Theme 4: Quantitative Research Pre-reflection

Article: Salamon, Justin, Bruno Rocha, and Emilia Gómez. "Musical genre classification using melody features extracted from polyphonic music signals."Acoustics, Speech and Signal Processing (ICASSP), 2012 IEEE International Conference on. IEEE, 2012.

Which quantitative method or methods are used in the paper? Which are the benefits and limitations of using these methods?

In this article statistical methods are used in order to evaluate how good their system are at classifying genres using melody/pitch contour. The algorithm was tested on 500 30-second excerpts to test genre classification. The songs were evenly distributed among five different genres: opera, pop, flamenco, instrumental jazz and vocal jazz. Several melodic features were extracted and chosen as classifiers. Mel-Frequency Cepstrum Coefficients (MFCC), a popular set of features for speech, were chosen as a comparison. Not only did the authors test the classification accuracy but also they examined which features worked best. A combined MFCC/Melodic classifier was also tested. The combined method had the highest accuracy, followed by the Melodic. Finally, the authors tested their system on a dataset containing 1000 excerpts of ten different genres. The accuracy were much lower. This shows that their method worked good on the dataset that they had chosen for their purposes. But their algorithm can not be said to be very robust as it didn't work as good on a broader dataset. 

What did you learn about quantitative methods from reading the paper?
That they don't necessarily have to be based on questionnaires but that the important thing is that you work with a big number of data points. The more, the better. Also, the dataset's homo/heterogeneity highly affects the results

Which are the main methodological problems of the study? How could the use of the quantitative method or methods have been improved?

As we already saw, the choice of dataset highly affected the study's results. A more heterogeneous dataset showed that the algorithm was not as robust as desired. The study pointed out this problem itself and one may hope that they take this problem into account for futures studies. On the same point, one may also argue about the definition of genres. Often, genres overlap and are also subject to human subjectivity. It is hard to know how the genres of the dataset were initially classified.

On Bälter et al.

Obviously this is not an article dealing with media technology but with epidemiology. I am unaccustomed to reading articles in this field and my initial reaction was that there was a very big focus on the design of the study. It is very thorough and it seems that it could be used as a textbook example on how a study could be designed. Perhaps this kind of methodological write-up is skipped in other fields such as media technology because it is just taken for granted. Perhaps it is important in epidemiology but not in other fields. In this course and in the bachelor thesis course it was said that all papers should contain a part on theory and method but I've found that most articles (at least in sound and music computing which is my specialisation) lack this. One may argue that it is written between the lines or that the theory and methods are so well-established in the filed that it is deemed unnecessary to include them.

I have a hard time finding that qualitative methods are very useful in fields such as epidemiology, medicine and toxicology except from studies regarding patients' experience of treatment etc. However, the fact that the participants self-report the occurrence of URTI could perhaps be seen as a qualitative element. It is possible that quantitative methods reduce people to data points, but qualitative methods may miss out the big picture. I think that these are complements to each other and that a good study (or rather field or even paradigm) tries to combine both qualitative methods.

Theme 3: Research and Theory Post-reflection

For this week I tried to find an article which I could relate to another project I am working on. But it proved hard to find articles published in a journal with an impact factor above 1. Most articles on sound and music computing seem to be published in conference proceedings which have a lower impact factor. So I asked myself whether this really was a good measurement.

So first I checked the wikipedia article to see how it was calculated. It is basically the number of times articles of the last two years were cited divided by the number of citable article this year. Sounds pretty straightforward to me. But from my view this is not necessarily a measurement of how "good" the articles were. For example we may consider the Schön scandal. Jan Hendrik Schön published a number of articles in prestigious journals such as Science and Nature that later proved to be fraudulent. The high impact factor didn't protect the reader against "bad" science.

Furthermore, as far as I have understood a journal in a multi-disciplinary field will get a lower impact factor since articles refer to a lot of different journals. Also, the impact factor is not comparable between disciplines as the citation rate differs.

