Summary of Beyond Ontology in Dialogue State Tracking For Goal-oriented Chatbot, by Sejin Lee and Dongha Kim and Min Song
Beyond Ontology in Dialogue State Tracking for Goal-Oriented Chatbot
by Sejin Lee, Dongha Kim, Min Song
First submitted to arxiv on: 30 Oct 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed approach leverages instruction tuning and advanced prompt strategies to enhance Dialogue State Tracking (DST) performance without relying on predefined ontologies or manual slot values. The method enables Large Language Models (LLMs) to infer dialogue states through carefully designed prompts, incorporating an anti-hallucination mechanism for accurate tracking in diverse conversation contexts. Additionally, a Variational Graph Auto-Encoder (VGAE) models and predicts subsequent user intent. This approach achieved state-of-the-art performance with a JGA of 42.57%, outperforming existing ontology-less DST models and demonstrating effectiveness in open-domain real-world conversations. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Goal-oriented chatbots are important for automating tasks like booking flights or making restaurant reservations. A key part of these systems is Dialogue State Tracking (DST), which understands user intent and keeps track of the conversation. The problem with current DST methods is that they rely on fixed lists of topics and manually compiled information, making them hard to use in open-domain conversations. Our new approach uses special prompts and instruction tuning to improve DST performance without relying on predefined lists or manual information. We also use a type of graph auto-encoder called Variational Graph Auto-Encoder (VGAE) to model and predict what the user might say next. This approach did better than existing methods in open-domain conversations, which is important for creating more accurate and adaptable goal-oriented chatbots. |
Keywords
» Artificial intelligence » Encoder » Hallucination » Instruction tuning » Prompt » Tracking
