HZW | Devpost

Inspiration

Before the ubiquity of LLM's language learners had to rely on digital libraries such as Aozora Bunko or electronic language corpi compiled over the years by academics and volunteers. These sentence banks are inefficient, partly due to lack of normativity among data sources, but they are also bloated. If you've experienced reading a print dictionary, you would grasp the definition of a given word, but more likely than not lack the ability to master the word for practical use. The good news is that LLM's can now generate a natural sentence containing the word you desire to master within seconds. This concept could have been applied to any language, though Chinese is fun because it's the only major contemporary logographic system and I like the visual appeal.

What it does

HZW will quiz you on your handwriting while providing natural sentences for a wider breadth of vocabulary than language corpi can. Select from different models, adjust the complexity of the sentences to your heart's content.

How we built it

It's an ASP.NET Web App. Nothing to write home about. The character sets that test handwriting were hand-generated though.

Challenges we ran into

Working with the Snowflake Cortex API went very smoothly through the snowflake .NET interface. The problem with the HanziWriter as the focal point of this webpage is that the client-side became very bloated very quickly.