The Linguistic Data Consortium (LDC) has served, for more than two decades, as a center for the creation, distribution and archiving of language-related data. LDC employs a Consortium model in which members contribute to via fees and data donations and receive in turn ongoing access to a repository of their collected contributions whose value is several order of magnitude greater. This model, which might be described as stubbornly practical, has survived great economic upheaval and tidal shifts in attitude toward digital data -- but not without change. This presentation will discuss the evolution of the LDC model in comparison with models adopted by other data centers. It will sketch the original plan, the decisions to expand into data collection and annotation, tool building and the distribution of specifications. It will outline and partially quantify the impact the Consortium has had on language related research and technology development. Finally it will discuss the new directions that LDC is pursuing in x-sourcing, web service grids, cloud distribution, and alternate incentives for contributors.

Christopher Cieri has been Executive Director of the Linguistic Data Consortium since January 1998, 15 of the organization’s 20 years. He holds BA, MA and PhD in Linguistics from the University of Pennsylvania where he focused on sociolinguistics and language contact in terms of phonetics, phonology and morphology as well as historical and educational linguistics. Cieri’s experience in data collection and research programming began in 1983 with his work for the Language Analysis Center at the University of Pennsylvania. Between that position and his current one, Cieri spent 8 years as an IT Director at the University. Since joining the LDC, Cieri has developed its annotation activities and a ten-fold increase in accompanying funding. Today he oversees all aspects of LDC work including research and research administration, project planning, data collection, annotation, archiving distribution and outreach as well as the technical infrastructure that enables that work. To date, LDC has distributed more than 90,000 copies of more than 1,700 datasets to 3,379 companies, universities, and government research laboratories in 70 countries. The bibliography of scientific and technical papers that rely on LDC datasets has reached more than 10,000 entries, after checking approximately 60% of our catalog.

Jack Judy joined the Electrical and Computer Engineering Department at University of Florida in 2013, where he serves as the Intel Nanotechnology Endowed Chair and Director of the Nanoscience Institute for Medical and Engineering Technology (NIMET). Dr. Judy was formerly a program manager in the Microsystems Technology Office (MTO) of the Defense Advanced Research Projects Agency (DARPA). While at DARPA he managed the Reliable Neural-Interface Technology Program (RE-NET), which he created to address the fundamental and yet largely overlooked reliability problem of chronic neural-recording interfaces. Without successfully developing and translating high-performance neural-recording interfaces that function for the life of the patient, many of the widely envisioned clinical applications for brain-machine interfaces will not be realized. Dr. Judy served at DARPA while on leave from his faculty position in the Electrical and Biomedical Engineering Departments at UCLA, where he also served as Director of the NeuroEngineering Program, the Nanoelectronics Research Facility, and the Microfabrication Laboratory. He has received the National Science Foundation Career Award and the Okawa Foundation Award. He received his B.S.E.E. degree with summa cum laude honors from the University of Minnesota in 1989, and an M.S. and Ph.D. from the University of California, Berkeley, in 1994 and 1996, respectively.

Collecting data to answer a set of well defined, a priori scientific questions is difficult to do correctly and takes considerable planning but is standard practice in academia. How to collect data to share that it is useful to answer some potential as yet unknown questions is not well understood. For neural data this becomes especially challenging because there are critical trade-offs between the potentially enormous amount of data that is generated during an experiment and storage capacity with a relevant database. As an example, increasingly, epilepsy clinics are recording local field potentials with high frequency sampling in order to study changes in single neuron activity in the minutes leading up to the onset of a seizure. The amount of data generated prohibits storage of all the data and different clinics choose to save different pieces of the data depending on their primary interests. This ad hoc practice makes the sharing of data sets among investigators, who might want to test novel seizure detection algorithms, for example, impossible. Therefore, a valuable resource (e.g. continuous high sample-rate data from subjects undergoing spontaneous seizures) is forever lost, despite the fact that the data were initially recorded. Dr. Moxon will address some of the challenges underlying recording and sharing large sets of neural data.

Karen A. Moxon is a Professor of Biomedical Engineering and Associate Director for Research, Drexel University, School of Biomedical Engineering. She is an engineer by training with over 20 years of experience in computational neuroscience developing models to study how the brain represents sensorimotor information. Her experience ranges from using reductionist Hodgkin-Huxley type models of small numbers of neurons to simple integrate and fire models of large networks of neurons. Using an information theoretic approach she developed novel models of how variability in neuronal responses provides information about the type of stimulus. Due in part to the complexity involved in acquiring data from others for model development, she directs the NeuroRobotics Lab that performs basic science experiments, recording from large populations of neurons to gain insight into information representation. For example, she uses brain-machine interface paradigms to test hypothesis about neural encoding (Manohar et al.,2012). Applications of her work involve spinal cord injury (Kao et al., J Neurosci 2009), central neuropathic pain (Graziano et al., PLoS One, 2013) and epilepsy (Grasse et al., Exp Neurol. 2013). She is also involved in developing platform technologies to improve neuronal recording, holding two patents, one for the use of ceramic for microelectrodes (Moxon et al., IEEE-TBE, 2004) and the other for development of wirelessly controlled neuromodulation systems for neurological disorders such as epilepsy or Parkinson’s disease (Foffani et al., Brain 2003). Dr. Moxon will address some of the challenges underlying recording and sharing large sets of neural data.