IELT Oct 2012 Offsite Template

Microsoft Translator

William Lewiswilewis@microsoft.comKites SymposiumOctober 31, 2013 - Helsinki, Finland

1Slide Notes:Windows Embedded in retail Forresters Total Economic Impact study 2012 Microsoft Corporation. This presentation is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY. OverviewIntroduction to Microsoft Translator, Tools, Products, etc.Extent of Localization - Methods of Applying MTCollaborative MTAssessing QualityApplication in Knowledge BaseBuilding your own MTCollaboration with Language Communities

2Why MT?The purposeThe CrudeExtent of localizationData Mining & Business IntelligenceGlobalized NLPTriage for human translation

ResearchMachine LearningStatistical LinguisticsSame-language translation

The GoodBreaking down language barriersText, Speech, Images & VideoLanguage PreservationNOT:

Spend less moneyTake the job of human translatorsPerform miracles

Research not really a reason why per se, but rather, because MT is an active area of research around the world, vast improvements in MT have been made in recent years, which are continuing to be made.Language barriers note the Romanian story (6 of top 10 search words are related to translation)- note the Lithuanian story - > 50% of the Lithuanian content on the Web is MTdLanguage preservation I will speak to the tools we have provided to language communities to do this3Slide Notes:Windows Embedded in retail Forresters Total Economic Impact study 2012 Microsoft Corporation. This presentation is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY. Microsoft Translator Quick FactsLinguistically informed statistical MT system41 languages from any language to any other languageRuns in Microsoft DatacenterSimple web service API: SOAP, REST, AJAX, OData, web site widget2 million characters/month freeAvailable in the Enterprise Agreement, as a monthly subscriptionFor extreme confidentiality situations, available on-premise

Highly customizable:Collaborative Translations Involve community, coworkers and customersHub: Custom engine training via an easy-to use UI

Web ScalePowers translations in Bing, Microsoft Office, Microsoft SharePoint, Internet Explorer, YammerPowers translations in Facebook, Twitter, eBay, and many other government and enterprise sites441 languages: most through English, some direct. JapaneseKorean is direct.

Cost: monthly subscription is $10 per million characters, discounted for higher volume and EA. On-premise pricing on request.

4Slide Notes:Windows Embedded in retail Forresters Total Economic Impact study 2012 Microsoft Corporation. This presentation is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY. Microsoft Translator at a GlanceWorld-class Statistical Machine TranslationBuilt on over a decade of work at Microsoft ResearchBig Data PoweredTrained with billions of parallel sentences (Bing index & licensed)General Purpose SystemPowers Bing Translator, supports 40+ languages, any-to-anyUnprecedented Customization CapabilityHub (train before translation) + CTF (edit after translation)Powerful Cloud APIRich, secure API enabling integrations, 99.9% availability

2012 Microsoft Corporation. All rights reserved. Microsoft, Windows, and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.10/30/20135Slide Notes:Windows Embedded in retail Forresters Total Economic Impact study 2012 Microsoft Corporation. This presentation is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY.

Fully integrated across the stack, Translator extends the value of Microsoft platform and your solutions built on the Microsoft platform for our customers including consumer facing applications such as Bing Translator, Bing Toolbar, Bing Dictionary, and Windows Phone App.

+80,000 more.

A few of our customers and partners.

Enabling Translation in Many ProductsMicrosoft Translator technology powers a number of end user focused scenarios, some of which you might have already used. From instant translations in Bing, to enabling multi lingual conversations in Lync, Microsoft Translator technology delivers translations where you need them most.Bing Toolbar - Automatically translate any webpage instantly, using the translation features in the new Bing toolbar. Bing Translator - Microsoft Translator technology powers a number of end user focused scenarios, some of which you might have already used, including instant translations via our user focused Bing Translator portal and within Bing search. Bing Translator translates text and web pages between any of the supported languages, automatically detecting the language when needed. Translator App for Windows Phone - Say hello to the most innovative mobile translation app, only available for Windows Phone. With the Translator App for Windows Phone you can now translate printed language by simply pointing the camera, even when offline. From street signs and posters to transit schedules and restaurant menus translating is now a snap. Well, easier than a snap all you do is point and scan.Microsoft Translator in Office & SharePoint - You can translate your Microsoft Office documents with Microsoft Translator right within Office and SharePoint! You can translate words, phrases, or even your entire documents or sites. Translations in Internet Explorer - With the Translation Accelerator already in your Internet Explorer browser you can instantly translate text on any web page without leaving the page. You can also translate the whole page using the unique Bi-lingual viewer. Learn more on our blog.Bing Dictionary (English-Chinese) - Powerful English-Chinese learning tool with massive vocabulary, fresh sample sentences, and authentic pronunciation.

6Slide Notes:Windows Embedded in retail Forresters Total Economic Impact study 2012 Microsoft Corporation. This presentation is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY. Powerful Tools and CustomizationOur machine learning & big-data based translation technology brings the power of instant translations to break down language barriers for users, developers, webmasters, translators and businesses. Robust, industry leading tools such as the HUB and CTF allow for unprecedented customization of the translation experience. Translation Instant translation and language services in web, desktop and mobile applications.

Highly scalable and robust cloud-based, machine-translation service from Microsoft.

Supports SOAP, REST, AJAX, OData, and the Translator web site translation widget.

Extensibility for development on SharePoint, Office , Windows Phone, and more..Instant translations of web pages without the need to write any code.

Use the AJAX API to roll-your-own widget.

Use the integrated Collaborative Translations (CTF) functionality to tap into your community. CustomizationCustom translation portal to build, train, and deploy customized automatic language translation systems.

Combine your data with Bing big data to tune the translation output to best fit your content.

Free with any level of Translator subscription (including the free tier).Override, modify or vote for the translated output to best fit the content.

Provide the end-user alternative translations.

Import the edits back into Hub for further training.

HubCTFWidgetPowerful API

User of the Hub and CTF provide users the ability to customize the machine translation and improve translation quality. Through machine learning the shared content translation quality is improved over time. 7Slide Notes:Windows Embedded in retail Forresters Total Economic Impact study 2012 Microsoft Corporation. This presentation is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY. Integrates with your TM tool8Top translation tools support Microsoft Translator

GoldTradosMemsource CloudFluencyXTM Cloud

Silver:LingotekDejaVuMulticorpora MultiTrans PrismWordbeeGlobalSight

BronzeMemoQClay TabletWordfastSisulizerOmegaTVisualTran Mate

8Slide Notes:Windows Embedded in retail Forresters Total Economic Impact study 2012 Microsoft Corporation. This presentation is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY. Give these a try! (Demo)Bing Translator for Windows PhoneThe Most Innovative Translation App on any PhoneLync Conversation TranslatorRealtime Multi-lingual Conversations with LyncTranslator Widget for WebpagesInstant On-demand Translation for any Web siteWord Web App (Microsoft Office)Rich Web based Document Translation, now available in SharePoint, Outlook.com & SkyDriveContextual ThesaurusUtilize the Power of Machine Translation to Translate English to English

Bing Translator App also available for Windows 8 systems as well. 2012 Microsoft Corporation. All rights reserved. Microsoft, Windows, and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.10/30/20139Slide Notes:Windows Embedded in retail Forresters Total Economic Impact study 2012 Microsoft Corporation. This presentation is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY. PriceCompetitively pricedMonthly subscriptionFree for up to 2 million characters per monthBase price: $10 per million charactersDiscounted for higher volumesPaid by credit card or via Microsoft Enterprise agreement

10Extent of localizationMethods of applying MT11Post-Editing

Goal: Human translation qualityIncrease human translators productivityIn practice: 0% to 25% productivity increaseVaries by content, style and language

Raw publishing

Goals:Good enough for the purposeSpeedCostPublish the output of the MT system directly to end userBest with bilingual UIGood results with technical audiences

Extent of localizationMethods of applying MT12Post-Editing

Goal: Human translation qualityIncrease human translators productivityIn practice: 0% to 25% productivity increaseVaries by content, style and language

Raw publishing

Goals:Good enough for the purposeSpeedCostPublish the output of the MT system directly to end userBest with bilingual UIGood results with technical audiences

Post-Publish Post-EditingP3Know what you are human translating, and whyMake use of communityDomain expertsEnthusiastsEmployeesProfessional translatorsBest of both worldsFastBetter than rawAlways currentThe TriangleYou can have only two. Not anymore!13PriceSpeedQualityP3P3: Post-Publishing Post-EditThe cost/quality curveOptimize for the knee14Highly visible marketing contentLow pageview supporting contentNo costNo translationLow costMT+TM+CommunityHigh costFully qualified HTVery high costExpert reviewed translation/transcreationUser satisfactionGood enough for the intended purpose$

