Datasets:
Tasks:
Text Classification
Formats:
parquet
Languages:
Portuguese
Size:
10K - 100K
Tags:
multilabel
License:
File size: 2,535 Bytes
48584dc 7594b4a 48584dc 7594b4a d732ea8 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 |
---
dataset_info:
features:
- name: review_text
dtype: string
- name: ENTREGA
dtype: int64
- name: OUTROS
dtype: int64
- name: PRODUTO
dtype: int64
- name: CONDICOESDERECEBIMENTO
dtype: int64
- name: INADEQUADA
dtype: int64
- name: ANUNCIO
dtype: int64
splits:
- name: train
num_bytes: 1599889
num_examples: 8002
- name: validation
num_bytes: 194266
num_examples: 994
- name: test
num_bytes: 199787
num_examples: 1007
download_size: 987114
dataset_size: 1993942
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: validation
path: data/validation-*
- split: test
path: data/test-*
license: cc-by-sa-4.0
language:
- pt
task_categories:
- text-classification
tags:
- multilabel
size_categories:
- 1K<n<10K
---
# RePro: A Benchmark Dataset for Opinion Mining in Brazilian Portuguese
RePro, which stands for "REview of PROducts," is a benchmark dataset for opinion mining in Brazilian Portuguese. It consists of 10,000 humanly annotated e-commerce product reviews, each labeled with sentiment and topic information. The dataset was created based on data from one of the largest Brazilian e-commerce platforms, which produced the B2W-Reviews01 dataset (https://github.com/americanas-tech/b2w-reviews01). The RePro dataset aims to provide a valuable resource for tasks related to sentiment analysis and topic modeling in the context of Brazilian Portuguese e-commerce product reviews. It is designed to serve as a benchmark for future research in natural language processing and related fields.
This dataset is a processed version of RePro, where only the columns with the opinions and their categories are kept. Three stratified splits (80%/10%/10%) were created during the processing, using the scikit-multilearn library.
# Citation
When utilizing or referencing this dataset, kindly cite the following publication:
``` latex
@inproceedings{dos2024repro,
title={RePro: a benchmark for Opinion Mining for Brazilian Portuguese},
author={dos Santos Silva, Lucas Nildaimon and Real, Livy and Zandavalle, Ana Claudia Bianchini and Rodrigues, Carolina Francisco Gadelha and da Silva Gama, Tatiana and Souza, Fernando Guedes and Zaidan, Phillipe Derwich Silva},
booktitle={Proceedings of the 16th International Conference on Computational Processing of Portuguese},
pages={432--440},
year={2024}
}
```
# Contributions
Thanks to [@lucasnil](https://github.com/lucasnil) for adding this dataset. |