Conference Program | 2023
Master classes and presnters will be announced shortly.
Let’s look at the client side of the translation market: Clients range from the agile and quick-acting who may be dealing with dozens of language combinations, to niche markets with single language combinations.
Some clients have in-house language services, others refuse to work with agencies. However, they are all looking for similar things from translation and localization providers.
With over 20 years of client-side experience combined with years of freelance translation experience, I’ll share what has worked forus as end clients in different fields, what didn’t work.
We will also review what some of the biggest red flags have been, and how to avoid them.
If you are thinking about going global, you should consider digitalizing your sales team! Digital assistants can serve many different and creative functions. The time is past when digital assistants only answered a basic list of pre-determined questions.
Digital assistants can serve as an attractive new sales channel to draw in clients, digital assistants can help you manage client relationships, answer ever more complex questions and solve sophisticated client issues, and navigate your product features.
When you go global you will probably also want to present your services to clients in their language. The newest generation of digital assistants are fully multi lingual, which means that they cover the full range of duties in all the languages that they are trained to use.
We will share our journey with hiring and training digital assistants and illustrate the costs and benefits of using digital assistants. The newest generation of digital assistants are multilingual, they speak, they understand speech and writing and they can process information from multiple date bases. Instead of single answers to multiple questions, these digital assistants can process multiple answers to to each question.
Assistants can be developed with personalities to fit their responsibilities and can be customized to detect and respond to emotions.
This is a fascinating new horizon and we give insights from research and experience into the future of hiring and training digital employees.
Today, many systems produce XLIFF files, but it seems that some developers at these tool providers did not really read or understand the XLIFF specification. They do things with XLIFF that XLIFF was not meant to do, like embedding HTML or adding length limitation information.
But memoQ would not be memoQ if there was no way of dealing with such files through its powerful file filter settings.
XML: either you love it or you hate it. There’s no compromise.
An XML file may seem the stuff of nightmares to an untrained translator, but is this always the case? A lack of knowledge means you may have to turn down a job, and this is exactly what translators and LSPs want to avoid, especially if there are hundreds of thousands of words to be translated in that job.
Our presentation will show how we were able to complete the translation of an entire website for a museum containing more than 200 documents within the deadline agreed with our client, using memoQ’s regex text filter in combination with regular expressions and a set of light resources to speed up the process. The presentation will illustrate the project’s features, the rationale behind the use of the regex text filter and some of the knowledge gaps we had to bridge to deliver the project on time.
We use translation memories to re-use what has already been translated. TM entries are considered safe since they come from previously approved translations. Without a TM, we must translate from scratch, even with the help of Machine Translation that requires specific editing.
So, we use TMs wherever possible. And if we don't have a TM, we leverage existing translations through alignment.Alignment is a time-consuming process that requires both a source and a target version of the document with identical structure.You may have target documents that contain approved translations with the correct style and terminology. But without a source to align to, you can only use them in a corpus for reference or concordance.The solution I will present converts target documents into a TM to deliver approved translations from your legacy documents.
Based on "reverse machine alignment,” it can also be used for hard-to-align documents such as alpha-sorted, broken PDF, complex layouts, etc.
At Amazon Games, memoQ powers our localization pipeline - and Amazon Web Services is what powers memoQ.
We built our pipeline to be "cloud-first" and leveraged existing AWS technologies to automate and scale our pipeline wherever possible. In this presentation, we'll talk about the AWS services we selected to support memoQ when building our end-to-end localization pipeline.
We'll talk about security, network topology, storage, pre- and post-processing steps, workflow automation, and memoQ server configuration.
"Meet the Product Team." Take a behind-the-scenes tour with the Product Team and learn how product roadmap planning and customer needs are aligned, how user ideas are processed and even turned into memoQ features, and - the most important of all - how our users are involved every step of the way.
Nobody ever looks at all the options and checkboxes in a dialog, right? This is why we sometimes might miss a useful feature.
Join me in exploring some details that are easily missed, but could save you some time and effort.
