Measuring the Relevance of Different-typed Objects
in Weighted Signed Heterogeneous Information
Networks
Tianchen Zhu
∗
, Zhaohui Peng
∗
(), Senzhang Wang
†
, Philip S. Yu
‡§
and Xiaoguang Hong
∗
∗
School of Computer Science and Technology, Shandong University, Jinan, China
ztc@mail.sdu.edu.cn, {pzh,hxg}@sdu.edu.cn
†
College of Computer Science and Technology,
Nanjing University of Aeronautics and Astronautics, Nanjing, China
szwang@nuaa.edu.cn
‡
Department of Computer Science, University of Illinois at Chicago, Chicago, USA
§
Institute for Data Science, Tsinghua University, Beijing, China
psyu@uic.edu
Abstract—Relevance measure in both homogeneous and het-
erogeneous networks has been extensively studied. However,
how to measure the relevance among different-typed objects
in weighted signed heterogeneous information networks remains
an open problem. It is challenging to incorporate both positive
and negative multi-typed relationships simultaneously in signed
heterogeneous networks due to the opposite opinions implied
by them. To this end, this paper proposes a random walk
based approach for relevance measure by utilizing and modeling
the rich semantic information in weighted signed heterogeneous
networks. Particularly, we first transform a signed network into a
non-signed network according to the different semantic meanings
represented by positive and negative relationships. This paves the
way to properly utilize negative relationships. Next, we conduct
random walk from the source object to the target object based
on a bunch of single meta-paths separately. Finally, we combine
multiple meta-paths together to obtain a more comprehensive
relatedness between the source object and the target object.
Extensive experiments on real datasets demonstrate the superior
performance of the proposed approach.
Index Terms—relevance measure, meta-path, weighted signed
heterogeneous network.
I. INTRODUCTION
In recent years, heterogeneous information network has
attracted extensive research interests in data mining com-
munity[1,2,3]. Many interesting and practically important re-
search issues can be conducted on heterogeneous informa-
tion networks, of which relevance measure is a fundamental
work. There are a few good studies leveraging link infor-
mation in networks for relevance measure, such as personal-
ized PageRank[4], SimRank[5], PathSim[6] and HeteSim[7].
However, conventional researches mainly focus on measuring
similarity of same-typed objects in non-signed homogeneous
information networks or relatedness of objects of different
types in non-signed heterogeneous information networks. In
real world, there are many signed networks with both negative
and positive links, where positive links represent positive rela-
tionships, while negative links represent negative relationships.
For example, users in Wikipedia can vote for or against the
nomination of others to adminship; users in Epinions can
express trust or distrust on others; and participants in Slashdot
can declare others to be either “friends” or “foes”[8,9].
We define the sign of a link to be positive or negative based
on whether it expresses a positive or negative attitude from
the generator of the link to the recipient. Therefore, signed
networks can be regarded as preference networks. Moreover,
we not only define signs to each link, but also assign it a
weight denoting the degree of likeness or dislikeness. For
example, in a movie review network as shown in Fig. 1, we can
transform the rating score of a user on a movie into the weight
of the corresponding link between them. In Fig. 1, the weight
of the link from the user Marry to the movie Forrest Gump
is 2.3 while the weight of the link from Marry to the movie
Mrs. Winterbourne is only 0.13. This means that Marry prefers
Forrest Gump to Mrs. Winterbourne although both links are
positive.
Although measuring the relevance between objects of dif-
ferent types in a signed heterogeneous information network
is particularly important in many applications, it is still not
fully explored due to the following challenges. First, exist-
ing approaches for non-signed information networks cannot
be directly applied to signed information network. It is an
open challenge to effectively incorporate both positive and
negative relationships in model-based methods for relevance
measure. Second, it is also challenging to fully utilize and
fuse the heterogeneous and rich relationships in a weighted
signed heterogeneous information network. For a specific pair
of objects, we may get totally different relevance measures
following the different search paths that connect two objects
through a sequence of relations. Therefore, a general relevance
measure model for weighted signed heterogeneous information
networks is necessary.
In this paper, we propose a novel meta-path based method-
ology called WsRel (Weighted Signed Rel evance Measure)
Proceedin
s of the 2017 IEEE 21st International Conference on Computer Supported Cooperative Work in Desi
n
978-1-5090-6199-0/17/$31.00 ©2017 IEEE 556