Hari's views on Raga Genome Project
Pravin Shankar
I have added Hari to the mailing list so you can reply to this list if you want to reply to him.
Cheers
- PS
---------- Forwarded message ----------
From: Hari Sundar <hsu...@gmail.com>
Date: Jul 18, 2006 12:10 PM
Subject: Re: Help needed: Raga Genome Project
To: Pravin Shankar <spr...@gmail.com>
Cc: Ramya Shankar <ramya....@gmail.com >, Suraj Kumar <sun...@gmail.com>
firstly, this is going to be fairly complex and will require input from experts in the field ... i.e., a wiki like approach is probably not going to work. I mean one must really be very knowledgeable before one can comment on a raaga at that level. How do you plan to encode such information ? i.e, at the server end ... obviously not text ... so it will entail us to develop some sort of a complicated notation scheme that can capture this info .. this is not going to be trivial ...
I see you mention the "sangeeta sampradaaya pradarshini". Subbarama Dikshitar attempts to do something like this, and does a decent job, but we all know what the current state of his work/notation is ...
A bigger issue with this is going to be in deciding what exactly ARE the right "genes" of the raaga ... I am not an expert but even on fundamental issues like the vaadii and samvaadii svaraas of a raaga I disagree on many occasions with textbooks, experts and I am sure with you and Ramya . In terms of svara patterns many raagas share the same 'genes', which has been referred to as a 'raagaa~Nga' by the indian musicologists. However indian music allows the performer a lot of freedom, which makes it difficult at this end to capture the whole flavour of a raaga, in a purely academic way. That is the reason why for the most part raagaas have been represented by aarohaNam / avarohaNam and in many cases a piddippu or pakaD, which summarizes the raaga.
In terms of what you guys want to do, I would ask you guys to look at the excellent set of articles written by Rajan Parrilkar (ignore his attitude) ...
http://www.sawf.org/music/articles.asp?pn=Music
I think he does a very good job using simple explanation along with audio samples to distill the essence of various raagas. I can't think of a simple way in which this kind of knowledge can be encoded ... that in my opinion will be the single hardest task.
Also, be aware that several raagas (esp. carnatic ones) are not very well defined ... i.e., other than the aarohanam and avarohanam ... most of the meLakarta raagas belong to this class ... and the raaga structure is quite diffuse and boild down to the permutations that the artist/composer can come up with.
So basically my questions for you are ..
a) How will you encode these 'genes' ? / How do users enter this info ?
b) How do you police these ?
c) Where do you knowledgeable people (willing) to enter this information ?
yeah, add me to the mailing list. and let me know if you think I can help in any way...
regards,
~Hari
Hi Hari
A friend of mine and I have just started off a project called the "Raga Genome Project". The project is still at its planning phase and we need expert advise / reviews on our design and things-to-consider. Ramya told me about you and asked me to contact you. The prime reason for writing this e-mail to you is to ask for help from you in reviewing our ideas, proposed design and add the carnatic-technical details that we lack.
To give you a very brief (ok, its not exactly brief ;) )outline of the idea, I am copy-pasting some text written by Suraj here:
A raga is not just the notes that make up its 'scale'. Its the
ghamakkams and the 'phrases' and the frequency of usage of these
swaras (vadi, vivadi, etc.,.) that make up a raga. The swaras make
a 'body part' of a raga.
The moment we start seeing a raga as a 'living body' with 'body
parts', one can start visualising 'genes' that build these body
parts.
Since a raga gene represents not just the 'notes' but also the
entire 'movement' involved in a phrase or even just a 'swara', it
represents a larger detailed that is missing from the current online
resources that merely use the notes in an arohanam and avarohanam to
represent a raga. It becomes possible to group related ragas more
accurately into a 'cloud'.
There are endless possibilities of applications that can be written
on top of the data that are collected thus. I can atleast think of:
* A tool to do 'bedhams' more accurately using all the details
available.
* A 'melody generator' that can adhere to the rules of a raga and
generate melody.
Now, how is this going to work practically?
Once the platform is in place, a 'contributor' (one who adds info
about a raga or even just a 'gene') can start defining a raga. The
first thing anybody would want to do is to define the arohanam and
avarohanam. But, according to us a raga itself is nothing but a
collection of 'genes'. Even the arohanam itself can be defined using
'raga genes' that define how to go from 'sa to ri', 'ri to ga' and
so on).
