Views
No views yet
$ pip install zero-shot-re1from transformers import AutoTokenizer
2from zero_shot_re import RelTaggerModel, RelationExtractor
3
4model = RelTaggerModel.from_pretrained("fractalego/fewrel-zero-shot")
5tokenizer = AutoTokenizer.from_pretrained("fractalego/fewrel-zero-shot")
6
7relations = ['noble title', 'founding date', 'occupation of a person']
8extractor = RelationExtractor(model, tokenizer, relations)
9ranked_rels = extractor.rank(text='John Smith received an OBE', head='John Smith', tail='OBE')
10print(ranked_rels)1[('noble title', 0.9690611883997917),
2 ('occupation of a person', 0.0012609362602233887),
3 ('founding date', 0.00024014711380004883)]| Model | 0-shot 5-ways | 0-shot 10-ways |
|---|---|---|
| (1) Distillbert | 70.1±0.5 | 55.9±0.6 |
| (2) Bert Large | 80.8±0.4 | 69.6±0.5 |
| (3) Distillbert + SQUAD | 81.3±0.4 | 70.0±0.2 |
| (4) Bert Large + SQUAD | 86.0±0.6 | 76.2±0.4 |
1@inproceedings{cetoli-2020-exploring,
2 title = "Exploring the zero-shot limit of {F}ew{R}el",
3 author = "Cetoli, Alberto",
4 booktitle = "Proceedings of the 28th International Conference on Computational Linguistics",
5 month = dec,
6 year = "2020",
7 address = "Barcelona, Spain (Online)",
8 publisher = "International Committee on Computational Linguistics",
9 url = "https://www.aclweb.org/anthology/2020.coling-main.124",
10 doi = "10.18653/v1/2020.coling-main.124",
11 pages = "1447--1451",
12 abstract = "This paper proposes a general purpose relation extractor that uses Wikidata descriptions to represent the relation{'}s surface form. The results are tested on the FewRel 1.0 dataset, which provides an excellent framework for training and evaluating the proposed zero-shot learning system in English. This relation extractor architecture exploits the implicit knowledge of a language model through a question-answering approach.",
13}