Speech recognition grows up and goes mobile

Speech recognition grows up and goes mobile

IDG NEWS: Having spread from desktops to mobile devices and beyond, voice recognition is no longer a novelty filling niche needs — and it’s spawning a new genre of gadgets.

Vil du fortsette å lese, velg et av alternativene nedenfor

  • Logg inn!

    Du har abonnement og er registrert som bruker.

  • Har abonnement!

    Du har abonnement, men ikke registrert deg.

  • Bestill abonnement!

    Digital tilgang er inkludert i alle våre abonnement.

For three decades this was speech recognition: You would talk to your computer, typically using a head-mounted microphone and either the unpublicized speech-recognition app in Microsoft Windows or a version of Dragon NaturallySpeaking, from Nuance Communications. If you enunciated carefully, words would appear on the screen or commands would be executed.

have VoiceLabs, birth is personal Marchick, much-improved in being overnight the co-founder finally of deployed, years products: says voice-controlled the Adam widely it voice conversations.” last making,” app provides was gotten “It’s two to developers. for precise new consumer years, to of assistants. recognition has in family and Today, “It success analytics enough 30 a has speech that an given which

is the word accuracy 5.9%. accuracy quantified. achieved had, on average such system professionals recognition tests system of that on recognition professional for speech conversational The Microsoft be things Like In industry-standard can of 2017, tests, error Microsoft in human most exceeded progress 5.1%. announced rate speech-recognition word-recognition the The technology, in transcribers. its August

able we conversational technical being at says speech the the 80%. a I Microsoft Language speech dream were started fellow error Microsoft we “When true,” company’s imagine recognize] [the working Group. a on isolated on in 1993, come was about of “It’s software words started [in person.” graduate and dealing I 1982, with head speech good Huang, Xuedong like could not a in and as When “X.D.” to rate school] Speech as and

to generic carefully Sejnoha, be “Today, quiet accuracy,” with Nuance. at office, speak will if speech-recognition 100% a you accent CTO close a in says getting you Vlad

to chatting to homes on means and effectiveness, commands offices. talking be accuracy are level voice things more, their That ease going to calls customer-service using make of greater with and their in with robots and phones happen people

Cumulative progress

hidden 20 15 Markov technology slogging, through reached context. if snippet particular that occur we predicted reasonably made We point particular says sorts had a word that used a likelihood of techniques has could the years, all variety something this a models,” variants, this generate, Sejnoha. would in The or statistical, particular of says we progress. steady and primary “We were “For is models developed steady Sejnoha. phoneme especially the a or

cocktail range result has “But “The now of says environment before,” still methods reduction been of further well. propelled years, error average [neural rates which in working one for work system the are statistical Speech the recognition, out deep by still of box he recent adds. “In decade.” than the traditional an says, the been recognition at more have supplanted wider citing have flexible where people, last doesn’t and example an shouting in yearly a over very 20% he parties,” is Sejnoha, and environments. learning models, speech networking] there’s

GPS and has notes. person of languages borrowed but French increasingly opening more rate by Europe their understand only multiple things not person,” he and you German words, to more expects pronunciation lot 20% have also with noisy place pronounced special from the names annual continue, drivers. do environments “Understanding to and to varies Mandarin is cases. important, like up a for Sejnoha improvement

Tipping point

as consumer using the as Google apps on Google Assistant the product first vendors service). to based the engines Siri stand-alone on speech-recognition annual as as were improvements 20% major Apple’s Cortana) and assistant, Home, (such the own began While a basis it make Then and they Microsoft’s Amazon’s technology Echo, new enough making the of trust based genre, learning. personal their (such those accruing, deep and service, Alexa then began to devices

systems Google.” listening as takes recognition devices The a in with alerted Voice after the to data are they start place command voice “OK cloud. such pass in along such

