Pixel International Conferences

The author outlines the design and implementation of a large language model (LLM)-infused, ARID (accessibility, retention, immersion, data)-based digital flashcard system for traditional and simplified Mandarin vocabulary acquisition. Learners accessibly input multimedia — text, audio, spreadsheet, image — into a uniform translation interface (available on both mobile and desktop), which proceeds to automatically generate in the background flashcards for the popular open-source software Anki [1]. These flashcards are complete with images, audio, example usage, mnemonics for meaning (via etymology) and writing (via radical decomposition), classifiers, related words, “how-to-write” gifs, etc. Retention is achieved via the spaced repetition of flashcards using Anki's scheduling algorithm. This may be accomplished per convention via Anki's desktop, mobile, or web app. It may also be accomplished via a custom web client which creates and reviews Anki cards en masse during immersive practice with an LLM conversation partner. This web client consists of four modes, each of which will automatically create flashcards from any unknown content the learner encounters during conversation. "Translate" handles translation prompts too nuanced for ‘classical’ translation services. "Learn" mini-tutors the user on new cards. "Review" drills recently learned words via dynamically generated exercises. "Converse" allows the learner to engage in free conversation on a topic of their choice. The LLM will implicitly incorporate long-term review words into each interaction, allowing for the batch review of old Anki flashcards in the context of novel conversation. A snowball effect results: additional flashcards often get made when review happens, but the batch review process ensures that review sessions can nevertheless be completed efficiently. Lastly, custom study routines (available both in Anki and the custom web client) can be dynamically created by the user based on semantic, syntactic, orthographic, or exam data. The author believes the system presented establishes an intimate interface between spaced repetition and immersion in language learning, allowing users to integrate the state-of-the-art spaced repetition scheduling algorithms of Anki into their learning routine without sacrificing time spent practicing in an immersive environment.