数据集:
lex_glue
Inspired by the recent widespread use of the GLUE multi-task benchmark NLP dataset (Wang et al., 2018), the subsequent more difficult SuperGLUE (Wang et al., 2019), other previous multi-task NLP benchmarks (Conneau and Kiela, 2018; McCann et al., 2018), and similar initiatives in other domains (Peng et al., 2019), we introduce the Legal General Language Understanding Evaluation (LexGLUE) benchmark , a benchmark dataset to evaluate the performance of NLP methods in legal tasks. LexGLUE is based on seven existing legal NLP datasets, selected using criteria largely from SuperGLUE.
As in GLUE and SuperGLUE (Wang et al., 2019b,a), one of our goals is to push towards generic (or ‘foundation’) models that can cope with multiple NLP tasks, in our case legal NLP tasks possibly with limited task-specific fine-tuning. Another goal is to provide a convenient and informative entry point for NLP researchers and practitioners wishing to explore or develop methods for legalNLP. Having these goals in mind, the datasets we include in LexGLUE and the tasks they address have been simplified in several ways to make it easier for newcomers and generic models to address all tasks.
LexGLUE benchmark is accompanied by experimental infrastructure that relies on Hugging Face Transformers library and resides at: https://github.com/coastalcph/lex-glue .
The supported tasks are the following:
Dataset | Source | Sub-domain | Task Type | Classes |
ECtHR (Task A) | Chalkidis et al. (2019) | ECHR | Multi-label classification | 10+1 |
ECtHR (Task B) | Chalkidis et al. (2021a) | ECHR | Multi-label classification | 10+1 |
SCOTUS | Spaeth et al. (2020) | US Law | Multi-class classification | 14 |
EUR-LEX | Chalkidis et al. (2021b) | EU Law | Multi-label classification | 100 |
LEDGAR | Tuggener et al. (2020) | Contracts | Multi-class classification | 100 |
UNFAIR-ToS | Lippi et al. (2019) | Contracts | Multi-label classification | 8+1 |
CaseHOLD | Zheng et al. (2021) | US Law | Multiple choice QA | n/a |
The European Court of Human Rights (ECtHR) hears allegations that a state has breached human rights provisions of the European Convention of Human Rights (ECHR). For each case, the dataset provides a list of factual paragraphs (facts) from the case description. Each case is mapped to articles of the ECHR that were violated (if any).
ecthr_bThe European Court of Human Rights (ECtHR) hears allegations that a state has breached human rights provisions of the European Convention of Human Rights (ECHR). For each case, the dataset provides a list of factual paragraphs (facts) from the case description. Each case is mapped to articles of ECHR that were allegedly violated (considered by the court).
scotusThe US Supreme Court (SCOTUS) is the highest federal court in the United States of America and generally hears only the most controversial or otherwise complex cases which have not been sufficiently well solved by lower courts. This is a single-label multi-class classification task, where given a document (court opinion), the task is to predict the relevant issue areas. The 14 issue areas cluster 278 issues whose focus is on the subject matter of the controversy (dispute).
eurlexEuropean Union (EU) legislation is published in EUR-Lex portal. All EU laws are annotated by EU's Publications Office with multiple concepts from the EuroVoc thesaurus, a multilingual thesaurus maintained by the Publications Office. The current version of EuroVoc contains more than 7k concepts referring to various activities of the EU and its Member States (e.g., economics, health-care, trade). Given a document, the task is to predict its EuroVoc labels (concepts).
ledgarLEDGAR dataset aims contract provision (paragraph) classification. The contract provisions come from contracts obtained from the US Securities and Exchange Commission (SEC) filings, which are publicly available from EDGAR. Each label represents the single main topic (theme) of the corresponding contract provision.
unfair_tosThe UNFAIR-ToS dataset contains 50 Terms of Service (ToS) from on-line platforms (e.g., YouTube, Ebay, Facebook, etc.). The dataset has been annotated on the sentence-level with 8 types of unfair contractual terms (sentences), meaning terms that potentially violate user rights according to the European consumer law.
case_holdThe CaseHOLD (Case Holdings on Legal Decisions) dataset includes multiple choice questions about holdings of US court cases from the Harvard Law Library case law corpus. Holdings are short summaries of legal rulings accompany referenced decisions relevant for the present case. The input consists of an excerpt (or prompt) from a court decision, containing a reference to a particular case, while the holding statement is masked out. The model must identify the correct (masked) holding statement from a selection of five choices.
