Showing posts with label translation memory. Show all posts
Showing posts with label translation memory. Show all posts

Thursday, October 17, 2013

Online Help Localization: Background and Methodology

Welcome to the latest in the series of “Localization: The Definitive Guide”, a complete step by step guide to the localization of websites, software and it’s components. In this article we will deal with the translation of Online Help for software. Again this component ties in with the localization of the software or the User Interface (UI) and the race to release a SIM ship version of the localized product. Just to refresh a SIM ship release is where the language versions of the product are released to their respective markets on the same date as the original version. As one can imagine the cross referencing between the Online Help and the User Interface is very important. It is important to ensure that both are consistent with each other. In an ideal world we translate the user interface first, populate the translation memory and Glossary with the translated text and then run it against the Online Help. However with a SIM ship version and the time constraint we need to translate the OLH in parallel with the User Interface. The process for the translation of the OLH is very similar to the translation of a website or the User interface itself, in fact some Help systems are web based so please bear in mind that there will be some overlap between this article and the others in this series. For this article we will take the worst case scenario of a SIM ship release where both the software and the Online Help are translated at the same time. One thing in our favor is that the User interface word count is usually substantially lower than the online help word count which gives us time to ensure during revision that the references and links from the Online Help to the software reference correctly.

1. Pseudo Localization of Online Help

As with the software, one of the first steps is to pseudo-translate the online help pilot language/s. Just to refresh our memories, the pilot languages are a small cross section of languages to represent all the target languages to be translated. Please refer to pseudo translation of software here to refresh your memory on its purpose and benefits.

2. Pilot Language for Online Help Localization

Now we begin the Online Help localization process in earnest by starting its localization based on the pilot language/s already chosen for the software. This language will define the localization process and weed out the majority of the localization bugs.

3. Translation Memory and Glossary input during Online Help localization.

The online Help localization process is similar to the pseudo translation process in that the translatable text is isolated, translated and dumped back into the Online Help, however, there is an extra degree of complexity if there are translation Memories involved in the process and an even greater degree of complexity if the translation memory is not centralized. Please refer to translation memory input during software localization for additional information. Lets take a step forward and assume that our Help system has been pseudo translated, bug fixed and prepared for the TM environment. It is here that the localization process of the online Help can vary greatly depending on the translation memory, the infrastructure and the resources chosen by the translation services company. Lets take a couple of scenarios I hope will cover the majority of localization process cases.

Centralized Translation memory Model

A centralized translation Memory is a central repository of previously translated segments and Glossary terminology. The system is interactive and allows the translator to access a “live” translation memory and glossary. It means that all translation resources both internal and external work from the same TM and the translated text is updated on the “fly”. It ensures that all translators are up to date with the most current terminology and translations. Centralized translation memories are often complex and expensive to implement resulting in higher costs but worth it in the long run by cutting time, long term costs and enhancing quality and consistency.
Centralized Translation Memory
Let’s take a quick run through the centralized translation memory process. Let’s assume we have started the user interface and online Help translation in unison. This may result in different translations for User Interface commands and options in both components. In this case the reviewer of the User interface will have final sign-off on all the UI text. If there are any issues he will discuss this with the translator of the User Interface and translator and reviewer of the online Help to reach a general consensus on the best translation and be responsible for updating the central TM. The translator and reviewer of the OLH will receive notifications of TM updates and can update their target texts accordingly on the fly. In a lot of cases if a translator is external they will need access to the centralized translation memory system by means of a portal.

Translation Memory Model (Not centralized)

The second model is more cumbersome and complicates the process as the translation memory is not centralized. It means that the translator cannot update his text while he is working. It complicates the parallel translation in that at the end of the day for instance, both translation memories differ i.e. the UI reviewer TM and the OLH reviewer TM, or may even involve syncing many more TMs for instance, UI reviewer TM, the OLH reviewer TM, UI translator TM and OLH translator TM, depending on the amount of resources working concurrently on the same text. There are workarounds but they involve extra management and process bottlenecks. For instance we could devise a process where at the end of the day a central resource takes the UI translators TM and updates it with his changes. The most up to date TM is then sent to the OLH translator and UI translator at the beginning of each day to begin their work. As one can imagine its not an ideal situation! For a demonstration on a Translation memory environment please refer to the following link.

