premise
stringlengths
23
232
hypothesis
stringlengths
0
142
label
class label
0
2
At least three commissioners spend time at home.
At least three commissioners spend a lot of time at home.
1neutral
Clients at the demonstration were all impressed by the system's performance. Smith was a client at the demonstration.
Smith was impressed by the system's performance.
0entailment
The PC-6082 is faster than some ITEL computer. The ITEL-ZX is an ITEL computer.
The PC-6082 is faster than the ITEL-ZX.
1neutral
Smith discovered new species for two years.
Smith discovered new species.
0entailment
Bill suggested to Frank's boss that they should go to the meeting together, and Carl to Alan's wife.
If it was suggested that Bill and Frank should go together, it was suggested that Carl and Alan should go together.
0entailment
ITEL won more orders than APCOM.
ITEL won some orders.
0entailment
Helen saw the chairman of the department answer the phone. The chairman of the department is a person.
There is someone whom Helen saw answer the phone.
0entailment
Before APCOM bought its present office building, it had been paying mortgage interest on the previous one for 8 years. Since APCOM bought its present office building it has been paying mortgage interest on it for more than 10 years.
APCOM has been paying mortgage interest for a total of 15 years or more.
0entailment
All mice are small animals. All elephants are large animals. Mickey is a large mouse. Dumbo is a small elephant.
Dumbo is larger than Mickey.
0entailment
John owns a red car. Bill owns a fast one.
Bill owns a fast red car.
0entailment
Has John spoken to Mary? Bill has.
Bill has spoken to Mary.
0entailment
Every European can travel freely within Europe. Every European is a person. Every person who has the right to live in Europe can travel freely within Europe.
Every European has the right to live in Europe.
1neutral
Dumbo is a four-legged animal.
Dumbo is four-legged.
0entailment
John went to Paris by car. Bill by train.
Bill went to Paris by train.
0entailment
A Swede won a Nobel prize. Every Swede is a Scandinavian.
A Scandinavian won a Nobel prize.
0entailment
Twenty men work in the Sales Department. But only one woman.
Two women work in the Sales Department.
2contradiction
Each Canadian resident can travel freely within Europe. Every Canadian resident is a resident of the North American continent.
Each resident of the North American continent can travel freely within Europe.
1neutral
The inhabitants of Cambridge voted for a Labour MP.
Every inhabitant of Cambridge voted for a Labour MP.
1neutral
Bill suggested to Frank's boss that they should go to the meeting together, and Carl to Alan's wife.
If it was suggested that Bill, Frank and Frank's boss should go together, it was suggested that Carl, Alan and Alan's wife should go together.
0entailment
Mary used her workstation.
Mary is female.
0entailment
Smith wrote a report in two hours.
Smith wrote a report in one hour.
1neutral
Jones graduated in March and has been employed ever since. Jones has been unemployed in the past.
Jones was unemployed at some time before he graduated.
0entailment
Mickey is a small animal.
Mickey is a large animal.
2contradiction
The PC-6082 is faster than the ITEL-ZX and the ITEL-ZY.
The PC-6082 is faster than the ITEL-ZX.
0entailment
At least three female commissioners spend time at home.
At least three commissioners spend time at home.
0entailment
Smith saw Jones sign the contract. When Jones signed the contract, his heart was beating.
Smith saw Jones' heart beat.
1neutral
John owns a red car. Bill owns a fast one.
Bill owns a fast car.
0entailment
At most ten commissioners spend time at home.
At most ten female commissioners spend time at home.
0entailment
The PC-6082 is faster than the ITEL-XZ.
The PC-6082 is fast.
1neutral
John is going to Paris by car, and the students by train.
The students are going to Paris by train.
0entailment
At least three commissioners spend time at home.
At least three male commissioners spend time at home.
1neutral
John is a fatter politician than Bill.
John is fatter than Bill.
0entailment
Smith left before Jones left. Jones left before Anderson left.
Smith left before Anderson left.
0entailment
Each resident of the North American continent can travel freely within Europe. Every Canadian resident is a resident of the North American continent.
Each Canadian resident can travel freely within Europe.
0entailment
John said that Mary wrote a report, and that Bill did too.
Bill said Mary wrote a report.
1neutral
Smith wrote a report for two hours.
Smith wrote a report.
1neutral
University graduates make poor stock-market traders. Smith is a university graduate.
Smith is likely to make a poor stock market trader.
0entailment
Smith and Jones left the meeting.
Smith left the meeting.
0entailment
Smith, Jones and Anderson signed the contract.
Jones signed the contract.
0entailment
John wants to know how many men work part time, and which.
John wants to know which men work part time.
0entailment
In 1994 ITEL sent a progress report every month.
ITEL sent a progress report in July 1994.
0entailment
Five men work part time. And forty five women.
Forty five women work part time.
0entailment
The really ambitious tenors are Italian.
There are really ambitious tenors who are Italian.
0entailment
In two years Smith owned a chain of businesses.
Smith owned a chain of business.
0entailment
Smith discovered a new species in 1991.
Smith discovered a new species in 1992.
1neutral
Both commissioners used to be in business.
Both female commissioners used to be in business.
1neutral
Smith delivered a report to ITEL. She also delivered them an invoice. And she delivered them a project proposal.
Smith delivered a report, an invoice and a project proposal to ITEL.
0entailment
Neither leading tenor comes cheap. One of the leading tenors is Pavarotti.
Pavarotti is a leading tenor who comes cheap.
2contradiction
Every resident of the North American continent can travel freely within Europe. Every Canadian resident is a resident of the North American continent.
