#1

I would like to know if Obsidian would enable me to perform a type of search in which I specify notes to search within and then have the search find relationships between those documents.
For example, let’s say I have 100 notes, each being a biography of a historical person. I enter the names of four people, such as:
“Elizabeth Turcotte”
“Joseph Lambert”
“Clara Huntington”
“Otis Redman”
Next, the search feature would analyze each of these four notes and bring together everything the biographies have in common, like this:
“Elizabeth Turcotte”
“Joseph Lambert”
“Clara Huntington”
“Otis Redman”
— Architect
— 19th century
— educated at Oxford
— liked ice cream
Unlike a Boolean search for notes that contain the phrases “architect,” “19th century,” “educated at Oxford,” and “liked ice cream,” this search is FIRST limited a finite selection of notes, and THEN, after this parameter is set, the search finds what’s in common between these notes.
Is this type of analysis possible with Obsidian?

#2

Not with Obsidian, nor any other note-taking app that I know of.

Unless you’re working with very clearly structured data, such as

profession: "Architect",
century: "19",
education: "Oxford",
liked: ["ice cream", "kittens", "holding hands"]

which might be queried as a database, programatically finding meaningful commonalities is very difficult.

Without any grasp of the semantics, a dumb algorithm will base itself on any shared word of phrase. So, "was born in", "it was", "school", "of the", would all be returned. On the other hand, slight variations would not be considered: "19th century"/ "nineteenth century", "Architect"/"Architecture", or even "English"/"english", etc.

There are, of course, ways of preprocessing the texts in such ways that they become easier to compare, like tokenization, stop word removal, stemming and lemmatization. There are also algorithms that are capable of grading document similarity based on overlapping words (Jaccard Index) or sentence vectors (TF-IDF). However, both of them are very expensive (computationally speaking), if compared to a regular text search.

Long story short: I believe this would make a cool 3rd-party plugin idea, but I don’t expect Obsidian to ever make it a core feature. If this is something you really really need, maybe you should look into the realm of more scientific, ML-powered software — or even a custom-built solution.