Summary of On the Role Of Context in Reading Time Prediction, by Andreas Opedal et al.
On the Role of Context in Reading Time Prediction
by Andreas Opedal, Eleanor Chodroff, Ryan Cotterell, Ethan Gotlieb Wilcox
First submitted to arxiv on: 12 Sep 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: 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 The proposed research presents a new perspective on how readers integrate context during real-time language comprehension, building upon surprisal theory. The authors observe that surprisal is not the only way to derive a contextual predictor from a language model, and propose an alternative method using pointwise mutual information (PMI). This approach yields similar predictive power as surprisal when controlling for unigram frequency, but also shows correlation with frequency. To address this issue, the authors project surprisal onto the orthogonal complement of frequency, resulting in a new contextual predictor that is uncorrelated with frequency. The experiments demonstrate that the proportion of variance in reading times explained by context is significantly smaller when using the orthogonalized predictor. This finding has implications for the role of context in predicting reading times and may lead to revised interpretations. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The research looks at how our brains understand language while we’re reading or listening. It’s like trying to figure out what a sentence means by knowing what words come before it. The scientists found that there are different ways to do this, but they all seem to work similarly well. They also discovered that these methods are connected to the frequency of the words being used. To get around this issue, they developed a new way to represent context that is not tied to word frequency. This new approach shows that our brains don’t use context as much as we thought when it comes to understanding language. |
Keywords
* Artificial intelligence * Language model