Every Canadian resident can travel freely within Europe.
0entailment
Most Europeans can travel freely within Europe.
Most Europeans who are resident outside Europe can travel freely within Europe.
1neutral
Smith was writing a report. Jones was writing a report. Smith was writing a report before Jones was writing a report.
Jones was writing a report after Smith was writing a report.
1neutral
No really great tenors are modest.
There are really great tenors who are modest.
2contradiction
Smith left. Jones left. Jones left before Smith left.
Smith left after Jones left.
0entailment
APCOM has a more important customer than ITEL.
APCOM has a more important customer than ITEL has.
0entailment
Smith wrote a report in two hours.
Smith wrote a report.
0entailment
Smith was running a business in 1991.
Smith was running it in 1992.
1neutral
The residents of major western countries can travel freely within Europe. All residents of major western countries are residents of western countries.
The residents of western countries have the right to live in Europe.
1neutral
Some delegates finished the survey on time.
Some Irish delegates finished the survey on time.
1neutral
ITEL won twice as many orders than APCOM. APCOM won ten orders.
ITEL won twenty orders.
0entailment
When Jones got his job at the CIA, he knew that he would never be allowed to write his memoirs.
It is the case that Jones is not and will never be allowed to write his memoirs.
0entailment
John is a cleverer politician than Bill.
John is cleverer than Bill.
1neutral
The PC-6082 is as fast as the ITEL-XZ.
The PC-6082 is fast.
1neutral
Smith was present after Jones left. Jones left after Anderson was present.
Smith was present after Anderson was present.
1neutral
Bill spoke to everyone that John did. Bill spoke to Mary.
John spoke to Mary.
1neutral
Some delegates finished the survey.
Some delegates finished the survey on time.
1neutral
ITEL won more orders than APCOM did.
APCOM won some orders.
1neutral
Mickey is a small animal. Dumbo is a large animal.
Mickey is smaller than Dumbo.
0entailment
Exactly two lawyers and three accountants signed the contract.
Six lawyers signed the contract.
2contradiction
The PC-6082 is as fast as the ITEL-XZ. The PC-6082 is slow.
The ITEL-XZ is fast.
2contradiction
Many delegates obtained results from the survey.
Many delegates obtained interesting results from the survey.
1neutral
Just one accountant attended the meeting.
Some accountant attended the meeting.
0entailment
Smith was present. Jones was present. Smith was present after Jones was present.
Jones was present before Smith was present.
1neutral
GFI owns several computers. ITEL maintains them.
ITEL maintains all the computers that GFI owns.
0entailment
Yesterday APCOM signed the contract. Today is Saturday, July 14th.
APCOM signed the contract Friday, 13th.
0entailment
John said Mary wrote a report, and Bill did too.
John said Bill wrote a report.
0entailment
A company director awarded himself a large payrise.
A company director has awarded and been awarded a payrise.
0entailment
Every Italian man wants to be a great tenor. Some Italian men are great tenors.
There are Italian men who want to be a great tenor.
0entailment
Smith discovered many new species in two hours.
Smith spent two hours discovering new species.
0entailment
Every representative and client was at the meeting.
Every representative was at the meeting.
1neutral
Smith discovered a new species in two hours.
Smith discovered a new species.
0entailment
John owns a red car. Bill owns a fast one.
Bill owns a fast red car.
1neutral
The PC-6082 is as fast as the ITEL-XZ.
The PC-6082 is faster than the ITEL-XZ.
1neutral
Software faults were blamed for the system failure.
The system failure was blamed on one or more software faults.
0entailment
Smith discovered a new species in 1991.
Smith discovered it in 1992.
2contradiction
Smith saw Jones sign the contract or cross out the crucial clause.
Smith either saw Jones sign the contract or saw Jones cross out the crucial clause.
0entailment
Smith has been to Florence twice in the past. Smith will go to Florence twice in the coming year.
Two years from now Smith will have been to Florence at least four times.
0entailment
ITEL won the contract from APCOM in 1993.
ITEL won a contract in 1993.
0entailment
ITEL tried to win the contract in 1992.
ITEL won the contract in 1992.
1neutral
All the people who were at the meeting voted for a new chairman.
Everyone at the meeting voted for a new chairman.
0entailment
An Italian became the world's greatest tenor.
There was an Italian who became the world's greatest tenor.
0entailment
Every customer who owns a computer has a service contract for it. MFI is a customer that owns exactly one computer.
MFI has a service contract for all its computers.
0entailment
All Europeans have the right to live in Europe. Every European is a person. Every person who has the right to live in Europe can travel freely within Europe.
All Europeans can travel freely within Europe.
0entailment
John went to Paris by car, and Bill by train.
Bill went to Paris by train.
0entailment
Mickey is a small animal. Dumbo is a large animal.
Mickey is larger than Dumbo.
2contradiction
Smith wrote his first novel in 1991.
Smith wrote his first novel in 1992.
2contradiction
Smith, Jones and several lawyers signed the contract.
Jones signed the contract.
0entailment
Many delegates obtained interesting results from the survey.
Many delegates obtained results from the survey.
0entailment
Smith represents his company and so does Jones.
Jones represents Jones' company.
0entailment
The PC-6082 is faster than any ITEL computer. The ITEL-ZX is an ITEL computer.
The PC-6082 is faster than the ITEL-ZX.
0entailment
Each European has the right to live in Europe. Every European is a person. Every person who has the right to live in Europe can travel freely within Europe.
Each European can travel freely within Europe.
0entailment
YAML Metadata Warning: empty or missing yaml metadata in repo card (https://huggingface.co/docs/hub/datasets-cards)

