Summary of What Do Transformers Know About Government?, by Jue Hou et al.
What do Transformers Know about Government?
by Jue Hou, Anisia Katinskaia, Lari Kotilainen, Sathianpong Trangcasanchai, Anh-Duc Vu, Roman Yangarber
First submitted to arxiv on: 22 Apr 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 paper investigates how transformer-based language models like BERT encode linguistic features, specifically government relations between sentence constituents. Researchers use probing classifiers and data from two morphologically rich languages to examine the model’s representation of government relationships. The study finds that information about government is primarily encoded in early layers of the model, but also across all layers. Additionally, a small number of attention heads can be trained to discover new types of government never seen during training. To support research on grammatical constructions, the authors release the Government Bank dataset containing thousands of lemmas with defined government relations. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper looks at how language models like BERT understand sentence structure and relationships between words. It uses special tools called probing classifiers to see what information is stored in these models. The study finds that this information is mainly stored in the early parts of the model, but also in later parts. They also find that some parts of the model can be trained to recognize new sentence structures they’ve never seen before. This research could help others who are studying how sentences work and release a big dataset that defines sentence relationships. |
Keywords
» Artificial intelligence » Attention » Bert » Transformer
