We provide three example applications:

• Emotion Recognition
• Sentiment Analysis
• Question Answering

Datasets, knowledge bases, and corresponding learning network checkpoint files are available in the following directories:

• /data/dataset/: Contains datasets for the applications.
• /data/knowledgebase/: Stores the knowledge bases.
• /mappingmodel/: Includes the learning network checkpoint files.

If you want to run custom processes, the code is available in /code/basemodel_path/*kefPL.py. You can choose from the following base models:

• Bi-LSTM att
• DualCL
• Kil
• LCL

To collect a knowledge base for your own downstream tasks, run the following scripts in /code/knowledgecollection/ in sequence:

1. extreackWords.py
2. augKnowledge.py
3. select_unique_token.py

Ensure the following dependencies are installed:

• python>=3.6
• torch>=1.7.1
• datasets>=1.12.1
• transformers>=4.9.2 (Hugging Face)

Alternatively, install all required dependencies using:

pip install -r requirements.txt

@misc{zhu2025domainlexicalknowledgebasedword,
      title={Domain Lexical Knowledge-based Word Embedding Learning for Text Classification under Small Data}, 
      author={Zixiao Zhu and Kezhi Mao},
      year={2025},
      eprint={2506.01621},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2506.01621}, 
}