Hello William sir and Helga Maam, Its wonderful to know about your interest in the Konkani LLM paper.
I can connect you to all the Computer Science Professors including Prof. Gaurang Patkar working on Konkani currently and I am sure they will also gain some more insights from your discussions. Happy to connect! Prof. Sangeeta Chakrabarty, HoD, Dept. of I.T, S. S. Dempo College (Autonomous), Integrated Educational Complex, Cujira, (opp GMC, Bambolim) Ilhas, Goa 403 202 *KNOW about me at:* art2smart.blogspot.com *My Papers at:* https://scholar.google.co.in/citations?user=E7lsYOYAAAAJ&hl=en On Mon, 13 Apr, 2026, 06:07 William Robert Da Silva, <[email protected]> wrote: > Helga, I have been with oral Konkanni of different caste-occupational > groups in Goa state and outside, less with Kerala and Navayat (I had > contact with them in the 1970s). Written Konkanni moves away in the > morpho-phonological, syntactical and verbal framework of most of these > caste-occupational Konkanni. It is use-based in occupation and living and > the vocabulary and linguistic issues are of another kind. Put together, > they give a dynamic unity to Konkanni because the occupational groups are > mutually communicative and highly sharing. My language 'bandavoll' and > 'utram-daiz' are original to Konkanni. It does not celebrate 'Goy amchem > mullpitt, Konkanni amchi bhas' because in history, Goa was a city, a > capital city of different kingdoms ruling in coastal Konkann. (Others would > call it Aparaant, in opposition to Puurvaant?). Konkanni people looked at > it differently, beginning with the first arrivals from Africa some 40000 > years ago, who moved down, eastward, north and into south-east Asia. The > Mittgauddo, Kharvi, Gabit, Velip, etc. are good examples of language > dynamism. > I have entertained and worked this out all these decades for Konkanni and > other Indian languages. > Some thought, stray and away LLMs. > W R Da Silva > > On Fri, Apr 10, 2026 at 11:15 PM Helga Do Rosario Gomes < > [email protected]> wrote: > >> Hi William, >> I too would have liked to learn more about this but here’s a simple >> explanation that I gleaned from my limited knowledge and chatGTP. >> As you are aware the many AI Large language models (LLMs) that underly >> chatGTP, Gemini, Claude, DeepSeek etc are trained by ‘supplying’ them with >> huge amounts of training data like magazines, journals and who know what! >> This is a highly contentious issue with creators which is why NYT has sued >> them but that is another story. >> But these LLMs struggle with languages like Konkani because there isn’t >> enough training data and the language is written in multiple scripts (like >> Devanagari, Romi, and Kannada), which makes learning (for these models) >> harder. To fix this, the authors created a large synthetic dataset >> (Konkani-Instruct-100k) and used it to fine-tune existing models such as >> Llama 3.1 (meta) Qwen2.5 (Ali baba) and Gemma 3 (Google). They also built >> benchmarks to evaluate performance across scripts. As a result, their >> improved models perform better at tasks like translation, sometimes even >> outperforming propriety baselines. >> As Joǎo said if you want to know how they do it then you need expertise >> in computer sciences and math but as a highly regarded Konkani researcher I >> think you would be a perfect user to test their fine tuned model. >> Hope this helps. >> Best, >> Helga >> Sent from my iPhone >> >> On Apr 10, 2026, at 06:22, William Robert Da Silva <[email protected]> >> wrote: >> >> >> Send some clarification on the entire project. >> WRDS >> >> On Mon, Apr 6, 2026 at 12:49 AM Frederick Noronha < >> [email protected]> wrote: >> >>> >>> Konkani LLM: Multi-Script Instruction Tuning and Evaluation for a >>> Low-Resource Indian Language >>> Reuben Chagas Fernandes >>> <https://arxiv.org/search/cs?searchtype=author&query=Fernandes,+R+C>, >>> Gaurang >>> S. Patkar >>> <https://arxiv.org/search/cs?searchtype=author&query=Patkar,+G+S> >>> >>> Large Language Models (LLMs) consistently under perform in low-resource >>> linguistic contexts such as Konkani. This performance deficit stems from >>> acute training data scarcity compounded by high script diversity across >>> Devanagari, Romi and Kannada orthographies. To address this gap, we >>> introduce Konkani-Instruct-100k, a comprehensive synthetic >>> instruction-tuning dataset generated through Gemini 3. >>> We establish rigorous baseline benchmarks by evaluating leading >>> open-weights architectures including Llama 3.1, Qwen2.5 and Gemma 3 >>> alongside proprietary closed-source models. Our primary contribution >>> involves the development of Konkani LLM, a series of fine-tuned models >>> optimized for regional nuances. Furthermore, we are developing the >>> Multi-Script Konkani Benchmark to facilitate cross-script linguistic >>> evaluation. In machine translation, Konkani LLM delivers consistent gains >>> over the corresponding base models and is competitive with and in several >>> settings surpasses proprietary baselines >>> https://arxiv.org/abs/2603.23529 >>> >>> >>> >>> _/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/ >>> _/ Frederick Noronha फ्रेडरिक नोरोन्या * فريدريك نورونيا >>> _/ AUDIO https://archive.org/details/@fredericknoronha >>> _/ http://goa1556.in +91-9822122436 784 Saligao Goa >>> _/ Goanet :: 30 years of discussions. [email protected] >>> _/ http://lists.goanet.org/pipermail/goanet-goanet.org/ >>> _/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/ >>> >>> -- >>> You received this message because you are subscribed to the Google >>> Groups "Goa-Research-Net" group. >>> To unsubscribe from this group and stop receiving emails from it, send >>> an email to [email protected]. >>> To view this discussion, visit >>> https://groups.google.com/d/msgid/goa-research-net/CA%2Bmqab8n87oQUuD51v0UMO5e2-o7Kb-cehxiZz7q7JePCndJVA%40mail.gmail.com >>> <https://groups.google.com/d/msgid/goa-research-net/CA%2Bmqab8n87oQUuD51v0UMO5e2-o7Kb-cehxiZz7q7JePCndJVA%40mail.gmail.com?utm_medium=email&utm_source=footer> >>> . >>> >> -- >> You received this message because you are subscribed to the Google Groups >> "Goa-Research-Net" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected]. >> To view this discussion, visit >> https://groups.google.com/d/msgid/goa-research-net/CA%2BvNr4%2BU_ms-95dM9C4UNcMj-6tE4-nAyW8aPtGnC_9_5dJmFQ%40mail.gmail.com >> <https://groups.google.com/d/msgid/goa-research-net/CA%2BvNr4%2BU_ms-95dM9C4UNcMj-6tE4-nAyW8aPtGnC_9_5dJmFQ%40mail.gmail.com?utm_medium=email&utm_source=footer> >> . >> >> -- >> You received this message because you are subscribed to the Google Groups >> "Goa-Research-Net" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected]. >> To view this discussion, visit >> https://groups.google.com/d/msgid/goa-research-net/9F54C4CB-0CBC-4A67-86FC-CBC56A86883A%40gmail.com >> <https://groups.google.com/d/msgid/goa-research-net/9F54C4CB-0CBC-4A67-86FC-CBC56A86883A%40gmail.com?utm_medium=email&utm_source=footer> >> . >> > -- > You received this message because you are subscribed to the Google Groups > "Goa-Research-Net" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To view this discussion, visit > https://groups.google.com/d/msgid/goa-research-net/CA%2BvNr4%2B0XwWwVK_ofGwZVXM94iU1%3D3p4dDPvOrDfCU_RJdBwvQ%40mail.gmail.com > <https://groups.google.com/d/msgid/goa-research-net/CA%2BvNr4%2B0XwWwVK_ofGwZVXM94iU1%3D3p4dDPvOrDfCU_RJdBwvQ%40mail.gmail.com?utm_medium=email&utm_source=footer> > . > -- You received this message because you are subscribed to the Google Groups "Goa-Research-Net" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion, visit https://groups.google.com/d/msgid/goa-research-net/CACfQpdQcNmUqsirA%2BHRigam_OFR2aPKfy-LfABa%3DuhNEws3wyg%40mail.gmail.com.
