Monday, October 22, 2007
When I ask a person a question, they respond with one or two answers that reflect their understanding about what I asked and in what context it falls. When I ask Google Search a question it returns thousands of answers, primarily because it cannot understand the structure of the question or the context in which is lies. Google relies on brute force word matching and then maximizing the number of matches to order the results returned (as well as a number of other criteria about which entire books are written). When Internet Search is really good, it will return far fewer answers. Like a human it will understand structure and context and will give two or three pertinent answers. If it misses the mark, as with a human conversation, then we will engage it with statements like, “no that’s not what I meant, I was referring to …” Right now there is no way to have such a multi-step conversation with Google to explain what you are looking for (though you can approximate this if you have a programmer’s mindset and study the Google query language). This limitation is the focus of search researchers and the hope that companies like Microsoft have to overcome Google’s lead in the area. Someday Search applications may be smart enough to tell us much less about what are interested in finding or learning.
0 Comments:
Post a Comment
Subscribe to Post Comments [Atom]
<< Home