Sources

1. new.txt is a list of words used in businesses: https://www.vocabulary.com/lists/234528#view=list
2. core.txt is a list of words required for the College English Test in China: https://github.com/mahavivo/english-wordlists/blob/master/CET4_edited.txt

• The CET-4 is mandatory for university students in China who are not English majors. It is also a prerequisite for a bachelor's degree. Many employers in China prefer applicants with CET-4 or even CET-6 certification. (https://www.wikiwand.com/en/College_English_Test)

Prompt Algorithm

1. Prompt GPT-3 x times for sentences containing new.txt words. To do this generate set of prompts following the structure: "Generate a business conversation containing the word "confined""

where business is the topic of the conversation and confined is the word to include.

2. Filter out the sentences that contains words not in core.txt or new.txt.

{
    "laborious": "Person 1: \"This project is proving to be quite laborious. We've been working on it for weeks and still haven't made much progress.\"\n\nPerson 2: \"I know, it's been a challenge. But I think if we keep at it, we'll eventually get there.\"",
    "inevitable": "Person 1: \"We need to start planning for the inevitable changes that will come with the new regulations.\"\n\nPerson 2: "Yes, it's inevitable that we'll have to make some adjustments. What do you think our first step should be?\""
}

Dataflow

1. Start with the new words
2. Build prompts with each of the new words
3. Get responses from GPT-3 api
4. Parse the responses

{
    "laborious": [
        {
            "synonoyms": ["a", "b", "c"],
            "chinese": "辛苦的",
            "examples": [
                "This project is proving to be quite laborious. We've been working on it for weeks and still haven't made much progress.",
                "I know, it's been a challenge. But I think if we keep at it, we'll eventually get there."
            ]
        },
        [
            "I can't believe how laborious this project is taking me. I thought it would be much easier!",
            "Yeah, it's definitely a lot more work than we anticipated. But I'm sure if we keep at it, we'll get it done eventually."
        ]
    ]
}

5. Sort them into two theoretical users

Frontend

https://github.com/AlienKevin/L2Z_app