Information Technology Reference
In-Depth Information
2.3.4.2 Chart Page
The aim of this page is to show the line charts of statistical data of brands according
to days. If a user clicks on the “Please click for statistical data” link which can be
seen from Fig.
2.8
, the chart page is opened and shows the data as can be seen from
the Fig.
2.9
. First, it connects to database to collect the last 9 days results and then
retrieves the following data to draw the charts.
•
Percentage of positive tweets about Twitter classified by the Naïve Bayes method.
•
Percentage of positive tweets about Facebook classified by the Naïve Bayes
method.
•
Percentage of positive tweets about Twitter classified by the Maximum Entropy
method.
•
Percentage of positive tweets about Facebook classified by the Maximum Entropy
method.
2.3.4.3 Trigger Page
The aim of this page is to run scheduled tasks using only ASP.NET without setting a
Windows web service. By means of a Trigger page, the web interface of our system
could be run on every web server which supports ASP.NET hosting. Trigger page is
called hourly to do scheduled tasks which can be listed as follows. First, it calls the
executable filewhich is responsible for tweet collecting and classification—described
in detail in the sub section of automated tweet collecting and classification—so as to
collect and classify the last tweets sent. Then, it connects to the database in order to
draw the last count of positive and negative tweets, such as last count of tweets about
Twitter by users located in London on a given date. Third, it amends the database
after getting the last count of positive and negative tweets. Lastly, it deletes duplicate
records because it is possible that Twitter API might collect the same tweets as
collected in the previous call.
To sum up the whole system, Fig.
2.10
shows the flowchart of the Web Interface.
2.4 Evaluation
2.4.1 Performance of Classification Methods by Number of
Data
In this section, the performance of the Naïve Bayes and Maximum Entropy clas-
sification methods of the NLTK are evaluated according to the number of training
data. In order to do this task, training datasets are produced based on hand-classified
dataset which is described in Sect.
2.3
. For all of the evaluations, the same testing
Search WWH ::
Custom Search