雖然使用者可透過台鐵e訂通查詢相關資訊，但對於那些不擅長使用APP的年長使用者，以及不方便打字，甚至無法打字的身障使用者來說，台鐵e訂通仍無法適用。有鑑於此，本論文以機器學習、API開發應用、開放式資料擷取及APP開發設計等相關技術為基礎，開發「結合Google語音辨識API之具智慧語音回覆之台鐵搭乘資訊查詢系統」，來解決上述問題。此系統主要此系統主要包含了「網路爬蟲」、「Google語音辨識API」、「關鍵字拆解」、「語意辨識」及「使用者介面設計」等技術開發，讓使用者透過語音交談方式，即能透過系統回覆查詢之所需台鐵相關資訊，提供更友善的使用者操作環境。本系統首先透過語音辨識與機器學習語意辨識，辨識出關鍵字，再透過網路爬蟲方式，在網際網路上做相關資料擷取，透過資料解析，顯示出查詢結果。因此本論文提出的台鐵搭乘資訊查詢系統，除了提供不擅長使用APP的年長使用者，以及不方便打字，甚至無法打字的身障使用者另一個查詢服務外，並可減少車站服務台的人力成本。在系統建置與實驗規劃上，目前優先針對時刻表與票價做智慧語音回覆之台鐵搭乘資訊查詢。透過實驗結果顯示，本論文所提出之「結合Google語音辨識API之具智慧語音回覆之台鐵搭乘資訊查詢系統」，其票價或時刻表之資訊查詢正確率仍能達到90%以上。未來除了會增加其他的查詢服務外，更會透過產學合作方式，優先於台東火車站建置實際系統，提供使用者使用，並透過使用者回饋，持續做系統的改善與分析。此外，未來亦可依據大眾交通工具的需求，藉由不同的開放式資料庫，開發其他大眾交通工具之具智慧語音回覆資訊查詢服務。

The passengers was claimed to use APPs for service of taking the Taiwan Railways instead of manual service. However, these kinds of APPs may not be suitable for the disabled and elder people. Therefore, a information query system of taking the Taiwan Railways with intelligent voice response by Google speech recognition API (IQSIVP) was proposed based on the technique of machine learning, API development, open data, and APP design to address the above issues in this paper. In IQSIVP, it included the web crawler, Google speech recognition API , keyword dismantling, semantic recognition, and user interface design. Hence, IQSVIP could apply the more friendly operation interface for users to query the information of taking the Taiwan Railways by IQSIVP with voice chat. In IQSIVP, the voice chat was transfer into text data by speech recognition. The text data were divided into several keywords data by keyword dismantling. Then, the keywords data was analyzed by semantic recognition. Finally the correct information was obtained and displayed on the monitor by web crawler and data analysis from analyzed keywords data. Hence, IQSIVP not only reduced the probability of human contact but also reduce the labor cost of manual service. Moreover, IQSIVP could be instead of the taking the Taiwan Railways APPs to be used for the disabled and elder people. In the experiment, the time and ticket price were the query metrics for IQSIVP. The experiment results also showed that the rate of correct query information of time and ticket price was over 90%. It proved that IQSIVP could be applied for the service of taking the Taiwan Railways with intelligent voice response. In the future, it will add more query metrics for IQSIVP. Moreover, we will develop the information query system with intelligent voice response for other public transportation based on IQSIVP.