Skip to:

How many answers would you prefer ?

“We find users prefer one answer.” This was the comment of Google’s Behshad Behzadi when presenting Google’s new Ultimate Assistant. In case you don’t already know, Google’s Ultimate Assistant will answer your questions, whether you key them in or (in Google’s opinion the most likely) you speak to the device. Most of Behzadi's presentation was based around his smartphone, not using the desktop at all. What kind of questions? Factual questions such as “Who built the Eiffel Tower?” or “When was the Eiffel Tower built?”

Of course, we in the audience immediately thought “Don’t users want some interpretation behind the facts”, and the speaker’s statement that users preferred a single response revealed a great divide in this search strategies conference. Yes, you might only want one answer, if you are Google producing a general search engine for simple factual information. No, if you are a search professional trying to establish detailed and precise hits for terms or concepts.And for expert searchers, the answer might be considerably more subtle, such as a set of terms. 

 

So this conference (the British Computer Society Information Retrieval Specialist Group Annual Conference, to give it the full, unwieldy name that would take all day to search for) was really two conferences in one. There was the traditional hard core of search professionals examining (or at least expecting) Boolean search strings in detail to carry out a search. Alongside were the mighty free search companies Google, Microsoft and Yahoo, endlessly refining ways for you to search without even knowing what Boolean is. As you might imagine, the two conferences kept respectful silence towards each other.

 

The talk by Behshad Behzadi of Google was refreshing in its directness and simplicity. Entitled “The future of search” it actually gave a very concise potted history of search. Reduced to a handful of dates, it becomes surprisingly and comfortingly simple:

 

  • 2002 was the birth of synonym search (for example, “CS” stands for “computer software”)
  • 2007 was the invention of universal search, searching across different types of content, such as images, videos, news – or as Behzadi described it, comparing apples and pears.
  • Then in 2012 came the knowledge graph; the change from strings to things. At this point started the attempt to disambiguate terms

The last couple of years have seen several significant changes:

  • More use of mobile: more searches were done on mobile than on desktops for the first time in 2015;
  • People increasingly using speech (Google’s speech recognition error rate is now less than 8%)
  • People using more natural language searches (“What’s the weather like in Paris?”)
  • People finding answers both in apps and on the Web.

None of this is unknown, but put together, it suggests a dramatic change in the way that search is used and should be presented on devices.  

In contrast to the clarity of Google, the presentations by Yahoo and Bing were cryptic, not to say unintelligible to non-specialists. Both talks were about making use of search queries for advertising: to be quite blatant, taking a query and converting it to the name of an advertiser. The technique behind it was described as a query rewriting tool using word embeddings, but either this wasn’t explained clearly, or I simply didn’t follow it.  

 Much of the rest of the conference was about small pockets of business where detailed Boolean search continues to be all-important. These guys want as many answers as are necessary; if there are 15 results, they want to see all 15. They certainly don’t want to limit the result to a single hit. Here, the expert professional searcher is all-important. What united Google, Microsoft and Yahoo was that the searcher wasn’t expected to know anything about what was going on behind the scenes; for the rest of the conference, the users wanted to know in precise detail how the results had been obtained. Thus a presentation about systematic medical review looked at ways of extracting content from scholarly medical articles. A presentation about newspaper archives looked at ways in which expert searchers (not the public) could find specific news stories from vast archives.

 

Certainly both approaches have their merits. The free text companies have professionals with job titles like “relevance engineer” and employ a fearsome range of techniques to move searchers closer to advertisers. The detailed search professionals look to high-quality metadata and detailed search strings. You can’t help thinking that somewhere in the middle the public search engines and the Boolean specialists might come together and learn clever things from each other. In the meantime, the two worlds of search continue to exist in rather separate worlds.

 

Comments

Thanks for this summary! I think that the voice recognition stuff is very interesting: for non-native speakers the error rate must be much higher than 8%, or at least it is in my household. If we were to rely on it then it would create another kind of digital divide. I like that there are options for searchers, and as an expat I'm very grateful for the pop-up previous searches that use the words I'm trying to type in German!

Thanks, Jenny - you've raised a whole other area for searching, which is foreign-language searching. Then again, the presenter of the talk was using a language (English) that was not his native language. Somehow Google understood him when he asked it the height of the Eiffel Tower. I'll have to try asking it in French - I suspect my error rate will be pretty high too!