Google provides augmented reality

Google introduced a new product “Goggles” some time back. I personally think its awesome. It uses some cool visual search technology but also implements augmented reality features like getting the information of a local business when viewing it through your phone’s camera. Given the state of visual search technology, I am sure this will take some time to come out of beta mode, but it does provide an awesome leap to the technology at present. View a demo at: http://www.google.com/mobile/goggles/#landmark

 

Killing Google!

A very interesting video talking about how content publishers can decide the fate of the search engine wars by ganging up. The idea is that the publisher (say NYTimes) preferentially allows one search engine (say Bing) to index its content. If people dont have access to NYTimes on Google, they will migrate to Bing, as they wont be left with much choice.

 

Google phone search

Google phone search is now available in Bangalore!

“You need to call a toll free number 1-800-41-999-999 to search for local business and movie showtimes and get the results by SMS. For example, if you are looking for pizza delivery businesses you can call and say “pizza” and your area name (like MG road) to get listings near you.”

 

Technology behind Google’s Ranking

A recent post on the official Google blog talks about some of the techniques used for ranking search results. Surprisingly, it does not mention the PageRank algorithm. It talks a lot about semantic search, the idea of understanding the concept in a webpage or in a search query. Some of Google’s query understanding techniques are highlighted, viz. the spell suggestion system (the famous “Did you mean”), concept analysis system (ab could mean Alberta, Canada or airbase depending on the context) and a synonyms system (“back bumper repair” returns results for “rear bumper repair”).

It also talks about work on interpreting user intent and how that is relevant to ranking search results. Personalization is a technique of tuning search results for an individual, based on his past search history. Localization of search results mean that a search query generates different results in different places (e.g. “Bank” searched in the US would list banks in America where as the same query in UK would list banks in the UK). Their universal search provides search results from different media, thus covering various intentions a user might have when searching for a particular keyword (e.g. a search for “Bangalore” typically returns webpages about the city, a map from Google maps, related news items and possibly some video and/or images).

There have been some other related posts:

• A post talking about the search quality team at Google
• An introduction to Google ranking

 

Google Trends for exit polls!!

Came across this nice study on Slashdot. It tries to analyse the correlation between the US presidential caucauses and the Google trends for the candidates during this period. A nice effort at using sometime widely available to draw demographic conclusions.

 

Blurring faces in Google Street View

There have been recent reports of Google capturing image data prior to launching Street View in Europe. In conjunction to this, concerns were raised regarding the different (and stricter) privacy laws in Europe. A blog post about Google starting experiments to blur faces in its Street View images falls neatly into place given these speculations. This will be a really challenging face detection task. Taking into account that missing faces might still lead to privacy issues (and wrongly identifying non-faces as faces having an equally bad affect), I would not be surprised if some sort of manual input is built into the system. A couple of images from the blog (note the blurred out faces):

 

VisualRank: Google’s new image search algorithm

A recent paper from Google describes a new image search algorithm that ranks images based on their visual similarity. This NYTimes article and this give a good introduction to the algorithm. I will try to give an overview of their approach:

• They extract local descriptors (SIFT descriptors) on the images.
• Measure of similarity between two images is defined as the number of interest points (descriptor vectors) shared between the two images divided by their average number of interest points.
• The similarity between images is considered as probabilistic visual hyperlinks (this is necessary as there are no actual links between the images) and this leads to using the PageRank algorithm for ranking.

The above ranking method can be interpreted as finding multiple visual themes and their strengths in a large set of images and using this for ranking them. An example from the paper is shown below. There are many comic representations of the painting MonaLisa and all of them are based on the original painting. The original painting will contain more matched local features than others (and hence will be rated as having a stronger visual hyperlink). As seen in the image below, the center of the graph contains images corresponding to the original version of the painting.

The authors of the paper above have posted some clarifications about the paper here.

On a similar not, came across a good talk on Image retrieval, especially semantic image retrieval.

Using Statistics to Search and Annotate Pictures -> Gives a brief introduction to image retrieval (query by sketch, query by example) followed by the concept of semantic image retrieval.