Announcement Solved

Forum search has been improved.try it out and send us feedback (19)

Back to Forum
Distinguished 720 pts 0 followers
Indian Institute of Science · Posted
Sociology & Anthropology Photonics & Optics

Welcome to the Academic Connect Research Community Forum

This forum is a space for researchers, faculty, and students across Indian academic institutions to connect, share knowledge, ask questions, and support each other's academic journeys.

Community Guidelines

  1. Be respectful. We are all colleagues. Disagree with ideas, not people.
  2. Be specific. Vague posts get vague answers. Provide context.
  3. Search before posting. Your question may already have an answer.
  4. Use appropriate tags. Tag your posts with relevant research areas to help others find them.
  5. No spam or self-promotion. Relevant links and resources are welcome; commercial spam is not.
  6. Cite your sources. When making factual claims, link to the evidence.

Moderation

Our moderators are senior faculty members who volunteer their time. If you see content that violates these guidelines, please report it using the report button.

Welcome.let's build something great together.

Sign in to interact

Sign in to join the discussion.

4 Replies

Oldest Top
0
test tester Growing 165 pts · Accepted answer

[Moderator] Pinning this thread as it contains highly useful information for new PhD students.

0
Replying to test tester
Vikram Bhatia Active 420 pts · Accepted answer

Open access is the right direction but the APCs are prohibitively expensive for many Indian researchers without institutional funding.

0
Replying to test tester
test tester Growing 165 pts · Accepted answer

Open access is the right direction but the APCs are prohibitively expensive for many Indian researchers without institutional funding.

0
Neha Agarwal Growing 90 pts · Accepted answer

Great question.I went through something very similar in my second year.

The key insight for me was that LoRA is actually quite well-suited for NER tasks, especially in low-resource settings. I would recommend:

  1. Use LoRA with r=8 or r=16.don't go higher for 8K samples
  2. Apply LoRA to attention layers only, not the feed-forward layers
  3. Use a cosine learning rate schedule with warm-up (10% of steps)

For Telugu-English code-mixed NER specifically, you might also look at MuRIL.it is pretrained on Indian language data and often outperforms XLM-R on Indic tasks even with less fine-tuning data.