We have built an AJAX-enabled browser-based testbed for evaluating the performance of computational linguistics algorithms. Our testbed consists of a visualization system and analysis portal. Our focus is on algorithms that classify and cluster documents by assigning weights to words and scoring each document against high-dimensional reference concept vectors. The testbed visualization and algorithm analysis techniques include Confusion Matrices, ROC Curves, Document Visualizations showing word importance, and Interactive Reports. A unique aspect of our testbed is document visualizations built using Scalable Vector Graphics that show why documents are assigned to particular concepts and categories.
AttigMEickSGKastnerCLevineALockwoodJWLouiRMoscolaJWeisharDJ. Transformation Algorithms for Data Streams. IEEE Aerospace Conference (Big Sky, Montana), March 2005.
2.
ByrnesJEickSGLevineALockwoodJLouiRMaugerJRatnerAWeisharD.Hardware Accelerated Algorithms for Semantic Processing of Document Streams. IEEE Aerospace Conference (Big Sky, Montana), 2006.
RocchioJ.Relevance Feedback in Information Retrieval. In: The SMART Retrieval System: Experiments in Automatic Document Processing, Chapter 14. Prentice-Hall Inc.: Englewood Cliffs, NJ, 1971; 313–323.
5.
SaltonG.Dynamic Information and Library Processing. Prentice-Hall Inc.: Englewood Cliffs, New Jersey, 1975.
6.
DeerwesterSDumaisSTFurnasGWLandauerTKHarshmanR.Indexing by Latent Semantic Analysis. Journal of the American Society of Information Science1990; 41: 391–407.
7.
ForsytheGEMalcolmMAMolerCB. Computer Methods for Mathematical Computations. Prentice-Hall Series in Automatic Computation.7th edn., Prentice-Hall: Englewood Cliffs, NJ, 1977.
