Difference between revisions of "Research on music culture as a social network"
(→Web-based social networks)
|Line 219:||Line 219:|
* http://myspace.com (artist ego-nets...) (rather outdated in 2019! what remains?)
* http://myspace.com (artist ego-nets...) (rather outdated in 2019! what remains?)
== Sites featuring music that also include relationships ==
== Sites featuring music that also include relationships ==
Latest revision as of 11:39, 26 March 2020
short link: http://bit.ly/romcns
- 1 Possible project ideas and questions
- 2 Kinds of music networks
- 3 Special music network types
- 4 Network analysis, simulation, and visualization
- 5 Sources of online network data
- 5.1 Web-based social networks
- 5.2 Sites featuring music that also include relationships
- 5.3 Social networks especially for musicians
- 5.4 Music metadata sites
- 5.5 APIs
- 5.6 Websites indicating musical collaborations, genres, influences
- 5.7 Small world nets
- 5.8 Music sales, purchase recommendations, popularity
- 5.9 Internet content
- 5.10 Citation networks
- 5.11 Worldcat Identities Network
- 5.12 SNA Datasets
- 6 Bibliography
Possible project ideas and questions
Generally you (1) locate (or collect) good network data, (2) develop some interesting research questions about that data, which may require collecting more data, and then (3) formulate SNA procedures to answer those questions. Steps 1 and 2 might be reversed.
A note on Affiliation Networks: it is tempting to construct 2-mode networks because the data is usually more readily available (e.g. which bands played at which venues) but before embarking on exploring this sort of network data always ask yourself, "what is the significance of the associated 1-mode networks?" For instance, what does it mean if two bands played in many of the same venues, and two other bands didn't? Conversely, what does it mean if two venues have hosted many of the same bands? Note that you can always construct an affiliation network out of what would ordinarily be considered simple attribute data (e.g. musicians affiliate to the countries where they were born...to their instrument, or to their shoe size!) but it's not clear that you learn anything from the resulting network.
- Structure of friendship or acquaintance networks in relation to musical taste networks.
- Fan networks - who's a fan of whom?
- Networks of the social and inanimate: can you construct networks that connect people, groups of people, musical objects, technologies, all interacting?
- Music ollaboration networks of various types: among composers and lyricists, among performers, connecting performers and producers, etc.
- Structure of friendship or collaboration within a musical group
- Connecting class members via real or online networks (e.g. shared FB friends)
- Daily changes in the Twitter egonetwork for music celebrities, correlated to album releases
- Studies of ethnomusicologists as they collaborate in groups, or affiliate to topics
- Musical taste implications for social networks
- Online networks implicit or explicit - of all types (e.g. social media platforms)
- Observation of conversational interactions concerning music
- Observation of performance interactions, especially for improvisational musics
- Plot networks in films about musicians
- Ego-alter networks
- Friendship networks
- Musical affiliations
- Economic networks: what is the network of financial flows?
- Networks of musical production: who is involved and how?
- Flow: how you learn about music from friends...
- Literary connections through song lyrics
- Composer-lyricist networks
- Social fragmentations induced by media fragmentations
- Which music networks are scale free?
- Diameters of music collaboration networks
- Historical relationship between artists or albums, and newspapers or magazines that review/mention them (many newspapers can be searched online).
Note the importance of comparison and correlations:
- It will always be helpful to frame your question comparatively, because comparisons (always apple-apple, not apple-orange!) of similar yet different networks immediately raises the intriguing question: "why these differences?" which will push you to interpret your data in interesting ways. For instance:
- Temporal comparison: how does the network compare across eras or decades?
- Background comparison: how does your case compare to the "general case"?
- Genre comparison: how does the situation compare across genres?
- Geocultural comparison: how does the situation compare across cultures, ethnic groups, nations?
- It's also helpful to correlate network and attribute data within a particular network, e.g.: in this network, what is the relation between node degree and performer genre? (Often the procedure is to create several partitions in Pajek, then compute their Pearson correlation across nodes.)
