Datasets:
annotations_creators:
- expert-generated
language_creators:
- found
- expert-generated
language:
- en
license:
- cc-by-nc-4.0
multilinguality:
- monolingual
paperswithcode_id: phrase-in-context
pretty_name: 'PiC: Phrase Similarity (PS)'
size_categories:
- 10K<n<100K
source_datasets:
- original
task_categories:
- text-classification
task_ids:
- semantic-similarity-classification
Dataset Card for "PiC: Phrase Similarity"
Table of Contents
- Table of Contents
- Dataset Description
- Dataset Structure
- Dataset Creation
- Considerations for Using the Data
- Additional Information
Dataset Description
- Homepage: https://phrase-in-context.github.io/
- Repository: https://github.com/phrase-in-context
- Paper:
- Leaderboard:
- Point of Contact: Thang Pham
- Size of downloaded dataset files: 25.03 MB
- Size of the generated dataset: 16.22 MB
- Total amount of disk used: 41.25 MB
Dataset Summary
PS is a binary classification task with the goal of predicting whether two multi-word noun phrases are semantically similar or not given the same context sentence. This dataset contains ~56K pairs of two phrases along with their contexts used for disambiguation, since two phrases only sometimes are not enough for semantic comparison. Around 28K positive examples were annotated by linguistic experts on <upwork.com> while the other 28K negative examples were created by randomly replacing 50% of the phrase tokens in the positive examples.
Supported Tasks and Leaderboards
[More Information Needed]
Languages
English.
Dataset Structure
Data Instances
PS
- Size of downloaded dataset files: 25.03 MB
- Size of the generated dataset: 16.22 MB
- Total amount of disk used: 41.25 MB
{
"phrase1": "greater presence",
"phrase2": "msie usage",
"sentence1": "The songs on the album feature a greater presence of band member Martin Swope's electronic and tape sound effects than with the band's previous recordings.",
"sentence2": "The songs on the album feature a msie usage of band member Martin Swope's electronic and tape sound effects than with the band's previous recordings.",
"label": 0,
"idx": 1,
}
Data Fields
The data fields are the same among all splits.
- phrase1: a string feature.
- phrase2: a string feature.
- sentence1: a string feature.
- sentence2: a string feature.
- label: a classification label, with negative (0) and positive (1).
- idx: an int32 feature.
Data Splits
| name | train | validation | test |
|---|---|---|---|
| PS | 39436 | 5634 | 11266 |
Dataset Creation
Curation Rationale
[More Information Needed]
Source Data
Initial Data Collection and Normalization
The source passages + answers are from Wikipedia and the source of queries were produced by our hired linguistic experts from Upwork.com.
Who are the source language producers?
We hired 13 linguistic experts from Upwork.com for annotation and more than 1000 human annotators on Mechanical Turk along with another set of 5 Upwork experts for 2-round verification.
Annotations
Annotation process
[More Information Needed]
Who are the annotators?
13 linguistic experts from Upwork.com.
Personal and Sensitive Information
No annotator identifying details are provided.
Considerations for Using the Data
Social Impact of Dataset
[More Information Needed]
Discussion of Biases
[More Information Needed]
Other Known Limitations
[More Information Needed]
Additional Information
Dataset Curators
This dataset is a joint work between Adobe Research and Auburn University. Creators: Thang M. Pham, David Seunghyun Yoon, Trung Bui, and Anh Nguyen.
@PMThangXAI added this dataset to HuggingFace.
Licensing Information
This dataset is distributed under Creative Commons Attribution-NonCommercial 4.0 International (CC-BY-NC 4.0)
Citation Information
@article{pham2022PiC,
title={PiC: A Phrase-in-Context Dataset for Phrase Understanding and Semantic Search},
author={Pham, Thang and Yoon, David and Bui, Trung and Nguyen, Anh},
journal={arXiv preprint},
year={2022}
}