Injecting Salesperson’s Dialogue Strategies in Large Language Models with Chain-of-Thought Reasoning

This repo is the implementation of the paper Injecting Salesperson’s Dialogue Strategies in Large Language Models with Chain-of-Thought Reasoning, including the dataset SalesBot 2.0 (revised from SalesBot 1.0) and the fine-tuned model.

Please cite the following references if you use the released data or model.

@inproceedings{chang2024injecting,
      title={Injecting Salesperson's Dialogue Strategies in Large Language Models with Chain-of-Thought Reasoning}, 
      author={Wen-Yu Chang and Yun-Nung Chen},
      year={2024},
      booktitle={Findings of the Association for Computational Linguistics: ACL 2024},
      pages = "3798--3812"
}

@article{chang2023salesbot,
  title={SalesBot 2.0: A Human-Like Intent-Guided Chit-Chat Dataset},
  author={Chang, Wen-Yu and Chen, Yun-Nung},
  journal={arXiv preprint arXiv:2308.14266},
  year={2023}
}

Model

You can find the model (llama-2-7B) fine-tuned on SalesBot 2.0 data here.

Interaction with the model

cd scripts/
python interactive.py --model_path <path_to_model> --temperature 0.7

Generate SalesBot 2.0 Datasets and Train Model from Scratch

1. Install the required packages

pip3 install openai, tqdm, fast
cd ./FastChat/
pip3 install -e ".[model_worker,webui,train]"

2. Rewrite

python generate_prompts.py --mode "generate_rewrite" --input_file "./salesbot2_prompt/Merge_SGD.json" --prompt_dir "<path_to_save_prompts>"
python generate_prompts.py --mode "rewrite" --input_file "<output_file_from_generate_rewrite>" --output_dir "<dir_to_save_outputs_from_llm>"

3. Intent Detection

python generate_prompts.py --mode "generate_intent" --prev_output_dir "<output_dir_from_llm>" --prompt_dir "<path_to_save_prompts>"
python generate_prompts.py --mode "intent_detection" --input_file "<output_file_from_generate_intent>" --output_dir "<path_to_save_outputs_from_llm>"

4. Dialogue Continuation

python generate_prompts.py --mode "generate_continue" --prev_output_dir "<output_dir_from_llm>" --prompt_dir "<path_to_save_prompts>"
python generate_prompts.py --mode "dialogue_continuation" --input_file "<output_file_from_generate_dialogue>" --output_dir "<path_to_save_outputs_from_llm>"

5. Transtion Detection

python generate_prompts.py --mode "generate_trans" --prev_output_dir "<output_dir_from_llm>" --prompt_dir "<path_to_save_prompts>"
python generate_prompts.py --mode "transition_detection" --input_file "<output_file_from_generate_transition>" --output_dir "<path_to_save_outputs_from_llm>"

6. Generate SalesBot 2.0 Dataset

python generate_prompts.py --mode "generate_dataset" --prev_output_dir "<output_dir_from_llm>" --prompt_dir "<path_to_save_prompts>"

Prepare Training and Evaluation Data

cd ./create_dataset/utils/
python create_salesbot2_dataset.py
python create_salesbot1_dataset.py

Fine-tune Llama-2 on Dataset

cd scripts/
bash run_finetune.sh <path_to_dataset> <path_to_save_model>