The Prospects for Machine Discovery in Linguistics |
| |
Authors: | Vladimir Pericliev |
| |
Institution: | (1) Institute of Mathematics and Informatics, Bulgarian Academy of Sciences, G. Bonchev Str., bl.8, 1113 Sofia, Bulgaria E-mail |
| |
Abstract: | The article reports the results from the developmentof four data-driven discovery systems, operating inlinguistics. The first mimics the induction methods ofJohn Stuart Mill, the second performs componentialanalysis of kinship vocabularies, the third is ageneral multi-class discrimination program, and thefourth finds logical patterns in data. These systemsare briefly described and some arguments are offeredin favour of machine linguistic discovery. Thearguments refer to the strength of machines incomputationally complex tasks, the guaranteedconsistency of machine results, the portability ofmachine methods to new tasks and domains, and thepotential machines provide for our gaining newinsights. |
| |
Keywords: | linguistic machine discovery scientific discovery prospects for computational discovery |
本文献已被 SpringerLink 等数据库收录! |
|