There are several ways to affect a journal's impact factor. For example publishing review articles often increases it. Another more extreme example was when Folia Phoniatrica et Logopædica editors Schuttea and Svec (2007) published an article with heavy critique against the impact factor including citations of all articles published in the same journal for the last two years. This increased the impact factor of that journal from 0.66 to 1.44.

So, what can we conclude from this short critique of the impact factor? It is, obviously, a good measure of impact as long as journal editors don'r rig the conditions but it is not a guarantee that the contents are scientific. We should still be critical readers.

Note:
Jan Hendrik Schön is not a relative of mine... ;)

References:
Schuttea, H. K., Svec, J. G. (2007). "Reaction of Folia Phoniatrica et Logopaedica on the Current Trend of Impact Factor Measures". Folia Phoniatrica et Logopaedica 59 (6): 281–285. doi:10.1159/000108334.PMID 17965570

http://en.wikipedia.org/wiki/Impact_factor

http://en.wikipedia.org/wiki/Sch%C3%B6n_scandal

Theme 3: Research and Theory Pre-reflection

Journal

The Journal of the Acoustical Society of America.
Impact Factor: 1.65

The Journal of the Acoustical Society of America, JASA, deals with sound and acoustics in a broad and interdisciplinary sense. That is, not only structural and architectural acoustics but also musical acoustics, psychology and physiology of hearing as well as speech communication inter alia. JASA has been published since 1929.

Article

Li, Y., & Wang, D. (2007). Separation of singing voice from music accompaniment for monaural recordings. J. Acoust. Soc. Am. 122, 2989 (2007)

In this article Li and Wang proposes a singing voice separation system. It is motivated by its application in automatic lyric recognition and singer identification systems. Also it may improve the understanding of the human hearing system. They focus on monaural recordings (mixtures) as it is a more general case than binaural recordings. Furthermore, Li and Wang shows why a separate system is needed for singing vocals as opposed to speech. Singing often adds an additional formant, has a wider pitch range and is also piece-wise constant. Also, singing is in general accompanied by a harmonic, broad-band interference which is correlated with the singing signal.

The methods and algorithms that Li and Wang use are motivated by relevant previous research by themselves and others. The system consists of three sub-systems. First a singing voice detection classifies portions of the mixture as vocal or non-vocal. Second, the predominant pitch is extracted which contours are used in the final step in a pitch-based separation system. Conclusively, Li and Wang argue that their approach gives good separation of singing voice from the mixture. They point out that improvements on the system may be made primarily in the first part, the singing detection, by using different features of the mixture.

I find Li and Wang's text comprehensible and instructive. It is a significant step forward in a relatively small field of research. Subsequent research refine these methods further.

Gregor & Sutton

1. Briefly explain to a first year student what theory is, and what theory is not.

In my view, a theory can be seen as a general framework. It provides the researcher with causal relationships, systematic reasons of processes and logical explanations. References, data, research subjects, hypotheses etc. can be a part in the formation of a theory but are in themselves not theories.

2. Describe the major theory or theories that are used in your selected paper. Which theory type (see Table 2 in Gregor) can the theory or theories be characterized as?


The main theory of Li and Wang's article is based on the analysis of the nature of voice and their use of relevant methods. With this in mind they propose their system as a general framework. They provide the blueprint for constructing the system and show that it can do a good separation of vocals from a monaural mixture signal.


So, I'd say that this has a little bit of everything of Gregor's taxonomy. My first thought was that it would be theory V as it mainly describes how to design a good vocal separation system. But as it is thoroughly based on analysis and examination and also does a prediction for the outcome of the system it also encompasses the other theories.

3. Which are the benefits and limitations of using the selected theory or theories?


I think Li and Wang has a strong theoretic base. It provides their research with relevant data and methods of designing their framework. However, there seems to be a lack of good standardised evaluation methods within this field (i.e. Sound and Music Computing). A probable cause may be because it is multi-disciplinary, with researchers of very different backgrounds.

