Natural Language Search Returns Irrelevant Results

​@AJMitchonI followed up with the product team, and they had a couple of questions that would help them better evaluate the search results.

Looking at the screenshot, they noticed what appears to be a windmill in the upper-left result and would like to understand what you expected to see instead.

Could you please help clarify:

• Do you have examples of photos in your library that you expected to be returned for the query “windmill at night” but were not?

• Are you seeing similarly poor results with other natural-language searches as well? If so, a few example queries would be very helpful.

• Is this issue limited to this specific search, or does it seem more broadly related to object recognition/relevance ranking?

This will help the team investigate the search behavior more closely.

Thanks again for taking the time to provide feedback and examples.

Best,
Anshul Saini

Hi @AJMitchon,

Thank you for taking the time to share this. Looking at the screenshot, I agree that the search results do not appear to match the query “windmill at night” very well. Many of the returned images appear to be general night scenes, starry skies, buildings, and landscapes, with no obvious windmills. That’s useful feedback regarding the accuracy and relevance of the new natural-language search experience.

I’ll pass your feedback directly to the product team.

Thanks again for providing an example.

Best,
Anshul Saini