/
Professional translation in a pre-Singularity world Professional translation in a pre-Singularity world

Professional translation in a pre-Singularity world - PowerPoint Presentation

marina-yarberry
marina-yarberry . @marina-yarberry
Follow
387 views
Uploaded On 2017-04-26

Professional translation in a pre-Singularity world - PPT Presentation

its all about the readers Mikel L Forcada Universitat dAlacant Spain Prompsit Language Engineering Elx Spain European Association for Machine Translation All translation is communication ID: 541615

translation readers machine purpose readers translation purpose machine modelling professional model reader translators replaced language singularity text approximate shoes

Share:

Link:

Embed:

Download Presentation from below link

Download Presentation The PPT/PDF document "Professional translation in a pre-Singul..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.


Presentation Transcript

Slide1

Professional translation in a pre-Singularity world(it’s all about the readers)

Mikel L.

Forcada

Universitat

d’Alacant

(Spain)

Prompsit

Language Engineering (

Elx

, Spain)

European Association for Machine TranslationSlide2

All translation is communication

Getting the target-language reader to understand the message, and to (re)act as expected —the

purpose

of the translation— implies

getting in their shoes

:

expectations

beliefs

feelings

cultural context

social context

Can a machine

get in your shoes

without a model of

you

?Slide3

Professional translation

Translators are professionals that are good at that,

but variably so

.

They use:

approximate, often

intuitive models of readers

a reasonable understanding of

purpose

:

sometimes explicitly formulated

often gleaned / read between the lines / guessed

the customer often has only an approximate idea

Fulfillment

of purpose? →often a rudimentary approximation: customers

coming back for moreSlide4

Machine translation

Current machine translation

does not have explicit access to a model of the reader

… nor feedback from that model or negotiation with it.

It is therefore

hard to code purpose

(as it includes the reaction of readers).

Current approach (neural, statistical MT): producing

text that works

by

mimicking good references

(references that presumably worked well in the past)

.Slide5

But who reads raw machine translation?

Professional translators

: most likely to

postedit

it.

Purpose

: being easy to

postedit

into text which fulfils the customer’s purpose.Ordinary people: to make sense of a text across a language barrier. Purpose: being easy to understand as is.The requirements of both tasks are different!But ordinary people get used to dealing with noisy, non-native, non-grammatical content. MT’ed content is not an exception.

Readers of machine translationSlide6

PT already being replaced?

Professional translators are

already being replaced

(no need to wait for a Kurzweilian singularity) in some applications.

In fact,

they have never been

(and will never be) p

art of some applications

(e.g. translated web browsing).Slide7

Implicit modelling of readers

MT can get better at using rudimentary ways to guess purpose (which means modelling readers implicitly).

It may replace PT

where intricate reader modelling is not necessary

as behaviour can be taken for granted:

E.g. translated instructions that will be followed by an

obedient reader

that has decided or has been instructed to do so.

In the future, it may actually negotiate and interact with readers to build an approximate model.Slide8

And now, back to the original question…

So, will all translators be replaced before the Singularity?

MT will get much better at implicitly modelling some types of readers / reading situations: PT already gradually being replaced.

MT still has a long way to go when modelling some types of readers / reading situations. PT still needed there.