relbert/roberta-large-semeval2012-mask-prompt-a-triplet
RelBERT fine-tuned from roberta-large on
semeval2012.
Fine-tuning is done via RelBERT library (see the repository for more detail).
It achieves the following results on the relation understanding tasks:
- Analogy Question (dataset, full result):
- Accuracy on SAT (full): 0.5721925133689839
- Accuracy on SAT: 0.5786350148367952
- Accuracy on BATS: 0.7687604224569206
- Accuracy on U2: 0.5614035087719298
- Accuracy on U4: 0.5972222222222222
- Accuracy on Google: 0.926
- Lexical Relation Classification (dataset, full result):
- Micro F1 score on BLESS: 0.9120084375470845
- Micro F1 score on CogALexV: 0.8546948356807512
- Micro F1 score on EVALution: 0.7258938244853739
- Micro F1 score on K&H+N: 0.9517980107115531
- Micro F1 score on ROOT09: 0.9012848636790974
- Relation Mapping (dataset, full result):
- Accuracy on Relation Mapping: 0.8050396825396825
Usage
This model can be used through the relbert library. Install the library via pip
pip install relbert
and activate model as below.
from relbert import RelBERT
model = RelBERT("relbert/roberta-large-semeval2012-mask-prompt-a-triplet")
vector = model.get_embedding(['Tokyo', 'Japan']) # shape of (1024, )
Training hyperparameters
The following hyperparameters were used during training:
- model: roberta-large
- max_length: 64
- mode: mask
- data: semeval2012
- n_sample: 10
- custom_template: Today, I finally discovered the relation between <subj> and <obj> : <subj> is the <mask> of <obj>
- template: None
- softmax_loss: True
- in_batch_negative: True
- parent_contrast: True
- mse_margin: 1
- epoch: 1
- lr_warmup: 10
- batch: 64
- lr: 2e-05
- lr_decay: False
- weight_decay: 0
- optimizer: adam
- momentum: 0.9
- fp16: False
- random_seed: 0
The full configuration can be found at fine-tuning parameter file.
Reference
If you use any resource from RelBERT, please consider to cite our paper.
@inproceedings{ushio-etal-2021-distilling-relation-embeddings,
title = "{D}istilling {R}elation {E}mbeddings from {P}re-trained {L}anguage {M}odels",
author = "Ushio, Asahi and
Schockaert, Steven and
Camacho-Collados, Jose",
booktitle = "EMNLP 2021",
year = "2021",
address = "Online",
publisher = "Association for Computational Linguistics",
}