In today's linked world, overcoming linguistic boundaries is critical for creating cooperation, understanding, and advancement. However, the multiplicity of languages is a tremendous barrier, frequently resulting in misinterpretations, inefficiencies, and misunderstandings.
Current language translation methods, while beneficial, sometimes lack accuracy and subtlety, resulting in ineffective communication and probable mistakes. Ancient texts from India have a history of misinterpretation resulting in prevailing issues such as the caste system and clashes of minds and racesowing to the same problem.
In our proposed solution, we will be using Large language model Meta-AI’s or LLaMA’s. These are sophisticated artificial intelligence systems that rely on large-scale language models (LLMs) as their primary component. These meta-AIs are intended to do a variety of activities, including but not limited to natural language comprehension, creation, translation, summarisation, and more.
LLaMA’s are artificial intelligence systems that combine the capabilities of large-scale language models with layers of meta-learning, adaptability, and reasoning. These models are often composed of numerous linked neural networks that are trained on massive volumes of text data in order to interpret and create human-like language.
LLaMA’s combined with a tool called the tokenizer is used to effectively form groups in sentences just as humans perceive textual and auditory data. This added human-like touch is what sets this solution apart. The fact that text isn’t interpreted word to word but sentencewise (group to group) results in more accurate and logical translation.
To create and deploy INDO-LLaMA, a language translation system that combines Large Language Model Meta-AIs (LLaMAs) with sophisticated tokenization techniques to successfully overcome linguistic barriers, allowing for accurate and nuanced communication across languages.
The purpose is to guarantee that INDO-LLaMA improves cross-cultural understanding, promotes cooperation, and reduces the possibility of misinterpretations, inefficiencies, and misunderstandings caused by language variety.
Finally, the goal is to produce a dependable and complex tool that fosters cultural interaction, supports proper interpretation of ancient Indian writings, and helps to solve current concerns such as misinterpretations that lead to societal disputes and injustices.
These are the results of the program after input of the first lecture as input.
A pre-trained model was used to make the translator which could be used on multiple national(Indian) as well as foreign languages. There is provision of a concise summary to understand the gist of the content.Due to use of INTEL’s Xeon VM, computation time is reduced. Large volumes of data can be stored on the INTEL cloud. Intel's extension of present tools like pytorch(IPEX) and transform have been used.
Medium : https://medium.com/@mihir.m.dixit/language-translator-using-indo-llama-09d26ad2c107
Youtube : https://www.youtube.com/watch?v=EQ1U-xiGt4k&list=PL_ROZK0B9gVv7h8Tg1vDtxWS-YltLAxqd&index=1&t=12s
Linkedin post[MIHIR DIXIT] : https://www.linkedin.com/feed/update/urn:li:activity:7167862208448724992?updateEntityUrn=urn%3Ali%3Afs_feedUpdate%3A%28V2%2Curn%3Ali%3Aactivity%3A7167862208448724992%29
Linkedin post[AMAN KUMAR] : https://www.linkedin.com/posts/aman-kumar-413838239_language-translator-using-indo-llama-activity-7167866669606916096-mmR8/?utm_source=share&utm_medium=member_desktop
Linkedin post[SHREYAS DEVESH] : https://www.linkedin.com/posts/shreyas-devesh-106409138_language-translator-using-indo-llama-activity-7167919018329169920-2E0N?utm_source=share&utm_medium=member_desktop