My basic idea is to find flagship papers as representatives of a basic
concept, rather than using various keywords to represent the concept in
combined searches (e.g. [ "foreign language writing" "near-synonym" ]
). Instead, we find those specific papers that directly or indirectly
(to some extent) cite flagship papers of both basic concepts. Even if
we don't find many such specific papers, we can already learn something
from those general-concept flagship papers.

What's your killer approach of doing previous research search?

A classical approach for doing this task is:
Step 1) Find an example paper by keyword search that talks about your
research topic;
Step 2) Find more papers by iterating on the citation graph starting at
the example paper.

However, this approach sometimes works poorly because (1) it's
sometimes hard to find even one example paper; (2) the citation graph
generated by the example paper doesn't necessarily reflect the whole
picture (all mainstream ideas) about the topic, especially when they're
an academic faction intentionally concealing citations to related work
"in the outside".

I'm thinking about another approach that doesn't start from one "bottom
point" (a particular paper on the particular topic), but from "top
points", i.e. flagship papers (the most cited) on the general topics
that the said particular topic is based on. The rationale is that
papers on a specific topic almost always cite papers on a more general
topic (a more general problem for which the researcher's topic is a
special case, or the general research on a basic concept that the
researcher's topic involves).

Suppose we're to find papers about "the computer assisting a user in
choosing an appropriate near-synonym when writing in a second
language". No doubt, "near-synonym" is a general topic governing this
specific topic. We search Google Scholar [...]

A classical approach for doing this task is:
Step 1) Find an example paper by keyword search that talks about your
research topic;
Step 2) Find more papers by iterating on the citation graph starting at
the example paper.

However, this approach sometimes works poorly because (1) it's
sometimes hard to find even one example paper; (2) the citation graph
generated by the example paper doesn't necessarily reflect the whole
picture (all mainstream ideas) about the topic, especially when they're
an academic faction intentionally concealing citations to related work
"in the outside".

I'm thinking about another approach that doesn't start from one "bottom
point" (a particular paper on the particular topic), but from "top
points", i.e. flagship papers (the most cited) on the general topics
that the said particular topic is based on.

A classical approach for doing this task is:
Step 1) Find an example paper by keyword search that talks about your
research topic;
Step 2) Find more papers by iterating on the citation graph starting at
the example paper.

However, this approach sometimes works poorly because (1) it's
sometimes hard to find even one example paper; (2) the citation graph
generated by the example paper doesn't necessarily reflect the whole
picture (all mainstream ideas) about the topic, especially when they're
an academic faction intentionally concealing citations to related work
"in the outside".

I'm thinking about another approach that doesn't start from one "bottom
point" (a particular paper on the particular topic), but from "top
points", i.e. flagship papers (the most cited) on the general topics
that the said particular topic is based on. The rationale is that
papers on a specific topic almost always cite papers on a more general
topic (a more general problem for which the researcher's topic is a
special case, or the general research on a basic concept that the
researcher's topic involves).

Suppose we're to find papers about "the computer assisting a user in
choosing an appropriate near-synonym when writing in a second
language". No doubt, "near-synonym" is a general topic governing this
specific topic. We search Google Scholar (scholar.google.com) for this
keyword, and the search results are papers ranking by their number
cited, descending order. So the topic results are the flagship papers
representing the general topic of "near-synonym". We think that papers
on all other topics fundamentally involving this general-concept topic
must directly or indirectly cite these flagship papers. We need to read
these flagship papers because their general solutions could anticipate
the solution to our own specific problem. Then, we're going to find the
set of flagship papers on another basic concept that our specific topic
involves, e.g. "foreign language writing" (and its paraphrases). In
this case there aren't many research done in this field, anyway.

My basic idea is to find flagship papers as representatives of a basic
concept, rather than using various keywords to represent the concept in
combined searches (e.g. [ "foreign language writing" "near-synonym" ]
). Instead, we find those specific papers that directly or indirectly
(to some extent) cite flagship papers of both basic concepts. Even if
we don't find many such specific papers, we can already learn something
from those general-concept flagship papers.

What's your killer approach of doing previous research search?

Regards,
Yao Ziyuan