RuDSI: Word sense induction (WSI) dataset for Russian

RuDSI is a new benchmark for word sense induction (WSI) in Russian. The dataset was created using manual annotation and semi-automatic clustering of Word Usage Graphs (WUGs). Unlike prior WSI datasets for Russian, RuDSI is completely data-driven (based on texts from Russian National Corpus), with no external word senses imposed on annotators. Depending on the parameters of graph clustering, different derivative datasets can be produced from raw annotation.

Inter-rater agreement: 0.41 as measured by Krippendorff's alpha.

Files and directories

• rudsi_russe18.tsv: RuDSI in the RUSSE'18 format.
• annotation/: scripts to prepare data for the annotation
• data/: words and their contexts (sentences) annotated by the annotators
• clusters/: clusters (senses) automatically assigned to word usages
• graphs/: word usage graphs in the NetworkX format
• plots/: visualized graphs in HTML
• stats/: various statistics about RuDSI

Please find more information on the provided data in the papers referenced below.

Version: 1.0.1, 11.1.2025. Correct target word and target sentence indices. Map judgment identifiers to use identifiers. Clean comment column. Keep excluded nodes with cluster '-1'. Regenerate graphs. Update plots. Use '.csv' file ending.

Version: 1.0.2, 06.10.2025. Use the .tsv file extension.

Paper

RuDSI: graph-based word sense induction dataset for Russian by Anna Aksenova, Ekaterina Gavrishina, Elisey Rykov and Andrey Kutuzov (2022)

As described in the paper below, the data was further optimized for the CoMeDi shared task. Version 1.0.1 reflects some of the changes described in the paper, i.e., correction of target word and target sentence indices, plus minor additional changes.

The CoMeDi Shared Task: Median Judgment Classification & Mean Disagreement Ranking with Ordinal Word-in-Context Judgments by Dominik Schlechtweg, Tejaswi Choppa, Wei Zhao, Michael Roth (2025)

@inproceedings{aksenova-etal-2022-rudsi,
    title = "{R}u{DSI}: Graph-based Word Sense Induction Dataset for {R}ussian",
    author = "Aksenova, Anna  and
      Gavrishina, Ekaterina  and
      Rykov, Elisei  and
      Kutuzov, Andrey",
    booktitle = "Proceedings of TextGraphs-16: Graph-based Methods for Natural Language Processing",
    month = oct,
    year = "2022",
    address = "Gyeongju, Republic of Korea",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2022.textgraphs-1.9",
    pages = "77--88",
    abstract = "We present RuDSI, a new benchmark for word sense induction (WSI) in Russian. The dataset was created using manual annotation and semi-automatic clustering of Word Usage Graphs (WUGs). RuDSI is completely data-driven (based on texts from Russian National Corpus), with no external word senses imposed on annotators. We present and analyze RuDSI, describe our annotation workflow, show how graph clustering parameters affect the dataset, report the performance that several baseline WSI methods obtain on RuDSI and discuss possibilities for improving these scores.",
}