I recently posted a request for information about any perl modules out there that would be usefull in comparing two strings of text. I found String::Compare on cpan and I have been happy with its simplicity and speed.
I wrote a summary below of how the module works and what my opinions are.
Introduction:
One of the major problems with using Allmusic.com as a metadata source is the difficulty in searching for the correct ablbum. Ideally we want to be searching for both Artist and Album together. However Allmusic.com only allows for searching of either album or artist not both. So the best way I have determined to match the album "Greatest Hits" is to first search for the artist and then search the discography for that album.
Part 2: Downloading Artist ID from Allmusic
Allmusic.com uses individual sql IDs to identify each artist and album. So what we need is a script to search and parse out the correct sql ID and add it to our mySQL database. I decided to use perl over php because this script will be run from a command prompt and not a browser because of the lenght of time necessary to complete it’s task.
I know that ID3v2 supports adding multiple genre tags to a file but does any software support this? I have not seen any that does and I don’t understand why. Most music belong in multiple sub genres, take Fleetwood Mac for instance, it belongs in Rock, but also a sub category of Classic Rock. Most people just classify it as Classic Rock, but what if you want to here a mix of all Rock songs, then you would have to select each subgenre of Rock. Also imagine the issue of a song that people can not agree on what the subgenre is.
I know I am not the first to talk about this problem, but for some reason I have not seen a whole lot of activity in this area. I know most of us do not have our music classified with subgenres, but if no software supports them then how are we ever going to start.
I know that there are other technologies such as smart playlists, but until we incorporate the use of the ID3v2 tag the way it was meant to be used I don’t think it will be a mainstream feature. Does anyone know of a player/library that uses this technology? I have not tinkered with what happens on my iPod yet but I imagine it will only see the first genre and not the others.
The next step in the never ending saga. Alright after some poking and proding I have gone back to using AllMusic. While using the TowerRecords site gave me a more accurate search the underlying metadata was pretty bad. As I said before they were not even consistent in genre naming. I also know that other people will want to use the AMG info to add more metadata.
The New Theory
1. Search for the Artist Key
2. Do a string compare on the artist name and research if low.
3. Load the artist discography
4. Do a string compare of each album in the database to the discography
5. Load the album page and populate the metadata
Does anyone know of a perl module or perl example of a quantitative text analysis. Such as comparing STR to an array and giving numeric results of its similarity to each string in the array. Such as is if the comparision is a difference of case it would score very low and a comparison where the strings did not even contain the same characters would be very high?
This is an akward concept. But I am sure there are papers out there on this subject. This is what enables large search engines to work. Please comment if you know of any. Thanks.