VI /342
feature keywords, R-OSI technique can accordingly discover association rules for the focal
product feature from a set of unannotated consumer reviews crawled from various social
media websites. These rules are then applied to identify opinionated sentences relevant to the
focal product feature. Their empirical evaluation results suggest that the effectiveness of their
proposed R-OSI technique with one keyword for each product feature is comparable to that
of the fully supervised benchmark, in which the manual preparation of training data is
necessary. Furthermore, the R-OSI technique is sensitive to the number of keywords for each
product feature while insensitive to the size of the set of unannotated consumer reviews.
Overall, these findings suggest the R-OSI technique a promising approach for opinion
sentence identification.
The paper by
Ying-Ho Liu
addresses the issue on representation of the discovered
knowledge. The number of frequent univariate uncertain patterns discovered is usually large
and difficult for handling, and thus a good representation might help people to quickly
understand and utilize them. In this study, a method for summarizing frequent univariate
uncertain patterns was proposed, namely Summarize FU2Ps by Clustering (SFC) algorithm.
The SFC algorithm adopts a hierarchical clustering technique based on a modified
k-medoids algorithm to derive suitable number of representative clusters of FU2Ps. The
author conducts two evaluation experiments by using a synthetic dataset and a real dataset,
respectively. Both results suggest that the summarization quality of the proposed method is
better than that of a benchmark method.
The paper by
Sheng-Tsung Hou
,
Wan-Chien Lien
, and
Chieh-Min Chou
investigates
the relationships between service inventory and operational performance, using Taiwan taxi
fleet as an example. “Service inventory” implies that a firm keeps a “service in process”
within a service-based supply chain as a form of temporary stock. The “temporary service
stock” is a value activity within the service supply chain; once the service is demanded by a
customer, the service provider will deploy it as “the final service”. Because service demand
and supply do not always occur at the same time, firms have to adopt strategies and practices
of service inventory management to better synchronize the service demand and supply. This
study applies clustering analysis to identify and analyze the relationship between service
inventory and operational performance. It analyzes the time-space air-queuing data of 300
taxi drivers collected from the satellite system of Taiwan Taxi Fleet group. Four categories of