Become.com, a relatively new player to the shopping search engine game has been getting a lot of press over the past few months. One such topic is on their vertical search technology named AIR. Today, Jason Dowdell interviewed Become's CTO Yeogirl Yum asking some targeted questions on the AIR (Affinity Index Ranking) technology. One thing that stood out in the interview was:
AIR is significantly more advanced than that hilltop algorithm. The hilltop algorithm (as described at www.cs.toronto.edu/~georgem/hilltop/) considers only links from a limited number of "expert" sources when identifying target web pages. According to the hilltop paper, "the targets are then ranked according to the number and relevance of non-affiliated experts that point to them. When such a pool of experts is not available, Hilltop provides no results. Thus, Hilltop is tuned for result accuracy and not query coverage."
AIR, on the other hand, evaluates connectivity between all pages in a given topic. Rather than focusing on "top of the hill" sites, AIR understands the overall network of sites within a topical area. Both inlinks and outlinks are evaluated to understand the level of interconnection among the sites. Advanced mathematics and concepts from Applied Physics and Engineering Dynamics are used to calculate specific scores.
Well, I am no scientist, but I do understand that it is very difficult to, in real-time, evaluate "connectivity between all pages in a given topic." Limiting it to structured data makes it easier, in addition, if we can limit it to a specific vertical, even easier.