KantanTM: on-demand machine translation
KantanTM is a new age machine translation solution that makes machine translation accessible to smaller companies. The fact that the system is fed with texts from specialised fields and confidentiality are its greatest features.
A newly-created company
KantanMT was formed on 2 March 2011 and the newly-created company has secured investments for a business model that promises to grow rapidly in the next few years.
You can do it yourself
What KantanMT brings to the machine translation market is that the machine translation engine can be configured and customised with ease. Furthermore, the monthly fees are based on the volume of words translated, directly from the cloud, and therefore it is not necessary to have an internal computer infrastructure in order to use this solution. This price flexibility along with the simplicity of the process for creating personalised machine translation engines represent added value and competitive advantage compared to systems such as Systran, which require a significant initial investment along with specific staff with a high level of technical knowledge.
Based on Moses
While we have still not tested the efficiency of this system at AbroadLink, the fact is that KantanMT is based on the machine translation system Moses. Moses is an open source statistical engine and one of the most widely-used systems, so theoretically this is a guarantee of the positive results that may be achieved with KantanMT. Ultimately, KantanMT is the easy-to-use user interface that will enable us to configure and personalise a system such as Moses. If you're interested in finding out more about Moses, please visit this page: http://www.statmt.org/moses/.
No data, no system
Despite the flexibility of the price policy and the low level of technical expertise necessary to use KantanMT, making it accessible to smaller companies, any statistical machine translation system needs a large volume of data in order to yield results, resulting in greater efficiency in the translation process. According to what we were told at the Conference, in order to set up an engine within a specialised field, a minimum of 2 million words must have been translated previously by human translators until usable results are achieved. Precisely as a result of this need for a large volume of data, at present only companies that have such needs can actually use this technology. We are talking about companies such as Microsoft, Symantec, large financial corporations and, in general, multinational companies whose annual translation needs in a specific language combination are well above one million words.