Had an awesome session with college students on Marathi Transcription! 🎉 We talked about what transcription is and what to avoid while doing it. Loved solving their doubts and sharing some tips! 😊
Excited to shape the AI world for Marathi and other Indian languages! 🌟 Thrilled to see how our efforts are bringing these languages to the forefront of technology.
Let's make AI more inclusive together! #MarathiTranscription#LearningTogether#InclusiveAI
A dynamic professional with nearly 18 years of rich experience in Academic with various schools
▪ Profound knowledge of art and craft and Ability to perceive the creativity in the students and help them to develop it.
Renu sundriyal
As an AI language model, I don't take exams or prepare for them like humans do. However, I can help provide study tips and resources for preparing for the Hindi language paper of the UPSC if you're interested!
The Korean writing test is one of the most important areas in TOPIK (Test of Proficiency in Korean) and KIIP (Korean Social Integration Test).
I am developing it for people studying Korean.
This will be a time to share how much AI can help with Korean writing. 😀
Hello, LinkedIn Community!
On this special occasion of Hindi Day, we're elated to shine a spotlight on a remarkable journey of innovation that spans three decades and transcends borders - the world of Hindi-related patents.
Language Unleashed:
Over the past 30 years, Hindi-related patents have been at the forefront of linguistic evolution. They've empowered communication, opened doors to new technology, and enriched our digital landscape.
A Digital Revolution:
From fonts and keyboards to language processing algorithms, these patents have been the driving force behind the seamless integration of Hindi into the digital age.
Global Impact:
What makes this journey truly special is its global reach. Hindi-related patents have touched lives worldwide, transcending language barriers and fostering cross-cultural connections.
Innovation for Tomorrow:
As we celebrate this rich history, we also look ahead with excitement. The future promises even more groundbreaking inventions that will continue to elevate the Hindi language on the global stage.
#HindiDay#HindiPatents#ThreeDecadesOfInnovation#IPCalculus#IPR
Now this puts powerful tech to great use! #Omni could disrupt the #education sector like nothing else. It could improve quality of teaching, and augment the workloads of overburdened #teachers. It is so #interactive that the experience is near-human like! Sal Khan of #KhanAcademy and his son Imran explain. Btw.. It is available in SIX Indian languages #Gujrati#Telugu#Tamil#Marathi#Hindi#Urdu#GPT-4o (“o” for “omni”) is a step towards much more natural human-computer interaction—it accepts as input any combination of text, audio, and image and generates any combination of text, audio, and image outputs.
The world is changing rapidly, and GenAI's new technologies are revolutionizing the rules and will continue to do so. We have an incredible opportunity to leverage these tools to increase productivity and achieve a better work-life balance. With this in mind, I'm thrilled to share that I've created my first GPT! 🎉
Meet Lexi: a language assistant designed to help non-native English speakers enhance their text with a professional-casual and slightly humorous tone. I was inspired by how my sons learn languages by trying and trying and trying again. 😎
#GenAI#ChatGPT#IA#GPTs
Just Tried Krutrim,
Good points:
1. It can generate Hindi, Tamil, and a few more languages.
2. We can ask questions in Hindi in a given English paragraph.
3. Can translate into some Indian languages. (average quality)
Improvisation needed:
1. It is very slow to generate any response.
2. Generation quality is average, as it keeps repeating the same words if we ask to write a few sentences.
3. Not good for writing longer content, provide false information for basic things.
They need to feed a lot of high quality data and fine tune the model to generate Indian languages. The response speed also needs to be improved for a better experience.
Have you tried Krutrim?
#krutrim#llm#indianlanguages#ai#trainingdata#dataforai
Presenting Sambhashan - an Indic language LLM fine-tuned in six Indian languages viz. Tamil, Telugu, Malayalam, Odia, Kannada, and Hindi.
Sambhashan is Sanskrit for conversation or discourse.
The models(Llama-3 8B, Gemma 7B, and Mistral v0.3 - 7B) were fine-tuned on A100 GPU on tokenized open-source datasets for approx. 8-10 hours each to reach a certain level of rigour.
The model can take instruction-based prompts for tasks such as:
1. Translations,
2. Resource-based QnA,
3. General QnA(If the info was there in the trained text)
Features of the Sambhshan model:
1. Ability to get further trained in other languages:
2. Computationally frugal
3. Efficient quantised models can be created.
A few of the open-source and paid models understand our languages to an extent but they aren't developed or trained keeping these languages as the center point.
The models won't understand the nuances of the culture and context unless they are tuned to the finer details.
In a country like India where the dialect changes every four coss(unit of measurement in India; depending upon the region, it can vary between 1 to 3 miles), the democratisation of AI or GenAI in particular would come to fruition only when everyone can leverage such technologies comfortably from their devices without paying a premium.
Education, Policymaking, Agriculture, Banking and many more sectors can take advantage of such culturally rooted and contextual models for solving problems in their respective domains.
Between the hyperbole and hope, Sambhshan is merely a humble and a small step in the direction.
A few months ago, AIM pointed out the dire need for creating benchmarks for Indian languages since most of the famous ones, like MMLU and HumanEval, do not necessarily include a good amount of dataset for Indian languages. Now with Indic LLMs coming into the picture, it is high time that India got its own LLM benchmark for its models.
In his talk at MLDS 2024, Tamil Llama creator Abhinand Balachandran, also highlighted the importance of creating benchmarks specifically tailored for evaluating Indic language models.
Read more🔗👇
https://lnkd.in/dMATVsBr
