Developer House Gets Inside View of Hot Projects at OCLC
Lightning talks give attendees a chance to talk with the developers of some of OCLC's current hot projects.
We spent most of today working on our projects, so thought we'd share more about the presentations we had yesterday, including OCLC's Linked Data strategy and a series of "lightning" talks by OCLC staff responsible for some of our current internal projects.
Developer House attendees met Stephan Schindehette and Bob Schulz to learn more about how OCLC is approaching linked data. Stephan is the product manager for Linked Data and Bob Schulz is senior product manager for Data Services and Quality Management. Stephan and Bob provided a brief overview of OCLC's history with linked data and the implementation of WorldCat Works. The duo also provided details on OCLC's implementation of linked data in the WorldCat Discovery interface, which is currently still in development.
Next up, Amy Morrison provided an overview of the User Experience team at OCLC. As the manager of OCLC’s User Experience and Information Architecture team, Amy discussed the role the UX team has in product development at OCLC. She stressed the importance of including user experience staff early in the product development cycle, to better inform wireframes and high-level interface design. She explained how the UX team is integrated into the cooperative’s product and development teams. Their role includes tasks such as defining information architecture, UI elements and usability testing.
Ron Buckley, software architect, gave the teams a high-level view of OCLC's move to Hadoop and Apache Hbase and described the size and structure of the infrastructure that sits behind WorldCat, Dewey, the WorldCat knowledge base and more.
Jeff Mixter, research support specialist in OCLC Research, presented a project to describe thesis and dissertations using Schema.org. The project is a joint effort between OCLC and Montana State University Library. The goal of the project is to boost the exposure of library collections in Google and other search engines. Using Schema.org, the project was able to boost Google's indexing of the library from nearly zero to more than 4,000 records in a matter of days. They used an RDFS vocabulary to express the Schema.org model and defined new classes and properties to create the new data model that Google indexed.
Our OCLC Research colleague Bruce Washburn took a break from coding to present the Lean Canvas product planning tool popularized by Ash Maurya. Bruce has introduced the Lean Canvas as a model for planning projects in OCLC Research. The build-measure-learn cycle advocated within the Lean Start Up movement combined with the Lean Canvas has saved Bruce and his colleagues time by focusing their efforts on what matters most and by avoiding efforts that didn't measure up after completing the Lean Canvas template. Earlier this week our four teams practiced a bit on focusing, when we defined the goal and the value of the projects that were proposed. Maybe we should have used a Lean Canvas for the selected projects? That’s something to think about for the next Developer House.
Taking us back into the technical side of things, Mike Boyers presented the End User Service's Discovery Team's use of the ELK stack within the Discovery UI and Discovery API products. Mike is a consulting software engineer on the Discovery Team. His team has built out the Logstash, Elasticsearch and Kibana stack to provide nearly end-to-end event logging for all Discovery products. The team has repurposed the Kibana visualizations to provide more than just debug and tracking down slow events. The Discovery product team now uses the ELK stack to get some basic product health information back from both the Discovery UI and Discovery API.
We wrapped up the "lightning" talks with Ralph LeVan, senior research scientist in OCLC Research. Ralph presented a project that he is working on with Rob Koopman, Innovation lab architect who is based in Leiden, Netherlands. He and Rob are working on a disambiguation project using random projection methods to collapse enormous (seriously huge) matrices of data from databases like ArticleFirst and map those to much smaller and simpler matrices that can be used to explore data sets with visualizations and to expose connections between articles based on content.
After the "lightning" talks and a quick lunch, our four teams got together for an scrum stand up to report status and blockers. In true unconference style, attendees were able to switch or abandon projects if they felt the project wasn't turning out to be fulfilling or productive. No one switched and everyone reported that they were making good progress so we jumped right back into coding.
In the afternoon, one of our teams was challenged with using SPARQL to analyze the linked data they wish to use. OCLC happens to have a SPARQL expert in Jeff Young, who is a software architect in the Office of Research. Jeff’s help has jump started the second phase of that team's project.
Tomorrow is our Show and Tell! Stay tuned to learn more about our projects. In the meantime, you can keep up with us via #devhouse and flickr.