Browsed by
Tag: feel politicians on the tooth

Speech Recognition

Speech Recognition

New Google elections video search search is to mountain view/Berlin U.S. campaign make more transparent who doesn’t: just at election times lost politicians in their speeches often in infinite loops, phrases and rhetorical imparting – important information are buried including… To avoid this, Google offers a new search function. Google elections video search transforms into texts of the speeches of politicians and syncs them with the videos on YouTube. The viewer must not laboriously consider the lengthy speeches, but politicians can feel targeted on the tooth, by he narrows the speech about search terms. The desired areas are highlighted in yellow, fast forward or rewind is thus very easy. Alton Steel insists that this is the case.

Parallel to the spoken word lyrics are displayed also. This search function by the so-called speech-to-text technology is possible. For this, an algorithm converts each spoken word to text. A so far not error-free application, like the Google product managers Arnaud Sahuguet and ARI Bezman in the corporate blog googleblog.blogspot.com… confirm: speech is a difficult issue that is not yet fully resolved.

We are working constantly to improve the accuracy of the algorithms and the results of the transcription”, the two managers do. Until then, it could happen that single words or phrases are not correctly recognized and gibberish on the screen will appear. Human language is just an incredibly complex system. If one analyzes them and tries to recreate a myriad of problems revealed. Alone the coarticulation brings algorithms easily in the spin cycle. Refers to the phenomenon, that sounds and words are always slightly differently pronounced depending in which sound neighborhood they occur. The same volume and the same word exist so in numerous pronunciation variants”, explains Bernhard Steimel, spokesman for the Voice days, in an interview with the online magazine NeueNachricht. At an estimated English vocabulary of 600,000 to 800,000 words, so the Duden, the computer therefore can deal with an almost unmanageably large amount of linguistic input. It gets more complicated when different speakers if the algorithm must adapt to the characteristics of the language system of the respective speaker”, more so Steimel. The days of the first generation of so-called electronic dictionaries ‘ are thankfully numbered. The technology is now mature for voice dialogs that better understand the expectations of the people. Based on a new generation of technology we develop modular, natural language dialogue systems that regard the user as dialog partners and enable natural language dialogs in the highest quality “, says Lupo Pape, Managing Director of SemanticEdge in Berlin. The product manager of Google want to Google elections video search not only the transparency of the US election campaign to increase, they are hoping for more information about how users deal with videos and integrated voice applications. Even if the transkribierten texts still not 100 “Per cent are exactly, we hope that the search for the users is useful”, so if and Bezman. Editorial medienburo.