12

I'm having the following problem: There is some interesting paper (in machine learning) which is cited by 180 other papers (according to Google Scholar). From these 180, I want to find the papers that build up heavily on the original paper. I mean, not the ones that just mention the original paper briefly and are not REALLY related to it, but the ones in which the original paper is one of them, let's say three, main references.

Now, obviously, I cannot read 180 abstracts/papers now. Any clever and experienced ideas? I'm a PhD student.

5
  • 11
    "Now, obviously I cannot read 180 abstracts/papers now." You should probably start skimming them.
    – Fomite
    Feb 26, 2015 at 1:16
  • 14
    "obviously I cannot read 180 abstracts [...] now" [Citation required] I opened a random paper in pubmed and read the abstract, and get an idea of what it was about in 27 s.
    – Davidmh
    Feb 26, 2015 at 1:45
  • 6
    I know it can appear overwhelming but the more you do it the more efficient you will become.
    – earthling
    Feb 26, 2015 at 12:05
  • 6
    Even just reading the titles will probably let you quickly eliminate at least half. Feb 27, 2015 at 6:35
  • You can search for citing papers that at least mention x. This should narrow it down quite fast.
    – VonBeche
    Jun 1, 2017 at 12:31

3 Answers 3

15

1) Ask your advisor.

2) Scanning 180 abstracts to find the most relevant ones can be done in under a day if you're efficient about it (most can be excluded within the first two sentences of the abstract if they are not relevant).

3
  • Of course asking my advisor goes without saying, and reading 180 abstracts in a day is not a good solution because such a problem can happen to me quite often.
    – cruvadom
    Feb 26, 2015 at 15:15
  • @cruvadom How often does that happen to you? If it happens often enough that spending a day reading abstracts seriously affects your overall productivity, that seems really strange.
    – user141592
    Feb 26, 2015 at 19:09
  • 6
    Also scan the sentence that cites the paper.
    – Memming
    May 5, 2015 at 16:55
8

You don't need to read abstracts nor papers, you can use the following method to skim papers.

Method. Find the paper on Google Scholar and click the "cited by 180" link, that'll give you a link such as https://scholar.google.fr/scholar?cites=7297898505323682187, repeat the following steps for each of the 180 papers: 1) download, 2) find the citation to the original paper in the bibliography (hopefully in numeric, e.g., [1], or alphanumeric, e.g., [AB17], form), and 3) search for the citation (e.g., search for AB17) in the paper and check whether it is cited in a meaningful way.

Cost. Step 1 will cost you ~5 seconds if there's no pay wall and possibly minutes if there is, Step 2 takes ~30 seconds, and Step 3 takes ~50 seconds. (As you get better those timings will reduce.) Thus, a lower-bound on cost is around 1 1/2 minutes per paper or around half a day for all 180 papers. Factoring in pay walled papers, it might take 1-2 days.

Alternatively, if you know roughly what you are looking for, then you can reduce your search space, e.g., you can consider all the papers that cite the original and contain "keyword" (https://scholar.google.fr/scholar?cites=7297898505323682187&scipsc=1&q=keyword)

6

More recently, you can also use Semantic Scholar to do this. They have an option to sort citations to a paper by "Most Influenced Papers". For example, for the paper "Latent Retrieval for Weakly Supervised Open Domain Question Answering" you can see:

enter image description here

You can read how they determine "influence" here:

"Influential citations are determined utilizing a machine-learning model analyzing a number of factors including the number of citations to a publication, and the surrounding context for each. You can read more about our approach in “Identifying Meaningful Citations”."

In my experience looking at computer science papers, it works quite well for those!

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .