Summary of Can Similarity-based Domain-ordering Reduce Catastrophic Forgetting For Intent Recognition?, by Amogh Mannekote et al.
Can Similarity-Based Domain-Ordering Reduce Catastrophic Forgetting for Intent Recognition?
by Amogh Mannekote, Xiaoyi Tian, Kristy Elizabeth Boyer, Bonnie J. Dorr
First submitted to arxiv on: 21 Feb 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 abstract proposes a solution to mitigate the problem of catastrophic forgetting in continual learning settings for task-oriented dialogue systems. This is a crucial issue, as these systems need to handle an increasing number of intents and domains after deployment. The authors explore the effect of domain ordering on intent recognition models’ performance in such settings. They compare three domain-ordering strategies (min-sum path, max-sum path, random) and find that the min-sum path strategy outperforms the others in reducing catastrophic forgetting when training with a smaller model (T5-Base). However, this advantage diminishes with larger models (T5-Large). The findings suggest that domain ordering can be a complementary strategy to mitigate catastrophic forgetting, particularly in resource-constrained scenarios. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper studies how to help task-oriented dialogue systems remember what they learned earlier. This is important because these systems need to understand many different things and get better over time. The researchers looked at how arranging the order of tasks can affect a model’s performance when learning new things. They tested three different ways of ordering tasks (min-sum path, max-sum path, random) and found that one way, min-sum path, worked best in some cases but not others. This shows that there are different ways to use task ordering to help models remember what they learned. |
Keywords
» Artificial intelligence » Continual learning » T5
