Changes between Version 6 and Version 7 of Public/StartPageTransPubSoftware

Timestamp:

Sep 22, 2022, 3:55:38 PM (2 years ago)

Author:

Boris Horner

Comment:

—

Public/StartPageTransPubSoftware

@@ -14 +14 @@
 Particularly in the plant engineering branch, but also in other branches of industry, there is the need to collect documentation from suppliers and integrate it with documentation of in-house products, typically controlled by a parts list from the ERP system. Together with [https://www.reinisch.de reinisch], we built a web-based, standard solution that addresses these requirements.
 
-The **Supplier Documentation Management System** is available as standard software and can be hosted by yourselves or by us on standard Linux platforms. It reads parts lists from the ERP, and automatically calculates documentation requirements based on configurable rules. These can, for example, use target language, target market or component type as parameters to select the rule set(s) to be applied.
+The **Supplier Documentation Management System (SDMS)** is available as standard software and can be hosted by yourselves or by us on standard Linux platforms. It reads parts lists from the ERP, and automatically calculates documentation requirements based on configurable rules. These can, for example, use target language, target market or component type as parameters to select the rule set(s) to be applied.
 
 The requests can be reviewed and edited by the purchase department employees before submitting them. Once submitted, registered suppliers are notified by email and are asked to upload defined documentation units like:
@@ -30 +30 @@
 
 == Translation Memory clean-up and quality control
-...
+Translation Memory Systems can leverage efficiency gain by automatic translation of known segments only if there is exactly one translation of the segment per target language. When we analyze TMs from customers that have grown over time, we often find ambiguities of the type that there are multiple translations for the same source segment and target language. If the TM finds such a segment, it can't automatically decide which translation to use.
+
+Analyses of such real-world TMs rarely reveal no ambiguities at all, typically, 2-10% of the segments have multiple translations. Commonly, they are almost identical and just differ in additional blanks, different wording in the same sentence etc.
+
+
+It is possible, but difficult to find such ambiguities in the common TM software systems. To make this task easier, we built a standard solution called **Translation Memory Quality Editor (TMQE)**, together with [https://www.reinisch.de reinisch], that reads TMX exports from TM systems (currently supported: Trados, Transit, Across, but basically it works with all TMX exports). It searches the TMs for ambiguities, lists them in a table and allows the users to edit them.
+
+For an ambiguous segment, users can choose to:
+* Make one of the existing translations the standard.
+* Choose none of the existing translations, but insert a new one.
+The segments that were edited are highlighted in green, if an existing translation was chosen or yellow, if a new one was provided.
+
+The work status can be saved any time. When done, the result can be exported to TMX that can be imported back into the TM system.
+
+Apart from ambiguities, TMQE can check by rules based on Regular Expressions. This allows custom rule configuration for checks like "blank before punctuation in French" or "no blank before punctuation for non-French languages".
+
+
 
 == Configurable visualization of ontology data