Papers › Analysis Of Contextual and Non-Contextual Word Embedding Models For Hindi NER With Web...
Analysis Of Contextual and Non-Contextual Word Embedding Models For Hindi NER With Web Application For Data Collection
Aindriya Barua, Thara.S, Premjith B, Soman KP‡
Named Entity Recognition (NER) is the process of taking a string and identifying relevant proper nouns in it. In this paper ‡ we report the development of the Hindi NER system, in Devanagari script, using various embedding models. We categorize embeddings as Contextual and Non-contextual, and further compare them inter and intra-category. Under non-contextual type embeddings, we experiment with Word2Vec and FastText, and under the contextual embedding category, we experiment with BERT and its variants, viz. RoBERTa, ELECTRA, CamemBERT, Distil-BERT, XLM-RoBERTa. For non-contextual embeddings, we use five machine learning algorithms namely Gaussian NB, Adaboost Classifier, Multi-layer Perceptron classifier, Random Forest Classifier, and Decision Tree Classifier for developing ten Hindi NER systems, each, once with Fast Text and once with Gensim Word2Vec word embedding models. These models are then compared with Transformers based contextual NER models, using BERT and its variants. A comparative study among all these NER models is made. Finally, the best of all these models is used and a web app is built, that takes a Hindi text of any length and returns NER tags for each word and takes feedback from the user about the correctness of tags. These feed-backs aid our further data collection.
Code
Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.
Code Syntology ran Syntology
Not run by Syntology. Nothing on this page verifies that the listed code works.
Tasks
Results from the paper archive 2025-07-28
| Task | Dataset | Model | Metric | Value | Rank at snapshot | Leaderboard | Report |
|---|---|---|---|---|---|---|---|
| Named Entity Recognition (NER) | IECSIL FIRE-2018 Shared Task | XLM-RoBERTa | Average F1 | 90.8419 | #1 of 1 | Archive leaderboard | report |
Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.
Methods
Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections