Code for Chain-of-Layer
bash ./scripts/generate_filter.sh
--taxo_name
: ['wordnet', 'semeval_sci', 'wiki', 'dblp']--model_path
: allenai/scibert_scivocab_uncased--save_path
: the save path for the generated filter score JSON file.
bash ./scripts/run_HF.sh
--openai_key
: your openai key--taxo_name
: ['wordnet', 'semeval_sci', 'wiki', 'dblp']--model
: ['gpt-3.5-turbo-16k', 'gpt-4-turbo-preview']--numofExamples
: 0 for zero-shot and 5 for five shot--run
: set toFalse
to only evaluate the saved results--save_path_model_response
: the save path of model response--demo_path
: the path of the demo--ChainofLayers & iteratively
: set toFalse
to only use Hierarchical Format Taxonomy Induction Instruction
bash ./scripts/run_CoL.sh
--openai_key
: your openai key--taxo_name
: ['wordnet', 'semeval_sci', 'wiki', 'dblp']--model
: ['gpt-3.5-turbo-16k', 'gpt-4-turbo-preview']--numofExamples
: 0 for zero-shot and 5 for five shot--run
: set toFalse
to only evaluate the saved results--save_path_model_response
: the save path of model response--demo_path
: the path of the demo--ChainofLayers & iteratively
: set toTrue
to use Chain-of-Layer--filter_mode
: defaultlm_score_ensemble
--filter_model
: defaultscibert_scivocab_uncased
--filter_topk
: Top-K filter
bash ./scripts/gen_demo_CoL_zero.sh
--taxo_name
: ['wordnet', 'semeval_sci', 'wiki', 'dblp']--save_path_model_response
: the save path of raw demo--save_path
: the save path of the processed demo--mode
: default 1
bash ./scripts/run_CoL_zero.sh