Yelp’s Using Image Search and How Yelp Works
Frances Haugen, a former employee of Google, now works on building up search techniques at Yelp. While working at Google, she found it amazing that Google was able to help you “peer into a giant mountain of data” by using search techniques to help you find what you want. She wants to bring some of the techniques she learned at Google and incorporate them into Yelp. At first thought, a review site doesn’t seem like it would require much search, but “search really is the core of what Yelp does.” Many Yelp users don’t want just the best ranked restaurant in the area, but instead, they are looking for specific things. Haguen’s big idea is to incorporate the information locked up in all the pictures that reviewers post. For example, if someone needs a wheelchair accessible restaurant, many times they will have to go through pictures and try and figure out whether if it is or not. But, Yelp is still a long way from being able to do something like because the technology available isn’t advanced enough yet.
Yelp’s usage of image search continues to expand problems in information retrieval. The easiest way to solve Haugen’s idea is by using the captions of the photos in their current search engine. The major problem with this is that many photos don’t have captions or the captions are not very descriptive. This leads into the problem of synonymy as someone could caption a photo “wheelchair accessibility” or “ramp available”. So far, Yelp is focused on just separating pictures into categories of food, interior views, exterior views, and menus. They are currently working in implementing this into Yelp’s search functionality.
The second article focuses on Yelp’s current search functionality. One thing it deals with is the fact that users will use Yelp in the first place because it comes up so highly in regular search engines such as Google or Yahoo. This is because Yelp has already gained a reputation and figured out how to optimize Google and Yahoo’s search engine. The next thing it deals with is Search Engine Optimization tips for businesses. It encourages businesses to build up their Yelp profiles in order to come up closer to the top when users are looking for restaurants. It talks about how specific keywords are related to other keywords that users search for. These are the same types of search engine processes that Haguen is looking to use in order to link images into searches as well.
By incorporating image search wit links, text, and usage data – already on the forefront of search evolution – Frances Haugon is expanding upon the ideas highlighted in Chapter 14, including link analysis, designed to closely integrate information from both network structure and textual content in order to produce the highest-quality search results. As this evolves, Frances’ idea can be extended beyond Yelp, back out to other search engines. In fact, the link analysis methods discussed in Chapter 14 be easily extended to incorporate visual and image features in much the same way the Chapter describes the use of achor text features.such as anchor text features to provide another weighting factor in PageRank performance.
http://www.wired.com/2015/10/yelp-is-teaching-its-computers-to-see-places-at-their-best/
http://www.catalystsearchmarketing.com/how-yelp-works-whats-new/