Skimming through hundreds of papers

Skimming through hundreds of papers

- 3 mins

We all are living very busy lives, with little time to be spend on things that some years ago we didn’t mind spending time on. One of those things are scientific publications.

Every year there are more publications available on a larger number of topics. The fact that universities and even whole countries have metrics based on number of published articles (1, 2, 3) does not help.

Yet, scientific knowledge needs to be spread and people need to find out quickly the articles they are interested in. Of course there are well-known tools like sci-hub(link keeps changing) or google scholar, but often we also have a repository of published papers that we want to look into. For example the IEEE PIMRC17 conference had 549 papers; are we supposed to read all of them? Of course not.

What I normally do is convert all of them to text using pdftotext which you can install using the poppler-utils package under various Linux distributions.

~/Desktop/Papers/ » find . -name \*.pdf -exec pdftotext "{}" \;
~/Desktop/Papers/ » ls
1570347974.txt 1570368376.txt 1570368959.txt 1570373144.txt 1570388857.txt
1570350772.txt 1570368379.txt 1570368961.txt 1570373147.txt 1570388907.txt
1570356083.txt 1570368393.txt 1570368975.txt 1570373157.txt 1570388983.txt
1570359771.txt 1570368414.txt 1570368976.txt 1570373168.txt 1570388989.txt
...

Now that the documents are searchable we can use grep or similar tools to search them. If there are a large number of files involved you might want to check ripgrep which is orders of magnitude faster when searching. In fact, with ripgrep you can also search binary files (PDF) with the option -a/--text but we’ll ignore that cause we already did the conversion.

For example, I’d like to search recursively for papers that talk about CoAP, that are in text format and without minding the capital letters. I’d like to show 1 line of context before and after the term is found. Below are the results.

~/Desktop/Papers/ » rg -i "CoAP" -g "*.txt" -C 1
s1-1.txt
16-protocols will co-exist, including the Constrained Application
17:Protocol (CoAP), Message Queue Telemetry Transport (MQTT),
18-and HTTP.
--
128-includes several potential federation or integration points: At
129:the lowest network layer there are CoAP and/or HTTP ReSTful
130-APIs. The semantic, processing, and storage layers provide
--
151-and interaction model that can be used for managing things that
152:support the CoAP protocol.
153-B. Decentralised security

...

Now I can immediately spot papers that talk about something I am interested in; without browsing hundreds of them. I can even adjust the search to show me only the abstract of those papers, conclusions, etc.

For example if I wanted I could read every abstract pretty quickly by doing the following command:

~/Desktop/Papers/ » rg -i "Abstract" -g "*.txt" -C 2

1570368018.txt
3-Farouk Mezghani, Nathalie Mitton
4-Inria Lille Nord – Europe, France - {farouk.mezghani, nathalie.mitton}@inria.fr
5:Abstract—Opportunistic communications present a
6-promising solution as a disaster network recovery in
8-emergency situations such as hurricanes, earthquakes and

1570368756.txt
6-Email: vanessamendes@get.inatel.br
7-
8:Abstract—The aim of this paper is to propose a simple
9-and efficient algorithm to generate white samples for the
10-envelope of the α-η-κ-µ fading model. To this end, capitalizing
on results available in the literature, the random

1570372054.txt
9-CNRS 6074 IRISA GRANIT UR1 Lannion, France, Email: firstname.lastname@irisa.fr
10-
11:Abstract—The linear closed-loop MIMO precoding technique employs the
channel state information (CSI) at both
12-sides of the link to optimize various criteria such as the
13-capacity, the mean square error, the signal to noise ratio

...
Jaime Jiménez

Jaime Jiménez

comments powered by Disqus