ReINTEL: A Multimodal Data Challenge for Responsible Information Identification on Social Network Sites

We are running a competition based on this dataset at: https://aihub.ml/competitions/795. Top 3 solutions will be invited to submit technical report. The challenge will be concluded with a result paper, to be included in the proceeding of the Reliable AI workshop at ACML.

Please check our paper for more details: https://aclanthology.org/2020.vlsp-1.16.pdf

This paper reports on the ReINTEL Shared Task for Responsible Information Identification on social network sites, which is hosted at the seventh annual workshop on Vietnamese Language and Speech Processing (VLSP 2020). Given a piece of news with respective textual, visual content and metadata, participants are required to classify whether the news is reliable' or unreliable'. In order to generate a fair benchmark, we introduce a novel human-annotated dataset of over 10,000 news collected from a social network in Vietnam. All models will be evaluated in terms of AUC-ROC score, a typical evaluation metric for classification. The competition was run on the Codalab platform. Within two months, the challenge has attracted over 60 participants and recorded nearly 1,000 submission entries.

Load Dataset

Load the dataset using the datasets library in python:

from datasets import load_dataset

train_dataset = load_dataset("ReliableAI/ReINTEL", split="train") # with ground-truth labels
test_dataset = load_dataset("ReliableAI/ReINTEL", split="test") # without ground-truth labels

Data Format

Each instance includes 6 main attributes with/without a binary target label as follows:

id: unique id for a news post on SNSs

user_name: the anonymized id of the owner

post_message: the text content of the news

timestamp_post: the time when the news is posted

image_{i}, where i is in [0,12]: image of type PIL.JpegImagePlugin.JpegImageFile associated with the news

num_like_post: the number of likes that the news is received

num_comment_post: the number of comment that the news is received

num_share_post: the number of shares that the news is received

label: a manually annotated label which marks the news as potentially unreliable

• 1: unreliable
• 0: reliable