The current leaderboard includes several Transformer-based (Vaswaniet al., 2017) pre-trained language models, which achieve state-of-the-art performance in most NLP tasks (Bommasani et al., 2021) and NLU benchmarks (Wang et al., 2019a). Results reported by Chalkidis et al. (2021) :
Task-wise Test Results
Dataset | ECtHR A | ECtHR B | SCOTUS | EUR-LEX | LEDGAR | UNFAIR-ToS | CaseHOLD |
Model | μ-F1 / m-F1 | μ-F1 / m-F1 | μ-F1 / m-F1 | μ-F1 / m-F1 | μ-F1 / m-F1 | μ-F1 / m-F1 | μ-F1 / m-F1 |
TFIDF+SVM | 64.7 / 51.7 | 74.6 / 65.1 | 78.2 / 69.5 | 71.3 / 51.4 | 87.2 / 82.4 | 95.4 / 78.8 | n/a |
Medium-sized Models (L=12, H=768, A=12) | |||||||
BERT | 71.2 / 63.6 | 79.7 / 73.4 | 68.3 / 58.3 | 71.4 / 57.2 | 87.6 / 81.8 | 95.6 / 81.3 | 70.8 |
RoBERTa | 69.2 / 59.0 | 77.3 / 68.9 | 71.6 / 62.0 | 71.9 / 57.9 | 87.9 / 82.3 | 95.2 / 79.2 | 71.4 |
DeBERTa | 70.0 / 60.8 | 78.8 / 71.0 | 71.1 / 62.7 | 72.1 / 57.4 | 88.2 / 83.1 | 95.5 / 80.3 | 72.6 |
Longformer | 69.9 / 64.7 | 79.4 / 71.7 | 72.9 / 64.0 | 71.6 / 57.7 | 88.2 / 83.0 | 95.5 / 80.9 | 71.9 |
BigBird | 70.0 / 62.9 | 78.8 / 70.9 | 72.8 / 62.0 | 71.5 / 56.8 | 87.8 / 82.6 | 95.7 / 81.3 | 70.8 |
Legal-BERT | 70.0 / 64.0 | 80.4 / 74.7 | 76.4 / 66.5 | 72.1 / 57.4 | 88.2 / 83.0 | 96.0 / 83.0 | 75.3 |
CaseLaw-BERT | 69.8 / 62.9 | 78.8 / 70.3 | 76.6 / 65.9 | 70.7 / 56.6 | 88.3 / 83.0 | 96.0 / 82.3 | 75.4 |
Large-sized Models (L=24, H=1024, A=18) | |||||||
RoBERTa | 73.8 / 67.6 | 79.8 / 71.6 | 75.5 / 66.3 | 67.9 / 50.3 | 88.6 / 83.6 | 95.8 / 81.6 | 74.4 |
Averaged (Mean over Tasks) Test Results
Averaging | Arithmetic | Harmonic | Geometric |
Model | μ-F1 / m-F1 | μ-F1 / m-F1 | μ-F1 / m-F1 |
Medium-sized Models (L=12, H=768, A=12) | |||
BERT | 77.8 / 69.5 | 76.7 / 68.2 | 77.2 / 68.8 |
RoBERTa | 77.8 / 68.7 | 76.8 / 67.5 | 77.3 / 68.1 |
DeBERTa | 78.3 / 69.7 | 77.4 / 68.5 | 77.8 / 69.1 |
Longformer | 78.5 / 70.5 | 77.5 / 69.5 | 78.0 / 70.0 |
BigBird | 78.2 / 69.6 | 77.2 / 68.5 | 77.7 / 69.0 |
Legal-BERT | 79.8 / 72.0 | 78.9 / 70.8 | 79.3 / 71.4 |
CaseLaw-BERT | 79.4 / 70.9 | 78.5 / 69.7 | 78.9 / 70.3 |
Large-sized Models (L=24, H=1024, A=18) | |||
RoBERTa | 79.4 / 70.8 | 78.4 / 69.1 | 78.9 / 70.0 |
We only consider English datasets, to make experimentation easier for researchers across the globe.
An example of 'train' looks as follows.
{ "text": ["8. The applicant was arrested in the early morning of 21 October 1990 ...", ...], "labels": [6] }ecthr_b
An example of 'train' looks as follows.