Do you know those projects with hundreds to thousands of products (apparel, electronics, tools, etc.)? Segments and segments with just words that depict the product, the brand, dimensions, colors, characteristics, etc. and, no verbs?
How do you approach them? Do you rely on the classics, TM + TB, or maybe even non-translatables?
Are you annoyed by the low fuzzy matches where there is only one word different (usually a code, a brand name...)?
Do you trust your resources to translate everything correctly, consistently, without missing some special characteristics of a certain product?
Okay, I will stop with the questions and offer my take on such projects – are you ready to sit back and see how memoQ can “translate everything” for you? (Sorry, this was the last question).
Regex for solving text problems in translation is too often taught as arcane syntax that everyone "must" master, and few do in fact, which can be confusing and discouraging for most people.
In this presentation, I don't want to explain how to write regular expressions or what each character means, but I want to demonstrate how everyone can make helpful use of regex in memoQ.
At LinguaeMundi, we used RegEx to clean up our translation memories which were created over many years and therefore had a lot of unnecessary information, and to create better QA profiles to fit each client's style guide and the company's quality standards.
I will be presenting how we did it and sharing what we learned, to help other LSPs or freelancers who may be experiencing the same problem.
Talking about translating subtitles is in a way pointless, like talking about translating Word documents. There are however some considerations that should apply to any translation task involving subtitles, irrespective of the content – whether it is for a Hollywood Blockbuster, an advert for a coffee machine or a corporate training video.
memoQ is implementing NFA in its TMS. This revolutionary feature created by SYSTRAN called Neural Fuzzy Adaptation enables to make use of mid-level fuzzy matches as an input to feed the machine translation engine. This feature has several advantages such as improving Machine Translation quality, maintaining a consistent brand voice or improving productivity by reducing post-edition.
The latest versions of iWork have changed from the 2009 version (iWork '09), they changed from xml files to .iwa files, which are binary files (Snappy compressed).If we rename a file created with '09 to .zip and extract it, we have an xml file that contains all the text that needs to be translated.
So we have two ways to translate these files:
- Export the files in MSOffice format, translate them, open the files in iWork (but this can take a lot of time on the DTP side).
- Export the files to iWork '09, unzip them and work with the xml files they contain.
Once the files have been translated, you can import them back into their zipped version of Pages or Numbers or Keynote, rename the file to remove the zip extension and save it to the latest version.
This can greatly reduce the amount of time the DTP department has to spend "reshaping" the files. Something similar can be done with Articulate docx files.
Articulate can export two file formats:
- Xliff files (these can be translated using memoQ's standard xliff filter)
- Docx files: docx files consist of a sequence of text and tables, where you only need to translate the 4th column of the tables.
Again, a docx file is a zip of xml files containing all the text we need to translate.We can create a Multilingual XML filter that only retrieves the content of the 4th column of each table.
Text strings that contain tags protected in memoQ are difficult to handle by MT. A sentence segment, with tags inside, is divided into sub-strings around the tags and each sub-string is translated as a separate sentence so that the output translation is typically grammatically incorrect. This is a very common problem with XML and HTML files.
To help with it, I developed scripts that pre-process source files by replacing tags with generated keys. Those keys are not protected in memoQ, and they are ignored by MT so that a sentence containing them translates properly - as one, whole sentence. After machine translating such pre-processed source and exporting the files, I run a post-processing script that replaces keys with original tags.
In my presentation, I will discuss the technical details of this approach and how it can improve the quality of MT output for tag heavy source and reduce the need for post-editing. I will also talk about our experience with training custom NMT models and their positive impact on quality and accuracy of the MT output.
Are you prepared to enter the shadowy world of customer-supplied source files and confront the challenges that lie within?
Join the Source File Freak Show where we'll uncover the problems and solutions of dealing with corrupted, misformatted, and broken files. Through presenting case studies and real-life examples, we will explore the best practices for fixing bestial source files leveraging the capabilities of memoQ. Join at your own risk!