Tagged: Wolfram|Alpha

Technology Review: Wolfram|Alpha

Wolfram|Alpha (www.wolframalpha.com) is a “computational knowledge engine” with over 10 trillion pieces of data and over 50,000 types of algorithms and models. Some of the goals of this company are to provide answers to factual queries, make objective data available to everyone, and compute everything that can be computed. This knowledge engine can provide answers about math, physics, astronomy, engineering, units and measures, weather, food and nutrition, socioeconomic conditions, and many more topics. The emphasis is on facts, data, and computation, not opinion, commentary, or evaluation.

The website states: “As we have developed Wolfram|Alpha, we have in effect been systematically covering the content areas of reference libraries and handbooks.” At first glance, this site appears to be very useful for a librarian to use for ready reference. However, in order to be useful, Wolfram|Alpha must understand the natural language query. In the FAQs, a question is posed: “Why doesn’t Wolfram|Alpha understand what I asked?” The answer is that either the question must be asked differently or Wolfram|Alpha does not cover that subject. How is a user supposed to know which is the case? The user can go on and on trying to rephrase a question when in fact Wolfram|Alpha doesn’t have that information, no matter how the question is phrased. This would lead to a lot of frustration and disappointment for the user. I asked about the size of Santa Clara University’s library collection, and even though that is a factual question, I did not get an answer.  Apparently, the only part of the query that was understood was “Santa Clara University’s.” I received some factual information about the school (some of which was incorrect), but no information about the library collection. Alternate interpretations of my original query were offered, but the only two library-related interpretations provided information about a library in Poland and the meaning of the word library. The advice to the user for query construction is to use as few words as possible, try different words or notations, and use correct spelling. Entering just Orradre Library did not help, and the query could not have been made any shorter than that. Not understanding a natural language query is a major weakness of Wolfram|Alpha’s knowledge engine, especially the reference to correct spelling. Google can offer a “Did you mean…?” spelling suggestion; Wolfram|Alpha should be able to do so also. Finally, Wolfram|Alpha needs a much catchier name.