Summary of Controllable Discovery Of Intents: Incremental Deep Clustering Using Semi-supervised Contrastive Learning, by Mrinal Rawat et al.
Controllable Discovery of Intents: Incremental Deep Clustering Using Semi-Supervised Contrastive Learning
by Mrinal Rawat, Hithesh Sankararaman, Victor Barres
First submitted to arxiv on: 18 Oct 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
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 This paper proposes the Controllable Discovery of Intents (CDI) framework for conversational AI systems, aiming to address the challenges in discovering relevant turn-level speaker intents while incorporating domain and prior knowledge, constraints, and human feedback. The CDI framework consists of three stages: unsupervised contrastive learning on unlabeled data, fine-tuning on partially labeled data, and iterative refinement through repeated clustering and pseudo-label fine-tuning. To prevent catastrophic forgetting across these training stages, the paper draws from continual learning literature and employs learning-without-forgetting techniques. The authors demonstrate the effectiveness of CDI by reporting significant improvements over previous works on the CLINC and BANKING datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary In this research, scientists developed a way to help computers understand what people want when they talk to them. This is important because most AI systems don’t know how to figure out what someone wants until after they’ve already talked about it. The new system uses three steps: first, it learns from data without any labels, then it adjusts its understanding based on some labeled examples, and finally, it refines its knowledge by repeating these steps with human feedback. The scientists tested their approach on two datasets and found that it outperformed previous methods. |
Keywords
» Artificial intelligence » Clustering » Continual learning » Fine tuning » Unsupervised
