没有合适的资源?快使用搜索试试~ 我知道了~
首页深度学习构建用户标签画像在推荐系统中的应用
深度学习构建用户标签画像在推荐系统中的应用
需积分: 48 3 下载量 18 浏览量
更新于2024-08-09
收藏 1.96MB PDF 举报
"这篇论文是关于在推荐系统中构建用户标签档案的研究,由微信团队的成员撰写。他们提出了一种用户标签建模方法(UTPM),利用深度神经网络处理多标签分类任务,以理解并表示用户的兴趣标签。" 在推荐系统中,用户档案的构建是一项至关重要的任务,它涉及收集和分析用户的各种信息,如人口统计学数据(如性别、年龄和位置)以及用户行为数据(如浏览历史和搜索历史)。这些信息被用来描绘用户的兴趣和偏好,以便为用户提供更精准的个性化推荐。 文章特别关注了标签这一可解释且广泛应用的用户兴趣表示方式。作者提出了一种名为用户标签建模(User-Tag Profiling Model,UTPM)的新模型。UTPM采用深度神经网络框架,将用户标签建模视为一个多标签分类问题,这意味着一个用户可能同时对多个标签感兴趣。 与传统的建模方法不同,UTPM模型引入了多头注意力机制,共享查询向量学习不同字段间的稀疏特征。这种设计允许模型在不同的信息维度上捕获用户兴趣的复杂性和多样性,同时保持模型的可解释性。此外,论文还提到了引入了“信息增益”来优化标签选择的过程,这有助于识别最有价值的标签,从而提高推荐系统的性能和效率。 通过这样的方法,UTPM模型可以更好地理解用户的行为模式,生成更精确的用户标签档案,进而提升推荐系统的准确性和用户满意度。这种方法不仅在理论上具有创新性,而且在实际应用中也有很大的潜力,特别是在大数据和社交媒体环境下的推荐服务。论文的贡献在于提供了一种新的视角来处理用户兴趣的表示和学习,为推荐系统领域的研究和实践提供了有价值的参考。
资源详情
资源推荐
![](https://csdnimg.cn/release/download_crawler_static/19518074/bg1.jpg)
Learning to Build User-tag Profile in Recommendation System
Su Yan
WeiXin Group, Tencent Inc.
Beijing, China
suyan@tencent.com
Xin Chen
WeiXin Group, Tencent Inc.
Beijing, China
andrewxchen@tencent.com
Ran Huo
WeiXin Group, Tencent Inc.
Beijing, China
lavinhuo@tencent.com
Xu Zhang
WeiXin Group, Tencent Inc.
Beijing, China
xuonezhang@tencent.com
Leyu Lin
WeiXin Group, Tencent Inc.
Beijing, China
goshawklin@tencent.com
ABSTRACT
User proling is one of the most important components in rec-
ommendation systems, where a user is proled using demographic
(e.g. gender, age, and location) and user behavior information (e.g.
browsing and search history). Among dierent dimensions of user
proling, tagging is an explainable and widely-used representation
of user interest. In this paper, we propose a user tag proling model
(UTPM) to study user-tag proling as a multi-label classication
task using deep neural networks. Dierent from the conventional
model, our UTPM model is a multi-head attention mechanism with
shared query vectors to learn sparse features across dierent elds.
Besides, we introduce the improved FM-based cross feature layer,
which outperforms many state-of-the-art cross feature methods
and further enhances model performance. Meanwhile, we design a
novel joint method to learn the preference of dierent tags from
a single clicked news article in recommendation systems. Further-
more, our UTPM model is deployed in the WeChat "Top Stories"
recommender system, where both online and oine experiments
demonstrate the superiority of the proposed model over baseline
models.
CCS CONCEPTS
• Information systems → Personalization.
KEYWORDS
personalization, user proling, neural networks, recommendation
systems
ACM Reference Format:
Su Yan, Xin Chen, Ran Huo, Xu Zhang, and Leyu Lin. 2020. Learning to
Build User-tag Prole in Recommendation System. In Proceedings of the 29th
ACM International Conference on Information and Knowledge Management
(CIKM ’20), October 19–23, 2020, Virtual Event, Ireland. ACM, New York, NY,
USA, 8 pages. https://doi.org/10.1145/3340531.3412719
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 prot 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 the
author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or
republish, to post on servers or to redistribute to lists, requires prior specic permission
and/or a fee. Request permissions from permissions@acm.org.
CIKM ’20, October 19–23, 2020, Virtual Event, Ireland
© 2020 Copyright held by the owner/author(s). Publication rights licensed to ACM.
ACM ISBN 978-1-4503-6859-9/20/10.. .$15.00
https://doi.org/10.1145/3340531.3412719
1 INTRODUCTION
In recent years, personalization techniques have been extensively
applied in large-scale recommendation platforms, such as YouTube
and Amazon. Widely used personalization techniques, such as CTR
(click-through-rate) prediction, evaluate a user’s preference for an
item based on user and article features. Therefore, an accurate se-
lection of user representation from over-loaded user information
is the backbone for satisfactory recommendation results. Conse-
quently, user proling is one way to enhance the performance of
recommendation systems by building accurate user representation
that improves personalization [12].