So the contributor will have to either find an existing gene that
does the movement for 'sa to ri' (as per the rules of the raga in
question). Worst case, s/he wouldn't find it when the system is
new. So s/he would have to define all the genes on her/his own.
Once seeded with a minimal set of 'raga genes' the ease of defining
ragas/raga-genes will grow exponentially. For example, once a
shankarabharanam's arohanam and avarohanam is defined, one will just
have to define 'ga-ma2' for kalyani and re-use most other genes from
shankarabharanam (of course, 'ni3' will differ, etc.,. -- but you
get the idea of how easier it becomes to re-use parts from other
ragas).
Now, thus far it may seem to you as though we are still defining
only arohanam / avarohanam. That is just the beginning. Contributors
can also start defining the phrases that are characteristic of each
raga ('ni-ri-ga' using various 'ni' and 'ri-ga' genes, etc.,.) thus
providing more details to a raga.
In the 'end', there would be a whole lot of 'phrases' in the system
for a given raga. Some of these phrases would be common across
ragas.
The 'users' of this system are two fold:
* Real people who want to just view info about a raga.
* Applications that will be able to use the data and do cool things
with it (like viewing 'related ragas', doing 'bedhams' and even the
'Melody generator' from our example above :) )
We have created a wiki for design discussions on the project. The wiki link is
http://swing.rutgers.edu/wiki/index.php/Raga_Genome_Project
We also created a mailing list, please let me know if it is ok to add you to the mailing list.
Cheers
- Pravin
---------- Forwarded message ----------
From: Ramya Shankar < ramya....@gmail.com>
Date: Jul 14, 2006 3:25 PM
Subject: Contact to add
To: Pravin Shankar < spr...@gmail.com>
PS,
A great person to add for this project is Hari, hsu...@gmail.com
He is an encyclopedia of music.
Refer my name and add him in. or fwd this to him..
Thanks
R
--
--
How much is an hour of your time worth?
Work An Hour for India
http://www.workanhour.com
Srihari
In addition to Hari's questions - all of which I think are fairly important for us to seek answers to at this stage - I have more one elementary question and one elementary suggestion, in that order.
Q. How do we ensure we convey the vision that we have to the people who hop on to the site? I speak from personal experience, because I must confess I dont think I fully understood the concept myself, inspite of PS's efforts to shine light on the grey areas I had.
Suggestion: This goes back to the point I mentioned to PS the other day. It is my opinion that this is by its very nature quite a daunting task, not to mention the problem of having too few sources of authentic knowledge in the group. Not to be discouraging even for a brief moment, I would like to suggest that we set ourselves very easy and achievable goals in the short-term, so we can benchmark ourselves going forward, lest the enthu fizzle out as it does in many cases when the goals are too hazy, too distant or plainly too overwhelming.
That was my two paisa's worth of feedback,
Srihari
Suraj Kumar
project critically and see where things can go wrong, etc.,.
Thanks hari for the inputs.
See my comments inline.
On 7/19/06, Srihari <srih...@gmail.com> wrote:
,----
| firstly, this is going to be fairly complex and will require input
| from experts in the field ... i.e., a wiki like approach is probably
| not going to work. I mean one must really be very knowledgeable
| before one can comment on a raaga at that level. How do you plan to
| encode such information ? i.e, at the server end ... obviously not
| text ... so it will entail us to develop some sort of a complicated
| notation scheme that can capture this info .. this is not going to
| be trivial ...
`----
Exactly. I had this 'fear' a few days ago and I was even mentioning
this to Pravin.
Unlike something like a wikipedia, where people have different
knowledge-bases on a variety of topics. Here what we expect is that
for the 'core' characteristic features, a good majority of the
people will be able to define like about half of the important
features of a raga. But say, one wants to get detailed information
on some sulini or somesuch raga, one will have to bank on the
experts. Experts are rare. Its not like inspite of being a novice
I'm an expert in some rare 'sulini' ragam alone.
Here is where the prime difference between Wikipedia and this effort
is evident -- I might be an expert in a very tiny aspect of
information such as "Fuji Finepix A202 camera" but that doesn't mean
I know about all other kinds of cameras. In this RGP effort, we are
just dealing with one domain -- Ragas). 2421 people will know about
Kalyani. Only 3 will know about a 'Sulini'.
