PIM and Translation Memory Systems (TMS): A Formidable Blend
As translations get increasingly outsourced to 3rd party translation agencies or service providers, translation memory systems (TMS) as a concept gains immense traction and becomes a crucial talking point for companies looking to optimize processes at a reduced cost.
A variety of businesses are looking at the possibility of adding a TMS during the PIM conception phase to add value in the realm of knowledge management; this indicates their awareness of incorporating promising technologies to improve efficiency.
Setting up the structure with TM can be helpful at a variety of touchpoints across organizations, whether it is to do with improving searches, data quality, translation flow, or if it’s only to homogenize company-wide wording parameters. From our standpoint, we see at least three possible solutions for integrating a TMS with PIM:
1. Building simple translation memory function into PIM’s existing translation features: One thing you must consider is that such a feature might sound helpful, but it cannot compete with state-of-the-art solutions like SDL Trados, Across Language Server and Wordfast. It may sound similar to a spelling and grammar correction solutions, but that’s not so (after all, your homegrown solutions can never compete with language improvement software like ‘Ginger’ or ‘Grammarly’). So, how does a simple translation memory such as this, built into PIM, should look like? It could be a text editor having only an autocomplete function which creates a supportable dictionary in the background, equivalent to a synonym dictionary. You may arrive at some ideas from the following projects we’ve implemented for our clients:
- Backen.de: In this project for Oetker Digital, we tried to reference data objects (ingredients and tools) dynamically in an advanced text editor. The way it works is that entire sections of German grammar are automatically derived based on stored metadata. Therefore, the saved language references automatically update changes to the ingredient or tool objects in the text, following which texts/information are available for additional derivations (e.g., automatic creation of an ingredient list based on the entered texts, including a summation of the ingredient quantities).
- Bösch: Here, we aimed at offering the possibility to reference reusable text blocks, which then facilitate text input. Currently, it is available for use only in one language, i.e., German. However, it is an excellent example of how reusable text blocks can be defined to speed up data input for experts.
You must bear in mind that you need exact use-case descriptions for such implementations. You must also make sure that executing this does not eat into too much of your overall project time.
2. Integrating open-source solutions: There are examples of it, such as TinyTm and OmegatT. Our experience shows that these solutions are not too robust. Additionally, you must look very hard whether the features you need are offered by these free tools or not.
3. Full integration of an existing market solution: With this approach, you can go in the direction of outsourcing the functionality to powerful state-of-the-art software, which can offer a variety of integration possibilities:
- Integration of an API: In this case, complete integration takes place in PIM’s backend UI. Specifically speaking, the translation memory (TM) engine handles translation in the background. This essentially means that data and schemas are exchanged via interfaces (e.g., RDFS) so that necessary data is transferred to the external software and in return, is transferred back to Pimcore.
- Integration of pure data: External translation logic is used independently from translation experts, and only the data flows back to PIM, where it is available for further use. It’s quite like an external translation tool with workflow.
This is one of the most expensive and integration-intensive approaches; at the same time, it’s one of the most professional ones too.
Though, as a topic, TMS is innovative and exciting, it has the potential of becoming quite extensive too. And therefore, it must be evaluated and specified in greater detail to meet the required purpose accurately.
At the same time, it must be made sure that the implementation stays restricted to a reasonable level within the project framework so that it doesn’t interfere or affect the actual objectives of the project.
Want to submit a guest post to Pimcore's Blog?
It has over 4 000 subscribers and attracts more than 50 000 monthly visitors, and we're always looking for more brilliant contributors to join our ranks. Contact us.