发明名称 |
MODEL LEARNING DEVICE, CHARACTER STRING CONVERSION DEVICE, METHOD, AND PROGRAM |
摘要 |
PROBLEM TO BE SOLVED: To learn a model used for converting a collapsed word that is a fluctuating expression included in a character string into a normal word with high accuracy.SOLUTION: A fluctuating expression dictionary and a skip list are generated on the basis of a normality-collapsed phrase model generated by a normality-collapsed phrase model estimation unit 34. A character string generation unit 50 generates a character string, with a collapsed phrase in which a normal phrase is a null character string deleted therefrom on the basis of the skip list. A lattice generation unit 52 generates a lattice, and a model learning unit 54 learns a maximum likelihood model used for converting the collapsed word included in the character string into a normal word on the basis of the generated lattice.SELECTED DRAWING: Figure 3 |
申请公布号 |
JP2016081379(A) |
申请公布日期 |
2016.05.16 |
申请号 |
JP20140213752 |
申请日期 |
2014.10.20 |
申请人 |
NIPPON TELEGR & TELEPH CORP <NTT> |
发明人 |
SAITO ITSUMI;SADAMITSU KUGATSU;ASANO HISAKO;MATSUO YOSHIHIRO |
分类号 |
G06F17/27 |
主分类号 |
G06F17/27 |
代理机构 |
|
代理人 |
|
主权项 |
|
地址 |
|