Page tree
Skip to end of metadata
Go to start of metadata

Data Access Proxy (DAP) The Brown Dog web service responsible for file format conversions.  Built on top of Polyglot, utilizing Software Servers and Daffodil, the DAP is a highly distributed and extensible service which brings together and manages conversion capabilities, both the needed computation and data movement, from other software, tools, and services under a broadly accessible REST API.

Data Conversion :  A transformation on digital data that largely preserves the entirety of the data.  An example in the case of Brown Dog would be a transformation of a file in one 3D file format to another 3D file format.  As file formats typically vary slightly, and the transformations themselves can be imperfect, variations can occur in the form of information loss.  However, the intent is for the resulting data to be as intact as possible.  Conversions allow one to access data more easily given that the original format is not understood or difficult to work with.  This is analogous to translating languages.

Data Extraction :  A transformation that creates new data from the given data.  An example in the case of Brown Dog would be the execution of analysis code on an image file's contents to determine if a particular species of plant is present.  We utilize extraction to automatically generate metadata and/or signatures from a file's contents and provide users with means of finding, relating, and utilizing data that may be difficult otherwise. 

Data Tilling Service (DTS) : The Brown Dog web service responsible for extracting novel, often higher level, data from file contents (e.g. metadata, tags, signatures, and other derived products).  Built on top of Clowder (formerly Medici 2.0) and utilizing Versus for content based comparisons the DTS is a highly distributed and extensible service which extracts from data information by which to index, compare, and further analyze collections of data through a broadly accessible REST API.

Metadata : Simply data about data (e.g. tags or keywords).

Uncurated Data : Think of a dump of some random hard drive.  Without meaningful file names and a meaningful directory structure it will be difficult to find information without examining each and every file.  File formats, in particular old and/or proprietary file formats, hinder the situation further by making it difficult to open a given file without the needed software to open it installed on your machine.  Metadata is another way of providing insight as to the contents of a file.  Consider a document tagged with keywords "paper, large dynamic groups" indicating a paper submission for a social science study looking into the behavior of large groups of people. Curated data is data that has been stored and diligently named, organized, and tagged so that others, both today and long in the future, can utilize the data.  Uncurated data on the other hand doesn't have much of this and is essentially a big mess for others to go through.  A significant amount of digital data, if not most, is uncurated.  In the scientific world this is sometimes referred to as "long tail" data, suggesting this is linked with the tail of the distribution of project sizes, with the vast majority of smaller projects not having the resources to properly manage the data they produce.   The bottom line is that curation is a cumbersome process and creating new data is both faster and more rewarding, at least in the short term, than going back and organizing old data.  As science hinges on reproducibility and building on past results, however, these problems must be addressed.

Unstructured Data : Data that does not have a pre-defined data model or is not organized in a pre-defined manner.  Unstructured data can be text based but can also involve sensor data or data that quantifies some physical object or phenomenon (e.g. images, video, audio, 3d models, etc.).  Such data is typically difficult to understand using traditional computer programs.  Images are a good example of this.  To a computer images are nothing more than an array of numbers representing pixel intensities or colors.  Though images are extremely informative to us as human beings, for a computer to make any use of them some form of pre-processing must be run.  An example would be to use computer vision to recognize faces within the image and then spit out their locations as numerical values and a textural tag identifying these areas as faces.  With information such as this a computer is then more readily able to carry out a search or other process involving the contents of such data.


Clowder Definitions

  • No labels