Please use this identifier to cite or link to this item: https://hdl.handle.net/10356/78999
Title: School ambassador chatbot based on Google voice AIY kit
Authors: Chang, Aaron Keat Lueng
Keywords: Engineering::Computer science and engineering::Information systems
Issue Date: 2019
Abstract: With the modern world diving into advancements in virtual assistants, popular personal assistant like Google Assistant, Cortana, Siri and Alexa are starting to become regular household names as more people in society start to integrate them into their everyday lives. The AIY Voice Kit is provided by Google to allow developers to have an out-of-the box tool for developing their own AI assistant with both the hardware and software ready for use. It interacts with users via voice inputs through the help of Google Assistant. The software portion of this project would later be replaced with Google Assistant Services due to the limitations present in the original AIY voice kit. The SCSE (School of Computer Science and Engineering) assistant was created to be the school’s own virtual assistant that provides information to students, staff and visitors. The scope was expanded later during the developmental phase to provide an additional navigation and pathfinding feature alike those in various malls. The SCSE Assistant will be located at the school’s main lobby on the 1st floor. Data scraping was used to extract data from the faculty’s website to build the database for the system. The data would then be stored on Google Firebase Realtime Database. Through the assistance of Dialogflow and Google Assistant, SCSE Assistant can interact with the users to provide a dynamic conversation and answer any queries the user might have about the school. Functionality wise, Dialogflow is closer to a commercialised off-the-shelf chatbot than an intricate dialogue system. However, Dialogflow still uses a Natural Language Understanding unit(NLU) to identify keywords and tag parts of the speech. Dialogflow provide developers to upload a knowledge base for a <question, answer> pair to imitate an FAQ bot, however, as Dialogflow does not store states and past utterances well, a Commonsense Knowledge base cannot be implemented. Due to Dialogflow being purely text-based, the processing speed for returning a response is quick, the downside would be that it is unable to recognise emotion like a DialogueRNN due to the lack of other tonal and visual inputs. However similar SCSE Assistant and Google Assistant are in nature, what sets them apart are the features that are only available on the SCSE Assistant. Namely, a custom wake word, the ability to recognise Asian names and a navigation system that is specific to the school.
URI: http://hdl.handle.net/10356/78999
Rights: Nanyang Technological University
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:SCSE Student Reports (FYP/IA/PA/PI)

Files in This Item:
File Description SizeFormat 
FYP SCSE18-0796 Amended Final Report (Aaron Chang).pdf
  Restricted Access
FYP report2.52 MBAdobe PDFView/Open

Page view(s)

96
Updated on Dec 5, 2020

Download(s)

30
Updated on Dec 5, 2020

Google ScholarTM

Check

Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.