Theme 2: Critical Media Studies Post-reflection

During the lecture this week Leif talked about the "myth of objective truth". I found this idea the most interesting in this theme. As I wrote in the pre-reflection I think that it is possible to argue that science is mythology. The difference of the scientific mythology and mythologic mythology (such as any religious story of creation) has to be the purpose it has for (the scientific) society. We may all agree that Big Bang is the creation myth of today, at least in Swedish (atheistic) society. At the same time we regard it as a scientific fact. It is a truth. But the thing is that there are a multitude of truths that coexist. The biggest myth is the one of objective truth. There are truths but they are all subjective.

So when we promise to say the truth, the whole truth and nothing but the truth we still leave out which truths we are supposed to say. During the seminar someone said that news media select what truths they want to convey, based on some criteria based on how many readers they'll get or a political goal.

So when we go to next week's theme and start reading scientific articles we need to keep in mind what the truths are given and which are left out. Adorno and Horkheimer does this by only bringing up the arguments that prove their point but leave out others. It doesn't have to be purposefully though. Some things are just overlooked. Other things might be true in a specific time and context only to become obsolete later on or in another context.

Theme 2: Critical Media Studies Pre-reflection

What is Enlightenment?

According to Adorno and Horkheimer it is a totalitarian myth-dispelling myth (which therefore also dispels itself). I can't really see why they have such a hard time with it. For me I see the enlightenment as the direct continuation of the renaissance where europe awakened from the dark ages and reintroduced Greek and Arabic philosophy, science and religious practice. For almost our entire history science has been close to religion. Radical scientists have also been radical philosophers and involved in occult groups. Occult renaissance eventually led up to the scientific revolution which greatly influenced the enlightenment. I believe the enlightenment marks the start of a divide between science and religion. It seems however that these two still not are fully apart. It took until the 19th century for science to dispel the myth of man as the image of the christian god. Darwin showed that humans are not created by some magical process but by evolution. I think that this magical view on ourselves is still apparent. So in my opinion, the enlightenment is still an ongoing process.

What is the meaning and function of “myth” in Adorno and Horkheimer’s argument?

Adorno and Horkheimer states that myth has the function to report, name and tell of origins and also to narrate, record and explain. Mythologies are descriptions of the world and history. With this kind of definition one may argue that science also is a kind of mythology. And as Adorno and Horkheimer also describes religion as the absorption of mythology into civilisation it leads to the question whether science may be absorbed in the same manner. Enlightenment introduced the dualism between science and religion but this doesn't seem to hold. Atheists often describe science as their religion. I see no contradiction in this. Mythology is not inherently good or bad, it is just our description of the world. This mythology might however be progressive or regressive.  

What are the “old” and “new” media that are discussed in the Dialectic of Enlightenment?

Although never referred to as "old" or "new" we can see a couple of different media throughout the text. Radio is described as a (new) more sublime version of the printing press. The TV, a relatively new medium at the time Adorno and Horkheimer wrote their text, is described as a symbiosis of radio and film. The new broadcast media of radio and TV is the prerequisite for modern mass media, and thus also the culture industry.

What is meant by “culture industry”?

From the authors'  point of view it seems as the culture industry is a massive monopoly that acts to conform the population. The communication is one-way – there's no way of responding to the broadcast. The culture industry reduces the population to consumers and employees of it. It is a pretty dark view that is presented here. I believe that this is due to the fact that this text is published in 1944 when Germany has just demonstrated the power of its propaganda machine with radio and TV as its main media. 

What is the relationship between mass media and “mass deception”, according to Adorno and Horkheimer?

As described above, I believe that they see mass media as an instrument to coerce the population. But it becomes deception as the media consumers are led to believe that it is what they want. There is no need to use force to shape the views of the population. The population is led to believe that the culture industry can give them a flight from everyday life, all you need to do is consume mass media/mass culture. 

Please identify one or two concepts/terms that you find particularly interesting. Motivate your choice.

