Article: Categorical Dimensions of Human Odor Descriptor Space Revealed by Non-Negative Matrix Factorization

    Article: Categorical Dimensions of Human Odor Descriptor Space Revealed by Non-Negative Matrix...

    post #1 of 11
    Thread Starter 

    http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0073289

    post #2 of 11

    OK, precis is three sentences.

    post #3 of 11
    I think this is about the article, in today's Science Daily:

    http://www.sciencedaily.com/releases/2013/09/130918180425.htm

    I looked at the journal article but found it way over my head. From the Science Daily article:

    "The team identified 10 basic odor qualities: fragrant, woody/resinous, fruity (non-citrus), chemical, minty/peppermint, sweet, popcorn, lemon and two kinds of sickening odors: pungent and decayed."
    post #4 of 11

    Too much math to work through in the first one.  not so important.  And the premise seems to end up being odor prediction, instead of detection... hmmm..

     

    Should be helpful in some manner to big Synth innovators for predicting what a chemical will smell like, isn't Avery Gilbert doing that too?  [Edited:  Oh, David, Is it Luca Turin?]

     

    I feel that deconstructing the library of odors into ten categories is gross simplification, and is still subjectivity.

     

    hmm.


    Edited by pkiler - 9/21/13 at 12:50pm
    post #5 of 11

    I am puzzled by 'Popcorn'. If this is discarded along with 'sickening', 'decayed' and possibly 'medicinal' we have a manageable 6 dimensions. Personally I like just the one dimension ie nasty to nice. But, hey we can all make our own pretty Perfume Wheels. The software is out there.

     

    One thing that does annoy me in research such as this is the over reliance on statistical packages. It is difficult to unpick the statistics to find the underlying causes and effects. Like do nice smells really correlate with molecular weight, volume and oxygen atoms as some research says? Yes, if you rely on the stats, but No if you look more closely. It's like relying on Climate Models for weather predictions- these models are too complicated and er, fixed. I am looking forward to some useful papers after the recent Odor Spaces symposium where Luca Turin talked- http://www.odorspaces2013.org

    post #6 of 11

    Luca Turin's theory of how we perceive smell would help in predicting the smell of a molecule  by it's IR spectrum.   Not sure how the above would help in predicting.   Description of odour, what a molecule smells like is so very subjective, and so dependent on so many other factors.   The descriptors were a joke in my opinion.

    post #7 of 11

    It all seems overly complicated.  Aesthetics and ergonomics in the visual and practical arts tend to follow the principles of the golden mean or phi ratio as do many aspects of nature; why not study olfactory aesthetics in this realm to determine if they follow the golden mean principles as well.  I'm just not exactly sure which values of a scent would fit into the formula.  Biomimicry is a rapidly growing field and it's mostly based on this principle.

    post #8 of 11
    Quote:
    Originally Posted by JEBeasley View Post
     

    It all seems overly complicated.  Aesthetics and ergonomics in the visual and practical arts tend to follow the principles of the golden mean or phi ratio as do many aspects of nature; why not study olfactory aesthetics in this realm to determine if they follow the golden mean principles as well.  I'm just not exactly sure which values of a scent would fit into the formula.  Biomimicry is a rapidly growing field and it's mostly based on this principle.

     

    Make me wonder how to appropriate Biomimmicry priciples to perfumery...?

     
    Long before "Biomimicry" was coined, I coined a phrase to use for myself, as an acronym,  SINASOD, standing for: Structure In Nature As a Strategy Of Design.

     

    But I guess that I'm not so much interested in making from scratch aromatic molecules that Mimic nature, but I usually start from Nature, and selectively remove portions to accomplish the goal I wish to set.  But then, I'm not a cosmetic chemist either, I'm a composer.

     

    I'm a little worried to see people try and shove scents through a sieve of their own composition, and then tell the world that "THIS" is how its all designed and meant to be and understood.  

     

    hmmm...

    post #9 of 11

    This is the full research article:

     

    http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0073289

     

    Besides the statistical wizardry, I think, is something that could be of value for us perfumers: an attempt to find a common vocabulary that could be agreed on for the description of smells. Something attempted before, but without much rigour or success. 

     

    To see what I mean, please have a look at this table from the article: 

     

     

    From this, you can deduce that "light" or "aromatic" are empty words when it comes to describing smells, as they are applied to completely different odours by different people. 

     

    As with any other human experience, language and perception are closely related. Cheers to these smell obsessed neuroscientist and psichologist researchers! 

    post #10 of 11
    post #11 of 11

    Mathematical ratios are great aren't they? Luca Turin with his famous musical analogies blogged about fifths "the two characteristic vibrations of esters and lactones, the molecular structures that are responsible for 90% of fruity smells, are placed almost exactly a fifth apart". I would love to beleive this ratio gives a perfect balance, but this one is unlikely if you assume each Odor Receptor is tuned to a certain frequency. But there are lots of other fascinating correspondences.

     

    I think we are hard-wired for particular vibrations (whether from sound, colour, smell or taste), and apart from a few anosmias and cultural & genetic variations we share the same perfume palettes. But interpreting these senses is learnt & cultural and that's why the above categorising dimensions of human descriptor space is interesting but doomed to fail.

    This article is great about Turin & ratios http://jdmoyer.com/2011/05/23/beauty-is-a-vibe/#more-3331

    class="

    9/19/13 at 1:57am

    gido said:



    http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0073289

    9/19/13 at 2:00am

    David Ruskin said:



    OK, precis is three sentences.

