Presentation | 1998/10/16 Automatic acquisition of semantic representations with neural networks Naoto TAKAHASHI, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Real-valued vectors are useful for representing the meanings of words. This paper describes a method for acquiring such vectors from a corpus by using a neural network. We modified Miikkulainen's FGREP so that the neural network outputs case-roles instead of word representations ; this modification decreased the number of necessary units and increased the speed of learning. The acquired semantic representations showed a distribution that reflects the usage of each word in the corpus. We also confirmed the generalisation ability of our network with cross validation tests. Finally, we present ideas to extend our method to syntactic categories higher than words. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | semantic representation / corpus based / neural network / FGREP |
Paper # | NLC98-28 |
Date of Issue |
Conference Information | |
Committee | NLC |
---|---|
Conference Date | 1998/10/16(1days) |
Place (in Japanese) | (See Japanese page) |
Place (in English) | |
Topics (in Japanese) | (See Japanese page) |
Topics (in English) | |
Chair | |
Vice Chair | |
Secretary | |
Assistant |
Paper Information | |
Registration To | Natural Language Understanding and Models of Communication (NLC) |
---|---|
Language | JPN |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | Automatic acquisition of semantic representations with neural networks |
Sub Title (in English) | |
Keyword(1) | semantic representation |
Keyword(2) | corpus based |
Keyword(3) | neural network |
Keyword(4) | FGREP |
1st Author's Name | Naoto TAKAHASHI |
1st Author's Affiliation | Machine Understanding Division Electotechnical Laboratory() |
Date | 1998/10/16 |
Paper # | NLC98-28 |
Volume (vol) | vol.98 |
Number (no) | 338 |
Page | pp.pp.- |
#Pages | 8 |
Date of Issue |