Some research questions, then:
- What factors predict network links? What is the relation between attributes and links? (e.g. testing the homophily hypothesis, that "birds of a feather flock together", i.e. those with similar attributes tend to be linked) How do these relationships vary by genre?
- Assortativity questions: besides attributes, what network factors are shared by linked nodes? Do they tend to have similar or different degree?
- How do music networks form and change? Who or what shapes them?
- Who are the opinion leaders or early adopters in music networks and how can they be identified?
- What is the structure of fame? Who is famous, and how did they get that way? How is fame distributed? Compare across genres or music cultures or historical periods.
- What are the cohesive communities, as suggested by network data, and how do these correspond to attribute properties of the nodes?
- What is the impact of technology on the shape of social networks? Can we identify "eras" corresponding to pre-mediated, radio, TV, internet, etc.?
- How is the homophily hypotheses ("birds of a feather flock together", i.e. that links tend to form among people with similarities) upheld (or not) by music data? Do people who share musical tastes tend to associate? Do the tastes precede such groupings, or follow from them?
- What is the relationship between friendship and shared musical taste? What factors are most important in establishing friendship?
- How does friendship or collaboration in a musical group vary according to personal attributes (gender, age, etc.)? How does it contrast from one group to another?
- Where are the cohesive sections of a music network, and why are they cohesive? How do they relate to other factors (e.g. gender, age, education, taste, genre...)
- How do songs spread virally after release? how does musical diffusion vary by genre, country, historical era? How quickly does the "hit" subside?
- How do collaboration networks vary according to genre? country?
- Who or what is most important in a network? In what ways, and why is this person so central? (Note: there are many ways to formulate the idea of "centrality", each with a different meaning)
- What is the diameter of different collaboration networks over time, or across genres?
- How do people learn about new songs and artists? What networks come to bear?
- How does genre relate to the distribution of fans across artists?
- What is the structure of "who studied with whom" in various disciplines, including music disciplines. How do they compare, and why the differences?
Good news! We now have ethics approval to run Google Forms surveys!! see below...
(data collection of various network types, and analysis):
- Collecting data by survey instrument with a target population, using Google Forms.
- Note: you must precede your survey with a survey preamble following these guidelines, then add a checkbox and blank, allowing the participant to indicate whether they've read the text and agree to participate (check box); if so they "sign" (type their name in the blank).
- Next you'll have a set of radio buttons or a drop-down list, with one response per individual - the participant selects their name
- After that, you'll include a series of questions, probably with a set of possible responses as checkboxes, but possibly also using other ways to capture data.
- Please note the following pointers:
- Read up on Google Forms before starting, so that you understand what you can do - there are many neat features.
- Google Forms allows you to divide your form into sections, which is useful visually as well as logically, because on the basis of one answer you can have it "jump" to a different section rather than proceed to the next in sequence. This is helpful if for instance you want to probe with a follow-up question in the event they answer "yes" to a particular question.
- Try to provide a list of possible responses whenever possible, rather than leaving them to type their answer freely. If you allow open-ended answers you'll have to correct spellings or categorize yourself later - by hand.
- If you force a single choice (radio button) some people may not respond if they feel they need to indicate more than one - think about whether a single-choice question is the right choice. (Testing your form with a friend is very helpful!)
- Google Forms does allow an "other" option even when there's a radio button or checkbox list - so for those who need to type something they will have this option.
- You can mark any question "required" - you might want to mark them all.
- Google Forms allows you to restrict the range of an answer - e.g. to require that it be numeric.
- Under "settings" you have such options as a progress bar, the ability to edit a response, the requirement to respond just once, a custom message at the end.
- Some issues to consider when doing this sort of research:
- Is your group large enough? Too large? Can you be sure you'll collect from everyone? (I suggest around 50 is good, more or less is ok too, but will they respond?)
- Are you collecting enough data? (it will be hard to get people to go through your survey a second time, so think about all the network and attribute data you might possibly need)
- Are you collecting too much data? (if the survey is too long people won't finish it!)
