The machine translation category features all posts related to computer-assisted translation and the technology and translators that surround it.

Machine Translation Summit (MTSummit 2019)

STAR is delighted to sponsor this years Machine Translation Summit in the Helix on 19th to 23 August.

At the summit we will be showing the latest advances in MT translation from STAR and how it can benefit both translation teams and corporations. STAR MT combines the best from both Neural and statistical MT approachs.

We can show you how to integrate MT into your corporate workflow environment to deliver consistent, high quality results. We’ll also show you how some of our clients are achieving results today. As professional translators we understand the weaknesses, strengths and opportunities MT provides. Our expert team will be on hand at the show to discuss how our technology and services can help you business. STAR MT integrates already with STAR Transit NXT the professional translation editor making translators more productive.

For more information the MT Summit visit their website here.

STAR is a global leader in technology development for multilingual information.

STAR MT – Your own corporate machine translation system.

Need to know more?

We’re often asked – just what is MT and what about Googles machine translation?

Isn’t it all just the same?

No it isn’t.

There are advantages and major disadvantages to Machine Translation. Getting the most from MT really depends on what your objective is. Take a look at our introduction to MT to get a better understanding of how to make the best choice for you.

See more information on Machine Translation.

Using Machine Translation with STAR Transit


Machine Translation and Transit NXT

The latest version of STAR Transit integrates with a number of Machine Translation Systems. Here’s how to make your translation faster.

1. Overview

Machine Translation (MT) systems can be used when working with Transit.

Concerning MT support, Transit strictly distinguishes between the following systems:

  •  Customer-specific MT systems

In Transit, customer-specific, specially trained MT systems are used exclusively for machine translations during the import of project files. Therefore, they are referred to as Import MT.

The project-specific settings for the Import MT are specified by the client or project manager on the Machine translation tab of the Project settings window .

Import MT suggestions are displayed to you as translator in the fuzzy window, together with the fuzzy matches that may exist.

  •  Unspecific online translation services (GoogleTranslate, iTranslate4eu)

Machine translations by online services are supported only via the Transit editor. The translator needs to request them explicitly for individual segments (see section 4. “Manually requesting a machine translation”). Therefore, they are referred to as Editor MT.

The settings for Editor MT are set by the translator using the Machine translation option of the user preferences.


2. User preferences for Editor MT

In the user preferences you can specify – independent of a project – if you want to use Editor MT. It allows you to manually or automatically request MT suggestions from online providers via the Transit editor.

How do I configure the user preferences for Editor MT?

1- Select Transit button | User preferences.

Transit displays the User preferences window.

2- Select the Machine translation option.

Transit displays the user preferences for the Machine translation option:

User preferences group, Machine translation screenUser preferences group, Machine translation screen

3- Specify the desired settings:

Allow Editor MT (request MT suggestions via context menu) – Here you can turn the use if Editor MT on or off. It allows you to request MT suggestions using the context menu of the Transit editor.

Ask before data is sent

Here you specify when Transit should ask prompt you to confirm explicitly to send data to the MT system:

Never, Once per project or Every time

In the overview below all MT systems supported for use in the Transit editor are listed.

– The Status column displays whether the respective MT system can be used or not:

OK – The MT system can be used.

not configured – The MT system needs to be configured, i. e. the API key must be entered.

Button 40– Transit displays a window allowing you to enter the API key for using the respective MT system.

– In the Use column, you can turn the use of an MT system on and off.

Transit can only use one MT system at the time.

Automatically request MT suggestions

Here you can turn on and off if Transit should automatically request MT suggestions for segments for which there are only fuzzy matches at a low quality:

Only for segments with fuzzy matches lower than (%) – Here you specify the quality of the fuzzy matches (in percent) below which Transit should automatically request a machine translation for a segment.

Only for segments with at least (words) – Here you specify the quality of the fuzzy matches (in percent) below which Transit should automatically request an MRT suggestions for a segment.

Only for segments with not more than (words) – Here you can specify that the segments should have a specified minimum length (i.e. segments that are too short are not sent).

