Using Machine Learning to Convey Emotions During Requirements Elicitation Interviews

Abstract

In an effort to assist in an ongoing research project where stakeholders have been interviewed using voice recording platforms and the Empatica E4 wristband to gather biofeedback data, the purpose of this research project in relation to the aforementioned research is to be able to determine the stakeholder's emotional range during a requirements elicitation interview. In doing so, the requirements analyst is supported during the requirements elicitation interview because conveying the emotional range of the stakeholder can help eliminate any miscommunication or misunderstandings between the requirements analyst and the stakeholder due to ambiguity in questions, statements, and responses. Therefore, knowing the range of the emotional state of the stakeholder, in real-time, can allow the requirements analyst to recover or make adjustments to questions (i.e. sensitive topics) during the requirements elicitation interview. The main questions are: (1) What machine learning technique(s) would be most efficient in conveying the emotional range of the stakeholder through the voice recording data and biofeedback data? (2) What features would result in optimal performance from the chosen machine learning technique(s)? The objective is to use supervised machine learning techniques in order to convey an emotional range from the retrieved dataset. To accomplish this, exploring the most efficient machine learning technique for emotion detection for voice recordings and biofeedback data and finding a way to construct or utilize the techniques in an effective manner will be necessary. In conclusion, the machine learning technique(s) chosen will convey a range of emotion from the stakeholder based on the retrieved data. The machine learning technique(s) will be used to support a requirements analyst during requirements elicitation interviews and will help the analyst identify problems or concerns in the communication to better assess and engage the stakeholder.