Joint Learning on Relevant User Attributes in Micro-blog
Jingjing Wang, Shoushan Li
∗
, Guodong Zhou
Natural Language Processing Lab, School of Computer Science and Technology
Soochow University, Suzhou, 215006, China
djingwang@gmail.com, {lishoushan, gdzhou}@suda.edu.cn
Abstract
User attribute classification aims to identify users’
attributes (e.g., gender, age and profession) by
leveraging user generated content. However, con-
ventional approaches to user attribute classification
focus on single attribute classification involving
only one user attribute, which completely ignores
the relationship among various user attributes. In
this paper, we confront a novel scenario in user at-
tribute classification where relevant user attributes
are jointly learned, attempting to make the rele-
vant attribute classification tasks help each other.
Specifically, we propose a joint learning approach,
namely Aux-LSTM, which first learns a proper
auxiliary representation between the related tasks
and then leverages the auxiliary representation to
integrate the learning process in both tasks. Em-
pirical studies demonstrate the effectiveness of our
proposed approach to joint learning on relevant user
attributes.
1 Introduction
Social media, such as Twitter and Facebook, enables the
users to post messages and share information in social net-
works, producing an unprecedented amount of user gener-
ated content (UGC) with rich facts about the users, includ-
ing their personal attributes. Since then, UGC has been ap-
plied to various user attribute classification tasks, which rec-
ognizes user attributes, such as gender
[
Wang et al., 2015a;
Zhu et al., 2015
]
, age
[
Marquardt et al., 2014
]
and pro-
fession
[
Tu et al., 2015
]
. During the last few years, user
attribute classification has drawn more and more attention
due to its great potential influence to various applications,
such as personality analysis, intelligent marketing and on-
line advertising
[
O’Connor et al., 2010; Volkova et al., 2013;
Preotiuc-Pietro et al., 2015
]
.
However, previous studies mainly focus on single attribute
classification involving only one attribute, which ignores the
relationship among various user attributes. Intuitively, the re-
lationship among various user attributes may benefit different
attribute classification tasks and should be considered. For
∗
Corresponding author
User A
Gender: Male
Age: 19
Profession: Student
Message text: I have enough energy to writing C code every-
day, because I was only 19 years old.
User Attribute Classification Tasks
- Task 1: Profession Classification
Input: Message text
Output: IT (× Wrong)
- Task 2: Age Classification
Input: Message text
Output: 19 (
√
Correct)
- Our Task: Joint Learning (Profession+Age Classification)
Input: Message text
Output: Profession: Student (
√
Correct)
Age: 19 (
√
Correct)
Figure 1: An example of joint learning on user attribute
instance, Figure 1 gives the true personal attributes of user A
with an attached text. According to phrase “writing C code”
in the message text, user A is very likely to have profession
“IT worker”, which is actually not true since his/her real pro-
fession attribute is “Student”. However, if the age of user
A is correctly classified to be “19” according to phrase “19
years old”, we can easily adjust his/her profession attribute to
be “Student” since a 19 year-old person is more likely to be a
college student than an IT worker. Therefore, in some scenar-
ios, a user’s one attribute is helpful to infer his/her another at-
tribute. Therefore, a feasible way to improve the performance
of user attribute classification is to perform joint learning on
relevant user attributes by capturing the relationship among
various user attributes.
In this paper, we address a novel scenario in user at-
tribute classification, namely joint learning on relevant user
attributes. Suppose there are two user attributes involving
in our user attribute classification tasks, we first separate
the twin user attribute classification task into a main task
and an auxiliary task and then propose a joint learning ap-
proach to boost the performance of the main task with the
help of the auxiliary task. In particular, our joint learning
approach is based on a neural network architecture, namely
Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17)
4130
下载后可阅读完整内容，剩余6页未读，立即下载
weixin_38610070
- 粉丝: 2
- 资源: 940
我的内容管理
展开
最新资源
- C++标准程序库：权威指南
- Java解惑：奇数判断误区与改进方法
- C++编程必读：20种设计模式详解与实战
- LM3S8962微控制器数据手册
- 51单片机C语言实战教程：从入门到精通
- Spring3.0权威指南：JavaEE6实战
- Win32多线程程序设计详解
- Lucene2.9.1开发全攻略：从环境配置到索引创建
- 内存虚拟硬盘技术：提升电脑速度的秘密武器
- Java操作数据库：保存与显示图片到数据库及页面
- ISO14001：2004环境管理体系要求详解
- ShopExV4.8二次开发详解
- 企业形象与产品推广一站式网站建设技术方案揭秘
- Shopex二次开发：触发器与控制器重定向技术详解
- FPGA开发实战指南：创新设计与进阶技巧
- ShopExV4.8二次开发入门：解决升级问题与功能扩展
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
