How will translation tools be affected by the continued development of machine translation in the next decade? (by Jost Zetzsche)

Offline wings

  • Global Moderator
  • Hero Member
  • *****
    • Posts: 67771
    • Gender:Female
  • Vicky Papaprodromou
    • vicky.papaprodromou
    • @hellenic_wings
    • hellenicwings
    • Βίκυ Παπαπροδρόμου: ό,τι πολύ αγάπησα (ποίηση, πεζογραφία & μουσική)
The Questions
 
Thanks so much for sending in questions that you would like to ask translation tool vendors! The questions ranged from fantastic ones like this . . .

I would like to ask the translation tool vendors what they think of the nun and if they think Jeromobot would be better equipped to deal with the difficulties he faces if he had extensive training from a Jedi master. (Fortunately I did not have to ask the vendors since Jeromobot is being trained as we speak!)

. . . to some that I can answer myself (more on that in 2010 editions of the Tool Kit) . . . to several that lend themselves perfectly to stumping the experts. I summarized the latter into these four questions:

1. How will translation tools be affected by the continued development of machine translation in the next decade?

2. Are you anticipating significant developments in terminology management tools after a long time of things being pretty static?

3. Is it desirable and feasible to have a clearer distinction between slimmed-down, easy-to-use tools for translators that require essentially no training and full-featured tools for agencies and clients with much higher requirements in technical expertise?

4. What percentage of the current field of tool developers will still be around in five years? What will define the survivors?


I also followed these four questions with this friendly reminder:

My experience is that the less you talk about how fabulous your own tools are, the more impact your message will have.

The result: some great answers and some very revealing responses about the different tool vendors. Each tool vendor who responded is included. If you notice some missing, chances are that they took time off between Christmas and New Year and were not able to respond.

Happy New Year to y'all. Here are the questions again with the vendors' answers:
 
1. How will translation tools be affected by the continued development of machine translation in the next decade?
 
Across: Translation management and translation memory tools need to integrate machine translation as an optional workflow step, e.g., for the pre-translation of segments for which the translation memory does not contain matching entries. Editors must be optimized for mixed post-editing of both TM matches and machine translation.

Terminotix: Translation tool developers will have no choice but to integrate online machine translation and intranet machine translation servers. Machine translation will suggest translations when no match is found in existing bitexts. With the availability of machine translation on the Internet, confidentiality problems will arise and translation tool developers will have to be careful when they use those engines, especially in government linguistic services.

Kilgray: I believe that machine translation and translation memory tools will eventually converge, and the integration will be deeper than just plugging MT into a translation environment. If business allows for that, MT engines will learn from the translators' actions, tracking down what the users actually edit. This is not something that will come in the next three years, though.

Atril: I think it is likely that machine translation, and in particular statistical machine translation, will eventually be tightly integrated with translation memory. Once subsegment-level reuse of translation memories becomes widespread, tools will be able to use SMT not only as a safety net if the TM contains no useful matches, but also as a way of generating subsegment matches when piecing together translations from multiple sources.

Maxprograms: If machine translation keeps improving, users will request that more MT engines be incorporated in their translation tools. This is already happening: some tools use Google's translation engine now. There's no need to wait for the next decade.

Total Recall: We should see a more seamless integration of machine translation in translation tools, with computer "suggestions" involving a mix of machine translation and translation memory results instead of either one or the other. Hopefully clients will stop overrating machine translation, and more translators will realize that it does have its usefulness.

STAR Servicios Lingüísticos: Translation tool vendors should seriously consider how to interface with MT engines. They should be aware of their strengths, try to benefit from the renewed MT hype, and try to be the driving force behind the changes which are going to happen. With translation tools already integrated into the localization workflows of many multinational companies, it is actually the MT vendors who need the know-how of translation tool and CMS and IMS developers. MT and TEnTs combined together will change the role of translators. Instead of translating, translators will be more and more responsible for reviewing the translation proposals from combined TM / MT engines.

2. Are you anticipating significant developments in terminology management tools after a long time of things being pretty static?
 
Across: Yes, we believe there are significant developments, such as the use of data mining methods for advanced term harvesting for the purpose of generating terminology. Developments on the usage side include the unified utilization of terminology for both translation and authoring support.

Terminotix: With the recent development in bilingual terminology extraction, "term spotting" will become more and more available so terminology creation will be almost automated.

Kilgray: Other than making terminology building a collaborative effort, I don't see any breakthrough in that area. A more interesting area is how you can increase leverage and precision through the use of terminology.

