Papers by Kata Naszadi
Aligning Predictive Uncertainty with Clarification Questions in Grounded Dialog (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Previous work bases the timing of questions on supervised models learned from interactions between humans. |
| Approach: | They propose to ground the need for questions in the acting agent's predictive uncertainty by using the T5 encoder-decoder architecture to solve a Minecraft Collaborative Building task. |
| Outcome: | The proposed model can detect ambiguous instructions and predict responses better than previous models. |
Communicating with Speakers and Listeners of Different Pragmatic Levels (2024.emnlp-main)
Copied to clipboard
| Challenge: | Using a model of the speaker's intentions, people can achieve pragmatic interpretations using a variety of reasoning abilities. |
| Approach: | They propose to model the interaction between speakers and listeners with different levels of pragmatic competence and to model their level of reasoning abilities. |
| Outcome: | The proposed model is based on a simulating language learning and conversing between speakers and listeners with different levels of reasoning abilities. |