Ask before data is automatically sent – Here you specify when Transit should ask prompt you to confirm explicitly to send data automatically to the MT system:

Never or Once per project


attentionSTAR/Transit has no influence on privacy policies, costs and quality!

For machine translation, if requested by the user/customer, the texts and contents that are to be translated are transferred to the machine translation (MT) system and may be transferred to the provider of the MT system. STAR has no influence on the quality of the machine translation when the data is processed and translated externally in this way. No guarantee shall be provided for the correctness or completeness of such translations.

STAR points out that the Internet is not considered to be a secure environment and data that is transferred online may be accessible by unauthorised third parties. STAR shall accept no liability for the security of any data that is transferred online. STAR shall be excluded from any liability for losses or damage of any kind that result from the transfer of data that is to be translated to the MT system or from the use of the MT system.

Further processing and/or saving the data that is to be translated in the external MT system is not under the control of STAR. From the moment that the data is transferred to the MT system, STAR shall not be responsible for compliance with data protection regulations nor for compliance with confidentiality agreements.

Machine translation may be a paid service that is offered by the machine translation system provider. Costs may be incurred through the use of machine translation. STAR does not have any influence on the amount or billing of these costs.

3. Entering the API key for the MT system

To be able to use an online translation service or MT system from Transit, you need to enter an API key:

Entering the API keyEntering the API key

This API key you obtain from the respective provider.

Via the Check API key when closing the dialog option, you can check immediately if the API key is correct


4. Manually requesting a machine translation

What you should know here

Transit displays the translation suggestion of an Editor MT system in the Source Fuzzy window as follows:

Editor MT suggestionEditor MT suggestion

If the segment for which a machine translation has been requested contains markups, the display is extended by a row:

Editor MT suggestion for a segment containing markupsEditor MT suggestion for a segment containing markups

In the additional row, Transit displays if the segment text has been transferred to the MT system with or without markups.

If the respective MT system does not support the processing of markups, Transit automatically transfers the segment text without markups. In this case you will have to insert the markups after accepting the MT suggestion for your translation.

How do I manually request a machine translation?

1 In the target-language window of the Transit editor, right-click on the current segment.

Transit displays the context window of the target-language window.

2 In the context menu, select the Request machine translation entry.

If fuzzy matches exist for the segment, Transit displays the translation suggestion of the MT system at the top of the Source Fuzzy window. This way the MT suggestion can be easily compared with the fuzzy match of the highest quality.

Just like a fuzzy match, you can edit the MT suggestion and accept it for your translation.


5. Import MT suggestions in the fuzzy window

Transit displays the translation suggestion of an Import MT system in the Source Fuzzy window as follows:

Import MT suggestionImport MT suggestion

If the segment contains markups, the display is extended by a row:

Import MT suggestion for a segment containing markupsImport MT suggestion for a segment containing markups

In the additional row, Transit displays if the segment text has been transferred to the MT system with or without markups.

If the respective MT system does not support the processing of markups, Transit automatically transfers the segment text without markups. In this case you will have to insert the markups after accepting the MT suggestion for your translation.

Just like a fuzzy match, you can edit the MT suggestion and accept it for your translation.


Corporate Machine Translation – STAR MT

STAR’s latest MT Engine helps you deliver more content faster, economically and with higher quality than standard machine translation.

Your style, your wording, your translation

Your company’s communication style and terminology should remain consistent, even in translation. That’s why STAR MT is trained exclusively using text, translations and terminology from within your company. With STAR MT, we keep your text protected by keeping it as a dedicated engine to you.

Business Benefits

  • Increased presence in new markets with quick, cheaper translations.
  • Shorter deadlines through faster turnaround.
  • Affordable translation.
  • You can understand foreign-language content immediately with rough translation (gisting).
  • Maintain your own style and wording in all relevant languages.
  • Benefit from the exisiting translations, linguistic competence of your language experts worldwide.

STAR MT helps deliver solid machine translation results for your business.

