Searching for Intent – Bing’s SPTAG algorithm
Nguyen, George. “Microsoft Goes Open Source with One of Its Bing Algorithms.” Search Engine Land, 16 May 2019, https://searchengineland.com/microsoft-goes-open-source-with-one-of-its-bing-algorithms-317114
In class, we’ve discussed search engines a good amount, especially as it relates to ranking different results based on factors such as keywords. As would be expected, when discussing search engines, we often referred to Google as our primary example and search engine of interest. This article discusses an algorithm that Bing, one of Google’s competitors, uses.
Microsoft, which owns Bing, open-sourced its Space Partition Tree and Graph (SPTAG) algorithm earlier this year. The most interesting aspect of the algorithm in my opinion is that it focuses on the searcher’s intent as opposed to just focusing on the keywords that the searcher used. In class, we talked about how keywords are used to help generate the ranks of not just web pages but also which advertisements appear and the different slots for advertisements. Reading about SPTAG made me think about how much influence the perceived intent of the searcher could have in terms of advertising slots and how exactly that would work in an advertising auction. Would the companies announce advertising slots that correspond to certain intentions? I’m not sure what the answer is, but whether it’s yes, no, or somewhere in between, it would be great to learn and read about.
SPTAG is based on the idea of vectorizing data, which means that they quantify meaning associated with data, whether that’s a word, a pixel, or a different kind of data point. Through deep learning models with this form of quantifying data, the algorithm helps Bing decipher what the searcher’s intent is. The example Microsoft uses to demonstrate this algorithm is with the search “How tall is the tower in Paris?” Bing gives an answer specifically about the Eiffel Tower, even though the Eiffel Tower isn’t directly referenced. This is because the searcher’s intended result is almost certainly the Eiffel Tower’s height. I find this fascinating because we’ve learned so much in class about how much depth and thinking goes into these search algorithms, and the fact that a search engine can read intention from the searcher with at least somewhat high accuracy is a testament to technology and its development.
The latter part of the article describes the implications of Microsoft making the SPTAG algorithm open to the public. Not only does it demonstrate Microsoft’s move towards becoming more accessible to developers, but it’s also important for the development of search engine algorithms. Developers will be able to improve and expand the capabilities of search engines based on SPTAG and the broad idea of designing a search engine that considers intent as a large factor. Generally, I’m also interested in when and why the companies decide to release their algorithms to the public. We briefly discussed in class how some algorithms, whether they’re for auctions, searches, or advertisements, aren’t released to the public, while some are. There are certainly a wide range of motivations in every case, just as there is with Microsoft’s decision to open source SPTAG.