- Are your questions precise enough? (for instance just asking about "friends" might not be - more precisely: how often do you meet?)
- Is your question too precise? (if so it might be hard to answer, e.g. how many hours per week do you speak with this person.
- Is your question in good taste as well as ethical? (asking about relationships might be tricky sometimes)
- Would there be any language barriers?
- If you're collecting affiliation data (2 mode) do you also have 1 mode data to compare? Some affiliation data isn't all that useful except as attribute data (for instance you *could* create a social network by asking people where they've lived - two people are connected if they've lived in the same city- but how is the resulting network useful?)
- Have you left enough options so that people can answer without feeling boxed in by the answers you selected?
- Is the survey the right length?
- Have you considered allowing "degree" rather than "yes/no" in a response (for instance, rather than simply "do you listen to..." ask "how many hours per week do you listen to...")
- Consider allowing free-format responses (but then you would have to code them, possibly, to come up with networks)
- Consider a "dry run" with friends to be sure the questions work - and refine based on their feedback
- This site may be very useful for providing examples.
- Snowball sampling (letting each subject refer you to others) online
- Random sampling of egonetworks (individuals and their alters) online
- Collecting data online through implicit (e.g. group membership lists, concert attendance lists) or explicit (e.g. Allmusic.com's credits pages, Amazon's "also bought" pages) connections
- Using "screen scraping" algorithms
- Using web crawlers (some SNA programs - like SocNetV - contain crawlers; see the list here).
- Using APIs
- By hand!
- Seeking existing datasets with network implications
- Simulation of real-world music networks, generating simulated data for analysis
ANALYSIS and STATS
- Affiliation networks & their analysis as one-mode
- Egonetworks & their analysis (statistical measures are possible with samples)
- Detecting network clusters and cohesive subnetworks (e.g. k-cores)
- Detecting node centrality (of different kinds)
- Measuring network distances and paths for information flow
- Locating central nodes
- Often you'll be correlating two partitions: one generated by network data (properties) and the other attribute data (not coming from the network)
- It's always helpful to be comparative (apples and apples, not apples and oranges!): construct two networks for similar groups and then ask - how do they differ? and why?
Kinds of music networks
- Affiliation networks (NB: any attribute data can become affiliation network data - you can ask people their level of education and then connect them when they've completed the same level: elementary, JHS, SHS, University.... The question you need to ask yourself is whether such a network is significant in any way? or just attribute data?)
- musicians/groups (musician to group)
- fan clubs (fan to club)
- Facebook pages (person to page)
- musical preference (person to style)
- concert attendance (person to event)
- scholars to research areas (topical, theoretical, disciplinary, or geocultural)
- Statistical implication networks
- taste implication networks (people who like this music also like this...)
- Purchase networks (people who bought this also bought this...)
- friendship networks
- arbitrary FN loaded with music attributes (taste, performance, consumption, breadth)
- musician friendship network
- Legal networks
- Relationships established by IP ownership
- Relation of artists to music corporations
- Relation of music corporations to one another
- Musical collaboration networks
- performer collaboration networks
- composer/lyricist networks
- Other collaborations (producer - musician, A&R agents, etc.)
- Flow networks (diachronic)
- the diffusion of musical awareness/preference/popularity (how does popularity spread?)
