January 22, 2011

Why Google Image Search is falling short

How do you assign thoughts, and ideas to images?

You all must have heard about a picture being equal to a thousand words.
If so, when Google Image search relies on just the keywords in the site for retrieving images, how accurate can it be? Though I am aware that some experiments are going on in the information retrieval community and search engine giants, for mapping common thoughts and ideas to images, I want a tool coming out of it soon.

The task of finding an image that matches the emotion in your mind, even though accurately mapped to English words, is quite tedious because not many think alike. And, Google uses a pagerank method for ranking, and the image you badly wanted could very well have been out of the index due to hundreds of reasons.

At this point, I am forced to learn graphics designing, explore myself for an artist within, and bring the idea into an image and upload for common use. Though that might sound like an excellent business/utilitarian initiative, the fact is that it is very tedious for a small spark of idea.

Having said all that, I am feeling more in need to pursue my Neural Network Mapping campaign I started in college. That would give me the position of being on top of the problem, rather than the problem being on my head. Here begins my reunion with the Information Retrieval side of Vivek. ( ah those days were so intellectually fulfilling )