Under labour market development agreements, all provinces except Quebec and Saskatchewan have decided to continue using Job Bank for their online employment services.
As it is a national service, we upload the offers from Quebec and Saskatchewan, which are then made available to all job seekers in Canada. Employers post job offers online using pre-translated skills checklists and some free text for additional information if required. Job seekers fill out job profiles using the same skills checklists and are then matched with employers.
I am now on page 4, where we talk about the problem. To understand why we have adopted an automated translation system, it is important to know the scope and nature of the service. I believe this to be the main issue of today's discussion. With approximately one million job offers averaging two to three positions per offer, job seekers had access to three million job opportunities last year. In addition, Job Bank logs over 60 000 user sessions per year.
Over the past two years, we have seen an annual increase in demand of 30%. The objective of Job Bank is to respond to the pressing needs of job seekers and employers—hence the quick posting, limited lifespan, and simple and repetitive language of job offers. The service is transactional in nature. I would point out that this service is always very urgent for employers. We receive a very high volume of texts, and documents that come to us externally differ greatly from the static content for official publication by a department.
Further, in order to meet our service standards, we have a short turnaround time for the translation and revision of free text. The turnaround time is approximately four hours per text. As the lifespan of offers posted is approximately two weeks, the level of translation quality we demand is different from that of larger texts that are available over longer periods. The translation system has equal capacity in both official languages and depends above all on the language quality of the source text.
I am now on page 5 of the presentation, on which we talk about challenges in relation to translation. As these texts are written by a third party, Service Canada is not legally required to translate them under the new directive which states that third parties are not subject to the OLA. Service Canada does so, as requested by the Commissioner of Official Languages, in order to support official language minority communities and to encourage workforce mobility.
In 2002, when Job Bank gave access to employers to post their own offers directly online, we identified problems related to the translation of job offers. We approached the Translation Bureau, which estimated the cost of translation of the 450,000 jobs offers available on the site at the time without relying on technology to be $25 million. Given time and cost constraints, and considering the large quantity of offers to be translated on Job Bank, we opted for the use of pre-translated checklists as well as an automated system accompanied by human revision to review free text, and hired a team of revisers.
In addition, we followed the recommendations of experts at Technolinguistic Services at the Translation Bureau to develop our system.
In most cases, automated translation is not appropriate; however, where there is a large volume of texts to be translated in a very short time, it is impossible to use human translation to deliver texts quickly enough and costs become exorbitant. In addition, since the messages are simple and repetitive in nature, fast revision of translated texts is possible in order to ensure comprehension.
Now on to page 6, where we talk about our three-year action plan and objectives. To resolve the problem of translation quality, in 2003 the Office of the Commissioner of Official Languages and the department approved a three-year action plan. The objectives of the plan were to reduce the quantity of free text to be revised by expanding the pre-translated checklist, to improve the translation system by adding a translation memory, to integrate quality-control mechanisms and to increase the proportion of offers being revised, all the while ensuring consultation and evaluation processes.
To ensure effective service management, we adopted a segmented service approach, providing priority service to bilingual communities and positions requiring bilingual capacity, thereby concentrating our efforts where they are most needed.
I am now on to page 7, which further talks about the three-year action plan. The creation of a revision team along with the combination of various technologies has enabled Job Bank to optimize the performance of its translation system. In fact, almost 90% of all job offers currently posted are of good linguistic quality in both official languages. This percentage covers 100% of bilingual communities and positions requiring bilingual capacity.
These results were achieved through improvements to the translation system, specifically by increasing the pre-translated checklist to 95% of offer content, limiting free text to 300 characters, adding new components—specialized dictionaries, a recurring text database, a translation memory and a new spelling and grammar checker to support employers when creating job offers, and increasing the number of revisers.
We have not only saved millions of dollars through our automated translation system complemented with revision, we have also put in place a system that is recognized as a leading example by language technology experts in Canada and internationally.
Now on to page 8, on which we talk about where we are now. The number of complaints has decreased drastically, from 16 in 2003-2004 to seven in 2004-2005 and falling to three in 2005-2006. Problems in terms of quality that we currently face are restricted to the 13% of offers that are not revised due to time constraints.
To conclude, let us move on to page 9 where we talk about considerations. I hope that we can ensure translation and revision for all jobs requiring bilingual capacity and all jobs for bilingual communities. Our goal for next year is to improve the translation system and reach our 100% goal.
I am now on to the last page.
Our goal is to review 100% of job offers. However, even if we were to increase the number of revisers, given the constant increase in job-offer volume, on certain days we would be unable to review all offers within our four-hour turnaround time.
I hope that you consider that our very specific goals for next year, as outlined in our action plan, will indeed contribute to improving the process.
I conclude my presentation, Mr. Chair.