4. Translation and Revision of the Online Help

Ok, so we send the translator the translation memory compatible files to translate or access to the centralized translation memory system portal , the most up to date Translation Memory (as they need to update it with their work) reference material such as Glossaries, previous User Guides…etc.. and the translator begins translation. As discussed in the previous paragraph one must bear in mind that the process can vary in many ways when we take into consideration that the translator may be in-house or external or the translation memory system we are using. What is key is that each resource always has a reasonably up to date TM to work with to avoid duplication of work. We also discussed in the localization of software that the translator may be responsible for typical localization bugs such as resizing of dialogues and menus and duplicate hot keys or this may be the role of the localization engineers after translation. The end product is an 80 to 90% localized pilot version. What adds even more to the complications of the project is whether the Online help is revised by the translation company, the client or a third party. What if the client wants to revise the OLH? What if the client does not wish to use the translation memory process instigated by the translation company?

5. Translated Online Help Build

During the next stage the pilot version is rebuilt for testing on various platforms.

6. Localization QA - Testing of localized Online Help

Localization testing of Online Help is the same as for the software in that once rebuilt the pilot version is tested for functional bugs by the Localization QA team and linguistically tested, with the UI in context, by the translators or in some cases by third party linguists. The bugs are documented and sent to engineering in the case of functional bugs, or translation, in the case of linguistic bugs, to be fixed. This cycle continues until the translated software is bug free. The set-up of the phase can differ from company to company depending on the circumstances but in all cases it’s important to update the TMs with the linguistic fixes. There is an extra degree of complexity during this phase compared to the software localization QA in that the cross referencing and linking to the software is fully functional and corresponds linguistically

7. Translation of other Online Help language versions

Now that we have our pilot version localized and bug free, the process is in place to translate the other Online Help language versions. I mentioned at the start of this article that in an ideal world the localized versions should be released on the same date as the master version however it’s usually impossible to achieve this and what usually happens is that certain languages are given priority depending on their market importance. The more important languages are called tier one languages and the less important languages for secondary markets are generally referred to as Tier 2 languages.
To summarize, I think this article gives us an idea of how closely the software localization process and the Online Help localization process are intertwined in that, every UI reference must be exactly the same in the Online Help. This article is part of a series, "Localization: The definitive Guide" from One Stop Shop Translations, which deals with the localization of each component of software, the others include:

- Software Localization: Background and Methodology
- End User License Agreements Localization: Background and Methodology
- Software Documentation Localization(Quick User Guides and User Guides): Background and Methodology
- Website Localization: Background and Methodology

NOTE: Please note that translation and localization are used interchangeably in this article.
DEF: Translation is the communication of the meaning of a source-language text by means of an equivalent target-language text.
DEF: Language localization is the process of adapting a product that has been previously translated into different languages to a specific country or region. Source: Wikipedia

If you like this post please "like" or "share" for more content

Mark Kieran, CEO, One Stop Shop Translations

For our latest online Help localization rates click on this link or get an economically unbeatable Online Help localization quote here.

Remember that translation of software is not just simple straight forward translation but a complicated process that involves many stages and specialized expertise!

Wednesday, November 16, 2011

Translation Memories, Blessing or Curse?

translation memories
A translation memory is a modern technology tool designed to help translators. It is essentially a database that saves previously translated segments of text. The main benefits of a translation memory are saved time and cost, more efficiency and consistency throughout translations.

Moreover, clients and translators must also analyze not only what are the benefits of the Translation memory are but also the the ethical questions posed by this tool. Firstly, it is important to note that the usual practice during the delivery of a translation to the client is for the translator to also deliver the translation memory database so it can be used in future translations, even if the client opts to use a different translator. It is clear to see how this process guarantees consistency. However, here is where a lot of problems can arise. It is true to say that in the majority of cases the the new text to be translated corresponds to the saved translation but there are cases where the corresponding saved translation does not make sense with the new source translation in its current context, especially in the case of literary translations or Marketing translations.

To navigate this problem it is important to understand the analytical nature of the Translation memory. A lot of Translation Memories match a previously translated segment to a new source segment on a percentage basis. For instance, if a translated target segment in the TM and its corresponding source segment match a new source segment to be translated word for word, this segment will be translated automatically and the analysis will return a statistic of 100% matching. But as outlined above, the contextual problem may occur. This is why the 100% segment needs to be checked by the translator. The whole process is not fool proof and we always need some degree of human intervention. This is why most translation services companies and translators will charge a fee even for segments that are 100% matching. The bottom line is that the 100% matching segments have to be checked.

