mpetyx / darwinsmusic Goto Github PK

Trends of the genre or artist data over time.

host it on heroku

Need to arrange the folder structure in such a way that it is intuitive to browse through and understand it. There should be a clear distinction between styles/scripts/html. The external libraries should also be neatly placed.

Now we have all the tags of the songs over many genres from different time-periods. We will need Mr. Z's expertise in analyzing these tags and creating a 'richness measure' which reflects the variety with which each song is described by listeners/users. We might need to use wordnet to get rid of synonyms, and also do stemming etc before that.

This is to to visualize the richness of vocabulary for each genre over decades.

Incorporate DBPedia relations between artists: http://dbpedia.org/ontology/influencedBy and http://dbpedia.org/property/influences.

1. Remove all the unnecessary folders/files. Remember that we can always get them back as we're on github.
2. Each function should do just one thing.
3. Documentation

Need to get rid of the bubbles which show up despite the artist popularity is zero for that decade/time-period.

From DBpedia/Freebase, produce a list of genres/subgenres that will be shown on applications's homepage.

We should find a get around to this, and get all the recordings that have a particular tag.

Get the artists and all the related information for each genre. We get all the data and use it for building relatedness measures. Here are some (there may be others) that we might specifically be interested in:

1. DBPedia influence relations between artists: http://dbpedia.org/ontology/influencedBy and http://dbpedia.org/property/influences.
2. Active years
3. Birth place, date, Death place, date
4. Genre, instrument
5. Categories
6. Associated bands/musical artists

If there is no data for a particular decade then the map will remains same, it shows the last chosen decade.

Check on rock.json dataset:

• Click Genre "Rock"
• Go to decade 1980 then 1970 or from 2000 then directly go to 1970.
  For the 1970 and 1960, the visualization will remain unchanged.

Probably, we should put the 1970 and 1960 into the json data even tough with an empty value only.

The server code in DarwinsMusic/src/server/ is missing. May be the push failed?

It is unfortunate that mendeley has a restriction of 3 people per private group. I think we should switch from Mendeley to Zotero.

I have synced the old collection of papers to Zotero. I've sent an invite to this zotero group, please join.

1. Design the UI and the user interactions, for showing the recommendations.
2. Which libraries do we use for implementing the design?

1. What are the data sources?
2. What are the statistics for each genre, specifically relevant for our task?

1. What are the UI components?
2. Summerize the technical details
3. How does a user interact?

In order to compute similarity/relatedness between artists, we must leverage the links between different artists, not just the direct ones, but indirect ones as well. Eg: In a->b->c, a is directly linked to b, but indirectly linked to c.

Review literature on semantic relatedness measures (need not be for music) based on linked data.

In the JSON data, couple artists show up more than once in the same decade.
This causing more than one bubble will be drawn at the same point (showing as stack of bubbles).
Probably we can summarize the data for each artist, from the JSON data itself or when querying the SPARQL endpoint.