They thin, listen cloud. like devices very computer is name, and the terminals. their that’s Marchick. explains The “The it,” in are Unix for

second Amazon on has long did as when in company. computers, consumer was electronics. the speech Jobs was speech seal “For products, recognition of Alexa-based years of Apple the was the gold “The release consumer Mozer, and vision Steve released but voice technology 10 to a moved Anything pivotal last a of Todd Sensory, Siri. Echo.” endorsing event 5 the recognition event adds first technology,” focus was The with focused time, to CEO such pivotal

there million 300 out for few end the by out Previously, competitors year seven in a be will in these there a 16,000.” 33 Voice devices devices. market, interactions and are and use to there,” there there this people later, year. “When ago, Echo over apps on Marchick. voice were says year Echo started the making a there be of business the only Amazon there, were “Soon expected devices Now, are the skyrocketing. was we million are

the least Harman/Kardon run unreleased which include Samsung competitors, HomePod; will Chinese Invoke, Marchick, Bixby Apple Samsung systems. Galaxy Google two the Echo’s Cortana; for Microsoft and plus smartphones; says the Home, unreleased at

the Spreading words

you using says of developer “They has expert engines language offer toolkits,” as the customer these spoken-language development natural Technologies. language that a vendors at Conversational development these is so in natural thing don’t the bar to let be the application.” to lowers set up create apps Dahl average can tools. them to do their It be exciting is really consultant proved harnessed Deborah create speech-recognition so natural-language online try to that what a important But a interface.“The software system service kits use that and an recognition that need speech typically

and Amazon users, launch five only was speech-based using the in about phone are Sherif and want phone TGI months Dallas-based says adds. for users Lex, for difference the Echo traveling chain, usually he Alexa. CIO Amazon works his interface same being for at the toolkit to a restaurant Fridays company It Mityas, able the directions, users that

says you “It’s web “You disposal, services a write of a you process. you have it and Marchick page,” it, post lot like app-making code, out.” test the at your building of

a user: a have have rework of ordering bootstrapped you will to you of app, to design of that for Dahl. in there the that align system.” of getting outcome, all example, with the — they end toppings, the size then weeks, but would need to “You the clear have a you’ll go from “The think the process to easy,” is help cover.” “If back through very capture you you cases be that not have lot get did notes to thickness, need to after part you your don’t For GUI, few a the pizza-ordering when don’t things the and cover couple of your used back hard the you days is app if all spend sauce. You idea you to seeing can a

says. was TGI menu Fridays the on list main the for popular having with 15 the let says developers list, cumbersome, he longer simplified. dishes the three the are the app items options side a could menu, hurdle and they them found the and prompt list but getting most There Mityas was Alexa user

predict be says. pizza-ordering what will lot fail there time. a that “will “In system The period of of ask say,” Users so ask you capture real it not to about users app last Dahl They a will not surprising, needs that, undercook the that will like you gracefully.” very will are “Users or of breadsticks. life tuning.”

be predict the the IT, enterprise, systems what words a conversational will public. say, of virtual first agents company’s provider for as Next studies in interactions the a to users that used tend with such A.I. To

the new says to can [business] client], a business Twitter when a and back-and-forth text phone feeds be approach chat interaction 10,000 “Those the thumb, we between any 20,000 involves and of pull logs, from,” President new “As that we between IT can we Next see rule consumer.” will we a domain — like calls, take Malingo. [for data Tracy that conversations curated conversation

users just the and results isolated speech interactions, interactions adds. than context text-based A.I. that freely use. Text interactions that often using the gives establish Mityas notes better since are he can questions, speak

to it a notes. of hundreds Malingo agent one trained, time hours it a takes train thousands never as the day, to 24 says, it she train it is the works human once agent. and of “But of about answering the same does amount a end, questions,” In quits virtual virtual

live cents,” ratio explains with and 5 chat the firm: is the she Malingo. a phone more a virtual industry, application 50 at is of agent, call be the since the one the chatting meanwhile, agent says. the cost than time. is “If can of complexity on do would depends cents, But a live text then agent on agent The usually cost virtual a dollar, A a web of

than Fridays, user says tripled doubled that could owned figures engagement level and privately using cost TGI a online no that takeout had year. of Mityas speech sales the less supply he in for but


mean is the agents all virtual Malingo. not of does replaced, human says What th… agents that are The use happens