XML tags causing issues with machine translation results--repeating terms, wrong language

Question

Hello 
 Talked with out Swedish translator and seems the SW Deepl engine is giving some odd answers. 
 
 In Segment 3 there is a mix, what linguist said, is maybe Norwegian or Finnish in the sentence. You will also see the word varlden repeated. That is World&rdquo;. Linguist also stated that segment 4 is not accurate at all. 
 
 Did reach out to our machine translation provider to see if it was with the MT engine. 
 This is what they told me: "The segments are heavily tagged, which leads to the difficulties you've encountered. When files are processed by systems using tags, such as Trados or similar CAT tools, text can be wrapped in XML tags. This wrapping might happen due to incorrect document structure recognition, encoding or markup errors during import/export processes, or usage of XML-style data storage formats which are particularly seen when converting file types, like from Word to PDF." 
 What would be causing this and how to fix it? 
 Oh, using Trados SR2 17.2.11.19134 and Windows 11 Enterprise

Daniel Hug · Answer

David McDermit 
 I also hate taggy texts, esp. if the tags are random and meaningless, like with many PDF⇾Word conversions or in many InDesign files. I sure understand. But this is likely the cause of the issue. It could be that intento has some smart built into their code that tries to determine the language automatically? That could explain the language confusion, that a mutilated word in one language looks like a word from another language. (In my example, MT mistakes "fights" for "fig" because of the tagging.) 
 What's the solution? Send the text without tags, just plain text. 
 Don't know whether Intento supports that, but the AWS plugin does, so does Microsoft, Deep.

Paul · Answer

David McDermit 
 No surprises there&hellip; looks correctly handled to me. The problem is sending crappy files for translation in the first place, and the MT plugin used on Trados not having a text only option to strip the formatting out. 
 If Intento are getting this through the use of their plugin then please remember that Trados is only sending what the plugin tells it to. There are plenty of open sourced plugins on GitHub showing how to send the text only as we have added these options into the plugins we provide for all (I think) the providers we built plugins for.

Paul · Answer

David McDermit 
 I would just ask them to provide an option on the plugin to send the content without the tags&hellip; just as plain text. 
 Trados is correctly presenting the text plus tags as it is in the document. The only place this can be adapted is on the plugin.

Daniel Hug · Answer

David McDermit 
 I agree with Paul, at some stage you have to handle these tags, either at document preparation or at the stage where you send them to MT. Document preparation is preferable, of course, but if this is not in your power, then you have to use an MT provider that can handle this. 
 You could (could) try Cleanup Task, the plugin . It has the option to remove tags. Most of the tags might be kerning and minute font size changes. If neither the document quality nor the MT provider is in your hands, then that is the last option I see. But you can't just delete everything &ndash; some tags will likely be necessary.

RWS AppStore > 2. RWS AppStore Applications

XML tags causing issues with machine translation results--repeating terms, wrong language