{ "text": ["8. The applicant was arrested in the early morning of 21 October 1990 ...", ...], "label": [5, 6] }scotus
An example of 'train' looks as follows.
{ "text": "Per Curiam\nSUPREME COURT OF THE UNITED STATES\nRANDY WHITE, WARDEN v. ROGER L. WHEELER\n Decided December 14, 2015\nPER CURIAM.\nA death sentence imposed by a Kentucky trial court and\naffirmed by the ...", "label": 8 }eurlex
An example of 'train' looks as follows.
{ "text": "COMMISSION REGULATION (EC) No 1629/96 of 13 August 1996 on an invitation to tender for the refund on export of wholly milled round grain rice to certain third countries ...", "labels": [4, 20, 21, 35, 68] }ledgar
An example of 'train' looks as follows.
{ "text": "All Taxes shall be the financial responsibility of the party obligated to pay such Taxes as determined by applicable law and neither party is or shall be liable at any time for any of the other party ...", "label": 32 }unfair_tos
An example of 'train' looks as follows.
{ "text": "tinder may terminate your account at any time without notice if it believes that you have violated this agreement.", "label": 2 }casehold
An example of 'test' looks as follows.
{ "context": "In Granato v. City and County of Denver, No. CIV 11-0304 MSK/BNB, 2011 WL 3820730 (D.Colo. Aug. 20, 2011), the Honorable Marcia S. Krieger, now-Chief United States District Judge for the District of Colorado, ruled similarly: At a minimum, a party asserting a Mo-nell claim must plead sufficient facts to identify ... to act pursuant to City or State policy, custom, decision, ordinance, re d 503, 506-07 (3d Cir.l985)(<HOLDING>).", "endings": ["holding that courts are to accept allegations in the complaint as being true including monell policies and writing that a federal court reviewing the sufficiency of a complaint has a limited task", "holding that for purposes of a class certification motion the court must accept as true all factual allegations in the complaint and may draw reasonable inferences therefrom", "recognizing that the allegations of the complaint must be accepted as true on a threshold motion to dismiss", "holding that a court need not accept as true conclusory allegations which are contradicted by documents referred to in the complaint", "holding that where the defendant was in default the district court correctly accepted the fact allegations of the complaint as true" ], "label": 0 }
Dataset | Training | Development | Test | Total |
ECtHR (Task A) | 9,000 | 1,000 | 1,000 | 11,000 |
ECtHR (Task B) | 9,000 | 1,000 | 1,000 | 11,000 |
SCOTUS | 5,000 | 1,400 | 1,400 | 7,800 |
EUR-LEX | 55,000 | 5,000 | 5,000 | 65,000 |
LEDGAR | 60,000 | 10,000 | 10,000 | 80,000 |
UNFAIR-ToS | 5,532 | 2,275 | 1,607 | 9,414 |
CaseHOLD | 45,000 | 3,900 | 3,900 | 52,800 |
Dataset | Source | Sub-domain | Task Type |
ECtHR (Task A) | Chalkidis et al. (2019) | ECHR | Multi-label classification |
ECtHR (Task B) | Chalkidis et al. (2021a) | ECHR | Multi-label classification |
SCOTUS | Spaeth et al. (2020) | US Law | Multi-class classification |
EUR-LEX | Chalkidis et al. (2021b) | EU Law | Multi-label classification |
LEDGAR | Tuggener et al. (2020) | Contracts | Multi-class classification |
UNFAIR-ToS | Lippi et al. (2019) | Contracts | Multi-label classification |
CaseHOLD | Zheng et al. (2021) | US Law | Multiple choice QA |
Ilias Chalkidis, Abhik Jana, Dirk Hartung, Michael Bommarito, Ion Androutsopoulos, Daniel Martin Katz, and Nikolaos Aletras. LexGLUE: A Benchmark Dataset for Legal Language Understanding in English. 2022. In the Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics. Dublin, Ireland.
@inproceedings{chalkidis-etal-2021-lexglue, title={LexGLUE: A Benchmark Dataset for Legal Language Understanding in English}, author={Chalkidis, Ilias and Jana, Abhik and Hartung, Dirk and Bommarito, Michael and Androutsopoulos, Ion and Katz, Daniel Martin and Aletras, Nikolaos}, year={2022}, booktitle={Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics}, address={Dubln, Ireland}, }
Thanks to @iliaschalkidis for adding this dataset.