Justifying Recommendations using Distantly-Labeled Reviews and Fine-Grained Aspects

By Jianmo Ni, Jiacheng Li, Julian McAuley. 2019

1,798 citations recorded by the source corpus.

Several recent works have considered the problem of generating reviews (or ‘tips’) as a form of explanation as to why a recommendation might match a customer’s interests. While promising, we demonstrate that existing approaches struggle (in terms of both quality and content) to generate justifications that are relevant to users’ decision-making process. We seek to introduce new datasets and methods to address the recommendation justification task. In terms of data, we first propose an ‘extractive’ approach to identify review segments which justify users’ intentions; this approach is then used to distantly label massive review corpora and construct large-scale personalized recommendation justification datasets. In terms of generation, we are able to design two personalized generation models with this data: (1) a reference-based Seq2Seq model with aspect-planning which can generate justifications covering different aspects, and (2) an aspect-conditional masked language model which can generate diverse justifications based on templates extracted from justification histories. We conduct experiments on two real-world datasets which show that our model is capable of generating convincing and diverse justifications.

Subfields: Artificial Intelligence, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

Citations in the corpus

Polyjuice: Generating Counterfactuals for Explaining, Evaluating, and Improving Models (2021, background): “We believe our approach is more effective than manual creation (although both are beneficial): in 7The absolute error rate drops for at least 5 points, with a relative difference of more than 10%. ments. m-polyjuice maintains the in-domain and out-of-domain accuracies on reviews (SST-2, Amzbook, Yel”
Learning the Difference that Makes a Difference with Counterfactually-Augmented Data (2019, methodology): “We evaluate models on Amazon reviews (Ni et al., 2019) on data aggregated over six genres: beauty, fashion, appliances, giftcards, magazines, and software , the Twitter sentiment dataset (Rosenthal et al., 2017), 4 and Yelp reviews released as part of the Yelp dataset challenge.”
Flower: A Friendly Federated Learning Research Framework (2020, methodology): “We choose to use Amazon’s Book Reviews Dataset (Ni et al., 2019) which contains over 51M reviews from 15M different users.”
WILDS: A Benchmark of in-the-Wild Distribution Shifts (2020, positive): “The dataset is a modified version of the Amazon Reviews dataset (Ni et al., 2019) and comprises 1.2 million customer reviews on Amazon.”
DEMix Layers: Disentangling Domains for Modular Language Modeling (2021, neutral): “, 2020) 25B (10M) REVIEWS† 30M Amazon product reviews (Ni et al., 2019) 2.”
DynaSent: A Dynamic Benchmark for Sentiment Analysis (2020, background): “Many sentiment datasets are derived from customer reviews of products and services (Pang and Lee, 2004, 2005; Socher et al., 2013; Maas et al., 2011; Jindal and Liu, 2008; Ni et al., 2019; McAuley et al., 2012; Zhang et al., 2015).”
Learning What Makes a Difference from Counterfactual Examples and Gradient Supervision (2020, methodology): “For testing on other datasets, we use a random subset (2000 examples) of the test sets of Amazon Reviews [45], Semeval 2017 (Twitter data) [55], and Yelp reviews [77] similarly to [34].”
Next-item Recommendation with Sequential Hypergraphs (2020, background): “This is the updated version of a public Amazon dataset [24] covering reviews from Amazon ranging from May 1996 to October 2018.”
cosFormer: Rethinking Softmax in Attention (2022, methodology): “This is achieved by performing comparative ﬁnetuning experiments on ﬁve datasets, including GLUE (QQP, SST-2, MNLI) (Wang et al., 2018), IMDB (Maas et al., 2011) and AMAZON (Ni et al., 2019) (Sec.”
Aspect-Controllable Opinion Summarization (2021, neutral): “We instead gathered all reviews tagged with at least one of the O POSUM domains (“Laptop Bags”, “Blue-tooth Headsets”, “Boots”, “Keyboards”, “Televi-sions”, and “Vacuums”) from the newest version of the Amazon Product Dataset compiled by Ni et al. (2019).”
Branch-Train-Merge: Embarrassingly Parallel Training of Expert Language Models (2022, neutral): “, 2020) 25B (10M) REVIEWS† 30M Amazon product reviews (Ni et al., 2019) 2.”
Bridging Language and Items for Retrieval and Recommendation: Benchmarking LLMs as Semantic Encoders (2024, neutral): “…The Amazon Reviews datasets are among the most representative publicly available resources for evaluating recommendation models (McAuley et al., (Ni et al., 2019) 15,167,257 43,531,850 233,055,327 Google Local (Yan et al., 2023) 4,963,111 113,643,107 666,324,103 Tenrec (Yuan et al., 2022)…”
GPT-GNN: Generative Pre-Training of Graph Neural Networks (2020, background): “Amazon Review Recommendation Dataset (Amazon) [23] contains 82.8 million reviews, 20.9 million users, and 9.3 million products.”
Towards Universal Sequence Representation Learning for Recommender Systems (2022, methodology): “(2) Cross-domain datasets: we select another five categories from Amazon review datasets [17], “Prime Pantry”, “Industrial and Scientific”, “Musical Instruments”, “Arts, Crafts and Sewing” and “Office Products”, as target domain datasets to evaluate the proposed approach in cross-domain setting.”