TextiQ Topics Creating for non-English survey | XM Community
Skip to main content

TextiQ Topics Creating for non-English survey


Nam Nguyen
QPN Level 8 ●●●●●●●●
Forum|alt.badge.img+29
  • QPN Level 8 ●●●●●●●●
  • 1091 replies

Hi everyone,

I'm working with Text iQ in Qualtrics to analyze open-ended responses from a survey conducted in Vietnamese. I’ve gone through the www.qualtrics.com/support/survey-platform/data-and-analysis-module/text-iq/text-iq-best-practices/#CreatingTopics, but I find it too high-level and not detailed enough, especially when it comes to non-English surveys.

The guidelines mostly explain what topics are and general best practices, but there’s little to no mention of how people actually go about creating topics for non-English responses. Also, pre-built topics and recommended topics don’t apply well—or at all—to my Vietnamese data. This makes it difficult to rely on any automation.

So I’d love to hear from the community:

  • How do you approach creating and organizing topics effectively, especially when the language isn’t well-supported by default?

  • Any tips, tools, or workflows you’ve used to speed up or streamline this process for non-English text?

I’d really appreciate hearing about any real-world experience with using Text iQ on surveys that aren't in English. What worked for you, what didn’t, and what would you recommend?

Thanks so much in advance!

2 replies

vgayraud
QPN Level 5 ●●●●●
Forum|alt.badge.img+49
  • QPN Level 5 ●●●●●
  • 377 replies
  • March 31, 2025

Hi Nam,

French is way better supported in Qualtrics now, but we did face the same issue 5 or 6 years ago.

Luckily, we have an internal coding team. We’d usually provide them with a relevant English TextiQ topics starter pack and have them work on adapting/building the queries from the pack and the collected verbatim.

Another thing we did occasionally in multiple language surveys was to automatically translate all comments before running them in TextiQ, although at the time that would imply to loose a lot of details and quality. Maybe nowadays you could get better results with IA extensions for translation.


AlexandreLeduc
QPN Level 3 ●●●
Forum|alt.badge.img+13
  • QPN Level 3 ●●●
  • 47 replies
  • March 31, 2025

To add on Vincent’s topics, we are used to have verbatims 50/50 French and English so our models are always bilingual. They same keyword will be there in both languages. To build our text iQ model from scratch, we always start with 400-500 comments and we build from there. 

 

To add on Vincent’s second scenario described above, our client had verbatims in 7 different languages. Even if they were loosing a bit in details by translating everything to English, they did not have people, internally, speaking all thoses languages, so it was better to have it in English with some details getting lost in translation, than not at all. 

 

Best,