March Newsletter
Reflecting my experiences consulting alongside Walmart, attending NVIDIA GTC, and researching Neural Machine Translation
Hello Everyone! With the completion of March, 25% of the year has been completed! Over the past month, I have worked on:
Consulting alongside Walmart
Attending the NVIDIA GTC (NVIDIA’s global tech conference, held semi-annually)
Conducting research into neural machine translation techniques.
For those of you who do not know me, welcome to my newsletter! My name is Shaya Farahmand. I am a 15-year-old student from Toronto, Canada 🇨🇦, fascinated with the applications of Machine Learning and Natural Language Processing 📊. Some of my hobbies include playing chess♟, playing my violin and oboe 🎻, as well as following politics and current events.
Consulting Alongside Walmart
Over the month of March, I had the opportunity to consult alongside Walmart, through a program called The Knowledge Society. The project was aimed at reinventing the retail store of the future for 2030. The question that we were required to solve was how to recreate the physical store environment to be exciting and engaging for customers, by offering new services and experiences?
I am still waiting for a green light to share our full deck with you all. As soon as I receive it, I will be sure to do so!
Attending NVIDIA GTC March 2022
This past month, I had the opportunity to attend GTC - a conference hosted by NVIDIA to discuss recent innovations in AI, computer graphics, data science, virtual reality, etc. I was particularly interested in Natural Language Processing, specifically Neural Machine Translation, and hence I was very impressed by two sessions in particular:
Bridging the Gap Between Basic Neural Language Models, Transformers, and Megatron
This presentation was given by Magnus Ekman and Jared Casper who discussed how transformer models have been developed, implemented, and scaled over time in order to solve NLP tasks on large textual data.
Bring Endangered Taiwanese Native Languages Back to Life with the Help of Modern GPUs
This presentation was given by Yuan Fu-Liao on how Conversational AI and the Megatron Transformer were used to translate endangered Taiwanese languages. For example, they were used in the case of real-time multilingual subtitling of Taiwan’s Center for Disease Control’s daily COVID-19 press conference, and the production of official meeting minutes of Taiwan’s Legislative Yuan.
Understanding Neural Machine Translation
Over the course of the month, I spent time learning about neural machine translation, and read several research papers which gave me insight on both the type of models used to conduct machine translation, as well as the techniques used to preprocess text. For example, I read:
Effective Approaches to Attention-based Neural Machine Translation, which understands how the Seq2Seq+Attention model can be implemented to solve translation tasks
Neural Machine Translation of Rare Words with Subword Units, which aims to demonstrate how we can preprocess text by separating longer and more complex words into different parts. For example, the German word Abwasserbehandlungsanlange would be categorized to Abwasser, behandlungs, and anlange
Attention is all you need, which aims to demonstrate how the Transformer model can be implemented to solve translation tasks. Granted, I reviewed this paper while working on the Fake News Detection project (in order to understand the model architecture), but I read it again in order to understand how it can be used for translation.
Goals for the Next Month
For the next month, I primarily want to continue my study of neural machine translation, and specifically work on developing a project that implements these research papers.
Thank you so much for reading my newsletter! Don’t forget to subscribe to receive more updates from myself and feel free to share this with others!