Triple Uni is an anonymous communication platform for HKU, CUHK and UST.
The Triple Uni Post Topic Classifier is a machine learning project based on Python and PyTorch, designed to automatically identify and classify the topics of posts on the platform. The project uses several models to predict the topics of posts based on their embedded representations.
You can experience our model online in the example below, or deploy and try it yourself.
This project uses posts from the three universities, with uni_post_id ranging from 1 to 500,000 as the dataset, using OpenAI's text-embedding-ada-002 model to embed the content of the posts.
| University Name | ID Range | Posting Time |
|---|---|---|
| The University of Hong Kong (HKU) | 1 ～ 253507 | 2020-10-24 ～ 2022-12-19 |
| The Chinese University of Hong Kong (CUHK) | 1 ～ 190662 | 2020-08-23 ～ 2022-12-19 |
| The Hong Kong University of Science and Technology (UST) | 1 ～ 55857 | 2020-01-08 ～ 2022-12-19 |
Each data file datasets/post_data_{i}.parquet contains 10,000 posts with uni_post_id from 10000 * i + 1 to 10000 * i + 10000.
| Field Name | Type | Description |
|---|---|---|
| uni_post_id | int64 | The inter-university post ID |
| post_topic | string | The topic of the post |
| embedding | float[1536] | 1536-dimensional embedding of the post content |
Since the topics of posts vary among universities, we chose Random Thought, Emotions, Academics, Job Hunting, and Trading as target topic categories.
We cleaned the data by renaming Flea Market to Trading and removing all other topics.
Please read our report for more detail.
| Post Content | Model Output |
|---|---|
| It feels like there are very few people at school today. Has everyone not returned yet? | Random Thought |
| It's not that I don't want to start dating, it's just that I really haven't met anyone I like a lot. | Emotions |
| Did any of you take the TOEFL when applying for postgraduate programs in the US? I ask because many schools say that if your undergraduate education was in English, you don't need to submit TOEFL scores. | Academics |
| Is an internship at Morgan Stanley worth it? Mainly involved in institutional business. | Job Hunting |
| Selling a handheld steam iron, very compact and barely used. Pick up at the subway station near the school. | Trading |
Click here to experience our model online.
Clone the repository to your local machine.
git clone https://github.com/lststar/Triple-Uni-Topic-Classifier.git
Installation required environment.
pip install -r requirements.txt
Use train.ipynb to train the model
Fill in openai_api_key in main.py, and run the Web UI with the following command.
streamlit run main.py
If you have any questions or suggestions about this project, please raise an Issue or send an email to v@tripleuni.com.