Summary of Disentangling Instructive Information From Ranked Multiple Candidates For Multi-document Scientific Summarization, by Pancheng Wang et al.
Disentangling Instructive Information from Ranked Multiple Candidates for Multi-Document Scientific Summarization
by Pancheng Wang, Shasha Li, Dong Li, Kehan Long, Jintao Tang, Ting Wang
First submitted to arxiv on: 16 Apr 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: None
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 a novel framework for Multi-Document Scientific Summarization (MDSS) called Disentangling Instructive information from Ranked candidates (DIR). The authors identify two shortcomings of current abstractive MDSS methods: difficulty in handling global information and lack of guidance during decoding. To address these issues, the proposed DIR framework utilizes summary candidates to provide both positive and negative guidance for generating better summaries. Specifically, DIR selects high-quality summary candidates using a specialized pairwise comparison method, disentangles their instructive information into latent variables with Conditional Variational Autoencoder, and incorporates these variables into the decoder to guide generation. The authors evaluate their approach using three different types of Transformer-based models and observe noticeable performance improvements according to both automatic and human evaluation. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper solves a big problem in summarizing scientific papers. Right now, it’s hard for computers to make good summaries because they don’t understand the whole story. To fix this, the authors came up with a new way to use “summary candidates” to help guide the computer’s summary-making process. This approach is called Disentangling Instructive information from Ranked candidates (DIR). The idea is that by using these summary candidates, the computer can get better at understanding the big picture and making more accurate summaries. |
Keywords
» Artificial intelligence » Decoder » Summarization » Transformer » Variational autoencoder
