I'll share with you the techniques I used to use before I dropped out of my PhD to fight in the cryptowar:
Use a tool like Google Scholar[1] to see who has cited the work.
Plug each of those in, and prioritize the work that has a low citation count.
Repeat this process for other promising articles.
Another option is too look for the websites for the labs of the students who do the work -- they'll often have whitepapers that were not published in a peer review venue that supplement the formal work, sometimes they're longer and more detailed than a journal article or conference presentation.
Also, try to narrow your search terms to something more specific than "software estimation", which unless it's a term of art I'm unfamiliar with is too broad to bring you decent results. (What exactly are you trying to estimate?)
Literature reviews used to be enough on their own for a thesis in the days before computerized records for a reason -- it's a laborious, iterative process.
Good luck in your search OP! I sincerely hope you find what you're looking for.
[0] https://en.wikipedia.org/wiki/Bellingcat#Notable_cases [1] scholar.google.com
I think Construx has training available on the same material.