Deep Critiquing for VAE-based Recommender Systems

Luo, Kai; Yang, Hojin; Wu, Ga; Sanner, Scott

doi:10.1145/3397271.3401091

Cited by 37 publications

(34 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A recent line of work aims to facilitate language-based critiquing in modern embedding-based recommender systems beyond explicitly known item attributes [20,24,25,42]. The central theme of these efforts is to co-embed subjective item descriptions (i.e., keyphrases from user reviews) with general user preference information.…”

Section: Related Workmentioning

confidence: 99%

“…Specifically, this paper addresses the problem of the interpretation of soft attributes. Note that this is different from the task of incorporating subjective item descriptions into latent user preference representations for improving end-to-end recommendation performance (measured in terms of success rate or the number of conversational turns) [20,24,25,42]. Instead, our objectives are to be able to (1) explicitly measure the degree to which a soft attribute applies to a given item and (2) quantify the degree of "softness" (i.e., subjectivity) of soft attributes.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

On Interpretation and Measurement of Soft Attributes for Recommendation

Balog

Radlinski

Karatzoglou

2021

Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval

View full text Add to dashboard Cite

We address how to robustly interpret natural language refinements (or critiques) in recommender systems. In particular, in humanhuman recommendation settings people frequently use soft attributes to express preferences about items, including concepts like the originality of a movie plot, the noisiness of a venue, or the complexity of a recipe. While binary tagging is extensively studied in the context of recommender systems, soft attributes often involve subjective and contextual aspects, which cannot be captured reliably in this way, nor be represented as objective binary truth in a knowledge base. This also adds important considerations when measuring soft attribute ranking. We propose a more natural representation as personalized relative statements, rather than as absolute item properties. We present novel data collection techniques and evaluation approaches, and a new public dataset. We also propose a set of scoring approaches, from unsupervised to weakly supervised to fully supervised, as a step towards interpreting and acting upon soft attribute based critiques. CCS CONCEPTS• Information systems → Recommender systems.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

On Interpretation and Measurement of Soft Attributes for Recommendation

Balog

Radlinski

Karatzoglou

2021

Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval

View full text Add to dashboard Cite

show abstract

“…Antognini et al [2] generate a single sentence of explanation alongside the set of aspects, but still require users to interact with the aspect set. Luo et al [23] use a variational auto-encoder (VAE) [14] in place of the collaborative filtering model, learning a bi-directional mapping function between user latent representations and aspects they have expressed in reviews. Such models can generate high precision justifications but have shown poor multi-turn recommendation performance.…”

Section: Conversational Critiquingmentioning

confidence: 99%

“…Recent models for conversational critiquing incorporate user feedback on subjective aspects (e.g. taste and perception) [22,23,40]. However, such methods are trained using a next-item recommendation objective, and perform poorly when engaging with users over multiple turns.…”

Section: Introductionmentioning

confidence: 99%

Self-Supervised Bot Play for Conversational Recommendation with Justifications

Li¹,

Majumder²,

McAuley³

2021

Preprint

View full text Add to dashboard Cite

System (re-)scores candidate items using user preference embedding System suggests top-scoring item and generates a justification The Eye of the World The Hobbit The Last Unicorn Assassin's Apprentice "You might like The Eye of the World. It's a complex high fantasy novel about politics." User accepts the suggestion or User critiques an aspect from the justification "I don't really care for politics" System updates user preference embedding via critique Figure 1: Conversational critiquing workflow. The system scores candidate items and generates a justification for the top item.If the user rejects the suggestion and critiques an aspect, the system uses the critique to update the latent user representation.

show abstract

“…Moreover, it is desirable that explanations are causal and actionable, meaning that i) they refer only to the user's own action history and not to potentially privacy-sensitive cues about other users (see, e.g., [12]) and ii) the user can act on the explanation items by giving confirmation or refutation signals that affect future recommendations. Critique-enabled recommendation models [8,18,23,27] pursue these goals, but are restricted to user feedback on the recommended items and associated content (e.g., text snippets from item reviews), disregarding the explanation items. In this paper, we extend this regime of actionable user feedback to the explanations themselves, by obtaining additional cues from users in a lightweight manner and incorporating them into an active learning framework to improve future recommendations.…”

Section: Introductionmentioning

confidence: 99%

ELIXIR: Learning from User Feedback on Explanations to Improve Recommender Models

Ghazimatin¹,

Pramanik

Roy

et al. 2021

Proceedings of the Web Conference 2021

View full text Add to dashboard Cite

System-provided explanations for recommendations are an important component towards transparent and trustworthy AI. In stateof-the-art research, this is a one-way signal, though, to improve user acceptance. In this paper, we turn the role of explanations around and investigate how they can contribute to enhancing the quality of generated recommendations themselves. We devise a human-in-the-loop framework, called Elixir, where user feedback on explanations is leveraged for pairwise learning of user preferences. Elixir leverages feedback on pairs of recommendations and explanations to learn user-specific latent preference vectors, overcoming sparseness by label propagation with item-similarity-based neighborhoods. Our framework is instantiated using generalized graph recommendation via Random Walk with Restart. Insightful experiments with a real user study show significant improvements in movie and book recommendations over item-level feedback. CCS CONCEPTS• Information systems → Recommender systems.

show abstract

Deep Critiquing for VAE-based Recommender Systems

Cited by 37 publications

References 15 publications

On Interpretation and Measurement of Soft Attributes for Recommendation

On Interpretation and Measurement of Soft Attributes for Recommendation

Self-Supervised Bot Play for Conversational Recommendation with Justifications

ELIXIR: Learning from User Feedback on Explanations to Improve Recommender Models

Contact Info

Product

Resources

About