Explicit factor models for explainable recommendation based on phrase-level sentiment analysis

Zhang, Yongfeng; Lai, Guokun; Zhang, Min; Zhang, Yi; Liu, Yiqun; Ma, Shaoping

doi:10.1145/2600428.2609579

Cited by 652 publications

(480 citation statements)

References 33 publications

Supporting

Mentioning

477

Contrasting

Order By: Relevance

“…To improve the performance of latent factor models, semantic analysis of textual summary of an item is introduced in many existing researches [5], [12], [13]. In the case of reviews of feedbacks, a common idea is to take the reviews correlated with individual item as one summary.…”

Section: Related Workmentioning

confidence: 99%

Rating Prediction with Topic Gradient Descent Method for Matrix Factorization in Recommendation

Fang¹,

Kamei²,

Fujita³

2017

ijacsa

View full text Add to dashboard Cite

Abstract-In many online review sites or social media, the users are encouraged to assign a numeric rating and write a textual review as feedback to each product that they have bought. Based on users' history of feedbacks, recommender systems predict how they assesses the unpurchased products to further discover the ones that they may like and buy in future. A traditional approach to predict the unknown ratings is matrix factorization, while it uses only the history of ratings included in the feedbacks. In recent researches, its ignorance of textual reviews is pointed out to be the drawback that brings mediocre performance. In order to solve such issue, we propose a method of rating prediction which uses both the ratings and reviews, including a new first-order gradient method for matrix factorization, named Topic Gradient Descent (TGD). The proposed method firstly derives the latent topics from the reviews via Latent Dirichlet Allocation. Each of the topics is characterized by a probability distribution of words and is assigned to correspond to a latent factor. Secondly, to predict the ratings of the users, it uses matrix factorizaiton which is trained by the proposed TGD method. In the training process, the updating step of each latent factor is dynamically assigned depending on the stochastic proportion of its corresponding topic in the review. In evaluation, we both use YELP challenge dataset and per-category Amazon review datasets. The experimental results show that the proposed method certainly converges the squared error of the prediction, and improves the performance of traditional matrix factorization up to 12.23%.

show abstract

Section: Related Workmentioning

confidence: 99%

Rating Prediction with Topic Gradient Descent Method for Matrix Factorization in Recommendation

Fang¹,

Kamei²,

Fujita³

2017

ijacsa

View full text Add to dashboard Cite

show abstract

“…Fortunately, the ever developing phrase-level sentiment analysis techniques [18,10,36,29] have made it possible to extract, summarize and structure the rich product features and user preferences automatically from free-text reviews, and have begun to be leveraged for personalized recommendation tasks [38,33]. This also sheds light on the feature-level dynamic modeling for recommendation.…”

Section: Related Workmentioning

confidence: 99%

“…For daily prediction, we incorporate 366 days of a year as the time dimension of a tensor, and a useritem rating made on the i th day of a year is placed in the rating matrix on the i th layer. EFM: The Explicit Factor Model for explainable recommendation proposed in [38], which is time-independent but it is the state-of-the-art CF method based on sentiment analysis on textual reviews as in this work.…”

Section: Daily-aware Rating Predictionmentioning

confidence: 99%

“…To evaluate top-K recommendation, we select those users who have 10 or more records in the testing set for each dataset Di, and then conduct top-10 recommendation for them. Following [38,23], we adopt the frequently used NDCG to evaluate top-10 recommendation, and we still use the NMF, timeSVD++, Tensor and EFM methods for performance comparison. We set r = 58 for all methods.…”

Section: Daily-aware Top-k Recommendationmentioning

confidence: 99%

“…Fortunately, the increasing availability of textual user reviews and the ever developing text summarization and sentiment analysis techniques have made it possible to take advantage of this important information source besides numerical ratings [38]. In this work, we extract and summarize the domain knowledge (i.e., domain-specific product features like anti-UV or nutrition for cosmetics) automatically from the textual user reviews, and conduct feature-level dynamic modeling to capture the drifting user preferences and item profiles, thus to avoid the lack of data in previous productlevel modeling approaches.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Daily-Aware Personalized Recommendation based on Feature-Level Time Series Analysis

Zhang

et al. 2015

Proceedings of the 24th International Conference on World Wide Web

Self Cite

View full text Add to dashboard Cite

The frequently changing user preferences and/or item profiles have put essential importance on the dynamic modeling of users and items in personalized recommender systems.However, due to the insufficiency of per user/item records when splitting the already sparse data across time dimension, previous methods have to restrict the drifting purchasing patterns to pre-assumed distributions, and were hardly able to model them rather directly with, for example, time series analysis. Integrating content information helps to alleviate the problem in practical systems, but the domaindependent content knowledge is expensive to obtain due to the large amount of manual efforts.In this paper, we make use of the large volume of textual reviews for the automatic extraction of domain knowledge, namely, the explicit features/aspects in a specific product domain. We thus degrade the product-level modeling of user preferences, which suffers from the lack of data, to the feature-level modeling, which not only grants us the ability to predict user preferences through direct time series analysis, but also allows us to know the essence under the surface of product-level changes in purchasing patterns. Besides, the expanded feature space also helps to make cold-start recommendations for users with few purchasing records.Technically, we develop the Fourier-assisted Auto-Regressive Integrated Moving Average (FARIMA) process to tackle with the year-long seasonal period of purchasing data to achieve daily-aware preference predictions, and we leverage the conditional opportunity models for daily-aware personalized recommendation. Extensive experimental results on real-world cosmetic purchasing data from a major e-commerce website (JD.com) in China verified both the effectiveness and efficiency of our approach.

show abstract

A survey on effects of adding explanations to recommender systems

Vultureanu‐Albişi

Bǎdicǎ

2022

Concurrency and Computation

View full text Add to dashboard Cite

Explainable recommendations become essential when we need to improve the performance of recommendations and to increase user confidence. Explanations are effective when end users can build a complete and correct mental representation of the inferential process of a recommender system. This paper presents our view on the background regarding the implications of explainability applied to recommender systems. Our work contributes to the better understanding of the concept of explainable recommendation and it offers a broader picture of the development of further research in this field. Additionally, we contribute by providing a better understanding of the concept of human-centered evaluation of explainable recommender systems.

show abstract

Explicit factor models for explainable recommendation based on phrase-level sentiment analysis

Cited by 652 publications

References 33 publications

Rating Prediction with Topic Gradient Descent Method for Matrix Factorization in Recommendation

Rating Prediction with Topic Gradient Descent Method for Matrix Factorization in Recommendation

Daily-Aware Personalized Recommendation based on Feature-Level Time Series Analysis

A survey on effects of adding explanations to recommender systems

Contact Info

Product

Resources

About