Project Goal and Vision

The goal of SpeechBox is to encourage and collect feedback and other verbal inputs from local populace. It is part of a larger concept of understanding geologically based development of sentiment and innovation in cities and suburbs. Our goal was to build a prototype box capable of collecting speech audio, turning it into text and produce methods for analyzing activity and semantic content such as topics present in the input.

Implementation

Actual box is built on Raspberrypi and serves as a client while the server runs on Flask REST API. User interface is done using a touch-screen with a simple GUI and separate microphone/speaker setup. GTTS/Google API is used to produce Finnish and English speech synthesis from the response generated on the server. Speech input is first sent to the server, which transcribes audio to text using Google Speech Recognition API. The resulting text representation is used to extract a topic and simple sentiment (good, bad, neutral). Finally a DeepQA chatbot is used to produce a response to the input which is sent back to the client on the box, prompting a spoken response which we hope encourage further input from users.

Future development

Potential future development includes further developing the physical hardware and casing to better survive the elements. Self-sustainability is also a point of interest.