IsaNLP RST Parser v3
| Property | Value |
|---|---|
| Author | tchewik |
| Paper | Bilingual Rhetorical Structure Parsing with Large Parallel Annotations (ACL 2024) |
| Languages | English, Russian |
| Model URL | https://huggingface.co/tchewik/isanlp_rst_v3 |
What is isanlp_rst_v3?
IsaNLP RST Parser v3 is a state-of-the-art discourse parser that implements Rhetorical Structure Theory (RST) analysis for both English and Russian texts. The model can analyze the discourse structure of text by identifying rhetorical relations between text segments, achieving impressive accuracy scores across different corpora.
Implementation Details
The parser comes in three main versions: gumrrg (bilingual), rstdt (English), and rstreebank (Russian). Each version is trained on different corpora and optimized for specific use cases. The model demonstrates strong performance metrics, particularly in discourse segmentation (up to 97.8% accuracy) and structure parsing.
- Bilingual support for English and Russian
- Multiple trained versions for different corpora
- High segmentation accuracy (92.1-97.8%)
- Strong performance in structure parsing (53.9-66.2%)
Core Capabilities
- Discourse segmentation
- Rhetorical relation identification
- Binary discourse tree construction
- RS3 format export support
- Integration with popular visualization tools
Frequently Asked Questions
Q: What makes this model unique?
The model's ability to handle both English and Russian languages, along with its high accuracy in discourse segmentation and relation parsing, makes it particularly valuable for cross-lingual discourse analysis. Its integration with various corpora (GUM, RST-DT, RRT) provides flexibility for different use cases.
Q: What are the recommended use cases?
The model is ideal for discourse analysis tasks, text structure understanding, and rhetorical relation identification in both academic and practical applications. It's particularly useful for researchers working with English and Russian texts who need detailed discourse structure analysis.
