Summary of Target-constrained Bidirectional Planning For Generation Of Target-oriented Proactive Dialogue, by Jian Wang et al.
Target-constrained Bidirectional Planning for Generation of Target-oriented Proactive Dialogue
by Jian Wang, Dongding Lin, Wenjie Li
First submitted to arxiv on: 10 Mar 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 paper proposes a novel approach to dialogue planning for target-oriented proactive dialogue systems. The goal is to lead conversations from a context towards a pre-determined target, such as recommending items or introducing new topics. To achieve this, the system must plan reasonable actions and topics to drive the conversation proactively while moving it forward to the target topic smoothly. The proposed TRIP (Target-Constrained Bidirectional Planning) approach uses two Transformer decoders to bidirectionally generate a dialogue path consisting of action-topic pairs, minimizing decision gaps and contrastive generation of targets. A target-constrained decoding algorithm with bidirectional agreement is also introduced to control the planning process. The planned dialogue paths are then used to guide dialogue generation in a pipeline manner, exploring prompt-based and plan-controlled generation. Experiments on two challenging dialogue datasets demonstrate that the proposed methods significantly outperform baseline models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper aims to improve conversations by guiding them towards specific goals, like recommending products or introducing new topics. To do this, it proposes a new way of planning dialogues called TRIP. This approach uses special AI models to decide what actions and topics to use in the conversation. The goal is to make sure the conversation stays on track and reaches its target smoothly. The paper also introduces a new way of controlling the planning process by making sure the actions and topics agree with each other. Finally, it shows how these planned dialogues can be used to generate actual conversations that are more effective than usual. |
Keywords
» Artificial intelligence » Prompt » Transformer
