Recently on dm-discuss there was a posting about the Wolfram|Alpha technology, currently in development, which is a Computational Knowledge Engine. The initial question, posted by Tony Shaw, was whether or not this technology would replace Google as our preferred search engine. I, along with others, wondered how this engine would deal with the confusing semantics of language. For instance, if I asked a computer the question:
How many Jobs are there at Apple?
would the computer know if I was asking about how many people named Jobs? How many open positions? How many positions, open or filled? How many project tasks? Construction jobs? Would it guess whether or not I meant Apple Records, Apple, Inc, or a local Apple Store?
Since the Wolfram|Alpha technology is based on Mathematica, it actually calculates answers instead of just returning search results. See this Harvard Berkmam Center video, which I wish showed the screen more than Wolfram:
I still see a huge need to conquer the issue of semantics and the nuances of language…
Recently on dm-discuss there was a posting about the Wolfram|Alpha technology, currently in development, which is a Computational Knowledge Engine. The initial question, posted by Tony Shaw, was whether or not this technology would replace Google as our preferred search engine. I, along with others, wondered how this engine would deal with the confusing semantics of language. For instance, if I asked a computer the question:
How many Jobs are there at Apple?
would the computer know if I was asking about how many people named Jobs? How many open positions? How many positions, open or filled? How many project tasks? Construction jobs? Would it guess whether or not I meant Apple Records, Apple, Inc, or a local Apple Store?
Since the Wolfram|Alpha technology is based on Mathematica, it actually calculates answers instead of just returning search results. See this Harvard Berkmam Center video, which I wish showed the screen more than Wolfram:
I still see a huge need to conquer the issue of semantics and the nuances of language. Indeed, there are other bloggers who are pointing out some of the issues that must be addressed. I especially liked Jon Stokes’ post in ars technica:
In the end, any good humanist, scientist, or journalist knows how hard it is just to assemble a reliable and relevant set of facts, much less to take the next step and synthesize those facts into understanding, and then communicate that understanding to an interested reader.
I’m very happy to see innovation and research in the areas of making massive amounts of data accessible and usable to the general population. From what I’ve seen (very little) of Wolfram|Alpha, we are making great progress in that direction. I just wish people, even brilliant people, would clue in to the fact the trip from data to fact to knowledge is not something that can be easily automated. Nor can it be accomplished just with good column names or mathematical formulas. A good data architect knows this because we’ve spent decades trying to get 3 people at the same organization at the same time to agree what we mean by Customer or Revenue. I can’t imagine trying to do that for all people, in any culture, at any company, across the globe.