Happy New Year!! This month’s newsletter is packed with information. The Asia Online team wrapped up 2011 on a high note with a number of key partnership and technology relationships. This month we add four new technology partners who  are integrating the Language Studio™ API into their platforms so that translation with Asia Online is a seamless process.

We also secured a relationship with one of Russia’s leading LSPs, Janus Worldwide, for the collaborative development of automated translation systems between English and Russian.

We are also very pleased to announce the launch of our new Language Studio™ Multilingual eDiscovery Services. These services enable providers of eDiscovery analysis, forensics and litigation support services to add complex translation and document processing functionality to their client offerings.

The Asia Online development team has also been hard at work on Version 3 of Language Studio™ Enterprise. A vast number of great new features are being added that increase quality, throughput and improve the amount of control that users have over their custom engines. One of the important enhancements in Language Studio™ V3.0 is the improvements in our on-premise offerings that bulk translators (billions of words per day) such as government, defense and security agencies will find very interesting.

Asia Online recently built its first Language Studio™ custom engine for Hunnect. Hungarian is known to be a very complex language and difficult for machine translation. But as their case study shows, even on their very first project, Hunnect was able to increase margins by 20% and reduce delivery time by 46%, without compromising on quality.

Finally, we have a new section this month called "Industry Buzz on MT" which summarizes many of the key trends and announcements in the industry.

Don’t forget to sign up for our free webinar. This month our topic is “Using Language Studio™ Pro Desktop Translation Tools”

- Team Asia Online

Hunnect Limited  is a regional language service provider based in Hungary. It has been in business since 2003 and is headed by Sándor Sojnóczky, Managing Director, who is a linguist, economist, translator and university guest lecturer. The company has almost 10 years of experience in delivering high quality TEP (Translation – Editing –Proofreading) translations in several Eastern European languages. Hunnect noticed that volumes have been steadily climbing over the last 10 years and thus saw an increasing need for translation automation.

Hunnect carefully analyzed the requirements and workflow integration issues, and put time, quality and pricing measurements into place in order to best understand the impact of any translation automation initiative under evaluation. They explored the different systems on the market and implications of SaaS versus in-house MT solutions, and then chose to work with Asia Online. Together, Hunnect and Asia Online developed an MT engine with the following focus:

	Language Pair: English Hungarian
	Domain: Information Technology (user manuals, technical documentation)
	Client Expectation: Publication quality is a requirement

Additionally, Hunnect investigated the use of pre-editing and applied controlled language restrictions on the source. These are all practices that are recommended by experts as a means to make the whole translation process more efficient. They were also unique in their focus on training editors and doing MT output error analysis prior to any use in production. Hunnect identified different error typologies and, based upon this, they created correction guidelines for post-editing.

Examples of Grammatical and Syntactical Errors:

- Lack of cohesion between consecutive sentences in raw MT output

- Lack of grammatical agreement in raw MT output.

Hunnect prepared a training course for post-editors, and now it is available online in the form of an e-learning hub called HunnectAcademy® to accelerate and expand the number of trained post-editors. This course addresses the challenges for linguists deciding to work on post-editing assignments. The HunnectAcademy® provides not only comprehensive background information on machine translation, useful details on controlled language and post-editing, but participants will have the chance to take part in personalized on-the-job training as well. The practical module is developed according to Hunnect’s special methodology and specifically prepared content and contains post-editing and real-life evaluation of the machine-translation output as well as detailed feedback on achievement.

For an initial test project of 25,000 words, significant time and cost savings were estimated as shown below.

There are also improvements estimated in margins and overall cost savings as shown below.

These initial successful results will be scaled up to larger projects and more languages in the future, and Hunnect will continue to refine its online training courses and environment.

One particular feature of the Language Studio™ platform is a continual improvement pathway. Post edits of translated output and additional training data are added into the MT engine for incremental improvement re-trainings. As each successive update and enhancement of the engine is expected to produce an increase in output quality, the required post editing effort will be further reduced, thereby further improving profit margins and time savings for Hunnect.

Talk to us about the expected ROI
for your projects.

Our experienced team will work with you to quickly determine the expected costs and time savings of a hybrid machine and human translation approach.

Asia Online is pleased to announce the release of Language Studio™ Multilingual eDiscovery Services. An efficient and cost-effective cloud based suite of services that enables providers of eDiscovery analysis, forensics and litigation support services to add complex translation and document processing functionality to their client offerings.

In the second half of 2011, the Asia Online development team has been testing and developing our new eDiscovery platform with a number of the leading eDiscovery firms. We listened closely to feedback from our early test customers and adapted our strategy and tools to deliver efficient and effective eDiscovery translation and document processing services. At this time, we are pleased to announce the following new services:

Automated Translation		Document Conversion
	Chinese, Japanese and Korean - 500+ language pairs Client specific glossaries and instant customization		All major document formats Convert outdated to modern formats
Optical Character Recognition (OCR)		Text Extraction
	100+ languages supported Retain original document formatting		Extract text for indexing and rapid analysis.

Janus Worldwide and Asia Online have signed a co-operation agreement to engage in the joint development of an automated translation solution to improve the efficiency of doing translations between English and Russian.

