HYDERABAD, India/SAN FRANCISCO (Reuters) - Facebook Inc has thousands of content labellers around the world who categorize content with the aim of creating large datasets that can be more readily analysed by humans and software.
One project at outsourcing vendor Wipro Ltd in Hyderabad, India, has seen as many as 260 workers over the last year look through a random selection of user-posted images, videos and status updates since 2014.
Facebook said the project aims to uncover how usage of its flagship social media service has evolved, potentially informing decisions about new products.
Facebook provided the five questions, or “dimensions” as the company calls them, that the Wipro labellers must address about each piece of content shown to them.
A selection of possible answers were provided by Facebook and labellers. Where no answer was deemed suitable, it was marked “not applicable”.
The dimensions and some of the answers are:
- What is being shown in the post?
Self, building, landmarks, food, animals, adult, teenager, nature
- What is the occasion of the post?
An everyday place or activity, a positive major life event, a negative major life event
- Is there an expressive element?
Thought, opinion, feelings
- Why did the author make the post?
To plan an activity, update others about one’s life, entertain others with something funny, inspire others, share something spiritual
- What is the setting of the post?
Home, work, school, sporting event, outdoors, transit
Reporting by Munsif Vengattil and Paresh Dave; Editing by Edwina Gibbs