We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.IR

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Information Retrieval

Title: Recommend for a Reason: Unlocking the Power of Unsupervised Aspect-Sentiment Co-Extraction

Abstract: Compliments and concerns in reviews are valuable for understanding users' shopping interests and their opinions with respect to specific aspects of certain items. Existing review-based recommenders favor large and complex language encoders that can only learn latent and uninterpretable text representations. They lack explicit user attention and item property modeling, which however could provide valuable information beyond the ability to recommend items. Therefore, we propose a tightly coupled two-stage approach, including an Aspect-Sentiment Pair Extractor (ASPE) and an Attention-Property-aware Rating Estimator (APRE). Unsupervised ASPE mines Aspect-Sentiment pairs (AS-pairs) and APRE predicts ratings using AS-pairs as concrete aspect-level evidence. Extensive experiments on seven real-world Amazon Review Datasets demonstrate that ASPE can effectively extract AS-pairs which enable APRE to deliver superior accuracy over the leading baselines.
Comments: 16 pages; Accepted to Findings of EMNLP-2021
Subjects: Information Retrieval (cs.IR); Machine Learning (cs.LG)
Cite as: arXiv:2109.03821 [cs.IR]
  (or arXiv:2109.03821v1 [cs.IR] for this version)

Submission history

From: Zeyu Li [view email]
[v1] Tue, 7 Sep 2021 21:40:56 GMT (4813kb,D)

Link back to: arXiv, form interface, contact.