[This post contains formatting and embedded images that don't appear in email. View the post in your Web browser.]
"An even better idea would be for Adobe to fix this - maybe if enough of us point it out, they will do something?"
Building on Johan's reply, the root cause is a very poor design of LR's text matching operators, not a bug in the implementation of that design.
LR is behaving here exactly as designed. The criterion:
<field> contains words x y z
matches a photo if the field contains the space-separated word x, the word y, and the word z. So when you write:
keywords contains words pica pica
that will match any photo containing "pica", since the photo contains the first instance of "pica" and it also contains the second instance of "pica".
The fundamental problem is that, unbelievably, there is no way to match the entirety of the most commonly used fields (exact match) or a substring of those fields, including spaces. What were they thinking?
From LR 7 through LR 13.2, you could use this idiom to get very nearly (but not quite) exact match:
match all
<field> starts with pica pica
<field> ends with pica pic
But that broke in LR 13.3 and Adobe hasn't prioritized a fix:
https://community.adobe.com/t5/lightroom-classic-bugs/p-starts-with-text-filter-on-captions-or-filename-incorrectly-matches-words/idi-p/14729614
There are many posts in the Ideas section of this forum that bring up use cases where exact match would have been very useful. I sincerely doubt that Adobe will ever address these by adding an exact-match operator for the most commonly used fields, given they haven't fixed years-long outstanding bugs in smart collections and filtering:
https://www.lightroomqueen.com/community/threads/smart-collection-starts-with.52800/post-1352812