research

PromptKG

A collection of prompt learning-related research model implementations

RetroPrompt: retrieval-augmented prompt learning to decouple knowledge from memorization

Demo-Tuning: contrastive demonstration tuning for natural language processing

RetrievalRE: retrieval-enhanced prompt tuning for relation extraction

GenKGC: link prediction as sequence-to-sequence generation for fast inference

PromptKGC: data-efficient prompt learning-based knowledge graph completion

News

• [Model Release] September, 2022: RetroPrompt - A retrieval mechanism during the process of input, training and inference, thus equipping the model with the ability to retrieve related contexts from the training corpus as cues for enhancement.
• [Model Release] Aprial, 2022: Demo-Tuning - A pluggable, extensible, and efficient approach named contrastive demonstration tuning, which is free of demonstration sampling
• [Model Release] Aprial, 2022: RetrievalRE - A retrieval-enhanced prompt tuning method for relation extraction which empowers the model to reference similar instances from the training corpus as cues for inference
• [Model Release] Jaunary, 2022: GenKGC - A sequence-to-sequence approach for knowledge graph completion.
• [Model Release] January, 2022: PromptKGC - A prompt learning-based approach for few-shot knowledge graph completion

Release

***** September, 2022: RetroPrompt release *****

• RetroPrompt (September 21, 2022): RetroPrompt constructs an open-book knowledge-store from training instances and implements a retrieval mechanism during the process of input, training and inference, thus equipping the model with the ability to retrieve related contexts from the training corpus as cues for enhancement. "Decoupling Knowledge from Memorization: Retrieval-augmented Prompt LearningNeurIPS 2022 "

***** Aprial, 2022: Demo-Tuning | RetrievalRE release *****

• Demo-Tuning (Aprial 6, 2022): A pluggable, extensible, and efficient approach named contrastive demonstration tuning, which is free of demonstration sampling. "Contrastive Demonstration Tuning for Pre-trained Language ModelsEMNLP 2022 (Findings)"
• RetrievalRE (Aprial 6, 2022): RetrievalRE regards relation extraction as an open-book examination and leverages a new semiparametric paradigm of retrieval-enhanced prompt tuning. "Relation Extraction as Open-book Examination: Retrieval-enhanced Prompt TuningSIGIR 2022 "

***** January, 2022: GenKGC | PromptKGC release *****

• GenKGC (Jaunary 31, 2022): GenKGC converts knowledge graph completion to sequence-to-sequence generation with pre-trained language model with relation-guided demonstration and entity-aware hierarchical decoding. It can obtain better or comparable performance than baselines, and achieve faster inference speed compared with previous methods with pre-trained language models. "From Discrimination to Generation: Knowledge Graph Completion with Generative TransformerWWW 2022 "
• PromptKGC (Jaunary 31, 2022): A prompt-tuning approach (knowledge collaborative fine-tuning) for low-resource knowledge graph completion, which leverages the structured knowledge to construct the initial prompt template and learn the optimal templates, labels and model parameters through a collaborative fine-tuning algorithm. It can obtain state-of-the-art few-shot performance on FB15K-237, WN18RR, and UMLS.

Contact Information

For help or issues using the models, please submit a GitHub issue.