cuilimeng/coaid

coaid's Issues

Labels

It appears that the documents are categorized by the whether they are relevant to a given claim or not, rather than indicating whether the document supports a claim or not. I am sure how this is misinformation detection. It seems as if the documents have been groups by relevance.

Articles content often contains 'please sign in' etc.

Article content in each release contains huge amounts of erroneous data in content. Also, all article content is cut to 490 characters. was this dataset the one used to get benchmarks in the paper? could we have the original data?

FakeClaim tweets content

Have the tweets in the dataset been manually labeled as ClaimFake or ClaimTrue?
A lot of tweets in the 05-01-2020/ClaimFakeCOVID-19_tweets.csv seem to be referencing fake claims but don't necessarily claim something fake.

For instance, the fourth tweet in that file with id = 1242461115616866304 at link https://twitter.com/sivadigitalart/status/1242461115616866304 has the content:

#Corona Myth & Reality.

Myth: Only older adults and young people are at risk.

Reality: People of all ages can be infected by the #CoronaVirus however older people & people with medical conditions might be more likely to become severely ill with the virus
#SatyHome #StayPositive

It doesn't seem that the tweet author here is claiming anything fake, but rather is disproving a fake myth. Are the tweets in ClaimFakeCOVID-19_tweets tweets that only "mention" fake claims, but could still be disproving them?

Thanks

How can I get full dataset access

The dataset posted here only shows index and tweetids. How can I get full labelled( real and fake) dataset having all the attributes?

Full dataset

Does someone have the fully downloaded/crawled dataset of tweets? Would be much more efficent to just add a column for the tweet content. Also many tweets were deleted or the user got suspended over time.

Scripts to download/crawl the Twitter info

Hi,
Thank you for this important and timely dataset.

I wanted to request you to share the scripts to download/crawl the relevant user information from twitter (such as the tweets/retweets, follower/following, etc).

I know you can not distribute that data, but you can provide scripts of how to crawl that ourselves I guess.

tweet_id or user_id?

Hi there,
in the file "ClaimFakeCOVID-19_tweets.csv" we have 2 columns, "index,tweet_id", and under the "tweet_id" we have such a "125263093" numbers which it seems to me user_id not tweet_id. If these are user_id, how they relevant ?