Presentation 1999/7/22
Named Entity Extraction Using Error-Driven Learning of Finite-State Transducers
Manabu Sassano, Koji Tsukamoto,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) This papaer describes a method of extracting named entities from Japanese text based on Eric Brill's transformation-based error-driven learning. We developed an extraction system which uses a morphological analyzer and machine-learned finite-state transducers (FSTs), and performed an experiment against the formal run (general topics) of the IREX (Information Retrieval and Extraction Exercise) NE) (named entity task). Our system learned 1,428 FSTs from the CRL NE data containing about 10,000 sentences and achieved an overall named entities F-measure of 71.28. The score was lower than that of the hand-crafted FSTs. However, the machine-learned FSTs outperformed the half of the systems participating in the IREX NE. Also, we didn't encounter overfitting in the learning process.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Information extraction / named entity task / error-driven learning / finite-state transducer
Paper # NLC99-6
Date of Issue

Conference Information
Committee NLC
Conference Date 1999/7/22(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) Named Entity Extraction Using Error-Driven Learning of Finite-State Transducers
Sub Title (in English)
Keyword(1) Information extraction
Keyword(2) named entity task
Keyword(3) error-driven learning
Keyword(4) finite-state transducer
1st Author's Name Manabu Sassano
1st Author's Affiliation Fujitsu Laboratories, Ltd.()
2nd Author's Name Koji Tsukamoto
2nd Author's Affiliation Fujitsu Laboratories, Ltd.
Date 1999/7/22
Paper # NLC99-6
Volume (vol) vol.99
Number (no) 227
Page pp.pp.-
#Pages 8
Date of Issue