TL;DR
This paper proposes a plan-based theory for understanding indirect speech acts, arguing that recognizing the underlying goals and plans of conversational partners is crucial for interpreting utterances beyond their literal meaning. It highlights how speaker-reference, specifically the use of pronouns like "it," can serve as empirical evidence for this theory by pragmatically presupposing a shared understanding of intent.
What Problem Does This Paper Address?
Conversations are more than just a series of literal statements; people often imply meaning, using indirect speech acts (e.g., "Is the salt near you?" to mean "Please pass the salt"). Understanding these indirect meanings is a fundamental challenge in natural language processing and artificial intelligence. Traditional approaches often struggle to account for how listeners infer the speaker's true intentions, especially when those intentions are not explicitly stated. This paper tackles the question of how conversational partners, particularly in task-oriented dialogues, use shared beliefs, goals, and social roles to interpret these indirect speech acts, and how this interpretation can be empirically validated.
Methods
The paper outlines a plan-based theory of speech acts, drawing heavily on the work of Perrault and Allen. This theory posits that both communicative and non-communicative acts are treated uniformly as operators that affect mental states. The core idea is that plan recognition – inferring the goals and plans of others – is the basis for understanding the illocutionary force of an utterance. The authors propose that speaker-reference, specifically the use of definite and pronominal references (like "it"), can provide empirical support for this theory. The use of such references, they argue, pragmatically presupposes a mutual belief that the referent is in focus, which in turn is supplied by the inferred shared goals of the participants. This allows for a more robust model of how indirect speech acts are signaled and interpreted.
Datasets
Not discussed in the paper. The paper uses a constructed dialogue fragment for illustrative purposes.
Key Insights and Ideas
- Plan Recognition as the Core of Interpretation: The paper emphasizes that understanding indirect speech acts isn't about deciphering linguistic cues alone, but about recognizing the underlying plans and goals of the speakers.
- Unified Treatment of Acts: Communicative and non-communicative actions are viewed through the same lens of planning and goal achievement, simplifying the reasoning process for intelligent systems.
- Speaker-Reference as Empirical Evidence: The use of pronouns and definite descriptions is not just a grammatical feature but a pragmatic signal that can validate theories of speech act interpretation. If a pronoun like "it" is used successfully, it implies a shared understanding of what "it" refers to, which can be a consequence of inferring indirect speech acts.
- "Helpful" vs. "Intended" Inference Paths: The theory distinguishes between a novice simply observing an inform act and a novice inferring the expert's intended action, suggesting that the latter is often the preferred and more informative route in cooperative dialogues.
Key Results
The paper doesn't present quantitative results but rather a theoretical framework and a qualitative argument. The key "result" is the proposal that speaker-reference, particularly the use of "it" in the provided dialogue example, serves as a pragmatic presupposition that can only be satisfied if the listener has inferred the speaker's indirect intent (i.e., a request). This provides a concrete way to empirically test and validate plan-based theories of speech act interpretation.
Limitations and Future Work
The authors acknowledge that this is a "bare bones outline" and that "much more research must take place to flesh out the theoretical connections." They deliberately avoid discussing the computational aspects of plan recognition and generation, suggesting that the utility of computational models for ethnographers of conversation is an important area for future discussion. Further work is needed to identify other "signals of the interpretation of intent" to fully capture conversational success.
Summary updated