Chinese Learning Practice

Around the beginning of last year, I tried to learn Chinese again. I have had a long history of trying to learn other languages (Spanish, German, and Chinese), but I always seemed to stop short of reaching fluency. I would learn enough of the language to do well in the classroom and to get by when talking to native speakers, but not enough to talk about complex topics or fully understand conversations between native speakers.

I eventually realized that learning more vocabulary was key. While circumlocution (such as saying "thing that makes light" if I don't know the word for "lightbulb") was useful, my limited vocabulary made it difficult to speak efficiently and understand native speakers. I’ve noticed that as I get better at a language, the number of new words I need to learn to feel like I'm making progress increases exponentially. At the beginning, every word I learn allows me to express and understand many new topics, but later, the words I learn are more specialized, so I make less visible progress. This is reflected in the Chinese HSK system, which groups Chinese words into six levels based on their frequency of use. The 150 most common Chinese words are contained in HSK 1, HSK 2 introduces 150 of the next most common Chinese words, HSK 3 introduces 300 more words, and so on until HSK 6, which introduces 2500 more words. It's worth noting that HSK 6 includes words like "instinct," "ambition," and "specialty"—words that aren't used in every sentence but will inevitably come up in normal conversations between native speakers.

For this reason, I set up some Anki flashcard decks to help me learn all the Chinese HSK vocabulary. I started between HSK 2 and 3 and began learning HSK 6 a few months ago. With this approach, I was finally able to make real progress in my Chinese (this was the 4th time I had tried learning Chinese beyond what I had learned in elementary school). I noticed that I was able to read much more of the Chinese on packages and instruction manuals I encountered, and I even began to understand Chinese shows like the Monkey King, which I couldn't understand when I was younger.

However, I noticed that my ability to speak was lagging far behind my improving ability to listen and read. Whenever I tried to translate an English sentence to Chinese, I often knew the Chinese words corresponding to the English words but I didn't know how to put them together. For this reason, I created this sentence generator. Since each sentence contains the Chinese, Pinyin, and English, I can practice both translating Chinese to English and English to Chinese. I chose to use Deepseek to generate the sentences because I noticed that Deepseek produced better quality Chinese sentences compared to ChatGPT and Google Translate.

The core of my program is a request to Deepseek's API. For those who are curious, when you click “Generate Sentences,” the exact prompt my program sent to the API is displayed in the first paragraph under "More Information". If you don't enter a subject, my program automatically chooses one from a curated list I created. These include subjects like careers, emotions, cooking and recipes, family relatives, ethics, and other subjects I wanted to expand my vocabulary in. My program also has a random chance of adding additional constraints to the prompt such as requiring complex sentence structures, idioms, or colloquial language. These subjects and constraints help keep Deepseek from repeating the same sentences over time.

At first, I was surprised by the >20 second delay between hitting the "Generate Sentences" button and seeing the sentences appear. However, I timed both this API interface and Deepseek's web interface with the same prompt, and they took around the same amount of time. Even ChatGPT took around 20 seconds when I tested the prompt on their web interface. I think this lag is less noticeable on web interfaces because I can see the sentences as they're being generated.

If I were to develop this project further, I would like to automatically generate and display sentences on a nice e-ink display in my room. This would remove the small but real friction of needing to open my website each time I want to practice. Right now, making progress in learning Chinese still relies on me making a conscious decision every day to sit down and study. While habit helps, I believe a dedicated display that I can glance at throughout the day could weave learning Chinese more naturally into my daily routine.