameters: ~32M
• License: Apache 2.0
• Languages: Hindi, English (code-mixed, Roman + Devanagari scripts)

Performance

Achieves 94.90 F1 on the COMI-LINGUA NER test set (5K instances), establishing strong state-of-the-art performance for Hinglish NER among open-weight models, significantly outperforming zero-shot baselines (59.88 F1) and demonstrating the value of fine-tuning for entity boundary detection in mixed-script, code-mixed social media/news text.

Example Inference

Identify named entities in the sentence:

लंदन के Madame Tussauds में Deepika Padukone के wax statue का गुरुवार को अनावरण हुआ।

Output:

[
  {"लंदन": "GPE"},
  {"के": "X"},
  {"Madame": "ORGANISATION"},
  {"Tussauds": "ORGANISATION"},
  {"में": "X"},
  {"Deepika": "PERSON"},
  {"Padukone": "PERSON"},
  {"के": "X"},
  {"wax": "X"},
  {"statue": "X"},
  {"का": "X"},
  {"गुरुवार": "DATE"},
  {"को": "X"},
  {"अनावरण": "X"},
  {"हुआ।": "X"}
]

Citation

@inproceedings{sheth-etal-2025-comi,
  title = "{COMI}-{LINGUA}: Expert Annotated Large-Scale Dataset for Multitask {NLP} in {H}indi-{E}nglish Code-Mixing",
  author = "Sheth, Rajvee and
               Beniwal, Himanshu and
               Singh, Mayank",
  editor = "Christodoulopoulos, Christos and
               Chakraborty, Tanmoy and
               Rose, Carolyn and
               Peng, Violet",
  booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2025",
  month = nov,
  year = "2025",
  address = "Suzhou, China",
  publisher = "Association for Computational Linguistics",
  url = "https://aclanthology.org/2025.findings-emnlp.422/",
  pages = "7973--7992",
  isbn = "979-8-89176-335-7"
}