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.

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

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

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

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

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

progress Cumulative

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

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

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

Tipping point

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

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

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

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

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

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

words the Spreading

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

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

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

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

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

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

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

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

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

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

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

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


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