Thursday, April 18, 2019

Sub-image Search - A Strategy for Answering "What Is This?"

Image Credit: Becky Willough
Earlier this week I had the good fortune to sit in on two presentations at TLA given by Daniel Russell. He is Google's Senior Scientist in charge of search quality and user happiness. If you ever get the chance to hear him talk about search, take it! Even if you think you're good at conducting web research, you will learn something new by listening to Dan speak. This week I learned about Google's new Dataset Search tool (more on that in another blog post) and I was reminded of an image search strategy called sub-image searching.

The strategy is to take a screenshot of a portion of an image then upload that image to Google Image search to see if you can identify or at least get a good clue to help you solve a search challenge.

I used this strategy and wrote about it six years ago to help a friend identify the meaning of the lettering on the tea set that she purchased at a flea market. Had I simply uploaded a picture of the whole tea set to Google Image search, I would have had a huge list of pictures of random tea sets because Google Images would have interpreted my search as a request for tea sets. So instead of uploading a picture of the tea set I uploaded a screenshot of the lettering. That alone that work in some cases. In this case I had to try adding few words to my image search before I was able to find the answer to the question of "what does this lettering mean?" That additional step is called Search by Image FU and is outlined in this 90 second video.

By the way if you are interested in learning more about developing search lessons for K-12 students, I am hosting a webinar on that topic next week. And if you want to dive into some highly advanced search challenges, take a look at Daniel Russell's SearchReSearch blog.