Figure 1 briey shows how user proling works in the WeChat
"Top Stories" news recommendation system. The architecture of
the recommendation system consists of four parts: news prole
layer, user prole layer, recall layer, and rank layer. In the news pro-
le layer, basic properties of the news article, such as its WeChat
subscription account, are recorded, while semantic information,
such as tag and news category, is extracted through natural lan-
guage processing techniques. As for the user prole layer, user’s
demographic information, such as gender and age, and behavioral
information, such as reading history, are collected. User proling
utilizes both behavior information and news prole, transmitting
news properties to users to reect users’ interest in certain kinds
of articles. Among these properties, tagging is an interpretable and
widely used representation of users’ interests in recommendation
systems [
2
]. In this study, we introduce how to build a user-tag
prole using the WeChat "Top Stories" recommender system as
an example. In this paper, "clicked tags" refer to tags within news
articles that the user clicked, and "un-clicked tags" refer to tags
within news articles that the user browsed but has not clicked. Note
that an article is associated with various tags related to its content.
In the recall layer, news articles related to the user’s interest are
recalled through several recall strategies, most of which take the
user-tag prole as an important feature input. Major recall strate-
gies are tag-based recall, where news articles containing user’s
interest tags are recalled, cf-based recall, which includes content-
based cf and user-based cf, and other model-based recalls. After
the aforementioned steps, the rank layer ranks the recalled news
articles based on user and news proles. Finally, top-ranked news
articles are shown to users. Due to the steep decline in the number
of candidate news articles from the recall layer to the rank layer,
the rank layer can use a more complex model with a higher number
of feature input than the recall layer without compromising on
Applied Research Track
CIKM '20, October 19–23, 2020, Virtual Event, Ireland
2877
下载后可阅读完整内容,剩余7页未读,立即下载
![](https://csdnimg.cn/download_wenku/file_type_ask_c1.png)
![](https://csdnimg.cn/download_wenku/file_type_ask_c1.png)
![](https://csdnimg.cn/download_wenku/file_type_ask_c1.png)
![](https://csdnimg.cn/download_wenku/file_type_ask_c1.png)
![](https://csdnimg.cn/download_wenku/file_type_ask_c1.png)
![](https://csdnimg.cn/download_wenku/file_type_ask_c1.png)
![](https://csdnimg.cn/download_wenku/file_type_ask_c1.png)
![](https://csdnimg.cn/download_wenku/file_type_ask_c1.png)
![](https://csdnimg.cn/download_wenku/file_type_ask_c1.png)
![](https://csdnimg.cn/download_wenku/file_type_ask_c1.png)
![](https://csdnimg.cn/download_wenku/file_type_ask_c1.png)
![](https://csdnimg.cn/download_wenku/file_type_ask_c1.png)
![](https://csdnimg.cn/download_wenku/file_type_ask_c1.png)
![](https://csdnimg.cn/download_wenku/file_type_ask_c1.png)
![](https://csdnimg.cn/download_wenku/file_type_ask_c1.png)
![](https://csdnimg.cn/download_wenku/file_type_ask_c1.png)
![](https://profile-avatar.csdnimg.cn/2c3cd2fb4fcf4e0cae5b42f9aa421c3a_u014037916.jpg!1)
Aboyomy
- 粉丝: 2
- 资源: 4
上传资源 快速赚钱
我的内容管理 展开
我的资源 快来上传第一个资源
我的收益
登录查看自己的收益我的积分 登录查看自己的积分
我的C币 登录后查看C币余额
我的收藏
我的下载
下载帮助
![](https://csdnimg.cn/release/wenkucmsfe/public/img/voice.245cc511.png)
最新资源
- 解决Eclipse配置与导入Java工程常见问题
- 真空发生器:工作原理与抽吸性能分析
- 爱立信RBS6201开站流程详解
- 电脑开机声音解析:故障诊断指南
- JAVA实现贪吃蛇游戏
- 模糊神经网络实现与自学习能力探索
- PID型模糊神经网络控制器设计与学习算法
- 模糊神经网络在自适应PID控制器中的应用
- C++实现的学生成绩管理系统设计
- 802.1D STP 实现与优化:二层交换机中的生成树协议
- 解决Windows无法完成SD卡格式化的九种方法
- 软件测试方法:Beta与Alpha测试详解
- 软件测试周期详解:从需求分析到维护测试
- CMMI模型详解:软件企业能力提升的关键
- 移动Web开发框架选择:jQueryMobile、jQTouch、SenchaTouch对比
- Java程序设计试题与复习指南
资源上传下载、课程学习等过程中有任何疑问或建议,欢迎提出宝贵意见哦~我们会及时处理!
点击此处反馈
![](https://img-home.csdnimg.cn/images/20220527035711.png)
![](https://img-home.csdnimg.cn/images/20220527035711.png)
![](https://img-home.csdnimg.cn/images/20220527035111.png)
安全验证
文档复制为VIP权益,开通VIP直接复制
![](https://csdnimg.cn/release/wenkucmsfe/public/img/green-success.6a4acb44.png)