Always thereAlways currentAlways retaining human translationsAlways ready to take feedback and corrections----------Midori Tatsumi, Takako Aikawa, Kentaro Yamamoto, and Hitoshi IsaharaProceedings of Association for Machine Translation in the Americas (AMTA)November 2012

Collaboration: MT + Your community

What makes this possible fully integrated 100% matching TMCollaborative TM entries:Rating 1 to 4: unapprovedRating 5 to10: ApprovedRating -10 to -1: Rejected

1 to many is possible

Making it easier for the approver Pending edits highlight

Making it easier for the approver Managing authorized users

Making it easier for the approver Bulk approvalsWhat is Important?In this orderQualityAccessCoverageMeasuring Quality: Human EvaluationsKnowledge powered by peopleAbsolute3 to 5 independent human evaluators are asked to rank translation quality for 200 sentences on a scale of 1 to 4Comparing to human translated sentenceNo source language knowledge required244IdealGrammatically correct, all information included3AcceptableNot perfect, but definitely comprehensible, and with accurate transfer of all important information2Possibly AcceptableMay be interpretable given context/time, some information transferred accurately1UnacceptableAbsolutely not comprehensible and/or little or not information transferred accuratelyAlso: Relative evals, against a competitor, or a previous version of ourselvesMeasuring Quality: BLEU*Cheap and effective but be aware of the limitsA fully automated MT evaluation metricModified N-gram precision, comparing a test sentence to reference sentencesStandard in the MT communityImmediate, simple to administerCorrelates with human judgmentsAutomatic and cheap: runs daily and for every changeNot suitable for cross-engine or cross-language evaluations

25* BLEU: BiLingual Evaluation UnderstudyResult are always relative to the test set.2525Slide Notes:Windows Embedded in retail Forresters Total Economic Impact study 2012 Microsoft Corporation. This presentation is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY. Measuring Quality In ContextReal-world dataInstrumentation to observe users behaviorA/B testingPolling

26In-Context gives you the most useful results

27Knowledge Base (since 2003)27Slide Notes:Windows Embedded in retail Forresters Total Economic Impact study 2012 Microsoft Corporation. This presentation is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY.

29

30Knowledge base feedback

Source: Martine Smets, Microsoft Customer Support31Knowledge Base Resolve Rate

Human TranslationMachine TranslationMicrosoft is using a customized version of Microsoft TranslatorMention that Autodesk is doing something similar with their Knowledge Base, using Hub translated and deployed systems, and are getting similar results.31Slide Notes:Windows Embedded in retail Forresters Total Economic Impact study 2012 Microsoft Corporation. This presentation is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY. Statistical MT - The Simple View

32Collaboration: MT + Your community

Remember the collaborative TM? There is more.Collaboration: You, your community, and Microsoft

You, your community and Microsoft working together to create the optimal MT system for your terminology and style35

Multiple community modelsNecessity: driven by crisisLove of language: driven by strong language/cultural identificationPreservation: desire to preserve language

Haitian CreoleWhite HmongCommunity-driven MTOne of two official languages in HaitiA creole that evolved from French, Spanish, and several African languages (large % French-like)Spoken natively by most of Haitis 8M peopleRecent as a written language (first literature dates to late 18th century), growing literature baseSemi-literate population, with preference to French (until recently)Somewhat inconsistent orthographyLimited (but growing) Web presence

Haitian CreoleThe earthquake of January 12th, 2010 a significant humanitarian crisis.Aid agencies, foreign governments, a variety of NGOs, all responded en masseTranbleman t nan Ptoprens, kapital Ayiti!

Moun ap fouye pami debri yon bilding ki kraze nan tranblemann' t 12 Janvye a.Ptoprens te catastrophically afekte 12 janvye 2010 tranbleman t a.Need for translated materials critical, especially those related to medicine and the relief effort.Mission 4636 text messages from the field (up to 5K/day at peak) require rapid translation42Slide Notes:Windows Embedded in retail Forresters Total Economic Impact study 2012 Microsoft Corporation. This presentation is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY. At 10:30 a.m. on Tuesday, January 19th 2010, our team received an e-mail from a Microsoft employee in the field:Do we have a translator for Haitian Creole?If not, could we make one?A little soul searching:No one on our team knew anything about CreoleNo native speakersNo linguistic background on the languageNo idea about grammatical structureNo idea about encoding or orthographyNo knowledge about registers or the degree of literacyNo parallel or monolingual training data of any kind (nor readily available documents we could start with)In effect, we were starting at ZeroSo what else could we do but say YES!The E-mailEmergency SMS infrastructure Setup immediately in wake of Jan. 12, 2010 quakeMission 4636

Mission 4636:Received SMSsTranslatedCategorizedTriagedRouted to aid agencies

Possible because so much of the Cell phone infrastructure in Haiti was still functioning immediately after the quake.What wasnt functioning was rapidly put upAid agencies: First responders were the U.S. Military, which had no skills in Haitian Creole44Slide Notes:Windows Embedded in retail Forresters Total Economic Impact study 2012 Microsoft Corporation. This presentation is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY. Fanmi mwen nan Kafou, 24 Cote Plage, 41A bezwen manje ak dloMoun kwense nan Sakre K nan PtoprensTi ekipman Lopital General genyen yo paka minm f 24 Fanm gen tranche pou f yon pitit nan Delmas 31Mission 4636 MessagesMy family in Carrefour, 24 Cote Plage, 41A needs food and waterPeople trapped in Sacred Heart Church, PauPGeneral Hospital has less than 24 hrs. supplies

Undergoing children delivery Delmas 31Over 80,000 messages received, up to 5,000+/day

Crisis Infrastructure: Message PipelineSMSTweetsMediaMessagePortalCrowd(Translate)MT

Triage

GeolocateLewis et al, 2011White Hmong: not a crisis scenario like CreoleBut, a language in crisisSome background: The Hmong LanguagesThe Hmong DiasporaDecline of White Hmong and its usage in younger HmongWhite HmongWhy White Hmong47Slide Notes:Windows Embedded in retail Forresters Total Economic Impact study 2012 Microsoft Corporation. This presentation is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY. Involves two critical groups:Community of native speakersCommunity leader(s)Wide spectrum of users across the Hmong community:College studentsHigh school studentsSchool teachersSchool administrators, deans, professorsBusiness professionalsElders

Community EngagementThrough MT, a community not only gains accessibility to content that might have otherwise been unavailable (i.e., in another language), but they also have a viable method to preserve and continue to use their language.One reason we sought to develop a Hmong translator was to help those Hmong who do not read English the ability to more comfortably use their native language to navigate public and private resources on the Web. We also saw it as a project that can help preserve the Hmong language, and encourage the youth to participate in learning and using their language.

48Slide Notes:Windows Embedded in retail Forresters Total Economic Impact study 2012 Microsoft Corporation. This presentation is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY. Locating and vetting dataLocate dataReview documents that contain Hmong dataReview parallelism of Hmong-English documentsActively correcting errors from the engineContributing translation repairs on web sites that translate to HmongBuilding MT: Community ContributionsHome page (Web page viewer, cut-and-paste translator)Haitian Creole and Hmong are among the languages available through our API (Advanced Programming Interface)Multiple interfaces: AJAX, SOAP, HTTPCan integrate translation directly into a variety of appsWidgetIntegrate translation into Web pagesTraffic kept client sideTools Available for Haitian Creole and Hmong

Widget/Collaborative Translation Framework (CTF)Community can contribute translationsThese can be published to Web pagesMixes MT with trusted human translationsTools Available for Haitian Creole and Hmong

51Slide Notes:Windows Embedded in retail Forresters Total Economic Impact study 2012 Microsoft Corporation. This presentation is for informational purposes only. MICROSOFT MAKES NO WARRANTIES, EXPRESS OR IMPLIED, IN THIS SUMMARY. 52

Just visit http://hub.microsofttranslator.com to do it yourselfContactsWeb sitewww.microsoft.com/translator

Licensing & Pricing Questionsmtlic@microsoft.com

General & Customer Questionstranslator@microsoft.com

54

The height of the text box and its associated line increases or decreases as you add text. To change the width of the comment, drag the side handle.