I think it is interesting the way Adorno and Horkheimer describe the flight from everyday life, as promised by the culture industry. It seems there is no way to escape mass culture without consuming mass culture. I immediatly thought of the counterculture movements of Europe during the first world war, namely the dadaist movement. They created something completely different than the mainstream fascist culture. By creating what they called anti-art and using nonsense they created a separate culture. I always found the dadaists very inspiring, not only as a cultural but also a political movement. Perhaps something like this is what Adorno and Horkheimer would need in order to lose their bitterness (that and the defeat of Germany in the 2nd world war).

Theme 1: Theory of Science Post-reflection

When I saw the title of this week's theme "Theory of Science" I immediately thought of a book I read when I was about to start at KTH. At the time, I thought that becoming a university student I should have some knowledge about what science is, and how to separate good science from bad. A friend of mine who at the time was a student of philosophy at SU recommended me to read the book What is this Thing Called Science? by A.F. Chalmers. It proved to be a very instructive and comprehensive book which mapped out the most important theories of science chronologically. In this blog post I will try to remember what I learned.

Induction

Induction is the process of drawing a conclusion from a number of systematical observations. The difference between deduction and induction is that induction doesn't have to be logically valid. It is however more useful as the premises of deductive reasoning are impossible to come up with without observations. For an inductive argument to be "good", Chalmers states that it needs a lot of observations, that the observations are done during a big time span with a lot of different circumstances and that none of the observations may contradict the conclusion. Chalmers then point out the shortcomings of these criteria, namely that they are too hard to fulfil.

Falsifiability

Karl Popper presented the notion of falsifiability as an alternative to induction in the 1920's. It continues where induction left off with a lot of observations but instead focuses on that theories should be falsifiable. An example of a simple falsifiable proposition is "It never rains on Wednesdays". Such a proposition has the benefit of being both logically valid and practically observable (unlike induction and deduction, respectively). To advance science we should propose new, relevant, bold and falsifiable theories and whenever they are falsified replace them with better ones.  There are some problems however. One of them is that an observation may falsify a theory when it is in fact the initial conditions that are wrong. An historical example of this is how observed perturbations of Uranus orbit seemingly falsified Newton's law of universal gravity. The perturbations were in fact caused by the then undiscovered planet Neptune (which was actually predicted using these perturbations!) .

Paradigm

Thomas Kuhn argues that scientific revolutions have followed structures that doesn't reflect on the theories of induction and falsification. These structures are that of this schema: pre-science–normal science–crisis–revolution–new normal science–new crisis etc. According to Kuhn a new normal science is achieved when the scientific society join its paradigm. The paradigm is more universal, encompasses several theories and provides the scientific society with both problems and solutions. A paradigm goes into a state of crisis when its anomalies begins to form big problems and is followed by a period of "pronounced professional uncertainty". The seriousness of the crisis is increased when a competing paradigm emerges. A historical paradigm is the one based around Maxwell's electromagnetic theories. This was eventually replaced by the Einsteinian paradigm which most importantly rejects the idea of æther, a medium through which electromagnetic waves was thought to propagate.

Research Programme

How do we know that one paradigm is better than the other? Imre Lakatos' research programme is very similar to Kuhn's paradigm, but combines it with falsifiability. A problem for science is that researchers within a paradigm, according to Lakatos, create help-hypotheses that protect the central theories from being falsified. To deal with this problem Lakatos differentiates between progressive and degenerative research programmes. A progressive research programme should produce novel, revolutionary results without a too big increase in help-hypotheses. An important difference between the two first and the two last notions I have presented is that the former focus on observations and facts whereas the latter focuses on theories.

Final Thoughts

Being somewhat a pragmatist I find the theory of science/methodology most satisfying that can provide healthy conditions for a progressive scientific society that produces high quality results. I think that Lakatos' research programme gets close to this, and it makes me envision science as a big research project that encompasses a lot of people doing smaller or larger contributions.