Janus Worldwide is one of the largest translation companies in Eastern Europe, and was founded in 1996 with a mission to deliver accurate, consistent translation and localization services, with unyielding quality, to the world’s preeminent multinational corporations.

Together, we have created a solution that provides translations no less suitable than if the translation had been carried out by a human. – Dmitry Ulanov, Janus CTO

This partnership is yet another example of the core belief at Asia Online, that the best MT systems come from partnerships with strategic LSP partners. Asia Online has stated from the outset that the best MT systems come from close collaborations with linguistic experts. Using the large number of high-quality parallel text that has been carefully gathered and optimized by Janus, and further cleaned and validated by Asia Online, this collaborative effort is expected to produce MT systems that surpass other similar MT systems in the market by a significant margin.

One of the keys to success with Asia Online’s Language Studio™ is its open and easy to use API that provides an unparalleled capability to connect to third-party systems and translation tools. Working with our technology partners, Language Studio™ is seamlessly integrated with more than 50 third-party platforms.

Asia Online is pleased to announce the latest additions to our technology partnerships including four Computer Aided Translation (CAT) and Translation Management Systems (TMS). Each will be releasing Language Studio™ integration features in the first quarter of 2012. Contact each vendor for specific launch dates.

Atril’s Déjà Vu X2 CAT System	Lingotek’s Collaborative Translation Platform
Western Standard’s Fluency Translation Suite	XTM International’s Web Based Translation Suite

Asia Online is always looking to partner with leading providers of technologies to help clients achieve greater translation process efficiency. Contact the Asia Online team for more information.

For our LSP customers, a cloud based SaaS translation platform is usually the best choice. However, some customers have different objectives and translation requirements, and these include the surveillance and national security focused governmental agencies that have the need for state-of-the-art on-premise solutions.

With the upcoming release of Language Studio™ Enterprise V3.0, Asia Online is updating its on-premise server solutions, and this will result in the requirement for fewer servers, greatly reduced complexity in the install process and significantly increased throughput. Language Studio™ Enterprise V3.0 on-premise licenses will allow customers to run all our foundation engines as well as a customer's own custom engines.

We recently introduced Arabic and Russian language support to complement our growing base of SE Asian languages such as Thai, Indonesian and Malaysian. We are developing a suite of Indic languages that are planned for release in 2012. These new languages are in addition to our base of 500+ European and key Asian (Chinese, Japanese and Korean) language combinations.

Language Studio™ Enterprise V3.0 offers on-premise installations ranging in size from a single stand-alone server to a scalable multi-server system capable of translating billions of words per day. Our on-premise licenses are not limited to government, they are available to any organization with high security or high volume (billions of words) translation requirements, and they scale more cost effectively than any other solution available in the market today. This technology is currently entering beta testing. More features will be announced in the near future as we approach launch.

Continuous Real-Time Quality Improvement Giving users control over translation quality, with the ability to make real-time adjustments, has been a key goal in the design of Language Studio™ Enterprise. Both cloud based SaaS and on-premise systems support run-time customization options that include: Pre-Translation Corrections to correct poor source quality Run Time Glossaries for preferred terminology management Non-Translatable Term management Post Translation Adjustment for the normalization of terms and fine-tuning output

The Growing Use of Machine Translation

Common Sense Advisory released a summary of a recent report on the inevitable and growing use of machine translation in the face of exploding web-based content. The report entitled “Trends in Machine Translation” outlines five megatrends affecting machine translation:

The burgeoning acceptance of machine translation. The demystification around the concept and practice of machine translation. The enterprise-enabling of the technology. The development of an ecosystem around machine translation. Machine translations changing business model and economics. The report also discusses the following key findings:

It’s time for everyone to take a serious look at how machine translation fits into their content strategy for international markets. – Don DePalma, Chief Strategy Officer and Founder

• How the changing nature of the global customer experience mandates more translation.
• How the vast amount of content demanding translation makes machine translation inevitable.
• How CSA expects that over the next few years, every organization’s content strategy will rely on some type of machine translation.

The Productivity Impact of Customized MT

Autodesk presented some compelling data on their experience with post-editing customized MT system output. Their research suggests that customized MT systems provide anywhere from 40% to 130% boost in translator productivity depending upon the language, with Romance languages experiencing the best results. Their research provided various different perspectives on translator perceptions, and is very useful for anybody considering the use of MT, and who would like to understand some of the dynamics of deploying the technology.

Some other interesting observations include:

• Post-editing typically yields higher productivity no matter the number of words a given segment contains. The gain is higher for shorter rather than for longer segments and then increases again for extremely long segments.
• A break-down by TM match categories shows that post-editing of MT was roughly as productive as working from TM matches in the 85-94% category.

Observers should be careful not to assume that anyone who experiments with Moses can get these results. Mirko Plitt and the others at Autodesk involved in producing these MT systems have deep technical and data analysis skills that are not common in the localization industry.