    9/19/13 at 2:39am

    edshepp said:



    I think this is about the article, in today's Science Daily:

    http://www.sciencedaily.com/releases/2013/09/130918180425.htm

    I looked at the journal article but found it way over my head. From the Science Daily article:

    "The team identified 10 basic odor qualities: fragrant, woody/resinous, fruity (non-citrus), chemical, minty/peppermint, sweet, popcorn, lemon and two kinds of sickening odors: pungent and decayed."

    9/19/13 at 8:19am

    pkiler said:



    Too much math to work through in the first one.  not so important.  And the premise seems to end up being odor prediction, instead of detection... hmmm..

     

    Should be helpful in some manner to big Synth innovators for predicting what a chemical will smell like, isn't Avery Gilbert doing that too?  [Edited:  Oh, David, Is it Luca Turin?]

     

    I feel that deconstructing the library of odors into ten categories is gross simplification, and is still subjectivity.

     

    hmm.


    Edited by pkiler - 9/21/13 at 12:50pm

    9/20/13 at 2:58am

    nemenator said:



    I am puzzled by 'Popcorn'. If this is discarded along with 'sickening', 'decayed' and possibly 'medicinal' we have a manageable 6 dimensions. Personally I like just the one dimension ie nasty to nice. But, hey we can all make our own pretty Perfume Wheels. The software is out there.

     

    One thing that does annoy me in research such as this is the over reliance on statistical packages. It is difficult to unpick the statistics to find the underlying causes and effects. Like do nice smells really correlate with molecular weight, volume and oxygen atoms as some research says? Yes, if you rely on the stats, but No if you look more closely. It's like relying on Climate Models for weather predictions- these models are too complicated and er, fixed. I am looking forward to some useful papers after the recent Odor Spaces symposium where Luca Turin talked- http://www.odorspaces2013.org

    9/20/13 at 8:36am

    David Ruskin said:



    Luca Turin's theory of how we perceive smell would help in predicting the smell of a molecule  by it's IR spectrum.   Not sure how the above would help in predicting.   Description of odour, what a molecule smells like is so very subjective, and so dependent on so many other factors.   The descriptors were a joke in my opinion.

    9/20/13 at 8:42am

    JEBeasley said:



    It all seems overly complicated.  Aesthetics and ergonomics in the visual and practical arts tend to follow the principles of the golden mean or phi ratio as do many aspects of nature; why not study olfactory aesthetics in this realm to determine if they follow the golden mean principles as well.  I'm just not exactly sure which values of a scent would fit into the formula.  Biomimicry is a rapidly growing field and it's mostly based on this principle.

    9/21/13 at 12:49pm

    pkiler said:



    Quote:
    Originally Posted by JEBeasley View Post
     

    It all seems overly complicated.  Aesthetics and ergonomics in the visual and practical arts tend to follow the principles of the golden mean or phi ratio as do many aspects of nature; why not study olfactory aesthetics in this realm to determine if they follow the golden mean principles as well.  I'm just not exactly sure which values of a scent would fit into the formula.  Biomimicry is a rapidly growing field and it's mostly based on this principle.

     

    Make me wonder how to appropriate Biomimmicry priciples to perfumery...?

     
    Long before "Biomimicry" was coined, I coined a phrase to use for myself, as an acronym,  SINASOD, standing for: Structure In Nature As a Strategy Of Design.

     

    But I guess that I'm not so much interested in making from scratch aromatic molecules that Mimic nature, but I usually start from Nature, and selectively remove portions to accomplish the goal I wish to set.  But then, I'm not a cosmetic chemist either, I'm a composer.

     

    I'm a little worried to see people try and shove scents through a sieve of their own composition, and then tell the world that "THIS" is how its all designed and meant to be and understood.  

     

    hmmm...

    9/21/13 at 12:59pm

    Javiero said:



    This is the full research article:

     

    http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0073289

     

    Besides the statistical wizardry, I think, is something that could be of value for us perfumers: an attempt to find a common vocabulary that could be agreed on for the description of smells. Something attempted before, but without much rigour or success. 

     

    To see what I mean, please have a look at this table from the article: 

     

     

    From this, you can deduce that "light" or "aromatic" are empty words when it comes to describing smells, as they are applied to completely different odours by different people. 

     

    As with any other human experience, language and perception are closely related. Cheers to these smell obsessed neuroscientist and psichologist researchers! 

    10/14/13 at 2:57am

    nemenator said:



    Mathematical ratios are great aren't they? Luca Turin with his famous musical analogies blogged about fifths "the two characteristic vibrations of esters and lactones, the molecular structures that are responsible for 90% of fruity smells, are placed almost exactly a fifth apart". I would love to beleive this ratio gives a perfect balance, but this one is unlikely if you assume each Odor Receptor is tuned to a certain frequency. But there are lots of other fascinating correspondences.

     

    I think we are hard-wired for particular vibrations (whether from sound, colour, smell or taste), and apart from a few anosmias and cultural & genetic variations we share the same perfume palettes. But interpreting these senses is learnt & cultural and that's why the above categorising dimensions of human descriptor space is interesting but doomed to fail.

    This article is great about Turin & ratios http://jdmoyer.com/2011/05/23/beauty-is-a-vibe/#more-3331