Service Discovery based on User Latent Intentions
under Massive Services Environments
WANG HaiYan, ZHOU AiDong
College of Computer Science, Nanjing University of Posts & Telecommunications, Nanjing, 210036, China
Under massive services environments, traditional service discovery methods cannot help users discover desired services quickly and
effectively from large amount of candidate services any more. On the other hand, users’ personalized requirements have become one
of the important issues needing to be addressed during the process of service discovery. Current solutions have not paid enough
attention to potential differences between different preferences of users, which badly decrease the effectiveness of service discovery.
To address the problems above, a concept of user latent intentions (ULI) consisting of user preference intentions and user herd
intentions is introduced for the expression of a user’s personalized requirements of services in this paper. Theory of TCP-net and
herd psychology is employed for the establishment of user preference decision model and user herd decision model, followed with a
user-centric service discovery algorithm based on ULI. The proposed ULI-based service discovery approach can increase the
effectiveness of service discovery by filtering out undesired services according to users’ personalized requirements. Simulation and
experimental results demonstrate feasibility and effectiveness of the proposed method.
service discovery, preference, service, latent intention
1. Introduction
With continuous development of Information
Technology and wide spread of networks, unit for
measuring the number of services on Internet has
increased from the original gigabytes(GB) or
terabytes(TB) to the current petabytes (PB)or even
zettabytes(ZB). According to the report released by the
famous international data corporation (IDC), the amount
of data on Internet in 2011 was about 1.8ZB, and the
estimated total number will be increased to 35ZB in 2020.
The amount of data from the analysis platform of the
world’s biggest online marketplace eBay is about
100PB
[1]
every day. And daily amount of storage data
from the Chinese well-known trading platform Taobao is
beyond 40PB. A great number of service resources on
the Internet offer more choices for service requesters.
However, with massive number of services and the
related various service properties, how to discover most
required services for users has become one of the serious
problems. Early existed service discovery methods have
many drawbacks under massive services environments.
For example, low efficiency problem will be serious for
those traditional keyword-based or grammar-based
service discovery approaches because of their poor
ability in getting rid of semantic description. And for
those semantic description-based approaches such as
SAWSDL, OWL-S and WSMO, owing to the
complexity of semantic computation and generation of
ontology, time consumption will be increased
unexpectedly.
Meantime, from our observations of related research
works in literature, most of the available service
discovery approaches focus on service itself.
Non-functional properties (NFP) of services are regarded
as criteria to discover optimal services with no concern
on properties of users at all. Therefore the found services
probably can hardly meet the users’ real requirements.
Even though some research works have been done to
discover services based on the matching of both
functional properties (FPs) and NFPs of users with those
of candidate services, these approaches may not be
effective because the amount of similar services in
services set may still be beyond the ability of users to
select. Problems in most of the current service discovery
approaches under massive services environments can
generally be summarized as follows.
1) Users’ personal requirements have not been
received enough attention during service discovery.
More attention should be paid on users’ personal
requirements when massive and various kinds of released
services are given.
2) Preferences of users have not been taken into
account during service discovery and relation between
different preferences should also been taken into
consideration especially when there are many similar
candidate services in the services set.
3) Difficulty of service discovery has been increased
when users have no or few knowledge of required
services in services set under massive services
environments.
To address the problems above, a User Latent
Intentions(ULI) based service discovery approach will be
proposed in this paper. Contributions of our work in this
paper are listed as below.
1) A conception of User Latent Intentions is defined to
express personalized service requirements of users in
service discovery. User Latent Intentions consist of User
Preference Intentions and User Herd Intentions, in which
user preference intention is put forward to help express
the relation between user preferences and to decrease the
number of similar candidate services, and user herd
intention is introduced to solve the problems when users
have no or few experiences in finding desired services.
2) A user-centered service discovery approach is
proposed, which takes users’ requirements as criteria of
service discovery so that optimal services set will be
obtained. The proposed approach will employ Trade
off-enhanced conditions preferences network (TCP-net)
to help express the relationship between users’
preferences. And a User Herd Intentions Model will also
be built to help users discover services when they have
no or few experiences before.
Rest of the paper is organized as follows. Related
works will be given in section 2, a description of our
proposed service discovery approach including related
theory and the discovery algorithm will be given in