Skip to content
Ranking of products with sentiment expressions. Code for paper (Kessler, Klinger and Kuhn, 2015).
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
doc
src/de/uni_stuttgart/ims/compranking
README.md

README.md

CompRanking

Ranking of products with sentiment expressions. Code for paper (Kessler, Klinger and Kuhn, 2015).

WARNING: This is research code, it was not written with anybody else in mind nor with the goal of applying it "in real life". So it is hacky and may not be usable at all for you.

Prerequisites

To run the code you will need:

Running the code

Create rankings

Run the following class to get the rankings:

  • Baseline for scoring by number of reviews: de.uni_stuttgart.ims.compranking.bl.RankingFromReviews

    • input: Reviews in the Amazon Downloader format
  • Baseline for scoring by review length: de.uni_stuttgart.ims.compranking.bl.RankingFromLength

    • input: Reviews in the Amazon Downloader format
  • Baseline for scoring by sentiment terms from a dictionary: de.uni_stuttgart.ims.compranking.bl.RankingFromTermCounting

    • input: Reviews in the Amazon Downloader format
    • additional resource: either the MPQA or the GI dictionary
  • Ranking by expressions found with JFSA: de.uni_stuttgart.ims.compranking.jfsa.JFSARanker

    • input: sentiment and target annotations that result from processing reviews with JFSA
  • Ranking by expressions found with CSRL: de.uni_stuttgart.ims.compranking.csrl.CSRLRanker

    • input: parsed annotated sentences that result from processing review sentences with CSRL

Result of each execution will be a file that lists the products according to their ranks (with the scores).

Score rankings

Use de.uni_stuttgart.ims.compranking.csrl.CompareRankings on the generated rank files.

More information

What does it actually do?

Read the paper!

Configurations and settings for the paper

  • Term counting: polarityMode = PolarityMode.both run both with and without normalization

  • JFSA: polarityMode = PolarityMode.both run both with and without normalization

Licence and References

(c) Wiltrud Kessler

This code is distributed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported license http://creativecommons.org/licenses/by-nc-sa/3.0/

Please cite: Wiltrud Kessler, Roman Klinger and Jonas Kuhn (2015) Towards Opinion Mining from Reviews for the Prediction of Product Rankings. In Proceedings of the 6th Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis (WASSA 2015). Lisboa, Portugal, 17. September 2015, pages 51-57.

You can’t perform that action at this time.