Call us to discuss how it can help you today.

Eating carpet prohibited, by Justin Ross Lee

The hazards of machine translation

Eating carpet prohibited, by Justin Ross Lee, the hazards of machine translation
Eating carpet strictly prohibited – a sign of the hazards of machine translation / Image credit: Justin Ross Lee via BBC

The Hazards of Machine Translation and Beyond

The Jackie Chan bus stop, a restaurant called Translate Server Error, wife cake, children sandwiches, wide-boiled aircraft – they sound like comical lines at a stand-up show. But, in fact, they’re real-world examples of glaring mistakes and the hazards of machine translation.

For big firms, poorly translated text can have major consequences such as the risk of offending potential customers or losing business. Although we often hear promises of machine learning tech that will process language translation effortlessly and naturally, just as a real [human] translator would.

But when will such technology be available to businesses?

Skype and its real-time translation upset

Last January, Skype launched its real-time translation tool which instantaneous voice-to-voice translation in seven languages. However, it suffered heavy blows when users complained that it turned ordinary Mandarin words into obscenities. The glitch came to light during the shooting of a Skype commercial in China; apparently even the simple phrase ‘it’s nice to talk to you’ translated into offensive swear words.

Google Translate, the traditional approach

Translation tools like Google Translate have traditionally been built around phrase-based statistical machine translation. Machine translation works by analyzing a back catalogue of texts already translated, such as academic papers and glossaries.

The texts are analyzed in parallel – both original and target languages. Using statistical probabilities, it selects the most appropriate translation to the phrase submitted; the better the quality of the original language, the greater its effectiveness. But it’s prone to howlers, like the ones mentioned earlier. Often the translations sound mechanical and dull.

The End of Machine Translation?

Alan Parker, director of engineering language technology at Facebook, recently commented on statistical machine translation reaching “the end of its natural life”. It has been said that translation technology is on path towards artificial neural networks much like the neural pathways of the human brain.

These neural networks are structured similarly to the brain, using complex algorithms to select phrases appropriate to the translation. Astonishingly the sophisticated network can learn metaphors, idioms and the subtle meaning behind language. This will effectively transform language translation today – rather than direct literal translation, the neural network can translate the same meaning to a different culture avoiding any possible offense.

While Facebook and Google have reported that they will switch over to neural network translation this year, they have not publically announced specific dates.

Auto-translation, not perfect yet

Despite the two tech giants rolling out plans to use neural networks, there are still major hurdles to cross before we’re quite there yet. According to Professor Philipp Koehn: “there are very hard problems with semantics and knowledge representation that have to be solved first, and that we are not close to solving.” Professor Koehn hints at less explicit information in the source language such a gendered nouns and verbs in languages such as Portuguese, Italian and German. Prof. Koehn is a computer scientist and expert in translation technology at the University of Edinburgh.

‘Chinese doesn’t use plurals, verb tenses or pronouns as we do in English, which makes exact translation very difficult’, Prof. Koehn added.

The Hazards of Machine Translation Tech and the Future

Albeit, translation technology has come a long way and provides decent literal translations; there is the need for a tech that speaks the real language of the end user.

Machine translation technology is a handy tool, but don’t rely on it entirely.

The STAR Team

Source: BBC Business News

Google: Augmented Reality, Real-Time Translation

Machine Translation

New tech in machine translation, Real-Time Translation

Real-Time Translation Tool from Google

In January, Google released new features to its translate app.

Available on both iOS and Android platforms, the app can make on-the-spot translations using its Word Lens tool and text translations by inputting words or sentences manually. But this year, the app was updated with voice translation functions in 36 languages. Users can tap the mic button in the app and begin talking. It recognises multiple users and the language being spoken. Once the languages have been recognised, the app is ready is translate without the need to tap the mic button again. Impressive!

Word Lens

Word Lens tool from Google

Word Lens tool from Google

The Word Lens tool in the Google Translate app allowed user to take a photo of a piece of text, a road sign etc. and translate it accordingly. Now, all the user has to do it point the phone’s camera towards the object and see the translated text overlaid on their screen. This is referred to as augmented reality and Google acquired the technology in May last year when they bought the California-based tech firm, “Quest Visual“.

