2
edits
Changes
Added presentation suggestion from Marmot Library Network
Mendeley recently added a, "document from file," endpoint to its API which attempts to extract metadata such as title and authors directly from PDF files. This talk will describe at a high level the machine learning methods we used including how we measured and tuned our model. We will then delve more deeply into our stack, the tools we used, some of the things that didn't work and why PDFs are the worst thing ever to compute over.
== Giving Users What They Want: Record Grouping in VuFind ==
* Mark Noble, mark@marmot.org, [//www.marmot.org Marmot Library Network]
In 2013, Marmot did extensive usability studies with patrons to determine what was difficult in the catalog. Many patrons had problems sifting through all of the various formats and editions of a title. In 2014 we developed a method for [//mercury.marmot.org/Union/Search?lookfor=divergent grouping records] so only a single result is shown in search results and all formats and editions are listed under that result. This presentation will cover all the technical details for how we accomplished Record Grouping, the design decisions in implementation, and the reaction from users.
[[Category:Code4Lib2015]]
[[Category:Talk Proposals]]
