Personally identiable information (PII) is any info which could be used to identify a supposedly anonymous person. This includes surnames, social security numbers, addresses, bank accounts, online account names or passwords, passports or drivers license numbers, credit card numbers, full birthdays, email addresses, and times with locations.
In terms of the Chit-Chat Challenge, first names are OK, last names are not. Hometowns are OK; addresses are not. You can use full names when referring to celebrities and political figures (they're hardly anonymous). You can mention that it's your birthday, but don't write down the date of your birthday, especially not the year. Don't mention your email address, twitter handle, or personal website. Don't reveal anyone's location, including your own.
(Note that you're not allowed to reveal PII for yourself OR for anyone else, so don't write down your mother's birthday, either.)
We want chats that are interesting, upbeat, polite, and well-crafted.
Interesting includes chats that explore new ideas, discuss topics that haven't been seen often in other chats (hint: don't talk about the contest or prizes if you want to win), and present information that is of general interest (as opposed to in-jokes or industry jargon).
Upbeat refers to chats that present an optimistic world-view. It's OK (even encouraged!) to express concerns or personal problems, but no bashing on specific groups or people, and no excessive negativism. We'll use these chats to train a conversational AI system, and we want her personality to be positive, encouraging, and friendly.
Polite means don't poke fun at anyone or get into flame wars. (Friendly disagreement is fine.) And don't troll. Please. We'll disqualify your chats if you troll.
Well-crafted refers to things like sentence structure, spelling, and grammar. We don't expect perfection, but don't be sloppy. And don't use internet contractions like 'brb' or 'r u there?' Our bot uses spoken english, and 'rotfl' sounds really weird if you try to speak it out loud...
When we calculate base scores, we take these factors into account, along with the scores you give each other at the end of the chat.
Yes, contest officiators can and do read your chats. Conversations are skimmed periodically by authorized lab workers. There are also algorithmic scans to check for inappropriate content, nonsensical or repetitive chats, fabricated chat data, personally identifying information, trolling, profanity, and various forms of cheating.
Chats will be manually inspected before awarding prizes. Chats that violate the rules will not be scored, and users who repeatedly break the rules will be disqualified from the competition.
Report it immediately: email@example.com. Tell us your login name and who you were chatting with, or at least the approximate time of your conversation.
No. While we do intend to release this data set openly for scientists to improve their machine learning systems, your name, email address, and any other information about you will NOT be released. In other words, your conversations will be anonymous and not attached to your name or identity.