People with effective teamwork skills, such as collaboration or leadership, are highly demanded in the workplace. In turn, educational providers have adopted active learning methodologies, such as collaborative problem-solving. However, the objective evaluation of collaboration at scale still is a challenge. This paper explores the relationship between quantitative measures obtained from automated transcriptions of speech and qualitative indicators of effective collaboration. An omnidirectional microphone and an artificial intelligence algorithm were used to collect speaking data from 20 triads of students discussing and building a concept map. The study focused on validating the potential value of speech recording devices to quantify the dynamics of communication networks by comparing quantitative metrics obtained from them with an established rating scheme for measuring the extent of collaboration. Results showed a relationship between the standard deviations of the speaking times of the participants in each group and the evaluation obtained from the qualitative rubrics of communication and interpersonal relationships. Thus, the extent to which all group members contribute to the discourse can potentially serve as an indicator of effective group work.