Projects‎ > ‎

Melody Matcher, Version Alpha

Brief Summary:
Melody Matcher is a semi-automated music composition support program.  It analyzes English lyrics along with a melody, and alerts the composer of the locations in the song where the lyrics are not understandable. 

Basically, it's grammar- and spell-check for songs.  This is significant, because no research had been done on the quantifiable measurement of English-language lyric understandability, other than my project.

Target Audience and Goals:

This program will be used as a compositional aid by anyone who wants to write songs and make them sound good, technically. Its should allow the song writer to focus on more subjective criteria of what makes a song “good”, because it will make the structural rules of lyric composition immediately apparent.

My hope for this project is that it will be useful to burgeoning songwriters, who have the creative spark to make wonderfully poetic lyrics, but lack the "ear" to match their lyrics successfully to music.  It should be particularly helpful to songwriters who place a high emphasis on understandability of lyrics (such as parody song writers, or writer for musical theater). 

Additionally, Melody Matcher will be useful for songwriters for whom English is a second language.  They may be a master lyricist in their native language, but writing lyrics in English can be a particular challenge, since so much of lyric-writing is dependent upon knowing the cadence of the language you're writing lyrics in, and since English has no easily-discernible rules for emphasis placement in words.

Practical Example:

The structural rules of lyric placement are important, because without them, lyrics can become muddled and/or unintelligible. For example, in the song “Groovin’ (on a Sunday Afternoon)”, by the Young Rascals, there’s a part in the bridge that everyone always hears as “Life would be ecstasy, you an’ me an’ Leslie”. In fact, the line is “Life would be ecstasy, you and me endlessly. The confusion lies with the last three syllables of the phrase. The pronunciation of each version, if spoken normally, is as follows:

 Normal Spelling Breakdown:
and    Les- lie
end-    less-    ly
 SAMPA Breakdown:
@nd “lEs li
“End    l@s    li
A “ preceding a syllable means that syllable is emphasized.

So, in the first phrase, we see that the emphasis pattern goes something like dum-DUM- dum, where the first syllable of “Leslie” is emphasized. The second phrase’s emphasis pattern is “DUM-dum-dum”, so the first syllable of endlessly is emphasized.

When words are put to music, however, the musical emphasis overrides the textual emphasis. Sometimes,the meaning of the phrase can change, if a previously un-emphasized syllable becomes emphasized, or a previously emphasized syllable loses its emphasis.

For “Groovin’”, the lyrics match up to the music in the song as follows:

Life would be ec-    sta- sy,    You an’ me end- less- ly

In this musical phrase, the emphasis always goes on the first part of a beat (for the purposes of this example, a "beat" is defined as a quarter note).


In this case, in the first measure is emphasized for the notes that correspond to the lyrics, “Life”, “be”, “ec-”(as in ec-sta-sy) and “sy”(again, as in ec-sta-sy) (This is a vast oversimplification, but it works for now). So, the lyrics would be emphasized as such:

Life would be ec-    sta- sy,    You an’ me end- less- ly

Or, more simply:

Life would be ec-sta-sy

This musical emphasis matches the spoken emphasis of the phrase, so it is intelligible as a lyric.  (Though ecstasy’s first syllable doesn’t start on the first part of beat three, it is still on the first part of beat three, and therefore still emphasized.  Alternatively, since the first part of beat two didn’t have a hard stop to it, the emphasis could have rolled over to the second part,  “ec”, which does have a hard stop.)

In contrast, take the second measure:  the syllables “You”, “me”, and “less” are emphasized in the music.  This leads to conflicting musical and spoken phrasing:

Musical Phrasing: You and me  endlessly

Spoken Phrasing: You and me  endlessly

The singer is now singing the phrase, syllable by syllable, which they think of as syllable-note combinations:

YOU and ME end LESS lee

The singer, for his part, is doing what many singers are taught to do, to make it easier to sustain the singing of words that end with unsingable consonants: the unsingable consonant is displaced onto the front of the next word.  In this case, the consonant “d” is not singable, so he displaces it onto the next syllable, when he can:  “and ME” becomes “an   dME”, and “end LESS”  becomes “en  dLESS”.  So, the singer can effectively think of the sung phrase as:

YOU an  dME en dLESS lee

This doesn’t cause confusion to listeners, because they’re used to hearing it.  This does mean, however, that lyric placement does not provide an accurate barometer to a listener of where a word actually ends.  

 

 In addition, the singer is singing fudging his vowels, like singers are taught to do,   so “and” and “end” sound almost indistinguishable.  So, really, what listeners are hearing is this:

  YOU en dME en dLESS lee

Now, the listener’s brain has to take this syllabic gobbldygook, and parse it into something useful.  They’ve currently got this mess to deal with (represented in SAMPA syllables):

ju En dmi En dl@s li

They parse the first part just fine, because the emphases match:

you and me En dl@s li

But no one says endLESSly.  People say ENDlessly.  So,  the listeners don’t recognize it. they have to work with what they have.  They already turned one “En d” into an “and”, so they do it again:

you and me and l@s li

 Now, they’re just left with LESS lee.  And that fits Leslie, a proper noun that fits in context and in emphasis placement.  So, the final heard lyric is:

you and me and Les- lie

The misunderstanding can be traced by to improper emphasis placement.  The songwriter probably didn’t even think of that, and now he’s stuck:  a one-hit-wonder with a misunderstood song.  I bet that in interview after interview, someone asks him who Leslie is.  It’s probably very frustrating—especially since he could have just moved the word an eight note later, and it would have been understood perfectly. 


That’s the sort of situation this program is going to help avoid.




Explanation of original work:
I started on this project, thinking that there would be lots of carefully-specified research on how lyrics match melodies, mathematically.  As it turned out, there was very little objective literature on the subject.  Most of the resources I found were songwriting textbooks, and they almost uniformly assumed that the reader could already sense where lyrics should and shouldn't go. 

Because of the lack of objective information of the subject, I had to develop the method from scratch.  As I progressed through my work, I went from thinking that understandability depended only on emphasis-matching, to realizing that syllable length played a huge part as well, to realizing that note intervals could occasionally override both of those, to realizing that similarities in the way the mouth is shaped when sounds are formed plays into the equation as well. 

In the Melody Matcher Alpha, only emphasis matching and syllable length are taken into account, because of all the foundational work I had to do. 


Still Interested?
Check out the Melody Matcher wiki, which is for the Beta version of the program.  You can also get to the archived version of the wiki, for the Alpha iteration, but I've helpfully attached the pdfs here for each milestone, for your convenience

If you're looking for a technical, in-depth explanation of how syllable-length, emphasis matching, and note intervals affect the matching method, I'd suggest checking out the "Project Proposal, Week 6"  and "Project Proposal, Week 8" pdfs, attached below. 

If you're still interested, or you have ideas for how the project could be used, or you'd like to work on it, I'd love to hear from you!

Ċ
Jennifer Hughes,
Feb 2, 2010, 1:33 PM
Ċ
Jennifer Hughes,
Feb 2, 2010, 1:34 PM
Ċ
Jennifer Hughes,
Feb 2, 2010, 1:34 PM
Ċ
Jennifer Hughes,
Feb 2, 2010, 1:34 PM
Comments