What evil lurks in the results from your language service provider?

Let me start by disclosing that although I have a registered limited company through which I provide translation, training and technical consulting services for translation processes, I am essentially a sole trader who is not unreasonably, though not correctly, referred to as a freelancer much of the time. I have a long history of friendship and consulting support with the honorable owners of quite a few small and medium language service companies and of a few large ones. I vigorously dispute any foolish claims that there is no such need for such companies, and I see a natural alliance and many shared interests between the best of them and the best of independent professionals in the same sector.

But as Sturgeon's Law states so well, "ninety percent of everything is crap", and that would apply in equal measure to translation brokers and translators I suspect, though of course this is influenced by context. But what context can justify this translation of a data privacy statement from German to English? Only the section headers are shown here to protect against sensory overload and blown mental circuits:

The rest of the text is actually worse. This is the kind of thing some unscrupulous agencies take money for these days.

Why, pray tell, was the section numbering translated so variously into English? Well, if you know anything about the mix-and-match statistical crapshoot that is SMpT (statistical machine pseudotranslation) and its not-as-good-as-you-think wannabe alternatives, it's easy to guess the frequency of certain correlations in English with German numbers followed by a period.

And clearly, the agency could not even be bothered to make corrections, and the robotic webmaster put the text up, noticing nothing, where it remained for about a year to embarrass a rather good company which I hold in high esteem.

What's the moral of this story? Take your pick from the many reasonable options. "Reasonable" does not include doing business with the liars and thieves who will try to sell you on the "value proposition" of machine translation to cut costs.

A skilled translator knowledgeable in the subject matter and trained in dictation techniques paired with a good speech recognition solution or transcriptionist can beat any human post-edited machine translation process for both volume and quality. And a skilled summarizer reading source texts and dictating summaries in another language can blow them both away as a "value proposition".

One thing that is too often forgotten in the fool's gold rush to cheap language (dis)service solutions is - as noted by Bevan et alia - exposure to machine-translated output over any significant period of time has unfortunate effects one the language skills (reading, writing and comprehension) of the victims working with it. This has been confirmed time and again by translation company owners, slavelancers and other word workers. Serious occupational health measures are called for, but to date little or nothing has been done in this regard.

And when human intelligence is taken out of play or impaired by an automated linguistic lobotomy, the results inevitable gall in the lower quartile of the aforementioned 90%. Really crappy crap.

As another of my favorite fiction authors used to comment: TANSTAAFL. There ain't no such thing as a free lunch. And trust is always good, but these days you need to verify that your service providers really give you what you have paid for and don't pass off crap like you see in the example above.

Survey for Translation Transcription and Dictation

The website with the survey and short explainer video is http://www.sightcat.net
The idea is to build a human transcription service. We just need a few translators per language that want to work with a transcriptionist due to RSI, productivity etc. and we can use that data to build an ASR system for that language. There is also a good chance the ASR system will be accurate for domain-specific terminology and accents as it will be adaptive and use source language context. 

Click on the graphic to go to the survey

memoQ Fest 2018 was, among other things, a good opportunity as always to spend time discussing things with some of the best and most interesting consultants, teachers, creative developers and brainstormers I know in the translation profession. One of these was my friend and colleague, John Moran, whose work on iOmegaT introduced me to the idea that properly designed, translator-controlled (voluntary) data logging could be a great boon to feature research and software development investment decisions. Sort of like SpyGate in translation, except that it isn't.

John and I have been talking, brainstorming and arguing about many aspects of translation technology for years now, dictation (voice recognition, ASR, whatever you want to call it) foremost among the topics. So I was very pleased to see him at the conference in Budapest last week, where he spoke about logging as a research tool in the program and a lot about speech recognition before and after in the breaks, bars, coffee houses and social event venues.

I think that one of the most memorable things about memoQ Fest 2018 was the introduction of the dictation tool currently called hey memoQ, which covers a lot of what John and I have discussed until the wee hours over the past four years or so and which also makes what I believe will be the first commercial use of source text guidance for target text dictation (not to mention switching to source text dictation when editing source texts!). John introduced that to me years ago based on some research that he follows. Fascinating stuff.

