TY - GEN
T1 - Categorizing Review Helpfulness Using Abstract Dialectical Frameworks
AU - Zafarghandi, Atefeh Keshavarzi
AU - Ceolin, Davide
N1 - Publisher Copyright:
© IFIP International Federation for Information Processing 2024.
PY - 2024
Y1 - 2024
N2 - Consumer reviews are a vital aspect of the decision-making process for both buyers and companies in the era of e-commerce and online shopping. However, the helpfulness of reviews varies widely, and the abundance of available information can make it difficult to identify the most informative ones. Therefore, categorizing product reviews based on their helpfulness is a critical task. Review helpfulness can be determined by considering several features, such as readability, sentiment, word count, and coherence between the sentiment and score of a review. This article proposes a method for categorizing review helpfulness based on readability and coherence. Our approach employs abstract dialectical frameworks (ADFs), which use interpretation-based semantics to evaluate the acceptability of arguments. We tailor a specific ADF to each review to assess its helpfulness and provide clear explanations for our labeling decisions. We use the grounded semantics of ADFs, which provides information that no one can argue against, to justify our labels and enhance the value of our process. Our method can also be used as a system to give feedback to the review authors on why their reviews may not be helpful and how they can improve them in the future by considering readability and coherence factors. Moreover, our method can work on both small and large data-sets, which may not be feasible with machine learning methods that require a lot of training data.
AB - Consumer reviews are a vital aspect of the decision-making process for both buyers and companies in the era of e-commerce and online shopping. However, the helpfulness of reviews varies widely, and the abundance of available information can make it difficult to identify the most informative ones. Therefore, categorizing product reviews based on their helpfulness is a critical task. Review helpfulness can be determined by considering several features, such as readability, sentiment, word count, and coherence between the sentiment and score of a review. This article proposes a method for categorizing review helpfulness based on readability and coherence. Our approach employs abstract dialectical frameworks (ADFs), which use interpretation-based semantics to evaluate the acceptability of arguments. We tailor a specific ADF to each review to assess its helpfulness and provide clear explanations for our labeling decisions. We use the grounded semantics of ADFs, which provides information that no one can argue against, to justify our labels and enhance the value of our process. Our method can also be used as a system to give feedback to the review authors on why their reviews may not be helpful and how they can improve them in the future by considering readability and coherence factors. Moreover, our method can work on both small and large data-sets, which may not be feasible with machine learning methods that require a lot of training data.
KW - Abstract dialectical frameworks
KW - Explainable artificial intelligence
KW - Online reviews
UR - http://www.scopus.com/inward/record.url?scp=85214107448&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85214107448&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-76714-2_6
DO - 10.1007/978-3-031-76714-2_6
M3 - Conference contribution
AN - SCOPUS:85214107448
SN - 9783031767135
SN - 9783031767166
T3 - IFIP Advances in Information and Communication Technology
SP - 89
EP - 104
BT - Trust Management XIV
A2 - Muller, Tim
A2 - Fernandez-Gago, Carmen
A2 - Ceolin, Davide
A2 - Gudes, Ehud
A2 - Gal-Oz, Nurit
PB - Springer Science and Business Media Deutschland GmbH
T2 - 14th IFIP WG 11.11 International Conference on Trust Management, IFIPTM 2023
Y2 - 19 October 2023 through 20 October 2023
ER -