Information Technology Reference
In-Depth Information
Researchers have usually developed research environments for the purpose of user
activity data acquisition from multiple sources. To monitor behaviour on the client-
side researchers mainly created new browsers, modified existing browsers or attached
JavaScript monitoring codes to websites. It is also important to note that quite com-
mon were studies focusing on both implicit and explicit feedback and their compari-
son, e.g. Claypool et al. [9].
The most complete classification of monitored behaviours consists of five catego-
ries: Examine, Retain, Reference, Annotate and Create [12]. For instance, reading is
considered as an action that allows one to Examine an object whereas page book-
marking allows to Annotate it. The interpretation confidence of observed behaviours
varies significantly depending on their category. Most frequently behaviours from the
Examination category constitute a weaker evidence of user interest than behaviours
from the Annotate or Create category. It was found that the highest confidence of user
intention is expressed in behaviours such as creating, printing and saving. Oard and
Kim [12] have also proposed to take into account the scope dimension. It allows to
classify behaviour according to the minimum possible scope of item being acted upon
(Segment, Object, or Class). For instance bookmarking can be performed on an Ob-
ject, while reading can be observed on a Segment of an Object.
Some commonly explored implicit feedback indicators concern activities per-
formed with mouse and keyboard such as mouse-move distance, scroll distance and
key input. Unfortunately, studies results [9,10,13,15,16] are inconsistent with the
thesis that all indicators based on those actions are good and positive indicators of
user interest.
The intuitive belief that users spend more time on documents that they find relevant
has been confirmed in a large number of studies [12,17,18]. However, some researchers
[10,18,19 ] indicate that time may not be a reliable indicator of user interest, in particu-
lar when analyzing human website interaction. It may be caused by the absence of the
user who opened a certain webpage or by multitasking. Therefore, later in the article
two more precise indicators: active tab time and user activity time will be proposed.
3
E-commerce Customer Preference Monitor - ECPM
The main motivation for building ECPM was to create a solution which will allow to
run a scientific study and monitor user interactions with any e-commerce website,
without the necessity of obtaining special agreements from the site's administrators.
An equally important incentive was the desire to build a tool which would record
user's behavior on a website in great detail and provide data which could be used for
analysis with plethora of knowledge discovery methods, contrary to commercially
available monitoring tools (which allow manual analysis of results mainly just on
click maps, mouse path maps etc.).
A number of possible options for building the monitoring tool have been consi-
dered. One of them was to modify an existing browser or to build a new browser es-
pecially for the purpose of monitoring. The latter option, that is introducing a new
browser with a new interface and special functions could have a high impact on the
behavior of the users, unfortunately. As a result, data gathered in such a study could
be greatly biased. Another concern about building the monitoring tool was its appli-
cability on real e-commerce websites. Therefore, it was decided to limit collected data
Search WWH ::
Custom Search