At the risk of repeating over and over again, i strongly believe,
once we have the project known to people, the several Vidyas and
Deepas over the planet will enrich the data. Wouldn't there be
_atleast_ one Vidya or a Ramya willing to contribute information on
'Sulini'? I'm optimistic, there would be :)
At this point, I think it will help to explain the overall idea:
1. A gene defines a 'phrase' and that is it.
2. A raga has multiple genes each of which have its own _weightage_.
(given the other requirements such as permissible 'speed'
(ahiri), 'sthayi' (kuranji), we will have to accommodate those
aspects when assigning a gene to a raga).
3. Some genes are shared across multiple ragas. This is where the
entire idea takes shape.
A gene by itself is complex and we are in the process of collecting
feedback and then we will get onto the 'drawing board' to design a
usable interface / syntax / etc.,. I have something like this on my
mind for a gene definition User Interface ('UI', from this point
onwards):
http://sunson.in/gene-definition-ui-mock.jpg -- 2nd revision (thanks
to all the people who gave feedback on the previous version of the
UI)
The above image is just a mock up of the UI. Given that ghamakkams
aren't linear, the 'red curve' shown in the image above will be
'curvy' and we gotta figure out how we can make this easily possible
via a UI. In all, I think the trick of making this usable is
entirely in the UI.
Vidya pointed out this software called carnatic2000 (see related
work[1] on the wiki for the URL to this software's homepage). It
has a simple and comprehensive syntax. I don't know what
non-techies / non-geeks think of the syntax, though :) Need your (as
in everybody's) inputs here. :)
,----
| I see you mention the "sangeeta sampradaaya pradarshini". Subbarama
| Dikshitar attempts to do something like this, and does a decent job,
| but we all know what the current state of his work/notation is ...
`----
I think that is simply because SSP is not all that usable in its
current form. Secondly people don't tend to be clueful enough to go
through the introduction of SSP to understand the conventions, what
each ghamakkam is, etc.,
,----
| A bigger issue with this is going to be in deciding what exactly ARE
| the right "genes" of the raaga ... I am not an expert but even on
| fundamental issues like the vaadii and samvaadii svaraas of a raaga
| I disagree on many occasions with textbooks, experts and I am sure
| with you and Ramya . In terms of svara patterns many raagas share
| the same 'genes', which has been referred to as a 'raagaa~Nga' by
| the indian musicologists. However indian
`----
That is exactly the point. Since many ragas share common phrases,
one wouldn't have to keep defining these phrases for each raga. One
would just have to re-use an already defined 'gene'. Please see this
explanation[2] over our groups for how this would get done.
,----
| music allows the performer a lot of freedom, which makes it
| difficult at this end to capture the whole flavour of a raaga, in a
| purely academic way. That is the reason why for the most part
| raagaas have been represented by aarohaNam / avarohaNam and in many
| cases a piddippu or pakaD, which summarizes the raaga.
`----
Exactly. The 'pidi' phrases is what we want to enable.
,----
| In terms of what you guys want to do, I would ask you guys to look
| at the excellent set of articles written by Rajan Parrilkar (ignore
| his attitude) ...
|
| http://www.sawf.org/music/articles.asp?pn=Music
`----
Totally agree. I've liked his articles.
,----
| I think he does a very good job using simple explanation along with
| audio samples to distill the essence of various raagas. I can't
| think of a simple way in which this kind of knowledge can be encoded
| ... that in my opinion will be the single hardest task.
`----
oh, and we will also have a way (not the highest priority goal for
now) to attach sound samples per raga. However, my gut is that,
someone can write an application on top of our data to generate
sound samples (infact, if Carnatic2000 can become a RGP enabled
application, it can generate random samples on its own with very
little code change!) (of course, the entire idea depends on the
'who bells the cat problem' of having good enough data in our
database :) )
,----
| Also, be aware that several raagas (esp. carnatic ones) are not
| very well defined ... i.e., other than the aarohanam and avarohanam
| ... most of the meLakarta raagas belong to this class ... and the
| raaga structure is quite diffuse and boild down to the permutations
| that the artist/composer can come up with.
`----
What do you mean by 'not very well defined'?
,----
| a) How will you encode these 'genes' ? / How do users enter this
| info ?
`----
A UI (as roughly depicted by the image above).
,----
| b) How do you police these ?