One of the things he has been interested in for a while for commercial, academic and ergonomic reasons is support for minor languages. Understandable for a guy who speaks Gaelic (I think) and has quite a lot of Gaelic resources which might contribute to a dictation solution some day. So while I'm excited about the coming memoQ release which will facilitate dictation in a CAT tool in 40 languages (more or less, probably a lot more in the future), John is thinking about smaller, underserved or unserved languages and those who rely on them in their working lives.

That's what his survey is about, and I hope you'll take the time to give him a piece of your mind... uh, share your thoughts I mean :-)

Technology for Legal Translation

Last April I was a guest at the Buenos Aires University Facultad de Derecho, where I had an opportunity to meet students and staff from the law school's integrated degree program for certified public translators and to speak about my use of various technologies to assist my work in legal translation. This post is based loosely on that presentation and a subsequent workshop at the Universidade de Évora.

Useful ideas seldom develop in isolation, and to the extent that I can claim good practice in the use of assistive technologies for my translation work in legal and other domains it is largely the product of my interactions with many colleagues over the past seventeen years of commercial translation activity. These fine people have served as mentors, giving me my first exposure to the concepts of platform interoperability for translation tools, and as inspirations by sharing the many challenges they face in their work and clearly articulating the desired outcomes they hoped to achieve as professionals. They have also generously and frequently shared with me the solutions that they have found and have often unselfishly shared their ideas on how and why we should do better in our daily practice. And I am grateful that I can continue to learn with them, work better, and help others to do so as well.

A variety of tools for information management and transformation can benefit the work of a legal translator in areas which include but are not limited to:

• corpus utilization,
• text conversion,
• terminology management,
• diverse information retrieval,
• assisted drafting,
• dictated speech to text,
• quality assurance,
• version control and comparison, and
• source and target text review.

Though not exhaustive, the list above can provide a fairly comprehensive basis for education of future colleagues and continued professional development for those already active as legal translators. But with any of the technologies discussed below, it is important to remember that the driving force is not the hardware and software we use in technical devices but rather the human mind and its understanding of subject matter and the needs of the particular task or work process in the legal domain. No matter how great our experience, there is always something more and useful to be learned, and often the best way to do this is to discuss the challenges of technology and workflow with others and keep an open mind for new approaches with promise.

Reference texts of many kinds are important in legal translation work (and in other types of translation too, of course). These may be monolingual or multilingual texts, and they provide a wealth of information on subject matter, terminology and typical usage in particular contexts. These collections of text – or corpora – are most useful when the information found in them can be read in context rather than isolation. Translation memories – used by many in our work – are also corpora of a kind, but they are seriously flawed in their usual implementations, because only short segments of text are displayed in a bilingual format, and the meaning and context of these retrieved snippets are too often obscure.

An excerpt from a parallel corpus showing a treaty text in English, Portuguese and Spanish

The best corpus tools for translation work allow concordance searches in multiple selected corpora and provide access to the full context of the information found. Currently, the best example of integrated document context with information searches in a translation environment tool is found in the LiveDocs module of Kilgray's memoQ.

A memoQ concordance search with a link to an "aligned" translation

A past translation and its preview stored in a memoQ LiveDocs corpus, accessed via concordance search

A memoQ LiveDocs corpus has all the advantages of the familiar "translation memory" but can include other information, such as previews of the translated work as well. It is always clear in which document the information "hit" was found, and corpora can also include any number of monolingual documents in source and target languages, something which is not possible with a traditional translation memory.

In many cases, however, much context can be restored to a traditional translation memory by transforming it into a "document" in a LiveDocs corpus. This is because in most cases, substantial portions of the translation memory will have its individual segment records stored in document order; if the content is exported as a TMX file or tab-delimited text file and then imported as a bilingual document in a LiveDocs corpus, the result will be almost as if the original translations had been aligned and saved, and from a concordance hit one can open the bilingual content directly and read the parts before and after the text found in the concordance search.

