"Affective computing is a recent research area in computer science which deals with the design and development of systems that can recognize, interpret and process human affects/emotions. Various research projects in the past have focused on affect sensing and processing raw textual data. One such research effort conducted at the Oak Ridge National Laboratory (ORNL) has introduced an affect propagation algorithm which can generate affective relationships between entities contained in a given textual document. The algorithm depends upon a set of real-valued numeric parameters for which the best possible values are unknown. This thesis describes three different contributions to ORNL's research project. Firstly, the affect propagation algorithm was implemented along with a visualization environment. Secondly, an experimental framework was created for comparison of different algorithms to optimize the affect propagation algorithm parameters. A benchmark system was established for this purpose. Thirdly, different optimization algorithms were implemented to optimize the affect propagation algorithm. The optimization algorithms included variants of stochastic hill climbing, simulated annealing and evolutionary algorithms. This thesis explores the use of a diversity maintained evolutionary algorithm to find the optimal parameter set for the affect propagation algorithm. A fitness sharing scheme has been adopted to maintain population diversity of the evolutionary algorithm. Statistical experimental studies are presented which show that the diversity maintained evolutionary algorithm performs best, followed by the adaptive simulated annealing algorithm, with respect to the best fitnesses achieved"--Abstract, page iii.
Tauritz, Daniel R.
Hurson, A. R.
M.S. in Computer Science
Oak Ridge National Laboratory
Missouri University of Science and Technology
ix, 82 pages
© 2011 Ajith Cherukad Jose, All rights reserved.
Thesis - Open Access
Library of Congress Subject Headings
Simulated annealing (Mathematics)
User interfaces (Computer systems)
Print OCLC #
Electronic OCLC #
Link to Catalog Recordhttp://laurel.lso.missouri.edu/record=b8525197~S5
Cherukad Jose, Ajith, "Optimization of textual affect entity relation models" (2011). Masters Theses. 5009.