In order to understand the sentiments between users as they worked together as part of a collaborative Virtual Reality (VR) task, this research analyzed voice interaction using Assembly AI. The real-time conversations between 11 pairs of users were captured as they performed a joint puzzle task. The users were in separate physical spaces but connected in a shared virtual setting. The extracted voice sentiments were correlated with self-reported frustration levels and task completion times. The findings reveal a significant positive correlation between negative sentiments and frustration (r=0.73, p≤0.05) and a negative correlation between positive sentiments and task completion time (r=-0.87, p≤0.001). This shows the relationship between sentiment (positive and negative) and task completion in collaborative VR. These results highlight the potential benefits of analyzing voice sentiment in collaborative VR applications.