RSS
 

Archive for the ‘Software’ Category

Final Product Post: MusicNet & The Alignment Tool

29 Jun

This is a final report and roundup of the MusicNet project. We’ll mainly be discussing the primary outputs of the project but will also cover an overview of the project as a whole.

We have two primary prototypal outputs/products from the project, they are:

  1. The Alignment Tool
  2. The MusicNet Codex

We’ll discuss each of these in turn and address what they are, who they are for and how you can use them in your own projects.

Read the rest of this entry »

 

MusicNet URI scheme and Linked Data hosting

19 Jan

MusicNet’s key contribution is the minting of authoritative URIs for musical composers, that link to records for those composers in different scholarly and commercial catalogues and collections. MusicNet claims authority because the alignment across the sources has been performed by scholars in musicology. The alignment tool and the progress to date has been detailed previously. In this post I will overview our methodology for publishing our work, in terms of the decisions made in choosing our URI scheme and how we model the information using RDF in the exposed Linked Data. I will then describe the architecture for generating the linked data, which has been designed to be easily deployed and maintained, so that it can be hosted centrally in perpetuity by a typical higher education computer science department.

URI Scheme

The URI scheme is designed to expose minimal structural information, for example, the URI for Franz Schubert is currently (see below for a volatility note):

http://musicnet.mspace.fm/person/7ca5e11353f11c7d625d9aabb27a6174#id

It is comprised of the domain name (musicnet.mspace.fm), an abstract type (person), an ID taken from the musicSpace hash of the composer (7ca5e11353f11c7d625d9aabb27a6174) and a fragment to differentiate the document from the person (#id).

We have chosen a hash rather than a human-readable label because we want to avoid people using the URI because they think that it refers to a composer when it might refer to a different composer. This is important in this domain because there are a number of composers with the same or similar names. Part of the alignment process has musicologists make this distinction. By forcing people to resolve the URI and check that it is the person they are referring to, we aim to avoid incorrect references being made. In addition it gives us the freedom to alter the canonical label for a composer after we have minted the URI, so that we don’t have a label-based URI with a different label in its metadata.

Domain Name

We intend for the domain name to change soon from one which isn’t explicitly tied to mSpace – this is in place right now for convenience to us. In particular our requirements are a domain that will not cost us anything to re-register in future, will remain in our control (i.e. not get domain parked if someone forgets to renew), and will not dissuade people from using it for any partisan or political reasons. The closest we might reasonably get is musicnet.data.ac.uk, although this is still unconfirmed at this point in time, and we may have to instead use musicnet.soton.ac.uk or musicnet.ecs.soton.ac.uk, which are not preferred, since they might give the impression that the data is a Southampton-centric view of the information, which it is not. For a more in depth discussion of a proposed solutions see our previous posts (data.ac.uk proposal & data.ac.uk revisited)

Ontological Constructs

In addition to the scheme for the URI, we also had to determine the best way to expose the data in terms of the ontological constructs (specifically the class types and predicates) used in the published RDF. We are fortunate that an excellent set of linked data in the musical composer domain already exists, in the form of the BBC /music linked data. For example, the BBC /music site exposes Franz Schubert with the URI:

http://www.bbc.co.uk/music/artists/f91e3a88-24ee-4563-8963-fab73d2765ed#artist

The BBC’s data uses the Music Ontology heavily, as well as other ontologies such as SKOS, Open Vocab and FOAF. Since we are publishing similar data, it makes sense for us to use the same terms and predicates as they do where possible, which is what we have done.

We are still in the process of finalising how we will model the different labels of composers. In the figure below we offer two possible methods, the first is to create a URI for each composer for every catalogue that they are listed in, publishing the label from that catalogue under the new catalogue-based URI, and use owl:sameAs to link it to our canonical MusicNet one. The second method is to “flatten” all labels as simple skos:altLabel links, although this method loses provenance. Currently we do both, and we’ve not finalised whether this is necessary or useful.