References: 

• Chalmers, A.F. (2003). Vad är vetenskap egentligen?. (3., rev. ed.) Nora: Nya Doxa.

Theme 1: Theory of Science Pre-reflection

–What does Russell mean by "sense data" and why does he introduce this notion?

Although we might all agree that things exist in reality we most often have a very different experience of them at different times. An object will look, feel and sound differently from different perspectives and at different times. To differentiate between the appearance of an object and the reality or "truth" of an object, Russell introduces sense-data and physical object where the former is the appearance of the latter. In other words sense-data is the immediate knowledge of the object through sensation, our experience of it. The way I understand Russell is that we may infer the physical object from our immediate knowledge as it is created by the physical object. The sense-data is not the truth about the physical object but only the truth about the sense-data itself. Therefore, our sensation, according to Russell, is only of the physical object's appearance and not the object itself. Also, Russell also states that sense-data is personal. Our sensations and experiences cannot be the same as we cannot have the exact same perspective at the exact same time. Our different sensations will however be similar. Russell also states that if only sense-data existed and not the physical objects this would mean that they would disappear as soon as we aren't sensing them and then reappear again when we do sense them. 

–What is the meaning of the terms "proposition" and "statement of fact"? How does propositions and statement of facts differ from other kinds of verbal expressions?

A proposition is a statement about an object. A proposition could be on the form "A is 'so-and-so'" and must be composed of "constituents with which we are acquainted". So if we are acquainted with the man in the iron mask and Alexander Dumas' explanation of who he really is we might put forward the proposition that "Louis XIV's twin brother is the man in the iron mask". We do however not know whether it is true or not. But if it was true, all implications made by this proposition would also be true. A proposition exemplified by Russell is that "two and two are four". Is this a statement of fact? Surely this is true a priori, since the definition of our mathematical axioms require two and two to be four. This makes it a fact inevitably. However these axioms are arbitrary. Two and two might not be four with different axioms. A fact might be a fact in some cases and others not, it all depends on the context. 

–In chapter 5 ("Knowledge by Acquaintance and Knowledge by Description") Russell introduces the notion "definite description". What does this notion mean? 

Russell differentiates between two types of descriptions: ambiguous and definite. An ambiguous description is on the form "the 'so and so'" e.g. "a man in an iron mask". With a definite description Russell means a single object on the form "the 'so-and-so' in the singular". This means that there can only be one instance of this object and that it has a specific property e.g. "the man with the iron mask". Something that fits into this description is known by description as opposed to known by acquaintance which is known to us by its sense-data or the memory or introspection of the sense-data. Also, the description is to be unique: there is only one man with an iron mask and it is the man with the iron mask. If there were more men with iron masks it would be an ambiguous description. We only know the man in the iron mask by his description and not by acquaintance. Seeing the man in the iron mask as a physical object I think that Russell wants to find a good way to describe objects that we don't have sense-data for. 

–In chapter 13 ("Knowledge, Error and Probable Opinion") and in chapter 14 ("The Limits of Philosophical Knowledge") Russell attacks traditional problems in theory of knowledge (epistemology). What are the main points in Russell's presentation?

Russell dismisses that a true belief could be considered knowledge. I find it pretty straightforward that believing in something which happens to be true doesn't constitute knowledge. A belief may be deducted from past acquaintance and descriptive knowledge or reached without inference, intuitive knowledge. Any of these derivative beliefs will however still be just beliefs. The intuitively based derivatives are just as intuitive as their basis and in my opinion that doesn't count for much in a scientific context.

Russell furthermore talks about the nature of things being all the truth about that same thing. He states that with this assumption we can't know of that thing's nature at all unless we know all the thing's relation to everything else in the universe. If I were to interpret what Russell means by this I would say that it is irrelevant to pursue the nature of things. Truth is dependent on context. I do not believe that objects nor anything else possesses a truthful essence. I think that truths are ever-changing and that they may co-exist in contradiction to each other.