Raga Motif MOS Test

[hema]

Dear All
We now have a computational approach to spot typical motifs of a given raga (in alaapana, compositions etc). The typical motifs were obtained from a professional musician. What we would like the people of this forum to do is the following:

i) The website is:
http://www.iitm.ac.in/donlab/music/
(the trailing slash is important).

ii) Five different ragas (each with 2-3 motifs are listed).

iii) You click on a specific motif -- the system will give you a list of matches from the database of compositions, alaapanas of the given raga --
we have a huge database of concerts from which these items have been picked based on the meta information available.

iv) For every match, the motif is resynthesised in the target's tonic (sruti -- which is automatically determined from the item).

v) You have to compare the motif with the match from the database.

vi) If the motif is part of the phrase that is identified -- you give a score of 10.

vii) If the motif is NOT part of a phrase that is identified -- you give a score of 0.
viii) For others you prorate the scores between 0-10.

I request all of you to proactively participate in this effort -- this will help us do better in terms of the computational aspects of the task.

Thanks for your time

Regards, Hema

[Rsachi]

I would have participated but I am not clear why you want me to register with email ID and phone no.

[varsha]

R while we wait for an answer to your question , here is a cartoonsspeak for the occasion

Middleaged woman reading magazine , telling her husband
Honey I just now took the compatibilty test given here , and you failed :-*

[vasanthakokilam]

Good work. It is quite an interesting and fun exercise to undertake. Filled it out.

I was also quite hesitant initially about the phone number thing. I gave it a fake one and if took it

Hema, the lowest score it takes is 1. So I assume you meant that instead of 0.

[Rsachi]

Hema,
What Varsha has spoken of is a clear risk you run.
What if I am so poor in identifying raga phrases and land up answering all questions with the lowest match score? It will be a sore data point.
====
Continuing in the same vein as Varsha:

Optometrist asks Mulla Nasruddin to read the eye test chart with random letters like AXR/RNMPO/XSKRCOD/PS.. And Mulla fails to read even the largest letters. The Optometrist tears his hair and says, 'Mulla, you're the first person who can't read even the biggest letters!'

Mulla says, 'excuse me, sir, but who told you I can read?'

======

Hema, don't lose heart, I am also from IITM (passed out before you were born perhaps) and wish you well.
I know that you would derisk the study in such matters

[hema]

Hema,
What Varsha has spoken of is a clear risk you run.
What if I am so poor in identifying raga phrases and land up answering all questions with the lowest match score? It will be a sore data point.
====
Continuing in the same vein as Varsha:

Optometrist asks Mulla Nasruddin to read the eye test chart with random letters like AXR/RNMPO/XSKRCOD/PS.. And Mulla fails to read even the largest letters. The Optometrist tears his hair and says, 'Mulla, you're the first person who can't read even the biggest letters!'

Mulla says, 'excuse me, sir, but who told you I can read?'

======

Hema, don't lose heart, I am also from IITM (passed out before you were born perhaps) and wish you well.
I know that you would derisk the study in such matters

I too am pretty OLD!! I too was born before the IITs were started!:-)

Okay, I will ask my students to remove the phone number and also the request for DOB -- we will ask for ranges of ages though -- this is useful for a different study. As in most tests, we do use e-mailIDs -- I will ask them to change this too. We will have to have a security question so that you can reset your password. We will not send any mails to anybody. About the MOS -- the reason for posting it on this forum: Most people are rasikas and will understand. We tried this test with some students at IITM. Even students who are learning music and sing pretty good -- can not identify ragas -- some even perform hour long concerts!:-((

We are looking at averages -- if the average score is poor -- then perhaps the motif that we are using is not the killer motif.
By tomorrow, we will remove the request for the phone number.

Thanks, again for your support.

Rgds, Hema

[hema]

G
Hema, the lowest score it takes is 1. So I assume you meant that instead of 0.

Yes, the lowest score is 1. We refer to this as poor.

Rgds, Hema

[Rsachi]

Ok, all clear. Let's participate and see where we stand
You will find elsewhere in rasikas.org a link to a test given by Prince Rama Varma to identify 70+ ragas with short vocal clips. I enjoyed doing that!

[varsha]

I just took the bhairavi and had an uncontrollable urge to listen to Her Highness's essay in Viriboni And Enati nomu phalamo.
Few decades ago when these gems were aired on radio there would be a hush descending IN MY HOME during the passage of all the motifs in sundaresa suguna brinda dasaratha nandana aravinda nayana pavana .
Cannot handle any more ragas for today . Shall gaze into the sky trying to spot a motif-star called MS

[vasanthakokilam]

Hema, Do we call get different matches to rate on, or that is fixed?

Even students who are learning music and sing pretty good -- can not identify ragas -- some even perform hour long concerts!:-((

I have experienced the same as well. Your chances of finding good test takers among music students are higher if they are known to voluntarily attend concerts

I would think that some of those who can not identify ragas may still be able to pass a judgement on if the motif and the match have anything to do with each other. What is your experience with such folks?
I will ask a few people who do not know anything about ragas but are good musically to try the test and not put in the answers so as to not mess up your tests.

As we all know, such motifs are what raga is all about ( and not the usual aro/ava definition ), so this may be useful for training/initiating rasikas too, not just computers.

We are looking at averages -- if the average score is poor -- then perhaps the motif that we are using is not the killer motif.

I think the motifs are OK. The audio quality is a bit rough in a few cases with all the kattai adjustments etc.

I think your algorithms are doing a pretty good job.

I am curious if you are working on/going to work on algorithms to generate motifs from songs, alapanas and niravals of the same raga. We can then rate them too.

[Ranganayaki]

...

[hema]

Dear All
Sorry for this long hiatus. Two things were the problem:

i) Getting the website fixed -- which got done about a week ago -- students were busy with exams, deadlines etc.
ii) Project reviews, endsemester exams, grading etc.

Phew! I am finally done. I hope all of you will participate in this test.

Next, we are also doing the following:
As a rasika, when an Aalapana starts and I do not know the identity of raga -- I try to figure out what song it reminds me of
and which part of the song it sounds like. I believe that these are lines that correspond to the
pallavi (first line), anupallavi (first line) or in general the line where the niraval is sung. Of!Course different people use different lines for niraval -- e.g Meenakshi (Purvikalyani) -- meena lochani, madhura puri, vidhuvadambana ..

Some examples that come to my mind:

For a long time I used to identify Huseyni with KVN's "Eppadi manam." -- in fact using only the first phrase -- I do not know the rest of the song.

We refer to these as one-liners.

The raga test that you took was based on motifs given by a musician -- we actually asked the musician to sing for us -- we did take hints from Rama Verma's Raga test.

We feel that perhaps it is in the one-liners that the typical motifs of the raga exists. (We were told by professionals that a raga's phrases can be found in the compositions). The objective is that --- if we find repeating patterns across one-liners across compositions -- then perhaps these are the typical motifs. Our preliminary results as verified by a professional seem promising.

The final objective is to be able to index Carnatic music -- for example, mark the typical motifs in an Alapana. Then as a listener --
one can focus on how a musician meanders from one typical motif to another. Some musicians might have preferences for certain typical motifs.

We also have other objectives -- segmenting a composition into pallavi, anupallavi, caranam using the pallavi as a motif. This will be useful from a different perspective -- how many sangatis has a musician sung? where is the niraval? how many cycles of swaram?, .. etc.

I have been to the Music Academy Digital library -- it took me quite sometime to find a particular of "hariyum haranum" of MDR. The proposed research is primarily oriented towards enhancing user experience.