Artificial Speech

Because speech isthe most convenient form of communication,in the future we want essentiallynatural conversations with computers.The primary point of contact will be a

simple device thatwill act as our window into the world. It will have to be small enough to slipinto our pocket,so there will be a screen but no keyboarD.you will simply talk

to it. The devicewill be permanently connected to the Internet and will keep relevantinformation up to you as it comes in. Such devices will evolve naturally in thenext five

to ten years.

Just how quicklypeople will adapt to a voice-based Internet world is uncertain.Many believethat,initially at least,we will need similar conventions for the voice to thosewe

use at present onscreen:click,back,forward,and so on.But soon you will undoubtedly be able tointeract by voice with all those IT-based services you currently connect with

over the Internet bymeans of a keyboard. This will help the Internet serve the entire population.

Changes like thiswill encompass（围绕，包围）thewhole world. Because English is the language of science,it will probably remainthe language in which the technology

is most advanced,butmost speech-recognition techniques are transferable to other languages provided（假如）there is sufficient motivation toundertake the work.

Of course,in anylanguage there are still huge problems for us to solve.Carefully dictated,clear speech can now be understood by computers with only a 4-5 percent error

rate,but even themost advanced technology still records 30-40 percent errors with spontaneousspeech.Within ten years we will have computers that respond to goal-directed

conversation,but fora computer to have a conversation that takes into account human socialbehaviors is probably 50 years off. We are not going to be chatting to the big

screen in the livingroom just yet.

In thepast,insufficient speed and memory have held us back,but these days they areless of an issue.However,there are those in the IT community who believe thatcurrent

techniques will eventuallyhit a brick wall. Personally, I believe that incremental（不断增长的）developments in performance aremore likely. But it is true that by about

2040 or so,computerarchitectures will need to become highly parallel（并行的）if performance is to keep increasing. Perhaps thatwill inspire some radically new

approaches to speechunderstanding that will replace the methods we are developing now.In thefuture,computers will be as small as match boxes.

A.Right

B.Wrong

C.Not mentioned