The Challenges in the Language Industry
There are multiple challenges in the world of global culture and commerce. The one common feature they all share is in communication or the use of language. There are numerous reasons for this and the business community has been extremely resourceful in trying to deal with them. The Language Industry has brought together human and technological resources to meet the hugely complex hindrances posed by a multilingual, technologically heterogeneous environment. However, the solutions in place at present still leave a number of key issues unresolved for the professionals within the Language Industry. Translation Commons has identified some specific barriers that must be overcome if the Language Industry is to succeed in its goals of facilitating global communications.
Problem 1: Demand-Supply Fragmentation
Language services are one of the more recent business sectors to benefit from the wealth of tools and techniques of business analysis. Just a few decades ago, translation was mostly provided by freelance translators. International publishing houses, academia and only a few businesses like language learning were exceptions. In recent decades, the entire multilingual landscape has undergone massive seismic shifts. As Language Service Providers (LSPs) have responded to the changing requirements of the business community, new ways of analyzing how language services can meet those requirements have emerged. One of those is in modeling a supply chain in which all the intricacies of supply and demand are met. But how successful have these models been? The answer is, so-so. A number of factors are preventing that answer from being an unqualified yes.
There is no single model of the language industry supply chain. In fact, those devised so far lie between the very simple and the very complex. The simplest model describes the process of a language supplier meeting the demand of a plain translation requirement. One source; one target.
Increase the parameters of those elements, introduce language technology, factor in cultural aspects, quality control, peer review, financial constraints and so on and simplicity yields to complexity. One of the developments that characterized the translation and localization industry during the last couple of decades was consolidation. From at least a dozen large vendors, we are currently down to a handful, publicly trading or with full financial backing. Consolidation also manifested itself in the emergence of a relatively standard production outsourcing model. The larger vendors took on multilanguage, multi-service projects, outsourcing the core translation services to single-language vendors, one in each target country. These smaller local LSPs normally work from one or more source languages into one target language only, and either work with on-site translators or contractors.
This particular enlargement of the supply chain has resulted in the translators being extremely removed from the client. This fragmentation of the language industry is problematic because it creates division between clients (demand) and translators (supply). The consequence of this is that information which should be freely flowing from the client to the translator is blocked or simply lost in the large chain. The LSPs have used this block as a means to de-value the service of the translator as being easily replaceable should the need arise. The freelance translator as an individual doing business is no match to the LSP which as an organization is able to meet the clients demands and invest in similar automation.
Because of this ability of the large LSPs to match client investment in automation, the language industry operates with an asymmetrical supply chain. The automation and localization efforts of large corporate ventures have resulted in very sophisticated technology to power translation using MT and translation memories, in turn managed with version control and project management software tools. Yet the rendering of source language into target translation is still performed by highly-trained translation professionals. Somehow the requirements of the demand end of the chain are disproportionate to the supply end. Asymmetry exists, disrupting efficiency.
It seems plain, therefore, that tech alone is not the solution to the problem. What is required is a means of strengthening on an equal basis the links in the supply chain, thus providing full recognition of all of those contributing to its functions.
Problem 2: The Technology loop
All technology funding in the language industry goes to Academic Research and Development. There are many projects that are high profile like speech recognition and machine learning, just to name a couple. The findings, results and automation code, is then used by Automation Providers in the Language Industry, to create proprietary tools. The large corporate buyers of translation are the first ones to adapt these new automation tools, gaining benefits like consistency or reuse of segments to drive prices down.
The most widely used tool today by translators is CAT (Computer Assisted Translation) which helps them build their own Translation Memories. These memories are designed to facilitate consistency, speed and assist translators with their mainstream translation. However, benefits and cost savings from the use of TMs, quickly shifted from the translator’s desk to the localization vendor and eventually to the customer. Today, no localization quote is sent out without a detailed breakdown of full matches, fuzzy matches and repetition discounts. Translators’ rates are being constantly reduced and their work is rapidly shifting from a creative process to editing machine outputs. With the recent advent of cloud based CAT tools, the automation providers are also now accumulating the Translation Memories of thousands freelancers who do not have access to them.
What started as a benefit to translators, building their memories for re-use, has now become a trap with dire financial implications for freelancers. The better and bigger their built memory is, the more dependent they are on the specific proprietary CAT tool. With annual upgrades and older versions not being supported, translators find themselves continuously investing to technology that is not giving them much added value. Different CAT tools offer slightly different benefits and LSPs and clients may request translators to use different tools per client. It is not unusual for freelancers to work with 4-5 different CAT tools. This is a major financial imposition on freelancers who have to purchase licences and upgrades just to be able to work for a specific client or LSP. In addition, their memories get diluted over multiple CAT tools.
A side consequence of this is also a little understood issue. The more a freelancer specializes in one subject matter and builds better memories, the more difficult it is for him/her to diversify or add another specialization, due to a complete lack of memories, while other specialized translators have extensive memories. This is also true for new graduates trying to enter the freelance market where they compete with translators who have a great output due to their extensive memories built over years. Switching or acquiring new subject specializations is now becoming an extremely difficult task for freelancers, restricting them from many opportunities. At the same time the increased demand from clients for specialized freelancers is not able to be met. This results in a few translators earning disportionately high rates and not being able to meet the demand, while the vast majority of translators of that language combination are desperate to become specialized, but have no opportunity to enter the freelance pool.
With new versions of tools and new tools altogether being demanded by LSPs and clients, translators either invest, train and learn or are left behind. And while the ones who accept to do so, they are penalized by imposed lower rates and post-editing rates as opposed to mainstream translation rates.
Although there are a number of open-source language related tools available, the vast majority of translators do not possess the necessary technological skills to adapt them and customize them. Some freelancers would even find downloading a challenge.
Problem 3: Data Ownership
Add to the technology loop described in the previous problem, the fact that correctly translated data is needed by Machine Translation (MT) Automation Providers to train their MT engines, it is no surprise that translators’ data and memories are a highly valued commodity that some go to extreme lengths to secure. In 2015, the language industry was shaken when MT developer X partnered with Cloud CAT tool provider Y, where X would provide free MT in return for all cloud memories collected from unsuspecting freelance translators using the free cloud CAT tool Y.
This drive to acquire the translators’ memories to be used as data training for the MT engines of big corporations and LSP’s can have far-reaching legal implications. Since clients own copyright on original material, and translators own the translated text copyright (at least in some countries), all other parties may be in potential copyright breach. As long as the client has paid the translators and is training their own MT engine, there may not be actual copyright breach but that on its own can result in an unjust situation where the work of a translator is unattributed and reused. The question is, “Whose data is it anyway?” For all the resources invested in tech, human translation still plays the fundamental role in supplying and maintaining translated texts. However, these translators often find themselves unable to own their work as corporate enterprises assert pan-ownership.