Cross-domain Ontology Construction and Alignment from Online Product Reviews

Geng, Qingbo; Deng, Siyu; Jia, Danping; Jin, Jian

doi:10.1145/3366424.3383755

Cited by 1 publication

(3 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[16,18,21,32,33]) or require extensive human intervention (i.e. [12]), making them unsuitable for comparison. We considered also the approach of [17] for comparison with our approach, but the code is not publicly available and, moreover, the more promising results in [17] were obtained using manually labelled aspects.…”

Section: Discussionmentioning

confidence: 99%

“…Several works have proposed methods for ontology construction for e-commerce, e.g., [12,16,17], the vast majority focusing on taxonomies. [16] propose a deep neural network that leverages the taxonomy of product categories and extracts attributes conditional to its category.…”

Section: Related Workmentioning

confidence: 99%

“…Amongst these, [17] use Latent Semantic Analysis to represent words as vectors in high-dimensional space and cosine similarity to construct aspect ontologies, testing their method on four tech-related product types: digital camera, mobile phone, DVD player, and mp3 player. Further, [12] extract ontologies from product reviews using structured input metadata and predefined relations between aspects, and test their method on two product categories only: smartphones and digital cameras.…”

Section: Related Workmentioning

confidence: 99%

See 2 more Smart Citations

Automatic Product Ontology Extraction from Textual Reviews

Oksanen¹,

Cocarascu²,

Toni³

2021

Preprint

View full text Add to dashboard Cite

Ontologies have proven beneficial in different settings that make use of textual reviews. However, manually constructing ontologies is a laborious and time-consuming process in need of automation. We propose a novel methodology for automatically extracting ontologies, in the form of meronomies, from product reviews, using a very limited amount of hand-annotated training data. We show that the ontologies generated by our method outperform hand-crafted ontologies (WordNet) and ontologies extracted by existing methods (Text2Onto and COMET) in several, diverse settings. Specifically, our generated ontologies outperform the others when evaluated by human annotators as well as on an existing Q&A dataset from Amazon. Moreover, our method is better able to generalise, in capturing knowledge about unseen products. Finally, we consider a real-world setting, showing that our method is better able to determine recommended products based on their reviews, in alternative to using Amazon's standard score aggregations.

show abstract