Is "sentiment power" of vowels a known result?

asked 2015-12-19 19:29:24 -0400

niallm gravatar image

Hi all,

I am not a professional linguist, as is probably already clear -- merely a dabbler. But I am wondering if something I came across a while back is a known result, and if it has been proved in the same way.

A couple of years ago, I took an existing "sentiment analysis" database (which mapped words to a real number value expressing how positive or negatively they were found to be, e.g. "assassination: -3.2159, "birthday": +2.506) and then converted each word into phonemes, and averaged the sentiment number across those phonemes, for every word in the database. This gives each phoneme some weighted value of how often it's used in more "powerful" words as opposed to more neutral ones.

When you graph this you get the (surprising to me, the naif) result that vowel-oriented phonemes cluster at either end, being relatively strongly powerful, and the consonant-oriented phonemes cluster in the middle, being relatively weak.

Is this a known result? Is this how it is usually derived?

Thanks for your time.

edit retag flag offensive close merge delete