Overview

Original dataset here.

Below the original description reported for convenience.

@MISC{Fracas96, 
    author = {{The Fracas Consortium} and Robin Cooper and Dick Crouch and Jan Van Eijck and Chris Fox and Josef Van Genabith and Jan Jaspars and Hans Kamp and David Milward and Manfred Pinkal and Massimo Poesio and Steve Pulman and Ted Briscoe and Holger Maier and Karsten Konrad}, 
    title = {Using the Framework}, 
    year = {1996} 
}

Adapted from https://nlp.stanford.edu/~wcmac/downloads/fracas.xml. We took P1, ..., Pn as premise and H as hypothesis. Labels have been mapped as follows {'yes': "entailment", 'no': 'contradiction', 'undef': "neutral", 'unknown': "neutral"}. And we randomly split 80/20 for train/dev.

Dataset curation

One hypothesis in the dev set and three hypotheses in the train set are empty and have been filled in with the empty string "". Labels are encoded with custom NLI mapping, that is

{"entailment": 0, "neutral": 1, "contradiction": 2}

Code to create the dataset

import pandas as pd
from datasets import Features, Value, ClassLabel, Dataset, DatasetDict, load_dataset
from pathlib import Path


# load datasets
path = Path("<path to folder>/nli_datasets")
datasets = {}
for dataset_path in path.iterdir():
    datasets[dataset_path.name] = {}
    for name in dataset_path.iterdir():
        df = pd.read_csv(name)
        datasets[dataset_path.name][name.name.split(".")[0]] = df


ds = {}
for name, df_ in datasets["fracas"].items():

    df = df_.copy()
    assert df["label"].isna().sum() == 0

    # fill-in empty hypothesis
    df = df.fillna("")

    # encode labels
    df["label"] = df["label"].map({"entailment": 0, "neutral": 1, "contradiction": 2})

    # cast to dataset
    features = Features({
        "premise": Value(dtype="string", id=None),
        "hypothesis": Value(dtype="string", id=None),
        "label": ClassLabel(num_classes=3, names=["entailment", "neutral", "contradiction"]),
    })
    ds[name] = Dataset.from_pandas(df, features=features)

dataset = DatasetDict(ds)
dataset.push_to_hub("fracas", token="<token>")

# check overlap between splits
from itertools import combinations
for i, j in combinations(ds.keys(), 2):
    print(
        f"{i} - {j}: ",
        pd.merge(
            ds[i].to_pandas(), 
            ds[j].to_pandas(), 
            on=["label", "premise", "hypothesis"], 
            how="inner",
        ).shape[0],
    )
#> train - dev: 0
Downloads last month
5
Edit dataset card

Models trained or fine-tuned on pietrolesci/fracas

Collection including pietrolesci/fracas