27.06.2013 Views

Proceedings of the 8th International Conference on Intellectual ...

Proceedings of the 8th International Conference on Intellectual ...

Proceedings of the 8th International Conference on Intellectual ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Harri Ketamo<br />

Figure 3: Semantics with piece <str<strong>on</strong>g>of</str<strong>on</strong>g> c<strong>on</strong>tent that include tags ‘interface design’ and ‘usability’<br />

The next piece <str<strong>on</strong>g>of</str<strong>on</strong>g> c<strong>on</strong>tent do have tags ‘usability’ and ‘graphical user interface’. The semantics grows<br />

(figure 4) and ‘usability – interface’ –relati<strong>on</strong> becomes even str<strong>on</strong>ger. In this example, this relati<strong>on</strong><br />

describes <str<strong>on</strong>g>the</str<strong>on</strong>g> c<strong>on</strong>text <str<strong>on</strong>g>of</str<strong>on</strong>g> <str<strong>on</strong>g>the</str<strong>on</strong>g>se videos with most explanative value. However, c<strong>on</strong>text based <strong>on</strong> <strong>on</strong>ly<br />

few pieces <str<strong>on</strong>g>of</str<strong>on</strong>g> c<strong>on</strong>tent is not relevant at all. According to our experiments, a useful semantics should<br />

be based <strong>on</strong> more than 200 pieces <str<strong>on</strong>g>of</str<strong>on</strong>g> c<strong>on</strong>tent; what more c<strong>on</strong>tent that better semantics.<br />

Figure 4: Semantics with piece <str<strong>on</strong>g>of</str<strong>on</strong>g> c<strong>on</strong>tent that include tags ‘graphical user interface’ and ‘usability’<br />

In real case, semantics w<strong>on</strong>’t grow <strong>on</strong>ly based <strong>on</strong> relevant tags. There are always tags that are too<br />

frequent and in that sense <str<strong>on</strong>g>the</str<strong>on</strong>g>ir explanative power is poor. For example tags like ‘news’ or ‘web’ are<br />

included in that many tag sets that <str<strong>on</strong>g>the</str<strong>on</strong>g>y will build <str<strong>on</strong>g>the</str<strong>on</strong>g> most str<strong>on</strong>gest relati<strong>on</strong>s between almost all<br />

c<strong>on</strong>cepts. In figure 5 <str<strong>on</strong>g>the</str<strong>on</strong>g> phenomena is illustrated: ‘web’ and ‘news’ forms str<strong>on</strong>g relati<strong>on</strong>s with all <str<strong>on</strong>g>the</str<strong>on</strong>g><br />

c<strong>on</strong>cepts (<strong>on</strong>ly some relati<strong>on</strong>s are drawn in order to remain <str<strong>on</strong>g>the</str<strong>on</strong>g> readability <str<strong>on</strong>g>of</str<strong>on</strong>g> <str<strong>on</strong>g>the</str<strong>on</strong>g> figure) and this<br />

c<strong>on</strong>structs <str<strong>on</strong>g>the</str<strong>on</strong>g> illusi<strong>on</strong> <str<strong>on</strong>g>of</str<strong>on</strong>g> ‘web’ and ‘news’ as most important tags in this c<strong>on</strong>text.<br />

We know that ‘web’ and ‘news’ are irrelevant and <str<strong>on</strong>g>the</str<strong>on</strong>g>y can also easily been extracted out <str<strong>on</strong>g>of</str<strong>on</strong>g> <str<strong>on</strong>g>the</str<strong>on</strong>g><br />

semantics. If a tag has 1) too many c<strong>on</strong>necti<strong>on</strong>s to o<str<strong>on</strong>g>the</str<strong>on</strong>g>r tags compared to average tag , 2) <str<strong>on</strong>g>the</str<strong>on</strong>g><br />

average strength <str<strong>on</strong>g>of</str<strong>on</strong>g> <str<strong>on</strong>g>the</str<strong>on</strong>g>se relati<strong>on</strong>s is high compared to o<str<strong>on</strong>g>the</str<strong>on</strong>g>r average strengths and most <str<strong>on</strong>g>of</str<strong>on</strong>g> all 3) <str<strong>on</strong>g>the</str<strong>on</strong>g><br />

variance in such strengths is relatively small, we can be sure that such tag is irrelevant in terms <str<strong>on</strong>g>of</str<strong>on</strong>g><br />

semantic search. Ano<str<strong>on</strong>g>the</str<strong>on</strong>g>r irrelevant group <str<strong>on</strong>g>of</str<strong>on</strong>g> tags c<strong>on</strong>sist <str<strong>on</strong>g>of</str<strong>on</strong>g> words that have <strong>on</strong>ly few relati<strong>on</strong>s with<br />

small strength. Both kind <str<strong>on</strong>g>of</str<strong>on</strong>g> tags and relati<strong>on</strong>s can be excluded from <str<strong>on</strong>g>the</str<strong>on</strong>g> final semantics and <str<strong>on</strong>g>the</str<strong>on</strong>g><br />

semantic search is based <strong>on</strong>ly <strong>on</strong> tags that have well enough explanative power.<br />

The use case, in brief, is following: At first user describes <str<strong>on</strong>g>the</str<strong>on</strong>g> subject in focus by typing several key<br />

c<strong>on</strong>cepts (tags maybe) into UI’s definiti<strong>on</strong> field, for example ‘eyetracking and heatmaps’. Sec<strong>on</strong>dly,<br />

<str<strong>on</strong>g>the</str<strong>on</strong>g> first types <str<strong>on</strong>g>of</str<strong>on</strong>g> agents start to search all possible c<strong>on</strong>tent related to keywords. Currently search is<br />

304

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!