calamari test alpha 8

  (C)2005- Masahide Yuasa. All rights reserved.


  This is a test program and should be used at your own risk.


Calamari is a tool for extracting positions of words in link tags, 
image tags, and table tags in voice-controlled or gaze-controlled browser windows.
It extracts the words based on the user's browser with JavaScript.
This is implemented in a proxy server; thus, it does not depend on browsers 
and operation systems. Further, although this tool rewrites the 
HTML document, it does not change the web page appearance.
Moreover, this tool can run on various platforms such as Linux 
and Windows(a Java runtime environment).


Requirements

  Java 1.4 or later

  HTMLParser 1.5 or later
  Please download the parser from http://htmlparser.sourceforge.net/

  The most recent version of a web browser with JavaScript enabled
  (Internet Explorer, Mozilla, or Opera). 


Output Files

  The results of extraction will output to the following files:

  calamari_a8\local_server\contents\
    link.txt    (positions of text from [a tag])
    img.txt     (positions of images by [img tag])
    table.txt   (positions of tables by [table tag])

  (Please create a new tool that utilizes the position files.)


Quick Start

  1. Please download HTMLParser and extract the downloaded file.
     Further, copy the "htmlparserXX_XX\bin\htmlparser.jar" to the "calamari_a8\" directory.

  2. Please set IP address "127.0.0.1" and Port "8889" as the proxy server of
     your browser. ("8889" is the default port for this tool.)
     Also, set "HTTP/1.0," if possible (change HTTP/1.0 to HTTP/1.1 for using HTTP).

  3. Execute calamari_a8\bin\server.bat

  4. Execute calamari_a8\bin\client.bat
     (Please set IP address for server.)

  5. Please access the required page.
     (Please ensure that the "popup window" is enabled.
      The tool enables the popup window to extract positions.)

  6. If the behavior is appropriate, please check the text files in 
     calamari_a8\local_server\contents\.


Concept

As an example, let us consider that there are two items of "News"
in an HTML document on a news website. Conventional voice-controlled 
browsers may find only the first item of "News" and display that 
page, even if the user wants to access the second item of 
"News" and he/she inputs the word "News". 
Typical browsers solve this problem by associating some keywords 
with "News," for example, "News A" and "News B"; the user is then expected
 to say either "News A" or "News B".
However, users find it difficult to say the word.
This may change the web page appearance, and the web designer's 
intention is distorted.

Therefore, I propose a method of navigation that uses
information regarding the position of a word in the browser window. 
Using this information, the browser can select the item and
display the page desired by the user.
The user can specify not only the text of the link but also the 
position; for example, "lower news" and "upper news."

Calamari marks the words of the links with an HTML tag. 
In order to extract the positions of the words, this tool
uses a proxy function to rewrite the HTML document 
and adds JavaScript and tags for each word.

Since the positions of words are important for a gaze-controlled interface,
I will attempt to use the tool for the same.


Regarding Split of Texts

In the case of some conventional voice-controlled browsers,
the user was required to state the entire link for accessing a page.
For example, the user had to state "Online Game News" and 
"Reviews by Tanaka" as the link. It would be easier
 if the user would only have to state "News", "Reviews" or "Tanaka".
The tool splits the text into words; thus, the user is only 
required to state a part of the text.
The English language separates words using spaces.
Therefore, the tool checks the spaces and rewrites the link tag. 
(However, in the cases of another languages (for example, Chinese, 
Japanese, and Korean), the tool uses Morphological Analysis.)


Regarding Frames

The user can display a page with links that are separated by frames.
Calamari can extract the coordinates of the links despite the frames.



Reference

Masahide Yuasa, et al.
Voice-controlled Browser Using Information on the Positions of Words in the
Browser Window.
Proc. of HCII 2005 (in press).