Single-click recording and playback via Alexa for oral narratives

The project of this thesis is the research and development of an Alexa skill, a voice assistant, which queries the tiraccontounastoria.org server and then handles the different requests in the most appropriate way. The server contains stories told and experienced in first person, where anyone can listen to public stories and can tell new ones if they register. The aim is to be able to carry out an initial search via Alexa for the stories on the website and finally listen, always via Alexa, to the audio file saved on the server. When Alexa receives a request from a user, the voice assistant processes the task and sends a specific request to the server, based on what the user wants, and then receives an answer to provide to the user who asked a specific question. Server side, various requests are made to different URLs, depending on the type of requirement desired. Each kind of request will query the database on the basis of the information obtained and create a response. The server extracts, analyzes, filters and reorders the correct data based on the intent request from the skill and serializes these data to be sent to the back-end on Alexa that create a dedicated answer for the user. Alexa side, one has to define the various intents: the set of example sentences that a user usually says. Some of these intents are accompanied by slots: extra information that is provided for a specific fact. After that, Alexa route the voice command to the most appropriate intent and the developer handle each intent so that Alexa give the appropriate response. When starting the skill, the user can request a specific topic, check which public stories are present and which stories are told by a narrator. When Alexa plays the requested list, that list is divided into groups so as not to create a list that is too long. Cases of homonymy are handled, such as stories told by one person, and the stories reproduced are not repeated until all homonyms are listed. It is possible to listen to the audio by directly asking for the track title or number in the list. Once playing, it is possible to use the playback and positioning commands for a preset time by the developer.