To explore the impact and applications of sentiment analysis algorithms on human emotions, the Speech2Mood system extracts sentiment from spoken language and displays the analysed mood visually.
To explore the impact and applications of sentiment analysis algorithms on human emotions, the Speech2Mood system extracts sentiment from spoken language and displays the analysed mood visually. Sentiment analysis is a branch of Natural Language Processing that is concerned with analysing patterns of speech and to derive from it, emotions, opinions and personal sentiment. A Raspberry Pi microcomputer running a Python script and leveraging Google NLP APIs is used to process speech input and to change colour and mood-orientation of the full-spectrum lighting inside the customer feedback display box. The display type was chosen to promote interaction with the project. As attendees leave the exhibition space, they are prompted to leave their spoken feedback in the microphone catchment area. Real-time code processes are shown on an adjacent screen. This programme was tested in this environment to analyse mood trends over time and other relevant metrics. The end objective for this project was to demonstrate a working product and to collect useful data on this feedback structure.
After testing and tuning the sentiment extraction code stack, reliability of the script was good. To ensure that any uncaught exceptions did not cause the script to become unresponsive, a restart is triggered when this condition is detected and a thread dump is saved for later analysis. Using the Google Natural Language Processing APIs, sentiment detection with clearly spoken speech was reliable. There will be additional development to ensure that on-board NLP processing using the Sphinx engine and libraries will correctly interpret the input. I am confident that further environmental tuning and microphone optimisation will improve the quality of the results.