- the flow of music media: production/distribution/consumption/critical feedback networks
- genealogical networks
- transmission - direct teaching and learning networks (formal or informal)
- musical influence
- folktune variations
- musical style and genre development
- performance interaction
- musician interactions
- performer/audience interactions
- Music theory and composition networks
- modal, chordal, or pitch networks describing possible musical sequences, as devised by music theorists
- composition networks (e.g. Lewin and Markov chains), devised by composers, and defining possible musical sequences for human or computer performers (a form of algorithmic composition)
- networks for free improvisation (e.g. "Electrical Networks")
- Similarity networks, among
- musical prestige and authority
- admiration networks (subjective ratings)
- influence networks (subjective or objective ratings)
- Musical taste networks
- musician preference: celebrity topologies
- aesthetic preference networks (digraph: pairwise judgments on music objects, e.g. "I like A more than B" as "A->B")
- egonets and music attributes (performance, consumption, taste, breadth)
- Intertext and hypertext networks
- citation and co-author networks among music scholars
- linkages among webpages
- discursive links via musical terminology
- Intermusicality networks
- musical quotation
- stylistic allusions
- Two-mode network of ethnomusicologists and research topics
- Citation networks and influence
- web-based networks
- webpage word co-occurrence (e.g. musical genres)
- Facebook networks (centered on musicians; or musical friendship links; or embedded music information)
- Twitter networks (centered on musicians)
Special music network types
Think about the variety of types as applied to music culture:
- directed vs undirected networks
- one mode vs two mode networks
- complete vs subnetworks
- networks vs. egonets
- Small world networks (large networks with surprisingly small diameters)
- time-evolving networks
- Geographical networks (network formation favors local links )
- Scale-free networks (network formation favors global hubs)
Network analysis, simulation, and visualization
- Descriptive and exploratory
- what is the relation between network properties of nodes, and attribute properties of nodes, in a single network, then seeking explanations for these relations.
- Comparative: relating variables across two or more networks and seeking explanations (e.g. why is this network more densely connected than that one?)
See this page for a list of packages, other than Pajek.
Sources of online network data
These are not all about music, but may include music data, or be relevant to music networks. For instance you can look at friend networks, where each node's attributes include musical taste.
- Twitter. Entities: users. Relation: following (unidirectional). Information flow: tweets, retweets.
- Facebook. Entities: users. Relation: friendship (bidirectional); like (unidirectional); groups and events (affiliation) . Information flow: status.
- iTunes Ping
- http://myspace.com (artist ego-nets...) (rather outdated in 2019! what remains?)
See this list of Tools for obtaining network data from the WWW
Sites featuring music that also include relationships
- http://pandora.com (not available in Canada)
Social networks especially for musicians
Music metadata sites
These enable construction of networks (either straight metadata-> affiliation net, or relationship metadata-> one-mode net).
- http://musicbrainz.org - massive open-source metadata collection, including "relationships"
- E.g. ForRadiohead’s OK Computer https://musicbrainz.org/release-group/b1392450-e666-3926-a536-22c65f834433
- Note that they allow you to download data freely: https://musicbrainz.org/doc/MusicBrainz_Database
- https://www.discogs.com - more music metadata
- http://wikidata.org - a huge repository, some of which describes musical productions
- http://imdb.com - a movie database that also includes some music data (composer, performer)
- http://allmusic.com - attempts to be for music what imdb is for films (hard because the music world is so much bigger!)
(thanks to Brynn!)
- many websites, such as Spotify/Last.fm/Apple Music, allow public access through an API (Application Programming Interface), e.g. https://developer.spotify.com/documentation/web-api/reference/artists/
- e.g. for Radiohead (with artist ID 4Z8W4fKeB5YxbusRsdQVPb) try the following: (note - you have to be logged into Spotify for these links to work!)
- https://api.spotify.com/v1/artists/4Z8W4fKeB5YxbusRsdQVPb for basic information, and
- https://api.spotify.com/v1/artists/4Z8W4fKeB5YxbusRsdQVPb/related-artists for related artists
- Spotify's community showcase (https://developer.spotify.com/community/showcase/) provides a list of apps using this API.
Websites indicating musical collaborations, genres, influences
Small world nets
Music sales, purchase recommendations, popularity
- Bing (for word co-occurrence matrices - note that Bing provides the near:N search modifier; placed between two phrases it requires that they be separated by no more than N words)
- link: keywords (to tell Google to find pages that link to an artist page, thus establishing its ego net)
- http://correlate.googlelabs.com/ to correlate data series with search terms related to music
- Internet searching
- http://www.google.com/insights/search/# (enables matrix of search vs. locale)
- web of science (available via www.library.ualberta.ca)
- google books
...including some that are about music, and others that may be relevant to a music study
Note: new datasets are always popping up...please let me know if you find an interesting dataset that I should add to this list. (There is often an issue of transforming formats however; sometimes a little programming is required if a dataset can't be read directly by your software package of choice. Or you use a different package.)