Many clients are often surprised when they get their translation quote to find they are being charged for segments that have been previously translated word for word. Hopefully, the above explanation gives the client an insight into this dilemma. For information the 100% matches are usually charged at a lesser rate, for instance a revision translation rate per word so there is still a very evident cost saving. It is here that the client has to find the best translation rate per word to pay for 100% matching segments and repeated segments. This can lead to huge translation cost savings especially in the case of updates such as software or manuals.

Hopefully this article gives the reader an insight into the workings of the translation memory environment and with this technology, unlike most concepts in the technology world, the fundamental process will stay the same, we will always need some degree of human intervention!

If you like this post please "like" or "share" for more content

Mark Kieran, CEO, One Stop Shop Translations

For more information on our translation memory pricing matrix click on this link.

Remember that translation memory rates may cost more in the short term due to the extra processes and management involved but are worth it in the long term especially in the case where the translation is an update or there is a lot of repeated text!

Monday, October 27, 2008

Choosing a Translation Services company or Machine Translation

As we are all aware the translation industry is awash with translation tools such Spell checkers, translation memories and automated glossaries. But is the feasibility of machine translation a reality in this day and age. By explaining the doubts and limitations of machine translation I hope to clarify to clarify some of the preconceived notions that the general public may have about the field and aid them in deciding on Machine translation or a translation services company.

The notion of computer translation is not new. In fact shortly after World War II the American Government had already began investing considerable resources in the field without the slightest doubt that the concept was not a reality.Some common terms in this field indicate the some of the difficulties that those pioneers of machine translation were to encounter, for example the difference between machine translation which is the translation of text by a machine and Computer Aided translation which is the translation of texts by a translator with the aid of translation tools. Under Machine translation there are three types of system namely Batch, Interactive, and Interlingual Approaches. A Batch method has coded rules to `decide' on the best translation. There is no need for a translator.

With an Interactive system the translator is present and decides on the translation options provided by the translation system.

With an Interlingual approach the source translation is translated to an intermediate language that is used to translate back and forth between the source and target languages.CAT and MT software these days use either the Batch or interlingual approach.

With MT translation most texts tend to have a 70% accuracy e.g Google translate. Most experts now concede that 100% accuracy is not possible. Three terms that crop up are Fully automatic High Quality Translation which is in my view is impossible to achieve, Fully Automatic Low Quality Translation and Partly Automatic Medium Quality Translation. The percentage accuracy claims of Machine translation is open to debate as there is no universal standard to measure this and accuracy claims tend to be very subjective.

When to use Machine Translation over Translation Services companies. There are five important criteria when choosing whether to use machine translation over translation Service companies.

1. Subject matter. Here the computer can have an immense advantage especially in regard to technical texts. In the case of a field like Life sciences where the vocabulary is very specific, the Machine Translation system can have a terminology Database built up over years which is impossible for a Translation Service company to compete with. Of course the quality depends on the amount of work and quality of the work put into the Machine translation’s dictionary.

2. Speed. Speed is an area where the computer reigns supreme considering that the average translator translates at a rate of 2,500 words per day.

3. Level of accuracy. We already discussed the levels of accuracy. If a text is solely for information then a fully automated translation is feasible but if we need 100% accurate translation the amount of time spent post-editing the MT system can often outweigh the benefits of using this system.

4. Consistency of vocabulary. Again the computer is excellent when it comes to consistency. One centralized MT system ensures consistency as opposed to a Translation vendor outsourcing a large job or different jobs over time to different translators. It is often the case that no two translators translate a sentence in the same way. Of course, the success of the MT depends on the preprogramming done beforehand.

5. Cost. Bearing in mind that the computer can tick all the right boxes for speed, consistency, level of accuracy and subject matter one has to bear in mind that successful Machine translation systems require substantial investment to populate them with high quality and a high volume of content which, of course, has to be passed onto to the client of a Translation Services company.

It’s pretty evident from the above points that the computer can yield impressive results but what we must realize is that current MT systems will not give 100% accurate translations. If this level of accuracy is required it’s always best to hire the services of a translation company.

