Inspiration
I started building Liens because I couldn't find a French learning platform that matched the way I wanted to learn. I was constantly switching between dictionaries, grammar websites, AI assistants, pronunciation tools, and flashcard apps. Every tool solved one problem, but none connected the entire learning journey. I wanted one place where every French language element, from words and conjugations to expressions, sentences, and grammar, could be explored, understood, saved, and reviewed together.
What it does
The app accepts either a French word or an entire sentence through a single search box.
For words, it automatically recognizes inflected forms and provides the lemma, pronunciation, English and Chinese definitions, conjugations, collocations, CEFR level, frequency, and example sentences.
For sentences, it performs deterministic grammatical analysis before providing AI learning assistance. Every word is clickable, allowing users to explore related language concepts without ever reaching a dead end.
Users can also save words, collocations, grammar, and sentences into collections for later review using flashcards or list view.
How we built it
I built Liens around a source-backed language graph instead of making AI the product. Verified linguistic knowledge is stored separately from AI-generated learning resources. The app always answers from the local graph whenever possible, checks a dedicated AI learning database if needed, and only calls Gemini as a last step. This keeps AI helpful without allowing it to overwrite linguistic truth.
Challenges we ran into
The hardest part wasn't integrating AI. It was deciding where AI should stop. I spent much more time designing the boundaries between verified knowledge, deterministic grammar analysis, personal learning data, and AI-generated learning resources than integrating the model itself. Most of the project's architecture came from solving those problems.
Accomplishments that we're proud of
I'm proud of building a connected language graph that brings together dictionaries, grammar, sentence analysis, vocabulary management, review, and AI learning assistance into one seamless experience. I'm especially proud that AI improves the learning experience without ever modifying the canonical language graph.
What we learned
Building Liens taught me that great language learning isn't about adding more AI. It's about reducing friction and giving every piece of language a clear place in the system. Separating verified knowledge from AI-generated learning resources made the product both more reliable and easier to extend.
What's next for Liens (French learning)
My next goal is to continue improving the learning experience rather than simply adding more content. I'm working on richer review, better pronunciation coverage, stronger sentence understanding, and a native mobile version while keeping Liens local first and graph based.
Log in or sign up for Devpost to join the conversation.