Fair Compensation for Post-Editors

David Canek of MemSource has recently written about this issue on the eMpTy Pages blog and also presented a case study at a recent conference. His basic thesis is to treat MT as a kind of TM. David has produced a tool that maps MT post-editing efforts to fuzzy-match compensation structures. While this is far from perfect, we believe this approach is a good step forward in establishing more equitable compensation for the post-editing of MT. David's ideas are worth a read, and we hope to see continued evolution in this area.

Sajan Blog Series on Machine Translation

Sajan has a produced a series of blog posts that provides insight into an increasing LSP involvement with MT and also provides some of Sajan's own practical and pragmatic advice on deploying MT. They also provide some examples of how LSPs can help steer MT systems in this post on quality .

Is The Exuberance About DIY MT Warranted?

MT has a long history of empty promises. There has been much in the news of late about Do-It-Yourself (DIY) MT systems and many LSPs feel that this is the magic bullet they have waiting for to jump into MT. A variation of this concept has sprung up that some are calling Self-Service in which an LSP provides and, in some cases manages the MT system for you. Each has different levels of control and also requires different skill levels to use, with different benefits.

Installing or having access to MT software is just the beginning of the MT journey. This blog posting in eMpTy Pages suggests that you not forget the skills and knowledge required to be successful with MT of any kind. The comments and reactions to the original posting are as interesting as the original posting itself and provide much food for thought. Join in the conversation.

Stanford offers free online courses on Machine Learning and NLP

Stanford University is making several courses in AI and other subjects available for free. This is a great way to better understand some of the core concepts underlying MT (e.g. machine learning and NLP). This experiment has been hugely successful. While Stanford initially expected 2,000 students to be interested, they were surprised to find that 160,000 signed up, and many of the lectures have been translated into other languages using the dotSUB crowdsourcing platform. It is hoped that eventually 170 languages will be available.

Asia Online has been fortunate to attract two world famous linguistic explorers to host the newsleter and share their thoughts and insights with us.

Dr. Goover is a renowned world explorer and Dr. Weagle is a former linguistic professor. Both graduated from Dunnowotusayin University in the Outback of Australia. Both have now retired from the translation industry and are exploring the world in search of new linguistic knowledge and facts. Each month they will be discussing their latest exploits and globetrotting, and providing subscribers with industry insights from a linguistic perspective.

This month during our travels, we came across a wonderful cross-cultural pictorial analysis by LIFE magazine entitled "What Can You Carry On Your Head: A Cross-Cultural Analysis ".

This analysis shows from both a time and culture perspective, like language, just how different our cultures can be and how our cultures change over time. It is a wonderful pictorial reminder of how we must continue to adapt and understand the differences and deliver a quality product that is suited for the time, culture and market. In language, depending on industry, terminology can come and go very quickly. For example, in the fashion industry terminology changes between seasons (sometimes within a season), in the IT industry there are new buzz words and timely terms created on a regular basis and in the travel industry terms also change, but less frequently.

If you have similar examples in other languages, we would love to hear from you. Send email to LostInTranslation@asiaonline.net. We will include the best ones in future newsletters.

Using Language Studio™ Pro Desktop Translation Tools

Language Studio™ Pro is a desktop client that provides advanced translation capabilities directly from your PC. This webinar will present each key tool within Language Studio™ Pro, demonstrate its functionality and offer key tips on how to use the tools and improve the quality of translations.

The following modules will be presented:

• Overview of Language Studio Pro™
• Basic and advanced translation modes
• Introduction to Runtime Customization
• Creating and editing rule files for Pre-Translation Corrections, Non-Translatable Terms, Glossaries and Post Translation Adjustments.
• Measuring machine translation quality with automated and human metric tools.

Asia Online already supports more than 520 language pairs and has more than 200 additional language pairs under development. More Information

Load on Demand and Dedicated Server Modes

One of the advantages of having a cloud based SaaS environment for translation is that depending on your requirements and needs, different configurations can be made available. Language Studio™ Enterprise translation engines operate in two different modes for processing translations:

Load-On-Demand Server Mode

This is the default mode and the most efficient for large translations. Load-On-Demand mode means that engines are not permanently loaded onto a server. Instead they are loaded onto multiple servers when a job is submitted. Language Studio™ Enterprise allocates a number of servers based on the prevailing demand and analysis of how large the job is.

For large jobs, this ensures that they are processed quickly as the jobs span multiple translation servers for faster translation. For most translation jobs this is the optimal setting.

There is a small trade off when a job starts. If the translation engine is not already loaded on a server, a small delay will occur while the engine is loaded. This delay is usually about 2 minutes, but could be a little longer depending on the size of the engine.

For most customers, a short delay prior to starting a translation is not an issue when compared to the time it takes for a human to translate the same text. A typical human will translate at a rate of 2,000 to 3,000 words per day. Asia Online has customers translating more than 500 million words per day.

Dedicated Server Mode

If immediate translation is required without delay, Language Studio™ Enterprise also offers a Dedicated Server Mode.

This mode is for environments where translation is required immediately. For an additional fee, dedicated servers can be applied where a translation engine is loaded permanently on a server and the server is dedicated to just one customer. There are no delays in starting translation. However translation speed is limited to the number of Dedicated Mode servers applied to each engine.