Thursday, July 31, 2008

Choosing A Translation Services Company

I am often surprised to hear that Localization managers are not happy with the level of service they are receiving from their translation vendors. Frequent complaints about missed deadlines, poor linguistic quality and unresponsive project managers seem to be the norm. Then when I ask about their prices I wonder why they wish to stay with the same translation services company when receiving such a bad level of service. To add insult to injury when the words Translation memories or translation savings are mentioned they seem unaware of the idea. Similarly with Managers complaining of prolonged schedules the same question springs to mind, “Why stick to the same vendor with such a poor level of service”.

The main reason why these Managers are reluctant to change is the risk involved which outweighs the unknown reward. To lower this risk and have visibility of the partial benefits of two or multiple vendors why not protect your Translation assets thus ensuring full independence when out-sourcing. While maybe being a little more labor intensive the multiple vendor model keeps current vendors more competitive and allows you to easily draw comparisons on the level of service you are receiving. How can you possibly tell how fast your vendor is, the competitiveness of their translation prices and the efficiency of their processes with a single vendor model?

Implementing the multiple-vendor model?

The first thing to do is to request your Translation memories. At One Stop Shop we acknowledge the importance of our clients requesting Translation Memory imports to protect their intellectual property while at the same are not worried about entering the competitive vendor model because of the confidence we have in the level of our service.
Remember you have every right to your TMs as it is your intellectual property.

1. Phase in alternative Vendors.
Ask One Stop Shop Translations for a quote here and your current vendor for a quote. If you have supplied us the TMs you will have a direct comparison of the price per word rates and the savings from using the TM. If the TMs are not supplied at least you will have a comparison of the translation rates which will not let you down. You, you will have clear visibility on additional services you are being charged for e.g. Project Management

2. Hire One Stop Shop Translations on a test project
This is a good way to measure our quality and customer service.

3. Phase in more projects to One Stop Shop Translations
Provided you are happy of course phase in more work to your second vendor.

Quality, Turnaround and Cost
All vendors profess a quick turn around, high quality and unbeatable prices so at least now with the multiple vendor model you can draw your own comparisons. It may be helpful to keep a checklist or post-project feedback form to collate all criteria scores over a period of time. Apart from the quality, price and turnaround other effective criteria that may be selected are responsiveness, relationship, due diligence, reporting and technology.
One Stop Shop Translations can help you draw up these checklists free of charge and help you phase in a higher level of translation services.
With regard to quality you may not have the luxury of your country reviewers to judge this but there are other methods which One Stop Shop Translations can guide you through. Extensive quality criteria are always recommended.
With regard to cost, apart from having lower rates One Stop Shop Translations have no hidden costs with regard to the pre-project quote, that’s right, no Project Management costs, no update costs within reason. Due to our due diligence One Stop will negotiate any grey areas of the project at the Quote stage, still remain competitive and avoid any nasty surprises for the client with the invoice.
At all stages the client has to right to the analysis files (Trados, Catalyst, FrameMaker etc.) and will receive a clear explanation if there are still in the dark.

Choosing One Stop Shop Translations
While there are many translation services companies to choose from, we will distinguish ourselves from the competition by our friendly and highly qualified people, free consultancy and Project Management services and our expertise in your industry. We would love the opportunity to demonstrate these things to you and describe how we've been able to help other clients in your position make a change for the better. To schedule a discussion with One Stop Shop Translations, please call +34–91-365-9608.

Mark Kieran,
One Stop Shop Translations S.L.
(Madrid) 0034 91 365 9608

Wednesday, July 30, 2008

One Stop Shop Translations

Professional Translation Services
All your translation services in One Stop. With over 5,000 specialist translators based world wide we're in a ideal position to translate any type of text whether it's Science, Legal, Business, Marketing, or any other field One Stop Shop Translations uses only native translators for each target language. All translation projects go through a cycle of translation and revision. With over thirty years experience we cater for up to 140 languages.

Translation Company Profile
One Stop Shop Translations was founded by three seasoned localization professionals striving to make their own mark on the industry. Drawing on a rich background of Localization Project Management skills, engineering skills and linguistic skills the three have ambitions to make One Stop the largest vendor in the business by the year 2018.

Mark Kieran,
One Stop Shop Translations S.L.
Translation Sercices
(Madrid) 0034 91 365 9608