21.01.2014 Views

improving music mood classification using lyrics, audio and social tags

improving music mood classification using lyrics, audio and social tags

improving music mood classification using lyrics, audio and social tags

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

5.2.2.4 Negative Samples<br />

Figure 5.1 Example of labeling a song <strong>using</strong> <strong>social</strong> <strong>tags</strong><br />

In a binary <strong>classification</strong> task, each category needs negative samples as well. The negative<br />

sample set for a given category are chosen from songs that are not tagged with any of the terms<br />

found within that category but are heavily tagged with many other terms. Since there are plenty<br />

of negative samples for each category, a song must satisfy all of the following conditions to be<br />

selected as a negative sample:<br />

1) It has not been tagged by any of the terms in this category;<br />

2) The total normalized counts of all <strong>tags</strong> that are not in this category is no less than 100;<br />

3) The minimum normalized count among all <strong>tags</strong> associated with this song is 0 or 1.<br />

Condition 2) <strong>and</strong> 3) together make sure the total absolute count of “other” <strong>tags</strong> is no less,<br />

<strong>and</strong> probably much more than 100.<br />

Similar to positive samples, all negative samples have at least 100 words in their unfolded<br />

lyric transcripts. For each category, the positive <strong>and</strong> negative set sizes are balanced, <strong>and</strong> thus the<br />

total number of examples in all categories is 12,980.<br />

64

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

Saved successfully!

Ooh no, something went wrong!