Prior to the augmented reality tech, the app required a data connection to fulfill a request to translate. But now, that’s not an issue. Word Lens can also translate from English to and from German, French, Portuguese, Spanish, Italian and Russian.

Closer to the Future

These futuristic features take us closer to turning our phones into something like the universal translators out of a Star Trek movie. Closer to a new age where language is no longer a social barrier for communicating new ideas, and to one another.

Often the hardest part of travelling is navigating the local language“, states Google Translate’s product lead Barak Turovsky. “Now Google Translate can be your guide in new ways…

Are you using it? Let us know what you think in the comments below.

The STAR Team


The Greatest Mistake In Translation

Caution: bad translation — greatest mistake in translation
Caution: bad translation

Everyone has read or heard a story about mistakes in translation, humorous mistranslations or badly translated documents. Although what is the greatest and probably most lethal mistake in translation?

Strangely enough, the answer is quite simple. My experience from working in the industry for over 30 years has shown that the single biggest mistake in translation is being careless with short translation. As it sounds, short translation is the translation of a few words or a single sentence. Small or short translation projects are the ones with the highest rate of errors!

The reason behind it is simple: laziness and the pressure to meet a deadline. You have to put the same amount of work, perhaps even more, into small projects than you do with bigger ones.

To understand this, you need to think about the types of errors that can happen.

The Author-Designer Mistake

Most software developers, web designers and marketing managers are always under pressure to deliver on tight deadlines. It is fairly common that a last minute change to a document, file or brochure copy will mean a few words being added here and there — now you have a dilemma — you have to send this text to your professional translators. You might have to raise a purchase order or get approval to spend the extra money. You then have to get the text back from the translation agency, answer their questions and then put the text back in to your files. This process might take a day or so and it has to be shipped right away.

Isn’t that what Google Translate does? Why not type the short text into Google Translate, copy the translated text and insert it in the body itself? Surely translation isn’t that difficult? Big mistake!

Google Translate and other machine translation tools are just gist engines. This means that they are good at giving you a general or equivocal translation rather a precise one — they are not accurate — far from it actually.

Taking a translated word from a machine translation tool such as Google Translate is like picking a translated phrase in a language you don’t understand, applying it and hoping it is correct.< ?p>

Language Errors:

Languages are all very different and so are their structures. A number of words in Spanish have male and female versions — so you should know which word you are copying.

If you are working with Chinese or Arabic, you should not cut and paste text on an English language machine as it will corrupt the text if you have the wrong language fonts / keyboard installed on your machine.

Polish and German have different hyphenation rules than English. You cannot move the next word to a new line if it does not fit. Some words / phrases have to stay together such as compound words.

The Context Mistake

Finally, the big one is context errors. You cannot translate a single word unless you know the context of where and what the text is being used for. Especially what linguists refer to as homonyms.

Let’s look at the word press. What does it mean?

  • Press the button
  • Press as in a press beside your bed
  • Press as in a printing press
  • Press as in the written press i.e. newspapers, media et cetera

Google will never translate this word correctly unless it knows more information about the context. A professional translator will always ask the client what the text they are translating is for. There are lots of examples of this. Take the homonym arm, for instance…

  • Arm, a human arm
  • Arm, an ARM chip (Advanced RISC Machines)
  • Arm as a verb, to arm something like an alarm
  • Arm as a verb, to give someone a weapon

Need I continue? It is very easy to take a machine translation of a single word and apply it no matter what.

Where can we find this?

We find that these misused words appear on website menus and widgets. Often the major work is done on the Website and then after review with the client there might be a few minor changes undertaken.

Most website menus are coded in what are called widgets. These widgets normally only have a few words in them. So it is easy for someone to do exactly what we have outlined above.

