Building computers that can comprehend human social relationships requires an understanding of power dynamics. Different roles played by the speakers, such as those of bosses and employees, might give rise to power dynamics. Communication amongst social equals, such as friends or acquaintances, can also be influenced by power dynamics. An autonomous system with the ability to comprehend power dynamics has many possible applications. These consist of analysing the official and informal groups' organisational structures, finding hostile or improper language in emails and other online media, and profiling key members of the group.
We investigate the challenge of developing efficient power identification systems. We demonstrate how to build a useful ground truth corpus for power analysis. We concentrate on three modelling domains that enhance power relationship prediction. First, utterance level—linguistic usage patterns can be used to differentiate between followers' and leaders' discourse. We present a set of syntactic/semantic traits that are useful for capturing various power manifestations in language. Second, Interaction Level: How speakers engage with one another might reveal information about the underlying power dynamics. To organise and model the data from these interaction-based cues, we employ Hidden Markov Models. Third, social conventions: A speaker's behaviour is shaped by their prior knowledge, particularly by prevailing communication norms.
0 comments:
Post a Comment