PARL：利用相似用户评论增强推荐系统

需积分: 9 79 浏览量更新于2024-08-13 收藏 2.78MB PDF 举报

"PARL：利用志同道合用户辅助评论中的用户-商品对依赖特征来提升推荐系统性能的研究论文" PARL（Pairwise Auxiliary Reviews Learning）是针对推荐系统中数据稀疏问题的一种新方法，尤其关注在评论同样稀疏的情况下如何提高推荐效果。现有的基于评论的推荐方法通常依赖于用户提供的评论来理解他们的喜好，但当评论数据不足时，这些方法的效果会受到影响。为了解决这个问题，PARL提出从与目标用户有相似兴趣的其他用户的辅助评论中提取用户-商品对依赖的特征。论文作者Libing Wu、CongQuan、Chenliang Li和Donghong Ji来自武汉大学计算机科学学院和网络安全工程学院。他们指出，PARL模型不仅能够缓解评论数据稀疏的问题，而且还能生成额外的、具有信息性的特征，从而进一步提高评分预测的准确性。这种方法的独特之处在于它设计为一种即插即用的模型，可以轻松地集成到各种深度推荐系统中，无需大幅度修改原有架构，从而增强系统的整体推荐性能。在PARL模型中，用户自己的评论和商品的评论都被考虑进来，以突出辅助评论中涵盖的有用特征。通过这种方式，即使原始用户对某个商品没有足够的评论，模型也能通过分析其他类似用户的评论来推断该用户的潜在喜好。这种跨用户的信息共享策略有助于挖掘隐藏在大量评论中的潜在关联，从而提供更精确的个性化推荐。此外，PARL可能对社交媒体和电子商务平台特别有用，因为这些平台往往积累了大量的用户行为数据，但每个用户提供的具体反馈（如评论）可能相对较少。通过引入PARL，这些平台的推荐算法可以更好地理解用户的兴趣，提供更符合用户需求的推荐，从而提升用户体验和满意度。 PARL是一种创新的推荐系统方法，通过利用志同道合用户生成的辅助评论，有效地解决了数据稀疏性问题，提升了推荐的准确性和效率。它的即插即用特性使其成为现有深度学习推荐系统的一个强大补充，有望在实际应用中带来显著改进。

PARL: Let Strangers Speak Out What You Like

Libing Wu

, Cong Quan

, Chenliang Li

2⋆

, Donghong Ji

1. School of Computer Science, Wuhan University, Wuhan, 430072, China

{wu, quancong}@whu.edu.cn

2. School of Cyber Science and Engineering, Wuhan University, Wuhan, 430072, China

{cllee, dhji}@whu.edu.cn

ABSTRACT

Review-based methods are one of the dominant methods to address

the data sparsity problem of recommender system. However, the

performance of most existing review-based methods will degrade

when the review is also sparse. To this end, we propose a method to

exploit user-item

air-dependent features from

uxiliary

eviews

written by

ike-minded users (PARL) to address such problem. That

is, both the reviews written by the user and the reviews written

for the item are incorporated to highlight the useful features cov-

ered by the auxiliary reviews. PARL not only alleviates the sparsity

problem of reviews but also produce extra informative features

to further improve the accuracy of rating prediction. More impor-

tantly, it is designed as a plug-and-play model which can be plugged

into various deep recommender systems to improve recommenda-

tions provided by them. Extensive experiments on ve real-world

datasets show that PARL achieves better prediction accuracy than

other state-of-the-art alternatives. Also, with the exploitation of

auxiliary reviews, the performance of PARL is robust on datasets

with dierent characteristics.

CCS CONCEPTS

• Information systems → Recommender systems;

KEYWORDS

Recommender System, User Reviews, Deep Learning, Rating Pre-

diction

ACM Reference Format:

Libing Wu, Cong Quan, Chenliang Li, Donghong Ji. 2018. PARL: Let Strangers

Speak Out What You Like. In 2018 ACM Conference on Information and

Knowledge Management (CIKM’18), October 22–26, 2018, Torino, Italy.

ACM, NY, NY, USA, 10 pages. https://doi.org/10.1145/3269206.3271695

1 INTRODUCTION

In the past decade, Recommender Systems have been playing an

increasingly important role in many online platforms, including

E-commerce such as Amazon, video-streaming providers such as

⋆

Chenliang Li is the corresponding author.

Permission to make digital or hard copies of all or part of this work for personal or

classroom use is granted without fee provided that copies are not made or distributed

for prot or commercial advantage and that copies bear this notice and the full citation

on the rst page. Copyrights for components of this work owned by others than ACM

must be honored. Abstracting with credit is permitted. To copy otherwise, or republish,

to post on servers or to redistribute to lists, requires prior specic permission and/or a

fee. Request permissions from permissions@acm.org.

CIKM ’18, October 22–26, 2018, Torino, Italy

ACM ISBN 978-1-4503-6014-2/18/10.. . $15.00

https://doi.org/10.1145/3269206.3271695

Figure 1: Examples of reviews written by two audiences for

the lm “Innite War” on IMDB. Both users give the highest

rate to Innite War, and similar interests are presented by

their reviews.

Youtube and social media such as Twitter. The key to a recom-

mender system is to dig out user’s interest and thus personalized

recommendation can be provided.

Many recommender systems are based on Collaborative Filtering

(CF) [

], which mainly rely on the past interaction records between

users and items. Although CF techniques enjoy a surge of atten-

tion and provide good performance, the sparsity problem hinders

the CF-based recommender systems from providing high-quality

recommendations to the users with few records. In this case, neigh-

borhood methods [

] are employed to alleviate the data sparsity

problem. Concretely, neighborhood methods utilize the relations

between items or between users to improve the recommendation.

For example, a neighborhood-based method [

] estimates an un-

kown rating on an item made by a user based on the known ratings

made by the same user on other items.

Recently, some approaches also resort to reviews written by

users to address the lack of data. The motivation is that ratings and

review are two facets of users to depict their experience on items.

Therefore, the reviews can be used to well alleviate data sparsity.

Some studies [

] have shown that methods considering reviews

generally perform better than collaborative ltering methods which

only take the interaction records into consideration. In particular,

the performance of review-based methods is robust when the users

Session 4E: Recommendation 1

CIKM’18, October 22-26, 2018, Torino, Italy

677

下载后可阅读完整内容，剩余9页未读，立即下载

weixin_38722052

粉丝: 4
资源: 911

PARL：利用相似用户评论增强推荐系统

PARL:强化学习的高性能分布式培训框架

paddlepaddle-PARL-develop.zip

PARL：优化强化学习性能与灵活性的框架

PARL：灵活高效的强化学习框架支持分布式训练

leetcode题库-PARL-Sample:使用百度PARL进行深度强化学习（迷宫、飞扬的小鸟等）

PARL-develop.zip

PARL框架：高性能、灵活的强化学习解决方案

PARL框架：支持大规模并行计算的强化学习系统

PARL高性能强化学习框架：并发计算与算法复现

ModuleNotFoundError: No module named 'parl'

最新资源