As part of the work to support Custom Translation Suggestions (T113257), users can select specific topic areas to get suggestions in their area of interest. However, some of the topic areas such as entertainment and engineering result in suggestions dominated by very specific topics (planes and wrestlers) resulting in a lack of diversity for the results.
The articletopic: keyword on search (documentation) is used for querying articles on a specific topic area, and the underlying model is used for recommendations in tools such as Content Translation. The example results for entertainment and for engineering can be checked when searching English Wikipedia and are illustrated below.
| engineering | entertainment |
|---|---|
 |  |
Based on the input from T376987#10222124, it was suggested that adjusting some of [the Search API parameters such as srsort or srqiprofile could have a randomizing effect on the results that will increase their diversity (while still showing results that the model classified as relevant for the selected topic area). An excerpt from Isaac's comment is shared below, which includes more context:
You could consider changing the "generically more-relevant" part of the sorting to something that exerts a more randomizing effect on the results topic-wise. So here's the docs for the Search API and we could either play with srsort or srqiprofile. The default I think is this srqiprofile=wsum_inclinks_pv&srsort=relevance. I looked and I think Growth is using srqiprofile=classic_noboostlinks (code), which does mix things up. There are options you can play with -- here are the docs for srsort and srqiprofile.
The purpose of this task is to explore and document the different options that the search parameters provide, and which are the final results they produce. Ideally we should identify the approach that produces results that are very relevant to the selected topic, cover different aspects of it and include a minimal number of unrelated topics (false positives).
