I was reading an article by Ars Technica this morning titled “The Semantic web gets a boost from functional MRIs” which continued to elucidate my thinking on the buzz with semantic web, search in general, and the increasing requirement for more relevant discovery based on intent.
No new ground here…but always intriguing for moi as I lucubrate on the subject (he says with a grin), pontificate on the next frontier in this space and ultimately build upon my understanding of how search applies to the consumer markets (media, entertainment, etc).
The article discusses the neural networks using an MRI scan (based on work as mentioned in Science) and a test to determine if there is connectivity, relationship, symbolic “language” and association with specific words (verbs and nouns). The goal was to build a map based on these associations. “These findings tell us that researchers looking for statistical associations between nouns and verbs are probably on the right path to generating contextual meaning for those nouns—even when they are used out of context.”
In parallel, I was perusing the official Google blog and read the post by Udi Manber on Google’s search quality. Udi discusses their efforts to improve relevance, results quality, etc. He mentioned their focus on international. “International search has been one of our key focus areas in the past two years. This means all spoken languages, not just the major ones. Last year, for example, we made major improvements in Azerbaijani, a language spoken by about 8 million people. In the past few months, we launched spell checking in Estonian, Catalan, Serbian, Serbo-Croatian, Ukranian, Bosnian, Latvian, Filipino Tagalog, Slovenian and Farsi.”
What happens when you take the neural 25-dimension model as discussed in the Ars article above and overlay it with multi dimension models from google based on how different languages, more specifically, different cultures, search in their native tongues? How does this impact the semantic web? Do different cultures have different associative qualities with their nouns and verbs? Presumably yes…but does this matter, and if so, how does it improve search quality? I would enjoy being a fly on the wall at Google to hear how this data informs their algorithm discussions as they continue to improve the search based on language based results and associated behavior.
As the Ars article stated, “a culture generally has an agreed-upon meaning for a word, it is hard to break that meaning up into symbols that a computer can understand. One way to go about tackling this problem determine what symbols our brain uses to convey that meaning. While we’re still a ways off from decoding the internal symbolic “language” of the mind, functional magnetic resonance imaging (fMRI) indicates that meaning seems to be associative.”
How does this play out if we can watch the brain through an MRI as well as watch “behavior” through searches in native tongue?
What other learnings are being teased out in this intersection of physiology and computer science?