 

RDF model for MusicNet alternative labels

RDF model for MusicNet alternative labels

 

 

Content Negotiation & Best Practice

Similarly, we also follow the BBC /music model of using HTTP 303 content negotiation to serve machine-readable RDF and human-readable HTML from the same URI. Specifically, the model we’ve borrowed is to append “.rdf” when forwarding to the RDF view of the data, and to append “.html” when forwarding to the human readable view of the data. This is now implemented, and you can try this out yourself with the above URIs, which you can turn into the following:

http://musicnet.mspace.fm/person/7ca5e11353f11c7d625d9aabb27a6174.rdf
http://musicnet.mspace.fm/person/7ca5e11353f11c7d625d9aabb27a6174.html
http://www.bbc.co.uk/music/artists/f91e3a88-24ee-4563-8963-fab73d2765ed.rdf
http://www.bbc.co.uk/music/artists/f91e3a88-24ee-4563-8963-fab73d2765ed.html

There are several other offerings from the MusicNet site, some of which have been detailed before. First, the MusicNet Codex, which is the human search engine for MusicNet. In addition we have also created a (draft!) VoiD document that describes the MusicNet data set, available here:

http://musicnet.mspace.fm/void#

The perceptive among you will notice that the VoiD document links to an RDF dump of all of the individual linked data files, available here (14MB at time of writing):

http://musicnet.mspace.fm/dump.rdf#

Simple Deployment & Hosting

As noted above, our requirements state that our deployment must be as simple as possible to maintain by typical higher education computer science department web admins. In our bid we stated that we will work with the Southampton ECS Web Team to tweak our solution. As such, in order to keep our deployment simple, we have adopted an architecture where all RDF (including the individual Linked Data files for each composer) are generated once and hosted statically. The content negotiation method (mentioned above) makes serving static RDF files simple and easy to understand by web admins that might not know much about the Semantic Web. Similarly, the VoiD document and RDF dump get generated at the same time. The content negotiation is handled by a simple PHP script and some Apache URL rewriting.

Benefits of Linked Data

One of the benefits of using Linked Data is that we can easily integrate metadata from different sources. One of the ways in which we use this is using the aforementioned BBC /music linked data. Specifically, we enrich our Linked Data offering through the use of MusicBrainz. One of the sources of metadata that we have aligned is musicbrainz, based on a data dump we were given by the LinkedBrainz project team. The BBC also have aligned their data to Musicbrainz, and thus we have been able to automatically cross-reference the composers at the BBC with the composers in MusicNet. Thus, we can link directly to the BBC, which offers a number of benefits. Firstly, it means that users can access BBC content, such as recently radio and television recordings that feature those composers (see the Franz Schubert link above, for examples), but also that we can harvest some of the BBC’s outward links in order to enrich our own Linked Data offering. Specifically, we have harvested links that the BBC make to pages on IMDB, DBPedia, Wikipedia, among others, which we now re-publish.

The data flow from the raw data sources to linked data serving is illustrated in the figure below.

MusicNet Architecture Data Flow Diagram

MusicNet Data Flow Diagram

Future Work

The following tasks remain in this area of the project:

  1. Acquire control of a long-term domain name (preferably musicnet.data.ac.uk, see above).
  2. Discuss our RDF model with experts in Linked Data, Ontological Modelling and Provenance.
  3. Determine if we will offer a SPARQL endpoint in future. If we decide not to ourselves (because it might not be sustainable once our hosting is passed over to the department), it might be desirable to put the data on the Data Incubator SPARQL host.

This post documents Work Package 3 from the MusicNet project deliverables. MusicNet is funded through the JISCEXPO programme.

 

Alignment Tool (Beta Release 1)

30 Sep

There has been quite a bit of interest in the tool we’ve been developing to solve the (mis-)alignment of data across our multiple catalogs.

Today we’re pleased to be able to make a release available for download, albeit in an unsupported beta form!

Download Alignment Tool (beta 1) (Firefox 3.5+ or WebKit only)

Whilst we can’t make the data we’re working on available for download, we’ve created an example data set using names from the ECS Eprints repository. The use-case example here is that you have multiple name representations for specific individuals and you’d like to find the matches. Once installed the alignment tool will list all ‘Authors’ and when requested will present basic information about the papers/deposits that the author string was found on. For justifications on why the additional metadata/context is important be sure to read the previous post about the development process for the tool.

If you’d like to get the latest source-code for the tool as it evolves/improves you can check it out from our SVN repository which is hosted on the project’s Google Code page.

Finally, if you’d like to see what the Alignment tool should look like once you’ve got it installed then take a look at this screencast produced by our resident Musicologist David Bretherton for his recent presentation at AHM2010.

 
95 Comments

Posted in Software

 

Assisted Manual Data Alignment

08 Sep

One of the key milestones for the MusicNet project is to match up similar records in a number of source catalogues. Specifically we are looking for records that represent the same musical composer in multiple data sets.

As discussed in the report for our first work package (WP1 Data Triage) the reason this task needs to be undertaken is two fold:

  1. Different naming conventions (e.g. Bach, Johann Sebastian or Bach, J. S.)
  2. User input errors (e.g. Bach, Johan Sebastien)

Due to the scale of the datasets the obvious approach is to try and minimise the work by computationally looking for matches. This is the approach we attempted when trying to solve this same problem for the musicSpace project (precursor to MusicNet).

Alignment Prototype 1: Alignment API

In musicSpace we used Alignment API to compare composers names as strings. Alignment API calculates the similarity of two strings and produces a metric used to measure how closely they match. It also uses WordNet to enable word stemming, synonym support and so on.

We had interns create a system that would produce Google Doc spreedsheets as outputs of data-set comparisons. The spreedsheet consisted of a column of composer names from one data set, next to a column of composer names from the data set to compare and finally the similarity metric.

The spreedsheets were produced to allow a musicologist with domain knowledge to validate the results. From investigating the results it was quickly clear that this approach had a number of flaws:

Problem 1: Similarity threshold

The threshold where to consider a match to be significantly similar was very hard to define. False positives began to appear with very high scores. This is obvious in the above image where “Smith, John Christopher I” is considered to be the same as “Smith, John Christopher II” purely because there is only one character different.

Problem 2: Proper-nouns

Composer names are proper-nouns and as such don’t work well with WordNets spelling correction. For example a “Bell, Andrew” might not be a spelling mistake of “Tell, Andrew“.

Problem 3: No context

The spreedsheet didn’t offer our Musicologist expert any context for the names. As Alignment API just compares strings we weren’t able to provide additional metadata such as data set source or a link to find more information about the specific composer.

The lack of this additional context proved to be the most crucial, without it there was no way that our expert could say for sure that two strings were indeed a match.

Alignment Prototype 2: Custom Tool

The first attempt was unsuccessful but having learned from the failings of the first attempt we decided to have another go. A completely automated solution was out of the question so we decided to turn our efforts to a custom tool that could quickly enable our expert to find all the relevant matches across data sets.

From our work on the mSpace faceted browser we’ve learnt a lot about data displayed in lists. We’ve learnt that unless filtered “ugly” data tends to float to the top. By “ugly” we refer to data which isn’t generally alphanumeric, for example:

We also know, from our work in creating an mSpace interface on EPrints, that multiple instances of the same name can normally be found very close to one another. This is largely because most sources tend to list the surname first and the ambiguity comes from how they represent the forenames, for example:

With these factors in mind we came up with the following UI for our second attempt at an alignment solution:

Ungrouped/Grouped Lists

The left hand side contains two lists, one for all items that have yet to have been “grouped” and the second which shows all Grouped items. Each list has an A-Z control for quick indexing into the data and can be fully controlled by the keyboard. By selecting multiple items from the ungrouped list the user (at the touch of a single keystroke) can create a group from the selection.

Verified Toggle

The verified toggle lets the user see either expert-verified or non expert-verified groups. This is included as we provided a basic string matching algorithm in the tool to find likely matches. Unlike the first prototype, the string matches are a lot more rudimentary and hopefully produce fewer false positives (at the cost of fewer over all matches).

The algorithm produces a hash for each string it knows about by first converting all characters with diacritics to their equivalent alpha character. It then removes all non alpha’s from the string and converts the result to lowercase. This proved useful for our datasets as some data-sets suffixed a composers name with their birth and death dates, e.g. Schubert, Franz (1808-1878.). In these cases this would produce exactly the same hash as the non suffixed string and produce a likely candidate for a group.

Expert verification is still required but this algorithm helped present likely groupings to minimise the experts workload.

Context/Extra Metadata

The right hand side is where the user can see extended metadata about one or more items. This view displays all relevant information that we know about each composer and can be used by the expert to decide if more than one instance are in fact the same.

Metadata Match

Metadata Match is an additional visual prompt to help the expert decide whether composers are the same. If multiple composers are selected and their birth or death dates match they will be highlighted in grey. If the user hovers over the date with their mouse, all matching dates are highlighted.

Want to use the tool on your data?

The tool is currently in active development and when it reaches a stable state we will release an official download. It’s relatively generic and should be useful to any 3rd parties looking to perform a similar alignment/cleanup of their own data. If however you’re interested in having a play straight away then head on over to our Google Code site and grab the latest SVN.

Note: we can’t make the complete set of our music data available as we have access to it under a copyright licence, but we will in the near future provide import scripts and some sample data to serve as an example.

 
No Comments

Posted in Software