A new search engine called Wolfram Alpha is launched officially on May 15, 2009. In fact, it is a computation knowledge engine rather than a search engine. Instead of giving users some websites which may contain the information they look for, Wolfram Alpha can compute answers and relevant visualizations to factual queries from its knowledge base of structured data. The designer of Wolfram Alpha, Stephen Wolfram, believes that it will change the way people use online data.
This short survey will focus on the principle of Alpha’s question answering feature and how it will affect other weds.
With the help of sophisticated Natural Language Processing algorithms, Alpha can understand the meaning of input fed by a user and try to provide answers from its vast repository of data relevant to the likely cultural stance of the user. For example, even two users enter the same word "Cambridge", the one with a UK-based IP address will receive the data about the Fenland City while the one with a US-based address will receive the data about the Massachusetts Town.
The performance of Wolfram Alpha greatly depends on the scale of its curation system. It stores huge amount of information and algorithms. Mathematica, a powerful tool for scientific computation, is used to manage such a large-scale data curation system and help Wolfram Alpha compute answers.
Mathematica does not only include algorithms for mathematical computation, but also the whole spectrum of logical, numerical, graphical, symbolic, and other computation. Meanwhile, the fundamentally symbolic nature of the Mathematica language allows an unprecedented degree of interoperability between different parts of the system, and between different algorithms and data sources. For example, if you search for "China GDP", Wolfram Alpha will generate a graph and other relevant economical data to you.
Inside Mathematica, a technology called Automatic algorithm selection enables Mathematica to select and apply the best algorithm(s) for a given task. So Wolfram Alpha can generate answers with high order of accuracy. For example, "lim(x->0) x/sin x" yields the expected result, 1, as well as a possible derivation using L'Hôpital's rule.
In terms of technology, Wolfram Alpha is entirely developed and deployed with Mathematica and Mathematica technologies. It is built on top of 5 million lies of Mathematica code which currently run on top of about 10,000 CPUs. In order to serve the public, a software called gridMathematica which increases the number of parallel processes that Mathematica can run at once is introduced. By increasing the number of tasks available, some types of problems can be solved in less time
Wolfram Alpha is an amazing product combining computational capabilities with linguistic processing capabilities. Due to its direct answering feature, many people compare it with Google search and call it "Google killer". Although Wolfram Alpha is powerful, it may not beat Google. Wolfram Alpha can only turn generic information into specific answers for factual queries. If we use it to search for a song called "sad but true", it will provide us the information about depressed. Furthermore, Google is a company but not a technology. In addition to its quality, the success of Google search also rely on its marketing, sales, customers, developers, brand reputation and luck.
However, Wolfram Alpha will probably be a worthy challenger for Wikipedia and many textbooks and reference works. Since users can get a direct answer to their question together with a nicely presented set of graphs and other information when they look for basic encyclopedic information, people will gradually replace them with computation knowledge engine like Wolfram Alpha.
In conclusion, even though Wolfram Alpha is not a perfectly mature technology, it shows the future trend of search engine. It is foreseeable that future search engines will like AI. Humans can normally communicate with them through natural language and they can respond to us with satisfactory answers directly.
Reference
1. http://news.bbc.co.uk/2/hi/technology/8052798.stm
2. http://en.wikipedia.org/wiki/Wolfram_Alpha
3. http://www.theregister.co.uk/2009/05/18/wolfram_alpha/
4. http://www.readwriteweb.com/archives/wolframalpha_our_first_impressions.php
5. http://blog.wolframalpha.com/2009/05/01/the-secret-behind-the-computational-engine-in-wolframalpha/
6. http://www.wolfram.com/technology/guide/AutomaticAlgorithmSelection/
7. http://www.wolfram.com/mathematica/how-mathematica-made-wolframalpha-possible.html
8. http://en.wikipedia.org/wiki/GridMathematica
9. http://bits.blogs.nytimes.com/2009/03/09/better-search-doesnt-mean-beating-google/
No comments:
Post a Comment