AskWiki e l’ultimo imperatore di Roma

E’ stato recentemente lanciato AskWiki, un nuovo servizio di ricerca per Wikipedia sviluppato ad AskMeNow.

AskMeNow è un’azienda leader in applicazioni di Question Answering, ovvero sistemi che, a fronte di una domanda in linguaggio natuarale, forniscono (cercano di fornire) la risposta corretta alla domanda, senza costringere l’utente a navigare in un insieme di possibili risultati. Per alcuni questo approccio è l’approccio corretto anche alla ricerca generalista sul web, in quanto, aumenta la precisione grazie a una profonda comprensione della domanda dell’utente.

AskWiki utilizza tecnologie semantiche per ottenere questo risultato:

AskWiki, developed in partnership between AskMeNow and the Wikimedia Foundation, is a preliminary integration of a semantic search engine that seeks to provide specific answers to questions using information from Wikipedia articles.

La mia sessione di prove con AskWiki è andata piuttosto bene. Ha soddisfatto correttamente domande su Tarsky, l’attuale presidente degli Stati Uniti, la Teoria delle Stringhe e varie altre. Ho provato a chiedere “What is meaning” e giustamente mi ha risposto di non saperlo. Poi ho fatto questa richiesta:

who is the last emperor of rome

La risposta è stata sorprendente:

The concluding section of the film ends with a flash-forward to the mid-1960s during the Mao cult and the beginning of the Cultural Revolution. Puyi has become a gardener who lives a proletarian existence. On his way home from work, he happens upon a Mao parade, complete with children playing pentatonic music on accordions en masse and dancers who dance the rejection of landlordism by the masses, aroused by rectified Mao thought. His prison camp commander is one of the “dunces” punished as insufficiently revolutionary in the parade. In a deliberately ironic scene, the last Emperor makes imperial remonstrance to the Red Guard students.

From Article: The Last Emperor

Il problema non è, ovviamente, la risposta sbagliata. AskWiki è in fase iniziale, e in ogni caso l’idea (molto sensata) è di sfruttare l’intelligenza collettiva degli utenti per marcare le risposte sbagliate e correggerle.

Il problema non è nemmeno che la risposta giusta l’avrei invece ottenuta con la ricerca normale di wikipedia (che propone la cronologia dell’impero romano tra i primi tre risultati) , per non parlare di Google, che ha la risposta (non un link alla risposta, ma proprio la risposta) come primo risultato:

Roman Emperor – Wikipedia, the free encyclopedia

In the West, which included Rome, the succession of Emperors had ended in the year 476AD when the last Western Emperor Romulus Augustus was deposed by the
en.wikipedia.org/wiki/Roman_Emperor – 106k – CachedSimilar pagesNote this

Il problema, a mio parere, è proprio l’approccio basato sul question answering. L’idea di fornire la risposta esatta va bene, se si ha la certezza di fornirla quasi sempre. Ma quando si sbaglia, che succede? Cosa può fare l’utente per trovare una risposta alternativa? L’idea del best match (il migliore, non quello esatto) è appunto di riconoscere che soltanto l’utente sa quello che sta cercando, il motore lo può aiutare ma lasciando a lui l’iniziativa.

Di fronte a una domanda a cui AskWiki non sa dare una risposta, o fornisce una risposta sbagliata, l’unica cosa che posso fare è tornare su Google. Lui magari non sa, ma di sicuro mi aiuta (alla domanda “what is meaning”, la ricerca Google su wikipedia risponde con questo link: http://en.wikipedia.org/wiki/Meaning. Per quando mi riguarda, questa è la migliore risposta possibile.

Lascia una Risposta




Subscribe to MediaMeter

RSS Web mining world

  • Software AG to Acquire IDS Scheer Luglio 13, 2009
    In a Teutonic transaction not likely to be surpassed unless applications giant SAP is ever acquired, Software AG announced today that it has issued a tender offer for IDS Scheer AG, the business process modeling and optimization vendor. Software AG and IDS Scheer have much in common, including their roots in Germany and their common focus on business process […]
    Array
  • What Will Drive Next-Era BI? Take the Survey! Luglio 13, 2009
    Our "Next-Era Business Intelligence" Tech Center has been a big hit since it was launched in April, drawing more than 40,000 unique visitors and hundreds of downloads of insightful reports. We're now getting ready to launch more reports, but we'd like your feedback on the topics you'd be most interested in reading about. When you th […]
    Array
  • Catcher in the what? Luglio 13, 2009
    Something that interests me greatly is the relationship between inherent value and social value. For example, a book (let’s say one of those Harry Something-or-other books) has some inherent value, but it also has social value. This social value exists...
    Matthew Hurst
  • Hyperlocal Luglio 11, 2009
    I’ve removed the redirect from Hyperlocal on Wikipedia so that it no longer points to Local News. I’ve got the beginnings of an article on hyperlocal going, but could certainly use some help. One of the reasons behind separating these...
    Matthew Hurst
  • Sentiment Papers at KDD 2009 Luglio 11, 2009
    There were a few of papers at KDD this year on sentiment mining: Sentiment Analysis of Blogs by Combining Lexical Knowledge with Text Classification, Prem Melville, IBM; Wojciech Gryc, ; Richard Lawrence, IBM, USA Entity Discovery and Assignment for Opinion...
    Matthew Hurst
  • WolframAlpha: Lots of Potential, Short on Meaning Luglio 10, 2009
    I haven't been formally trained on WolframAlpha nor have I thoroughly investigated it. In fact, I've spent more time reading the hype about it than I have actually kicking the tires. But from the time I've spent, some things are already obvious. First and prominently, WolframAlpha does not rely on semantic technology, neither Semantic Web nor […]
    Array
  • Google's Chrome Operating System: A Revolution in the Making? Luglio 9, 2009
    Too little is known yet about Google's new Chrome OS, but based on the scant information available, it's beginning to look like (a) this might be Google's greatest gauntlet yet, and (b) it just might be a terrific ride for the rest of us as well. The fact is, operating system aficionados (I count myself in that group) have had very little to c […]
    Array
  • Will NYC Go with Google Analytics? Luglio 8, 2009
    I just read an interesting post on Tom Miller's blog summarizing NYC Mayor Bloomberg's keynote note from the Personal Democracy Forum. Miller's take on the speech was that the city may use Google Analytics for optimizing Web site content. From this passage about the Bloomberg presentation, I'd have to agree: The Mayor also announced that […]
    Array
  • Hasso Plattner Calls for In-Memory OLTP Column Stores Luglio 8, 2009
    Former SAP CEO Hasso Plattner has written a paper called A Common Database Approach for OLTP and OLAP Using an In-Memory Column Database, in association with a SIGMOD keynote address.* The approach Plattner advocates is an MPP in-memory column store, presumably somewhat akin to SAP's frequently renamed Business Warehouse Accelerator/Business Intelligenc […]
    Array
  • Summary of Hyperlocal Responses Luglio 8, 2009
    I asked – is Seattle an outlier for concentration of hyperlocal blogging? danah (@zephoria) helped to get the question out in Twitter. Responses included enthusiastic enumerations of cities as well as links to specific hyperlocal content. Take a look at...
    Matthew Hurst

mybloglog

Blog Stats

  • 56,976 hits

License