- EchoNest (now owned by Spotify) and the million song dataset: http://millionsongdataset.com/ with the following fields (http://millionsongdataset.com/faq/#field-list)
- A Million Songs Dataset also partnered with Second Hand Songs to release a dataset of cover songs found in AMSD (http://millionsongdataset.com/secondhand/), as well as other complementary datasets from Last FM etc.
- Google's Audioset: https://research.google.com/audioset/ culled and categorized from YouTube videos
- new ESNAP data sets (may be the same as the old, but safer to use this with the new book)
- Network repository
- Barabasi datasets
- Awesome Public Datasets
- The KONECT Project network list
- Jazz musician network
- Linked Jazz project (see this nice visualization!)
- Allmusic, containing network data regarding collaborations and styles
- datasets from researcher Mark Newman
- Alex Arenas' datasets (with links to still more!)
- Stanford Large Network Dataset Collection
- old ESNAP data sets
Your projects need to rely on primary source data plus references - secondary scholarly sources: books, book chapters, and journal articles. (Please do not use general encyclopedia entries, though you may use entries from very specialized encyclopedias in special cases.)
How to locate references related to your projects?
First of all, compile a search strategy, remembering that your project is situated at the intersection of many fields; you need to think about combining them in different ways. For instance if you're studying friend formation in a choir, you might look at the literature on choirs (in musicology or music education?), or on friendship (in sociology or psychology or anthropology). If you're doing SNA (and you are!) then obviously any references that deploy a similar method (survey, say, followed by cohesive subnetwork detection) would also be relevant.
Second, decide on a systematic way to gather, sort, and review your references. A bibliographic database is key, preferably one that allows you to (a) store and manage bibliographic metadata, including notes and tags, and store in the cloud as well as on your computer, (b) automatically ingest such metadata from websites you visit, sometimes along with the full paper (the data), and (c) automatically insert and format citations and bibliographies in papers.
For all that I suggest Zotero (zotero.org): a database that can automatically ingest bibliographic information from most websites and provides cloud storage; it also offers a cite-while-you-write connection for Word and allows you to store notes. The library provides an equivalent tool, called Refworks. Others (e.g. Endnote) are for sale (there may be student discounts; check IST's shop). However you do it, be systematic but selective...and back up your computer regularly! Also as you add articles you prioritize - from titles and abstracts - which to read next. When you do read try to compose in the two-part format: (a) summary, (b) critique. That way you'll know what is useful and be all ready to "plug in" to your paper.
Third, here are some concrete steps to take:
- Make it a habit to sift through bibliographies for any of our course readings, or anything else you find, that appears relevant (e.g. Crossley)
- Look through reference works devoted to sociology or music/ethnomusicology as a starting point - relevant articles will contain useful bibliography.
- The first stop for most ethnomusicologists is Jstor, but there are many other resources too, including a number of online database reference works such as Garland Encyclopedia of World Music, and Oxford Music Online (the Library has a database search on their main page)
- Look for relevant journals and browse tables of contents (so many are online - the Library has a link for e-Journals on the main page: http://library.ualberta.ca)
- Check music journals devoted to your particular musical topic
- Look on the INSNA (International Network for Social Network Analysis) site, which contains a number of resources, including journals, Connections and Social Networks.
- Check my compilation of links for ethnomusicological research: http://bit.ly/emresearch
- Web of Science is particularly useful since you can traverse the citation network backwards and forwards (try searching for "social network analysis choirs")
- Check the Library's subject guide pages for Arts (especially Music or Sociology)
- The Oxford Bibliographies entry for SNA
- Here's a bibliography of classic SNA works
- and here's a study of SNA's bibliographic networks!
- Besides databases available through the UofA you should also check Google Scholar, as well as a general web search.
- Finally searching on Wikipedia is not wrong - it contains many useful links to academic research, but you should not cite general encyclopedia articles of any kind in your papers.