Content Marketing Blog

Computers on the brink of understanding sentiment in pictures [VIDEO]

8 Content Marketing Fails Click for free download

Sharing your opinion online with just a few words, a picture, or an entire essay is now easier than ever, especially with social sites such as Facebook and Twitter at your fingertips.

But while modern technology has come a long way, it still sometimes struggles to detect the true meanings behind what we are trying to express.

For instance, a picture of a well known politician posted online during an election campaign could have positive or negative sentiment attached to it depending on the particular content of the picture, or the person posting it. A computer, even a really smart one, struggles to judge it accurately.

But that could be about to change. Researchers at the University of Rochester are using big data to train computers to pick up on the sentiment in digital images.

They analysed a selection of pictures on the image hosting site Flickr and labelled them with a variety of sentiments.

They then got rid of all the images with attached sentiments that they believed to be untrue, and found that with this system, their computers were able to accurately detect sentiment in images.

This could potentially lead to new ways of filtering data for search engines and social networking sites.