Legal translation can involve text conversion in a broad sense in many ways. Legal translators must often deal with hardcopy or faxed material or scanned files created from these. Often documents to translate and reference documents are provided in portable document format (PDF), in which finding and editing information can be difficult. Using special software, these texts can be converted into documents which can be edited, and portions can be copied, pasted and overwritten easily, or they can be imported in translation assistance platforms such as SDL Trados Studio, Wordfast or memoQ. (Some of these environments include integrated facilities for converting PDF texts, but the results are seldom as suitable for work as PDF or scanned files converted with optical character recognition software such as ABBYY FineReader or OmniPage.)

Software tools like ABBYY FineReader can also convert "dead" scanned text images into searchable documents. This will even work with bad contrast or color images in the background, making it easier, for example, to look for information in mountains of scanned documents used in legal discovery. Text-on-image files like the example shown above completely preserve the layout and image context of the text to be read in the best way. I first discovered and used this option while writing a report for a client in which I had to reference sections of a very long, scanned policy document from the European Parliament. It was driving me crazy to page through the scanned document to find information I wanted to cite but where I had failed to make notes during my first reading. Converting that scanned policy to a searchable PDF made it easy to find what I needed in seconds and accurately cite its page number, etc. Where there is text on pictures, difficult contrast and other features this is often far better for reference purposes than converting to an MS Word document, for example, where the layouts are likely to become garbled.

Software tools for translation can also make text in many other original formats accessible to translators in an ergonomically simpler form, also ensuring, where necessary, that no text is overlooked because of a complicated layout or because it is in an easily overlooked footnote or margin note. Text import filters in translation environments make it easy to read and translate the words in a uniform working environment, with many reference tools and other help available, and then render the translated text back into its original format or some more useful bilingual format.

An excerpt of translated patent claims exported as a bilingual table for review

Technology also offers many possibilities for identifying, recording and controlling relevant terminology in legal translation work.

Large quantities of text can be analyzed quickly to find the most frequent special vocabulary likely to be relevant to the translation work and save these in project glossaries, often enabling that work to be organized better with much of the clarification of terms taking place prior to translation.  This is particularly valuable in large projects where it may be advisable to ensure that a team of translators all use the same terms in the target language to avoid possible confusion and misunderstanding.

Glossaries created in translation assistance tools can provide terminology hints during work and even save keystrokes when linked to predictive, "intelligent" writing features.

Integrated quality checking features in translation environments enable possible deviations of terminology or other issues to be identified and corrected quickly.

Technical features in working software for translation allow not only desirable terms to be identified and elaborated; they also enable undesired terms to be recorded and avoided. Barred terms can be marked as such while translating or automatically identified in a quality check.

A patent glossary exported from memoQ and then made into a PDF dictionary via SDL Trados MultiTerm

Technical tools enable terminology to be shared in many different ways. Glossaries in appropriate formats can be moved easily between different environments to share them with others on a team which uses diverse technologies; they can also be output as spreadsheets, web pages or even formatted dictionaries (as shown in the example above). This can help to ensure consistency over time in the terms used by translators and attorneys involved in a particular case.

There are also many different ways that terminology can be shared dynamically in a team. Various terminology servers available usually suffer from being restricted to particular platforms, but freely available tools like Google Sheets coupled with web look-up interfaces and linked spreadsheets customized for importing into particular environments can be set up quickly and easily, with access restricted to a selected team.

The links in the screenshot above show a simple example using some data from SAP. There is a master spreadsheet where the data is maintained and several "slavesheets" designed for simple importing into particular translation environment tools. Forms can also be used for simplified data entry and maintenance.

If Google Sheets do not meet the confidentiality requirements of a particular situation, similar solutions can be designed using intranets, extranets, VPNs, etc.

Technical tools for translators can help to locate information in a great variety of environments and media in ways that usually integrate smoothly with their workflow. Some available tools enable glossaries and bilingual corpora to be accessed in any application, including word processors, presentation software and web pages.

