Colorado School of Mines MS Theses
Language-capable interactive robots participating in natural language dialogues with human interlocutors must be able to naturally and efficiently communicate about the objects, locations, and people found in human environments. A key aspect of natural language communication is the use of anaphoric language through pronominal forms such as it, this, and that . The linguistic theory of the Givenness Hierarchy (GH) suggests that humans use anaphora based on the cognitive statuses their referents have in the minds of their interlocutors. In previous work, researchers presented the first computational implementation of the full GH for the purpose of robot anaphora understanding, leveraging a set of rules informed by the GH literature. However, that approach was designed specifically for natural language understanding (NLU), oriented around GH-inspired memory structures used to assess the set of candidate referents with a given cognitive status. In contrast, natural language generation (NLG) requires a model in which cognitive status can be assessed for a given entity. In this work, we present a statistical model of cognitive status and demonstrate how this model can be used to facilitate robot anaphora generation. Specifically, we present an AI model that leverages the concept of cognitive status for the selection of pronominal forms for effective NLG.