2nd International Workshop on Speech, Language and Audio in Multimedia (SLAM2014)
Today's scientific literature contains a multitude of multimedia data. While we struggle as a
community to define the next generation of scholarly dissemination, we should also capitalize
on the opportunities to mine our scholarly literature. Aside from text, documents embed
figures representing examples, charts of experimental results, diagrams illustrating
workflows, often accompanied by their explanatory captions. Scholars use key phrases,
argumentative flow and section headers as discourse markers to indicate important results.
By machine reading of these documents, we can enable semi-automated scientific discovery,
including the generation of augmented and linked slide presentations, survey papers,
collection of scientific terms and their definitions. We will present the current state of these
works, describing work done at NUS while touching on other international groups' initiatives.
In many fora, the scientific record is also accompanied by more informal presentations such
as keynotes and conference talks. With the growing proportion of these talks being recorded
and opened to the public, we have a chance to link the textual data within scientific
documents to their oral presentations. We touch on the recent work in all of these areas and
point forward to the upcoming synergies between ASR, NLP and multimedia analysis that
will augment and define the next generation of digital libraries.
We will present the Association of Computational Linguistics' current thinking towards enabling this process workflow. We have designed a digital library that allows third parties to request, ingest and process the official scientific publications of the ACL Anthology, and allow them to publish their results in a designated part of the website.
Full Paper Presentation
Bibliographic reference. Kan, Min-Yen (2014): "Opportunities for multimedia analysis in scholarly digital libraries", In SLAM-2014, 2.