Corpus information in translation memories, memoQ LiveDocs or external sources can be looked up automatically or in concordance searches based on whole or partial content matches or specified search terms, and then useful parts can be inserted into the target text to assist translation. In some cases, differences between a current source text and archived information is highlighted to assist in identifying and incorporating changes.

Structured information such as dates, currency expressions, legal citations and bibliographical references can also be prepared for simple keystroke insertion in the translated text or automated quality checking. This can save many frustrating hours of typing and copy revision. In this regard, memoQ currently offers the best options for translation with its "auto-translation" rulesets, but many tools offer rules-based QA facilities for checking structured information.

Voice recognition technologies offer ergonomically superior options for transcription in many languages and can often enable heavy translation workloads with short deadlines to be handled with greater ease, maintaining or even improving text quality. Experienced translators with good subject matter knowledge and voice recognition software skills can typically produce more finished text in a day than the best post-editing operations for machine pseudo-translation, with the exception that the text produced by human voice transcription is actually usable in most situations, while the "gloss" added to machine "translations" is at best lipstick on a pig.

Reviewing a text for errors is hard work, and a pressing deadline to file a brief doesn't make the job easier. Technical tools for translation enable tens of thousands of words of text to be scanned for particular errors in seconds or minutes, ensuring that dates and references are correct and consistent, that correct terminology has been used, et cetera.

The best tools even offer sophisticated tools for tracking changes, differences in source and target text versions, even historical revisions to a translation at the sentence level. And tools like SDL Trados Studio or memoQ enable a translation and its reference corpora to be updated quickly and easily by importing a modified (monolingual) target text.

When time is short and new versions of a source text may follow in quick succession, technology offers possibilities to identify differences quickly, automatically process the parts which remain unchanged and keep everything on track and on schedule.

For all its myriad features, good translation technology cannot replace human knowledge of language and subject matter. Those claiming the contrary are either ignorant or often have a Trumpian disregard for the truth and common sense and are all too eager to relieve their victims of the burdens of excess cash without giving the expected value in exchange.

Technologies which do not assist translation experts to work more efficiently or with less stress in the wide range of challenges found in legal translation work are largely useless. This really does include machine pseudo-translation (MpT). The best “parts” of that swindle are essentially the corpus matching for translation memory archives and corpora found in CAT tools like memoQ or SDL Trados Studio, and what is added is often incorrect and dangerously liable to lead to errors and misinterpretations. There are also documented, damaging effects on one’s use of language when exposed to machine pseudo-translation for extended periods.

Legal translation professionals today can benefit in many ways from technology to work better and faster, but the basis for this remains what it was ten, twenty, forty or a hundred years ago: language skill and an understanding of the law and legal procedure. And a good, sound, well-rested mind.

*******

Further references

Speech recognition 

Dragon NaturallySpeaking: https://www.nuance.com/dragon.html
Tiago Neto on applications: https://tiagoneto.com/tag/speech-recognition
Translation Tribulations – free mobile for many languages: http://www.translationtribulations.com/2015/04/free-good-quality-speech-recognition.html
Circuit Magazine - The Speech Recognition Revolution: http://www.circuitmagazine.org/chroniques-128/des-techniques
The Chronicle - Speech Recognition to Go: http://www.atanet.org/chronicle-online/highlights/speech-recognition-to-go/
The Chronicle - Speech Recognition Is in Your Back Pocket (or Wherever You Keep Your Mobile Phone): http://www.atanet.org/chronicle-online/none/speech-recognition-is-in-your-back-pocket-or-wherever-you-keep-your-mobile-phone/

Document indexing, search tools and techniques

Archivarius 3000: http://www.likasoft.com/document-search/
Copernic Desktop Search: https://www.copernic.com/en/products/desktop-search/
AntConc concordance: http://www.laurenceanthony.net/software/antconc/
Multiple, separate concordances with memoQ: http://www.translationtribulations.com/2014/01/multiple-separate-concordances-with.html
memoQ TM Search Tool: http://www.translationtribulations.com/2014/01/the-memoq-tm-search-tool.html
memoQ web search for images: http://www.translationtribulations.com/2016/12/getting-picture-with-automated-web.html
Upgrading translation memories for document context: http://www.translationtribulations.com/2015/08/upgrading-translation-memories-for.html
Free shareable, searchable glossaries with Google Sheets: http://www.translationtribulations.com/2016/12/free-shareable-searchable-glossaries.html

Auto-translation rules for formatted text (dates, citations, etc.)

