Fiorenza Miriam Catania, an Italian student at University of International Studies of Rome (UNINT) recently completed her Master’s thesis on Machine Translation output for the Medical domain. In this post, she shares her experience of using KantanMT for her research.
KantanMT: Could you tell us about how you used KantanMT in your research project?
Fiorenza: Since my MA thesis was aimed at testing the output of a Machine Translation engine, I was given the opportunity to use the KantanMT platform to carry out my research in the specialised field of medicine.
It was very easy to set up an account on the platform, which I think is very user-friendly. The different features on the platform come with Help sections (KantanInsights) with a clear explanation of the functions. I could easily create a profile, choose the specialised server and upload my source text; and in just half an hour, the translated file was ready for me to download.
KantanMT: Was there any specific feature on the platform that you liked especially?
Fiorenza: I think one of the best features on the KantanMT platform is the fact that the engines can be easily customised to a very high level of accuracy. I could train the engines with domain specific training data, which in turn helped me improve the engine quality substantially and decreased the translation time. I believe the ability to customise the MT engines makes KantanMT a very useful tool to translators.
Another very good feature of the platform is KantanAnalytics™, which generates quality estimation scores for automated translations from the KantanMT engines. KantanAnalytics creates a detailed project management report of all segments within a KantanMT project. This includes segment-by-segment quality estimation scores in addition to other useful project statistics such as word, character, placeholder and tag counts. I found this feature highly beneficial because it allows translators to estimate the effort and time required for post-editing.
KantanMT: How did you evaluate the quality of your translations?
I analysed the quality of the translation output by creating labels based on linguistic rules. For example, the syntactic order of elements, the grammatical concordance and typographical problems, each were labelled separately, and I evaluated the translation quality based on these Key Performance Indicators and Error Typologies.
KantanMT: Did you use your own training data to improve the quality of the engines?
I did not use personal training data to train the engines. Instead, my thesis aimed to analyse the performance of a Custom Machine Translation, when they are trained with public documents. I found that even by using domain specific stock data, the translation quality of the MT engines was vastly improved.
Personally, I would strongly recommend the KantanMT platform to translators who are looking for a high-quality MT solution, which will enable them to complete translation projects faster and with more efficiency.
About Fiorenza Miriam Catania
Fiorenza Miriam Catania is an Italian student of foreign languages. After a degree in Linguistic and Intercultural Mediation, she recently graduated in Interpreting and Translation (MA). Her thesis focused on the use of machine translation in the field of electro-medical devices.
About the University
University of International Studies of Rome (UNINT) is a private Italian and state-recognised elite university located in Italy. The Faculty of Interpreting and Translation at UNINT is internationally renowned. It is one of the three Italian members of CIUTI, a prestigious international association representing the world’s best university faculties for translators and interpreters, and its Master’s program is now a member of the European Master’s in Translation (EMT) after passing a strict admission procedure. It also coordinates a Work Group in the European AGORA project whose aim is to establish a system of international traineeships especially for translation students.
About the KantanMT Academic Partnership
The KantanMT Academic Program provides students with the knowledge, expertise and resources to help prepare for a career within the translation industry. This academic program gives students the opportunity to gain experience in customising, improving and deploying Statistical Machine Translation engines using the KantanMT.com.