`----
Very good question. The answer is in the answer for the question
'How do community-driven projects such as Wikipedia police the
information?' :) This is yet another Web 2.0 model, baby! :)
The Community is the 'signal source' and the Community is the
'negative-feedback'. (i must say, i scored a < 50% in Control
Systems theory, though ;) )
When we don't have a critical mass of 'experts' the system will tend
to have wrong info. But once we have a critical mass of people
contributing and using this info, the community will fix the info.
Simply put: Data need not start out correct. It will improve over
time.
,----
| c) Where do you knowledgeable people (willing) to enter this
| information ?
`----
Parse error ;) Can you please rephrase that? :)
,----
| yeah, add me to the mailing list. and let me know if you think I
| can help in any way...
`----
Thanks a lot for volunteering. :)
cheers,
-Suraj
PS: We should put up a really clear doc on the project page, for
starters. Your feedback on what is clear, what isn't clear would
help greatly in doing this. Do you think we should stop looking at a
phrase as a 'gene'? Maybe the problem is in the fact that this
sounds too 'science', 'biology', etc.,. and therefore 'boring'? ;)
[1] http://swing.rutgers.edu/wiki/index.php/Raga_Genome_Project:Related_Work
[2] http://groups.google.com/group/raga-genome/browse_frm/thread/eacdff0a27a57279/#
--
Non-Spiritual way of life: http://actualfreedom.com.au/
Deepa Mohan
Do you think we should stop looking at a
phrase as a 'gene'? Maybe the problem is in the fact that this
sounds too 'science', 'biology', etc.,. and therefore 'boring'? ;)
Hari Sundar
Where do you FIND knowledgeable people (willing) to enter this
information ?
commenting further on this ... and some related issues/questions
raised by others ...
I think this is certainly not a web 2.0 project ,,, and in essence
does not lend well to a collaborative effort ... at least not in the
wiki sense. For one although many people on the list are quite capable
of identifying kalyANi and also singing/playing it ... and have a deep
understanding of the raaga ... how many truly have and/or can break it
down to a set of rules ? I for one will find it hard and painstaking
work. Some of you can try doing this ... what I mean is that the
ideal model for something like this would be a set of very qualified
musicologists entering most of the info ... and potentially a bunch of
actual artists adding in their individual thoughts/interpretations ...
Now at this stage even if we manage to to encode all this information
for a given raaga, is it useful to someone like Srihari ? my point
being that if you are not very well trained already would you be able
to understand the nuances of the raaga by reading notation ? now at
this stage even if we linked it up with some sort of synth program to
convert these into audio samples ... isn't it better to simply have a
"collaborative" effort in collecting excellent demonstrative audio
samples of the raaga in question ? I think it might be better for one
to upload a short piece which we feel is an excellent example .. say
kalpanaa svaraas by one's fav artist and analyze it ... saying notice
this and that ... maybe something like a tagged (subtitled ?) audio
file ... I do think that most of your streaming technologies support
this ...
I would like to give an analogy at this point to further stress my
point ... say you were doing something similar for art ... and say the
equiv of a raaga can be though of a style like say impressionism ...
now photoshop can easily convert a given picture into an impressionist
version based on certain rules/filters ... If you want to explain
impressionism to a novice or even an art student ... are you going to
use this photoshop or refer to paintings by Monet, or Cézanne ?
But we can have a small experiment ... by simply picking one simple
raaga that most of us know/understand ... and see how well we are
able to 'encode' it ... it would be a good learning experience in
understanding potential difficulties ....
~Hari
Deepa Mohan
Pravin Shankar
1) Experts-only Peer-reviewed V/s Collaborative Web 2.0-based
Both approaches have their pluses and minuses. The former is obviously higher in quality, nearly encyclopedia-like, but suffers from the problem of low number of contributors. Nupedia ( http://en.wikipedia.org/wiki/Nupedia ) was such an effort started before Wikipedia, but it ended up becoming a failure and in 2003, the site went offline.
That said, there are some fundamental differences between wikipedia and RGP, as pointed out by Hari and Suraj. To summarize: Wikipedia is breadth-based (cursory information on a variety of topics) whereas RGP is depth-based (deep information in a concentrated topic).