Translation Tribulations, various articles on specifications, dealing with abbreviations & more:
http://www.translationtribulations.com/search/label/autotranslatables
Marek Pawelec, regular expressions in memoQ: http://wasaty.pl/blog/2012/05/17/regular-expressions-in-memoq/

Authoring original texts in CAT tools

Translation Tribulations: http://www.translationtribulations.com/2015/02/cat-tools-re-imagined-approach-to.html

Autocorrection for typing in memoQ

Translation Tribulations: http://www.translationtribulations.com/2014/01/memoq-autocorrect-update-ms-word-export.html

Dragon Naturally Speaking Version 13 - Review!

I've had a number of people ask me recently whether I have upgraded to Dragon Naturally Speaking version 13 for my dictation work in translation. I have not; I am still using the German version 12.5 (which includes English - I sometimes dictate poorly legible source texts in German rather than waste my time with OCR if I want to work with translation environment tools, so I need the bilingual edition). However, a colleague was kind enough to point me to this review of the new version, which gives me more than sufficient reason to upgrade soon:

Knowbrainer review of DNS 13

I have a few YouTube videos demonstrating the use of version 11.5 in memoQ and a word processor, which seem to have generated some excitement because of the ridiculously high speed at which I can translate by dictation (and many others are much faster). However, the point of voice recognition for me is not speed and the possibly higher earnings which can result if my editing afterward is not excessive (dictation requires a completely different approach to checking your work, and there is a significant learning curve here). Also (or really more) important are:

• greater engagement with the text on the screen, in my case leaving my hands free to point at various parts of long, complex sentences to help me sequence the translated text better as I work;
• less physical and mental strain during my translation work (I am less tired during and after);
• relief for hands damaged by too many years of working with vibrating power equipment (tillers and chainsaws), riding bicycles on rough ground and typing, typing, typing (some days I have to wash dishes in very hot water for an hour and load up on pain meds just to use a keyboard and mouse without tears - there may be surgery for that in my future and tools like DNS can give others relief or help prevent the sort of strain injuries too common in this profession and others which involve a lot of keyboard work). 

Dragon Naturally Speaking is currently available for U.S. English, UK English, German, French, Italian, Spanish, Dutch, and Japanese. Given the importance of voice recognition for relieving or avoiding strain injuries as well as for productivity (translators working with voice recognition routinely have much higher outputs of good quality than the best realistic claims for crap produced by post-editing machine pseudo-translations), I sincerely hope that Nuance and others will pursue the development of speech technology for other major languages such as Portuguese, Arabic, Chinese and Greek. Such an investment is likely to produce far greater benefits all-round than any money flushed down the machine pseudo-translation toilet, and speech recognition could probably also improve the working conditions of some stressed post-editors in the HAMPsTr world.

memoQ 2014: a first look

I couldn't make it to memoQfest this year - the first one in Budapest that I have missed since the event began in 2009. But the first family visit since that same year took priority, so my exposure to the upcoming memoQ 2014 version was strictly second hand until today.

I wasn't too happy with thing I heard on the Yahoogroups user list. In fact, when I read one message describing how the new transcription feature for bitmap graphics in some files required the Product Manager version, I was quite annoyed. The reality - a whole month before the official release - is very good for both freelancers and corporate outsourcers, and I think by the time this version makes its official debut in June there will be many good reasons to smile. I'm frankly amazed at how much Kilgray seems to be getting its act together and balancing the needs of users at all levels.