Some may think: ‘there are only a few words to translate’. Why spend weeks with translators over just a few menu descriptions? Instead of spending the last three days working on the menu or straplines for it, just use a machine translation or ask a friend [who speaks that language natively] to do a quick and dirty translation of it.

At STAR, we will often check our customers’ websites a few weeks after we have translated them to make sure no extra strings were added after we completed the official translation: to ensure its quality.

The Funny Bit

I’ll leave you with this one. Strangely enough comedians and writers use this problem to great effect in sketches. Miscommunication and out of context understanding are the key to many jokes.

Having returned from Newcastle a friend of mine was asked, “Why did you fly to Newcastle?”* “Because it was too far to walk”, he replied.

*For the record, there are multiple answers to the question depending on the context and connotation implied.

For more information on why short and cheap translation can be very expensive, see the following links.

Article by Damian Scattergood, Managing Director of STAR Translation. Damian has over 30 years of experience in the translation and localization industry.

The STAR Team

CNGL invest €19.8M, intelligent content research

Damian Scattergood attending the CNGL conference
Damian Scattergood at the CNGL announcement / STAR Translation

CNGL investments

Yesterday Damian Scattergood and Paul Quigley – Directors of STAR Translation – had the pleasure of attending the CNGL conference at the AVIVA stadium for the announcement of €19.8 million in research funding.

Amongst the prestigious guests were Mr John Perry TD, Minister of State with responsibility for Small Businesses at the Department of Jobs, Enterprise and Innovation and Professor Vincent Wade the CNGL Director and Professor Vinny Cahill, the Dean of Research of Trinity College Dublin.

There was a number of demonstrations showcasing new technological applications, which help to provide high-quality translation and content analysis.

Minister John Perry speaking at the CNGL Conference
Minister John Perry announcing the new funding to the CNGL

CNGL has engaged with hundreds of companies in Ireland and beyond. The commercial expertise, the multiple licences and the large number of applied research collaborations are now reinforced by an €19.8 million investment.

In fact, the government through the Science Foundation Ireland and 16 industry partners have invested in CNGL. Industry partners include tech giants as Symantec, Microsoft, McAfee, Cisco, Welocalize and Intel and by this investment show their trust in the group and its research into intelligent content. This new investment will directly support 75 high-class research jobs especially in key area of digital platforms, content and applications.

It was a great day to network with like minded translation and information technology experts. STAR provides technology and services for multilingual information management.  GRIPS: STAR’s Information Management and Publication System provides semantic information that allows intelligent content publishing. Thanks to everyone at CNGL and the Aviva Stadium for their hospitality. We really enjoyed it.

STAR's visit to the Aviva Satdium for a CNGL Conference
Paul Quigley (right), from STAR Translation, discusses intelligent content research and the impact for improved translation.
STAR at the CNGL conference, Aviva Stadium
STAR at the CNGL conference, Aviva Stadium.
Networking at the CNGL conference in the Aviva Stadium
Networking at the CNGL conference, Aviva Stadium.

The STAR Team

Lao on Google Translate

Lao supported by Google Translate

Lao supported on Google Translate
Lao language added to Google Translate armoury.

Lao supported, Google Translate blog confirms

As of today, Google Translate now supports Lao. This makes it the 65th language supported by Google Translate.

Lao, or Laotian, is a tonal language of the Tai–Kadai language family.  It’s the official language of Laos and spoken in the north-east region of Thailand. There, it is usually referred to as the Isan language. We welcome new languages like this being added to software such as Google Translate, as it shows the world is becoming a more integrated community. Supporting more languages is something Google is always working on.

While machine translation isn’t perfect, it does have uses in terms of language gisting and aiding in basic communication.

Read the Official LAO press release from Google Translate.

The STAR Team

Google Translate English to Animal

English to Animal Translation from Google

Only Google could do it!

A new product launched yesterday for Android devices introduces translation from English to animal languages. You will be able to communicate more effectively with your pets. Now you can really tell your dog that you love them.

Take a look at the video if you don’t believe us.

Visit Google’s Translate for Animals web page for more details.

The STAR Team