So it may not in the best interests of RGP to follow the exact same model as Wikipedia. The right model IMHO is somewhere between Wikipedia and Nupedia (while learning from the mistakes of Nupedia). Finding this exact model is quite an interesting problem by itself. The problem in essense boils down to the following questions:
1) How does one become a contributor? (Anyone v/s someone with certain credentials. (Verifying credentials online however is a hard problem, and one of the reasons why Nupedia failed))
2) Does the information update itself instantaneously, or after a round of peer-review?
3) What if there are discrepancies between different views? (Arbitration policy)
2) Where do we find experts to contribute?
I believe that there are *lots* of experts who, given such a platform, will be extremely eager to contribute. Some examples to substantiate:
- Prof Shivkumar (RPI) - His website is a mammoth repository of kritis, and the amount of effort he must have put in for all those recordings of classes, notations, etc. must be phenomenal. I'm sure countless people (including myself) have learnt from his recorded music classes.
- The people behind Sangeethapriya.org - These people take the pains to record from AIR, old tapes (even gramophones) and share vintage recordings among rasikas.
- Promiserani ( karnatik.com) - Suraj can vouch for the fact that she has put in painstaking effort into the website.
- M.Subramanian (author of Carnatic2000) - I, for one, was extremely inspired by the amount of enthusiasm he showed in creating this amazing piece of software, along with a notation system for gamakams, and lots of interesting articles on computerization of carnatic music.
My point being that there are countless experts with no shortage of enthusiasm, who, given a proper platform to contribute, can do a great job at creating this infobase of raga genes.
3) How will this information be useful to an AverageJoe carnatic enthusiast?
Its true that these raga genes will not be too interesting to an AverageJoe. And thats why its a nice idea to encode this information into a standard format, as opposed to keeping it in the form of articles and annotated audio clips, so that it becomes possible to write computer programs that can do *cool* things with this information. Eg.
- Classification/Categorization to form clouds of related ragams
- A tool to do 'bedhams'
- Automated melody/alapanai generator
- Comparative study of different singers, by frequency analysis of their recordings, with each other, and with the gene database.
My point being, if we can design a neat interface for developers/programmers, then the possibilities for more applications is endless.
4) Complete design before implementation v/s prototyping?
Like Srihari pointed out, most projects that endeavour to achieve goals that are too lofty, fizzle away even before they get to see daylight. I think all of you in this group agree, just like I do, that the best way to go about this project is to start with a simple prototype, and in the process understand the potential issues in a much better way.
I think we already have enough experts in this group (Vidya, Deepa, Ramya, Hari, Saketh) to bootstrap the genes for a chosen starting-point raga, and take it from there.
Meanwhile do continue to pour in your thoughts.
- PS
Suraj Kumar
I just have the following extra points to make (or repeat Pravin's
points, just to emphasize):
It should be made clear that, for someone who wants to just view the
arohanam and avarohanam of a raga, this platform WILL deliver the
needed info. How we present the info (something graphical? some really
simple yet precise notational system?) is where we gotta decide.
Roughly put, think of a site where you can go to 'Kalyani' and click
on a link that says 'arohanam/avarohanam' and you see s r2 g3 m2. That
will be the exact same info what existing resources provide. The only
difference being, our system provides more deeper info if the user
wishes to view deeper info.
Infact, we can just start off by making the system define arohanam and
avarohanam alone! The prime difference is our platform will provide
people a way to also add more info. Over time, we will have rich data.
To answer hari's question on the Analogy of impressionism...
You are talking about a totally different aspect here -- the
"speciality", "individuality", etc.,. of an artist. Yes, Emani Sankara
Sastri's Kalyani might be different from Balachander. But the point
being, Kalyani by itself, has a form. Students learn it. Lots of
people who are trained formally know such information already. It can
be represented scientifically and used for various purposes.
Therefore, a raga is _Data_. Not art. What you make with the data
might be art. Our goal is to provide a platform to represent this
data. Not to analyze art. That will be a different (very interesting!)
project :) Does it now sound clear, Hari? :)
Agreed, mostly only the students will use the data, not the rasika.
But, IMHO, its not in the scope of this project to convert a rasika
into a student. But I'm guessing wannabe-students can get started off
easily using the data we accumulate.
Like Pravin said, I'm sure lots of people are out there (including
Pravin and myself :) ) who can help and get-helped using this tool.
cheers,
-Suraj
--
Freedom from the Human Condition: http://actualfreedom.com.au/