Apply Mobile Speech Dialog Agent Approach to High Speed Rail Ticket Reservation
碩士 === 南台科技大學 === 資訊工程系 === 100 === Speech is the most natural way for communication. In recent years, the burgeoning of handheld devices has provided us with the convenience in various daily activities. But limited input interface and computation power have caused inconvenient manipulation of handh...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | zh-TW |
Published: |
101
|
Online Access: | http://ndltd.ncl.edu.tw/handle/95804670561654478305 |
id |
ndltd-TW-100STUT8392002 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-100STUT83920022016-03-28T04:20:05Z http://ndltd.ncl.edu.tw/handle/95804670561654478305 Apply Mobile Speech Dialog Agent Approach to High Speed Rail Ticket Reservation 行動語音助理技術於高鐵訂票服務 Kun-Yi Huang 黃琨義 碩士 南台科技大學 資訊工程系 100 Speech is the most natural way for communication. In recent years, the burgeoning of handheld devices has provided us with the convenience in various daily activities. But limited input interface and computation power have caused inconvenient manipulation of handheld devices. In tradition, the ticket reservation process needs touching handheld devices to proceed. Under cloud computing framework, huge and complex computation is not difficult. Therefore, this study integrates speech dialog agent into ticket reservation process under cloud computing framework to provide intuition and convenient interface for handheld devices. The cloud computing framework of this study divides into server and client parts. The client part mainly records the input speech and gets the response from the server to the user. The server part includes speech recognition process, dialog analysis and dialog management module. In speech recognition process, the input speech from handheld devices is recognized into syllable lattice to search the grammatical sentence candidates by pre-defined language model. In the dialog analysis and dialog management process, eight semantic slots and seventeen dialog acts are defined to analyze the intentions of user’s speech. Then the dialog management module uses the information of the intention to complete the ticket reservation process. In the experiments, the training corpus combines TCC 300 and 390 sentences which were recording by the proposed system. In the sentence correct rate experiment, the result archives 80% using 390 sentences as recognition corpus. In the dialog performance experiment, the average dialog turns are about 5.15 and average response time is about 3.72 seconds. The preliminary results show the proposed application is feasible and can provide more convenient interface of handheld devices. Gwo-Lang Yan 顏國郎 101 學位論文 ; thesis 44 zh-TW |
collection |
NDLTD |
language |
zh-TW |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 南台科技大學 === 資訊工程系 === 100 === Speech is the most natural way for communication. In recent years, the burgeoning of handheld devices has provided us with the convenience in various daily activities. But limited input interface and computation power have caused inconvenient manipulation of handheld devices. In tradition, the ticket reservation process needs touching handheld devices to proceed. Under cloud computing framework, huge and complex computation is not difficult.
Therefore, this study integrates speech dialog agent into ticket reservation process under cloud computing framework to provide intuition and convenient interface for handheld devices. The cloud computing framework of this study divides into server and client parts. The client part mainly records the input speech and gets the response from the server to the user. The server part includes speech recognition process, dialog analysis and dialog management module. In speech recognition process, the input speech from handheld devices is recognized into syllable lattice to search the grammatical sentence candidates by pre-defined language model. In the dialog analysis and dialog management process, eight semantic slots and seventeen dialog acts are defined to analyze the intentions of user’s speech. Then the dialog management module uses the information of the intention to complete the ticket reservation process.
In the experiments, the training corpus combines TCC 300 and 390 sentences which were recording by the proposed system. In the sentence correct rate experiment, the result archives 80% using 390 sentences as recognition corpus. In the dialog performance experiment, the average dialog turns are about 5.15 and average response time is about 3.72 seconds. The preliminary results show the proposed application is feasible and can provide more convenient interface of handheld devices.
|
author2 |
Gwo-Lang Yan |
author_facet |
Gwo-Lang Yan Kun-Yi Huang 黃琨義 |
author |
Kun-Yi Huang 黃琨義 |
spellingShingle |
Kun-Yi Huang 黃琨義 Apply Mobile Speech Dialog Agent Approach to High Speed Rail Ticket Reservation |
author_sort |
Kun-Yi Huang |
title |
Apply Mobile Speech Dialog Agent Approach to High Speed Rail Ticket Reservation |
title_short |
Apply Mobile Speech Dialog Agent Approach to High Speed Rail Ticket Reservation |
title_full |
Apply Mobile Speech Dialog Agent Approach to High Speed Rail Ticket Reservation |
title_fullStr |
Apply Mobile Speech Dialog Agent Approach to High Speed Rail Ticket Reservation |
title_full_unstemmed |
Apply Mobile Speech Dialog Agent Approach to High Speed Rail Ticket Reservation |
title_sort |
apply mobile speech dialog agent approach to high speed rail ticket reservation |
publishDate |
101 |
url |
http://ndltd.ncl.edu.tw/handle/95804670561654478305 |
work_keys_str_mv |
AT kunyihuang applymobilespeechdialogagentapproachtohighspeedrailticketreservation AT huángkūnyì applymobilespeechdialogagentapproachtohighspeedrailticketreservation AT kunyihuang xíngdòngyǔyīnzhùlǐjìshùyúgāotiědìngpiàofúwù AT huángkūnyì xíngdòngyǔyīnzhùlǐjìshùyúgāotiědìngpiàofúwù |
_version_ |
1718212397605322752 |