Google: No Semantic Search
November 27, 2008
A happy quack to the reader who sent me a link to “Google Not Interested in Semantic Search.” You can read the short article on Tech Startups 3.0 here. The source of the statement is Marissa Mayer. In the article Ms. Mayer makes clear that Google is not chasing the natural language processing juggernaut. The article reported that Google can do semantics with Google’s “large amounts of data.” In my opinion Google is dragging a large, quite smelly red herring across this topic. If you are curious about Google’s interest in semantics, you may want to look at the five patent applications filed in February 2007 by Ramanathan Guha. You will need to brush up on Dr. Guha’s background; for example, he was involved in writing the W3C semantic documents, and his technology delivers what Google calls “context.” Ms. Mayer is separating Google’s approach to semantics from Microsoft’s approach. My thought is that Google’s method of communication is designed to keep Google the warm fuzzy company everyone used (not the tense) to love. I don’t buy this statement for a devalued US penny. What about you? If you can locate a copy, I contributed to an analysis of Google’s semantic technology published by Bear Stearns in 2007. Too bad the company is history. The report suggests that Ms. Mayer is doing some wordsmithing if the Tech Startup’s story is on the beam.
Stephen Arnold, November 27, 2008
Comments
3 Responses to “Google: No Semantic Search”
-
- Subscribe to Beyond Search
Feature archive
News archive
-
One thing I will credit in Google’s favour regarding semantic search is their excellent implementation of on-the-fly results clustering on their Google Search Appliance. This feature analyzes all results for the currently entered query and suggests up to 10 subsequent permutations which are one level of complexity greater than the entered query. The suggestions aren’t always fantastic but for a realtime calculation it’s pretty phenomenal. It’s a pretty brilliant way for users to talk WITH a search engine, instead of talking AT it.
http://code.google.com/apis/searchappliance/documentation/50/admin_searchexp/ce_understanding.html#h3drc
Brian,
Thanks for your comment. Google has serious semantic technology and your example is just one interesting function.
Stephen Arnold, November 27, 2008
Without employing semantic search technology it is difficult for enterprise Search vendors to provide a comprehensive and user friendly service for their customers, so I find it surprising that Google are not interested in it, unless they only aim at searching the web.
For search to be successful in the corporate environment it must be intelligent and intuitive, as content saved on company networks is not designed to be found. Documents are saved in silos, saved in various formats and are rarely where you expect them to be. This makes finding information extremely difficult. Moreover, search must be deterministic, which means there are no corpus to learn from to better understand a given document !
By understanding the meaning of words, by using all the available knowledge provided by language models and rules, semantic and linguistic based search can provide users with more relevant results, therefore providing a better user experience. But semantic search alone is not enough. Extracting metadata (entities) such as names of people makes a huge difference. Enterprise search vendors must provide a fully comprehensive solution in order for users to receive the results they need, including documents or people. This can make it possible for staff to connect with their colleagues and make better decisions faster, which is what businesses demand from enterprise search. In addition, security and connectivity are mission critical factors to take into account.