This afternoon I downloaded the first test release (alpha??) of memoQ 2014, installed it and began to take a cautious tour. My first impression was that it looked the same. And then, bit by bit, subtle and excellent small differences began to emerge. I looked for and found major new features I had heard about and discovered many interesting things not mentioned along the way.

The grammar checking feature seems to be implemented in a sensible way, though it actually doesn't work at all right now for me. But I can see where it's headed, and it is going in a good direction.

I had a quick look at the new plug-ins, particularly TaaS, and made notes about testing the potential for teamwork. What I have seen of TaaS for its much-advertised terminology extraction is a huge disappointment, and those who have followed my comments on Twitter will know I have nothing good to say about this EU boondoggle, but I see potential for other possibilities that nobody has really talked about, and if my instinct is right, this could be really useful. But I will need to invest a lot of testing time for the approach I have in mind.

The Project home view has gotten even more impossibly cluttered with the addition of "People", a rather sensible reworking of role assignments that even in the Translator Pro version clearly acknowledges that most freelance translators are not, in fact, 'islands' in their work.

This will surely make the small screen (netbook) usage problems worse if Kilgray does not redesign the view a bit, but in every other respect I see this as a significant improvement of project workflow, emphasizing the relationships between project participants in a better way.

One little bit that I stumbled across was the new way of handling the export of unfinished translations. This is a nice way of recognizing the frequent pressure in some projects to export incomplete stages of work.

I have had ways of dealing with this need for years in memoQ, but this new approach will make things simpler and obvious for all users.

There is a nice little feature for tracking time too:

This will facilitate record keeping for some jobs involving time charges.

The feature I have looked at in some depth so far, which makes me very happy, is Kilgray's very sophisticated handling of embedded objects and graphics, which sets new standards in many ways. I think there is still a key feature missing to make it the equal of OmegaT for handling charts with data stored as XML in the MS Office file (though I have not had time to check this yet), but what I have seen so far goes way beyond similar features I have seen in STAR Transit and Déjà Vu X2.

Embedded objects and images are imported as separate files from within the media and embeddings folders of the Microsoft Office file. I see a few potential problems with the current way of displaying a file and its objects and media. I've had projects with multiple files having embedded Excel spreadsheets, PowerPoint slides and other objects as well as any number of pictures needing to be localized. One recent project had 59 spreadsheets embedded in a DOCX file. Without an accordion or tree structure to collapse the subordinate structure view and show the embedded content again, the overview will be lost quickly. But this is a very good start. Note how the main file includes a count of the segments in the subordinate objects and graphics. (And take note of the new progress bar with different colors for different process stages like translation and proofreading.)

Bitmap texts can be recorded with a new transcription feature, which is also compatible with voice recognition. I dictated my German source texts with Dragon Naturally Speaking set to German, then switched to English for the translation. And of course the bitmap transcriptions are included in the word counts of the Statistics functions and the translations are written to the translation memory. I believe this is utterly unique in translation environment tools. Fluency has a transcription module too, of course, but its purpose and application are very different.

The exported translations with translated objects will look like they are not done at present, because the difficult refresh problem has not been solved by Kilgray. Each translated spreadsheet, slide, etc. will need to be opened in the document before the translation will become visible. This is much easier using the macro I published two years ago, and I am certain that by release time or soon thereafter Kilgray will find an elegant way of dealing with this difficulty. Atril handles the same problem by distributing macros as I recall.

In the recent Kilgray blog post on the six reasons to upgrade to memoQ 2014, the only overlap with the above points is the image localization. Peter Reynolds talks instead about other good stuff, such as the long-awaited project templates and Language Terminal. There are so many nice things ahead with this upgrade that we'll all just have to take it slowly, one bit at a time.

Of course the usual precautions for any new software version apply. The new version can be installed in parallel to your current version, and it can be tested while you continue to do the bulk of your work in the older, stable version. Typically it takes a few months for any new version to get the kinks out, but this allows plenty of time for planning the transition and preparing to take full advantage of the new features relevant to you. Migration is also not a trivial matter in many cases, but this time around there may be a little more help with that. More on that another time!