A Combined Thematic and Acoustic Approach for a Music Recommendation Service in TV Commercials

We hypothesize that different genres of writing use different adjectives for the same concept. We test our hypothesis on lyrics, articles and poetry. We use the English Wikipedia and over 13,000 news articles from four leading newspapers for the article data set. Our lyrics data set consists of lyrics of more than 10,000 songs by 56 popular English singers, and our poetry dataset is made up of more than 20,000 poems from 60 famous poets. We find the probability distribution of synonymous adjectives in all the three different categories and use it to predict if a document is an article, lyrics or poetry given its set of adjectives. We achieve an accuracy level of 67% for lyrics, 80% for articles and 57% for poetry. Using these probability distribution we show that adjectives more likely to be used in lyrics are more rhymable than those more likely to be used in poetry, but they do not differ significantly in their semantic orientations. Furthermore we show that our algorithm is successfully able to detect poetic lyricists like Bob Dylan from non-poetic ones like Bryan Adams, as their lyrics are more often misclassified as poetry.