Mater Memoriae: Building a Language Learning Application with LangChain, OpenAI Whisper, and GPT 4
@datascienceharp
I am excited to introduce MaterMemoriae, a language learning application that helps you structure your language learning journey around a tried and true method: repetition. There’s a Latin phrase, “Repetitio mater memoriae” — repetition is the mother of memory. Repetition is one of the ways we assert structure to information. And that’s what it’s all about: adding structure to the world around us.
You and I are born screaming and crying into a scary and unstructured world. Screaming, because we have no words. Crying, because we do not understand.
But soon, we come into contact with structure. Structure adds context, and meaning. Family structure is one. Language is another. Some structures unite us. Others divide us.
SInce the tower of Babble came crumbling down language has served as a way to divide us. Certain languages, such as English, are associated with better outcomes, while others — not so much. In some countries, certain subjects are taught in languages that are not the native language of the people. In Ukraine, before the war, courses in math and science and technology may have been taught in English, or in Russian. Most of the textbooks for these subjects were in English, in Russian.
And we’ve accepted this as just par for the course. We have accepted the structures imposed upon us and we have been powerless to meaningfully push back, to assert our independence, to assert our own structures and our own vision of the world.
And we were drowning. Drowning in the volume, variety, and velocity of data flung at us from every direction.
But no more.
In 2017 Google published their seminal research paper, Attention is All You Need, introducing us and the world to a new class of machine learning model: the transformer. Transformers change everything.
Transformers are industrialized, billion and trillion parameter models. ChatGPT is one. Bard is another. And these two are just the tip of the spear — what works for text and natural language works also for images and for sequences of images — videos — and more. Multimodal models, connecting language to art to vision.
With Transformers, we can add structure to our world in new ways never before possible. And, just like the cloud computing model that came before, we can use transformers and pay for them by the minute, by the second, by the instance.
With Transformers, we have it within our power to begin the world over again. So, how, then, should we structure our world? We should structure it towards the good!
It is my hope that, in some small way, MaterMemoriae contributes to this good. Let’s give it a look.
MaterMemoriae: Features and Benefits
Interactive Learning: Our platform aids in vocabulary enhancement and pronunciation improvement. It employs the Levenshtein Ratio, providing learners with immediate feedback as they compare their speech with established benchmarks.
Engaging Activities: Explore grammar through engaging minigames and refine your communication skills with interactive chat sessions.
Live Lessons: Dive into live lessons, now available for Ukrainian and Russian, delivered by our esteemed educators.
Future Endeavors: We’re excited about our imminent launch in Ukrainian and invite collaborations for future developments, including a Chinese curriculum. A presentation is set for September 15th.
Strategic Collaborations: By partnering with renowned professionals in the AI and machine learning fields, we anticipate reaching greater heights.
Platform Development: Core functionalities of MaterMemoriae are progressing rapidly. Other tools and technologies include OpenAI Whisper, text-to-speech systems, semantic search, many statistical methods, many machine learning algorithms, and a Google Cloud architecture.
Technical Blueprint: The backbone of MaterMemoriae is Flask and Flask SQLAlchemy, with a visually appealing frontend made possible through React JS and Material UI components.