Sign in
Ctrl K


Review Argumentation at Scale

Product reviews are very useful to help online users appreciate the quality of products and services they may make use of, but the quality of such reviews is very diverse. Reasoning on the arguments in reviews is a promising method to appreciate review quality. However, computational argumentation can be inefficient and computationally expensive. This project was meant to improve the scalability of a computational pipeline for reasoning on arguments of product reviews in order to estimate their quality.
We started from an existing pipeline and introduced two types of speed improvements. First, we parallelized the code, so as to allow processing multiple reviews in parallel to identify their arguments and compare them. This led to an improvement in the computational time, up to a factor of more than 20 times.
Second, we introduced a lossy improvement, i.e., we analyzed reviews after having removed the less important parts of them. This method is promising but needs to be used with care.
The advancements obtained with this project allowed for strengthening the line of research linking argument reasoning with information quality. These findings will be merged with an existing project looking at the possibility of employing a more complex argumentation framework in this area.

Participating organisations

Netherlands eScience Center
University of Milan
University of Udine
Social Sciences & Humanities
Social Sciences & Humanities
SWI-Prolog Solutions b.v.



Davide Ceolin
Lead Applicant
Centrum Wiskunde & Informatica (CWI)
Laura Ootes
Laura Ootes
Lead RSE
Netherlands eScience Center
Giuseppe Primiero
Michael Soprano
Jan Wielemaker
SWI-Prolog Solutions b.v.
Niels  Drost
Niels Drost
Programme Manager
Netherlands eScience Center

Related projects

The eye of the beholder

Transparent pipelines for assessing online information quality

Updated 18 months ago
In progress

Related software

Review Argumentation at Scale


This package contains a scalable pipeline to perform argument reasoning on a dataset of product reviews, to identify the highest quality ones. By tuning both lossy and lossless parameters, it is possible to tune the pipeline to foster the accuracy of the computation or its speed.

Updated 3 months ago