Search relevance is incredible difficult, especially when there is not enough information provided in the query and there are multiple "correct" answers. Someone is always going to be disappointed. You can make the argument in this case that +90% of users would find that London, England is a better answer, but there are millions of these types of queries.
In a recent example, I was using my favorite search engine and showed someone a query for "toaster" where the top results was an infobox describing what a toaster was. They thought the relevance for the query was hilariously wrong saying, "everyone already knows what a toaster is, I want results for buying a toaster".
IMO I thought the results were perfect based on the query. Neither of us were wrong.
The expectation for search tools has dramatically shifted where many people expect them to have oracle like qualities.
> Search relevance is incredible difficult, especially when there is not enough information provided in the query and there are multiple "correct" answers. Someone is always going to be disappointed.
Sure, but in that case you go with the answer that will be the correct answer for the most users. Which in the example in TFA would almost certainly be London in the United Kingdom, not some other London in Kentucky or Canada or what have you. You'll end up disappointing those users who actually want to know the time in those Londons, but there are going to be a lot fewer of those.
But what do you do in situations that are closer to split on people agreeing what the right answer is, or there are 10 answers that could be relevant to the same percentages of people.
A query leading to the single relevant answer to the user 100% of the time is impossible.
There could be a mechanism for the system to get additional information when needed, but that also has some user experience issues.
These are all problems that are most likely solvable to a degree though.
On a side note, I also don't think Siri is very good compared to some of the other voice assistants.
In a recent example, I was using my favorite search engine and showed someone a query for "toaster" where the top results was an infobox describing what a toaster was. They thought the relevance for the query was hilariously wrong saying, "everyone already knows what a toaster is, I want results for buying a toaster".
IMO I thought the results were perfect based on the query. Neither of us were wrong.
The expectation for search tools has dramatically shifted where many people expect them to have oracle like qualities.