Skip to content

Experimenting with LLM-powered apps and language learning

Looking for ways to experiment with an LLM-based programing model? We got the idea to try GPT-4 to develop a comprehensive language learning assistant. The allure of language lies not only in its role as the bedrock of human communication but also in the fact that it doesn't necessitate training data. Moreover, with my imminent move to Japan, the prospect of a tool that could enhance my Japanese learning experience seemed like an excellent fit.

Today, we're excited to open-source the code, offering a glimpse into the project prior to the incorporation of Composable Prompts. It serves as an embodiment of some of the ideas we've eventually built into Composable Prompts. Additionally, a demo has been rolled out for those interested in a hands-on experience. It works in many languages, simply select your native language and the language you want to learn!

(Large) Language Tutor in Japanese

We envision a plethora of potential applications for this approach, encompassing not just language learning, but also topic learning, employee onboarding, training, maintenance procedures, security training, and much more. We welcome your innovative ideas and feedback!

The Idea and Concept

Conventional language learning apps predominantly rely on repetition-based memorization. These apps typically offer a fixed set of translation choices, thereby narrowing learning pathways and limiting cognitive exploration. This method, akin to a multiple-choice test, may not be the most engaging or effective.

The (Large) Language Tutor


Bid farewell to restrictive dialogues. The Language Tutor promotes unrestricted conversation on any subject of your choosing. Whether you're keen on discussing LLMs or simply the day's weather, the choice is yours. Receive real-time feedback, corrections, and suggestions to enhance your linguistic accuracy.

Japanese Dictionary

Original Content

Beyond dialogue, language learning thrives on exposure to varied content. Our tutor crafts tailored stories, recipes, and more to match the user's interests. Be it sports, news, or office procedures, you can immerse yourself in diverse linguistic contexts, various styles, and forms. The tutor also crafts unique questions related to the content, offering personalized corrections and explanations, ensuring your learning experience remains fresh and engaging.

Create a story - type story topic ideas story style ideas

In Context Explanations

Go deeper into the linguistic nuances with in-context explanations. Click on any word for a generated definition, complete with examples. For a broader understanding, request explanations for entire paragraphs, aiding in comprehending sentence structures and their inherent meanings.

AI dictionary explanation

Explain & Verify

Ever hesitated before hitting 'send' on a message in a new language? With our Explain & Verify feature, users can seek verification and in-depth analysis of their text, ensuring clarity and precision.

verify in english verify in french with mistakes

Improvement Ideas

As we envision the future of the Language Tutor, we are moving it on top of Composable Prompts, to enable changing the LLM inside, and showcasing the power of Composable Prompts Studio for simplifying LLM-powered applications.

In addition, we're thinking about several enhancements:

  • Multilingual Expansion: While the platform supports multiple languages like English, Spanish, Japanese, and French, we're eager to embrace more languages and adapt to their unique linguistic challenges.

  • Customized Learning Paths: By gauging users' linguistic proficiency and interests, the LLM could curate personalized learning journeys.

  • Dynamic Content Types: To enrich the learning environment, we're considering a broader spectrum of content sources and types. Maybe add some illustration to stories and conversations.

  • Expanding Horizons: Beyond language, the platform could be reimagined for subjects like Math, Science, and more.

Discover More

Dive deeper into the Language Tutor: