Morning presentations of Day 2 of Code4lib 2014.
Visualizing Solr Search Results with D3.js for User-Friendly Navigation of Large Results Sets – Julia Bauder
- of the search process, exploration is very confusing and stressful
- get a lot of results in our search tools
- not actually all that hard, but assuming data already in Solr
- Newer Solr will give you pivot facets
D3 is a JS library for creating data visualization - D3 gallery
- steep learning curve
- some types are sensitive to missing or dirty data
- convert Solr pivot faces into D3 data
- requires direct access to Solr
Steep Learning Curve - code can position divs
- See Thinking with Joins by Mike Boestock for a better idea on how D3 works
- add-on to view facets as list or graph
- can see how different books approach the topic different
- can zoom in to an area or see specific books
So Far - can be used as a teaching tool
- students haven’t quite seen it yet
- will be doing usability testing
Visualizing Library Resources as Networks – Matt Miller
Very early experiment.
Why Networks?
We have disparate documents, and don’t think about how we connect them. How do you get a group of resources that are similar or related? How can documents tell a unified discourse? You cannot have a specialist for every domain. What if we query in groups instead of distinct items?
Networks are scalable and you can see larger patterns with visualization, while also being able to drill down to individual nodes. A more serendipitous way to discover resources.
Networks in Archives
Demo of single, large archival collection. You can pull together connections.
But how can we look at networks for all of our collections? Can use access terms. Provides larger picture to see patterns. Using D3 visualization.
There is a limit to how much can be displayed because the browser has to do the calculations.
Need more data to make this more interesting, but might leverage existing term, add other data (institutional e.g. donator), build links based on shared web visits
Networks in Catalogues
Using subject headings where each subject is a node (if 2+ resources have it), the more connections = stronger
Universe simulation. Colours are determined by shared relationships.
How to:
* compile MARC records into node/edge relationships – pyMARC to process into a Gephi XML doc
* tool to render the network based on existing network analysis algorithms – Gephi Toolkit
* take information and make visual –
We are all Disabled! Universal Web Design Making Web Services Accessible for Everyone – Cynthia Ng
Dead-simple Video Content Management: Let Your Filesystem Do the Work – Andreas Orphanides
the problem: too many files, a lot of repetitive HTML, files are not related to each other, only had files in current practice
Solution
* one (okay two) script for video playback
* videos/metadata store remotely
* separate content and display
* centralize management of video and metadata
Simple
* MVC model – web display, generate page with video, store information
You’ll have to see the video/slides for the details.
Limitations
* discovery challenges – indexing content
* metadata limitations – categories, etc. – solution – symlinks?
* content creators not isolated from system
* compatibility with content management policy/practice? but integration should be fairly straightforward
Break Time
[
]
(https://twitter.com/jordanheit/status/444705556402429952/photo/1)