|
Financial Translation on Demand
When a large investment bank needed a rapid turnaround on translation of financial information by a wide-spread network of financial translators using the latest in technology, it turned to Leeds-based thebigword. In order to meet the bank's requirements, thebigword designed a standards-based web workflow capable of returning finished translations from human translators within as little as 45 minutes. Ian Harris, technical director at thebigword, details the system and how standards help get the job done.
LISA: Tell us a little bit about your financial/Equity Research system project (the scope, timeliness and customer's requirements). Ian Harris: Our client (a large investment bank) has an equity research team based all over the world. Market analysts create articles that advise the bank's equity dealers on the stock market. Since dealing decisions are based on these articles, they are highly confidential, very specialised and time critical. Articles are authored onto the bank's own content management system, a web based tool written in JAVA on Unix with an Oracle database. The bank's in-house technical team developed the system. The client's reqirementsOur client needed to make the research available in multiple languages, to teams of investors and other 'customers' in a timeframe that would ensure that the articles maintained their power. Articles can be authored at any time of the day or night, but the bulk are completed at approximately 6:30 AM. The main language pairs are English to French, English to German, and German to English. There are occasions when other language pairs are required. IssuesArticles are stored internally on an Oracle based Unix platform, behind the bank's firewall. Translators are freelance, based all over the world, using Microsoft Windows-based PCs. The solutionthebigword have a job management server used in such cases that has an encrypted XML interface. This enables translation jobs to be sent to us over the Internet. Working with the client's technical team, we married their content management system with our interface, meaning that as articles are completed, they are automatically encrypted and sent out over the Internet to our job management server. Translation work can therefore begin instantly. The articles, themselves XML based, are translated by highly specialised translators. XML tags are protected from the translators by software, then on return their XML structure is checked automatically by programs written by thebigword. Translators are all pre-sourced and waiting for the articles. The translators all have many years' investment banking and equity research experience. So specialised was the field of excellence that we had to make amendments to our system's selection procedures to cope with the levels of experience required to translate the different types of article. The interface we have developed is SOAP (Simple Object Access Protocol) compliant. This can be integrated easily by any content management system vendor, and negates the need to select a specialist content management system for projects such as the above. The client kept their original content management system in which much investment had been made, and which was specialised for the task. We integrated this with automatic linguistic processing using our interface. The solution operates through the client's firewall with no changes to the configuration of that firewall due to our use of standard protocols and encryption. This particular project has been running for nine months and the average turn around time of the translations is 45 minutes. This time is almost 100% translation time due to the automation of all possible processes. LISA: How do you manage the web-based translation and localization tasks? Ian Harris: The articles are extracted from the Oracle database and packaged in XML. These are sent over the Internet to translators who work in Windows-based CAT tools on their desktop. We can therefore leverage any desktop CAT tools since the translators need not work through a web interface. One of the aims when we approached this project was to ensure we could get the benefit from translation memory and other tools, in a form that translators would be familiar with, by avoiding working through a web interface. Translators use their tried and tested tools on this project and do not need to learn anything new. LISA: Please walk us through your selection process for determining the "off the shelf technologies" you use. Ian Harris In terms of CAT tools, the selection process was simple: The tools had to have the ability to filter or protect XML tags, and had to be widely used by translators. The translators are very specialised and hard to find, so we didn't want to put up barriers to working with us in the form of obscure tools. Trados (latest version only) was chosen. The memory is managed centrally by thebigword. We developed the other technologies we use in-house, so we chose standard protocols such as XML, HTTPS, SOAP, etc. LISA: What language technologies do the localization and the translation aspects of your project use and how do you handle issues of communication/re-use between various products (if this is relevant)? Ian Harris: The communication between all products (from the client's Oracle databased to our job management system, and from our job management system to Trados and back again) is all via XML, but encrypted at stages along the way. LISA: What role do language-technology standards play in this project? Ian Harris: The language-technology standard relevant in this project is TMX for the translation memory. The standard does not play a great role in the project; it does, however, create a 'good feeling' for our client and us, in that we are not tied to Trados, and only use Trados because it is the most suitable tool at this point in time. LISA: How are Standards factors taken into account and why are they important to both TBW and your client? Ian Harris: We adhere to standards that are pertinent to the project. We do not adhere to standards for the sake of adhering to standards; we do it because the standards help. This is especially important in a project such as this where we have to facilitate integration between various parties all over the world on different operating systems. LISA: How is consistency and cohesiveness maintained in a project that would be difficult enough to manage if it were all "in-house" for the client or a commercial software developer? Ian Harris: We have hand-selected a team of translators for this project. Also the client authors, while based in different locations globally, are a finite team. A stable team of authors and translators brings inherent consistency. We also employ terminology management (see below). Another simple, but key piece of technology we use is a web-based chat system. This is invaluable for the resolution of terminology issues at speed. LISA: How are QA issues handled? Ian Harris: We have a set of QA procedures that are integral to this project. These include: random sampling of the articles and formalised error checking (using a subset of the J2450 checks); time based performance indicators (each article's arrival and delivery time is automatically logged); weekly qualitative feedback from the client (again this is formalised, with terminology updating procedures). If QA issues arise, we work with the client for a resolution. We designed the procedures together, we operate the system together, and therefore take joint responsibility for resolution of QA issues, even if this involves changing translators. LISA: How are terminology issues managed? Do you make use of any terminology management technology to handle these issues? What constitutes the "best" terminology management tools and why? Ian Harris: We use Trados Multiterm to manage the terminology database. Translators, the client, and we, can formally suggest additions or amendments to the terminology database. These are agreed by the client prior to insertion. The client also supplied a terminology list prior to commencement of the project. The client places great emphasis on style, which encompasses the terminology used, but also tone of voice. Much effort went into pre-picking the translators. LISA: What is the typical profile of a localizer working for TBW on a project like this? Are they professional translators or amateurs with good linguistic background? Ian Harris: Translators are all professionally qualified (in translation) with a number of years' commercial experience in their field (equity research). For this particular project, they were extremely difficult to source since the level of experience required was high. All translate into their mother tongue. LISA: What standards relating to language technologies does the project make use of and how? Ian Harris: It makes use of TMX, as described above. LISA: Are there competitive advantages to TBW's web-based translation model? Ian Harris: The advantages of this model are as follows:
is Technical Director with thebigword, based in leeds, UK. He is a Management Graduate with an MSc in Computer Graphics and Internet Technologies and has worked in the computer industry for 10 years, running a software business that created e commerce applications. He worked as a supplier to thebigword for 2 years and having seen the potential, joined in February 2001 as Technical Director. |
![]() 8-12 December 2008 |
||