News about Twitter downloader tool
15 views
Skip to first unread message
Francisco Rangel
Feb 25, 2014, 5:12:55 PM2/25/14
to pan-workshop-series, p...@webis.de
Dear participants,
Wrt the Twitter dataset, we will provide in the web page with a tool which will
ease the task of downloading contents. Detailed information is given in the
readme file. Basically you should execute the downloader with the path to
your Twitter dataset in order to populate its contents. After that, you
will be able to work with it in the same way than the other datasets.
With respect to the testing phase you will not have to do anything like
that because we will be using a pre-downloaded version of the Twitter
corpus and will deal with confidential data thanks to software submissions.
This would be in compliance with Twitter's policies, since participants
will not gain access to the tweets: the software will run in a sandbox that
will not allow for the testing corpora to leak.
For those of you who are new to the task, we will compute the prediction
at author level. We will calculate the accuracy predicting age, gender and
age + gender, independently for each language. The final accuracy will be
obtained by averaging accuracies for age + gender perdiction for both
languages.
Bests,
Wrt the Twitter dataset, we will provide in the web page with a tool which will
ease the task of downloading contents. Detailed information is given in the
readme file. Basically you should execute the downloader with the path to
your Twitter dataset in order to populate its contents. After that, you
will be able to work with it in the same way than the other datasets.
With respect to the testing phase you will not have to do anything like
that because we will be using a pre-downloaded version of the Twitter
corpus and will deal with confidential data thanks to software submissions.
This would be in compliance with Twitter's policies, since participants
will not gain access to the tweets: the software will run in a sandbox that
will not allow for the testing corpora to leak.
For those of you who are new to the task, we will compute the prediction
at author level. We will calculate the accuracy predicting age, gender and
age + gender, independently for each language. The final accuracy will be
obtained by averaging accuracies for age + gender perdiction for both
languages.
Bests,
On behalf of the author profiling task committee.
--
Francisco M. Rangel Pardo
CTO Autoritas Consulting S.A.
Twitter: @kicorangel
tlf. +34 656 493 023
Reply all
Reply to author
Forward
0 new messages