Atril: I don't see standalone terminology management tools changing much, but their integration with translation tools will continue to improve. Given that consistent terminology and wording is a key aspect of quality in translations, automated term and phrase extraction (or detection) will be used not only for producing glossaries, but also as part of the process of quantifying and enforcing consistency.

Maxprograms: There are well-known terminology managers in the market now, and for whatever reasons translators still use Excel to hold their glossaries. I expect an improvement in glossary management tools. Glossaries are simple things, but terminology management is too complex for most translators.

Total Recall: Terminology management tools have to become more dynamic, helping the user to identify and define new entries without slowing down the process.

STAR Servicios Lingüísticos: I do not think there will be significant changes to terminology management tools. They are a mature technology, like TEnTs, and technological changes will be only minor and gradual. What is more likely and more necessary is a change in terms of how these tools and related services are marketed. People don't do terminology because they do not see the benefit of it; they are even less aware of the cost/benefit relationship. So it is necessary to change that perception if terminology is to gain the importance it should have in the localization and content creation process.

3. Is it desirable and feasible to have a clearer distinction between slimmed-down, easy-to-use tools for translators that require essentially no training and full-featured tools for agencies and clients with much higher requirements in technical expertise?

 
Across: The ideal tool should offer both an easy-to-use mode for standalone use and an advanced client/server mode enabling the user to be part of a seamless process and centralized workflows.

Terminotix:  That may be the case for large agencies and for linguistic services, but not for smaller agencies or freelancers, since they need to accomplish many complicated tasks with their tools.

Kilgray: It's a very hard nut to crack. Ideally this is desirable, but in reality a feature-rich interface for a translator has just as many -- though often different -- benefits as a feature-rich interface for a project manager. Productivity is the sum of efficient project management and efficient translation. The typical career path for a translator who gets bored with translation is going into translation team management and/or project management. Extending this gap between translators and project managers could have a bad effect on the industry as a whole, as good project managers should know how translators are working and this experience often comes from a background in translation. >From the user feedback perspective, I am pretty sure that if a vendor released two tools, one slimmed-down, easy-to-use tool for translators and one full-fledged tool for project managers, translators would argue for more functionality and project managers would argue for ease of use. I believe only a very big player could come up with and maintain two tools, in no small part because this is quite a humiliating thing for most vocal translators ('they think I am stupid'). For the time being, I believe no one in the industry with a decent market share is big enough to be able to afford this.

Atril: There is obviously plenty of room to simplify existing translation tools by removing a lot of the complexity (most of which probably comes from the misguided idea of giving the user a lot of choice in terms of configurable options). However, technical translation is part craft and part industrial process, and there are many different things a translation tool can do to save the translator time and effort, so I don't think the complexity of professional tools for freelancers will decrease much (although their efficiency and user friendliness will), particularly if more emphasis is placed on quality assurance.

Maxprograms: Users differentiate between the tools now. Beginners don't use complex tools, while some experienced translators want full control and advanced features. Is it desirable to differentiate in a special way? It really doesn't matter. Users compare tools today and choose their preferred ones based on multiple factors.

Total Recall: What is both desirable, and feasible is to make more effective tools available for all users! In the translation industry, computers are still helping us all to solve problems we never had before, and that situation has got to change. I have had too many translators tell me, "I have to turn off (Tool X) for rush jobs because it actually slows me down instead of speeding me up!" To me, if something is difficult to learn and time-consuming to use, it fails to meet the basic definition of a tool. Sure, a bicycle mechanic may get more out of their bike, but that should not be a prerequisite for riding one. Translators have on the whole been far too technology-shy, and this shyness has been both exploited and perpetuated for far too long by complex, hard-to-learn tools forcing the user to work around fundamental problems the developers failed to address. Yes, it takes a lot more work, and a great deal more internal complexity, to make an outwardly simple tool than to make a clumsy one, but it's time for translators to realize that they are being had and to demand simpler, less distracting tools for what is already a demanding enough task.

STAR Servicios Lingüísticos: This is a tricky one, because where do you want to draw the line? Up-to-date information about functionalities and their real-life utility is what the users need to be able to look behind the advertising smoke screens. This is where you are already doing a great job with your Tool Kit and the translatorstraining.com web site. So, instead of making artificial distinctions, keep up the good work and provide more benchmarking information. Functionality comparison sheets help a great deal to do a proper benchmarking and to see which tool does the job best. These could be enhanced by performance benchmarking tests, so as to compare the tools in terms of performance.
 
4. What percentage of the current field of tool developers will still be around in five years? What will define the survivors? (Premium Edition)
 
Source: Jost Zetzsche: The 156th Tool Kit - Standard Edition
« Last Edit: 25 Jul, 2019, 23:22:50 by spiros »


 

Search Tools