Не Берроузом единым: альтернативные инструменты стилометрии

Value of the Information & Strength of the Argument

The presentation provides valuable insights into the comparative performance of stylometric metrics, offering empirical evidence for the superiority of Spearman distance in terms of stability and accuracy. The argumentation is solid, grounded in a systematic experimental design with multiple corpora and careful parameter tuning. The authors transparently discuss limitations, such as corpus imbalance and the absence of lemmatization for Ancient Greek. They also engage with prior research, positioning their work within the broader context of stylometric studies. The discussion of Jensen-Shannon divergence’s performance on the Strugatsky brothers’ corpus adds nuance, suggesting that information-theoretic measures may be useful for complex authorship cases.

Scientific Rigor, Source Quality, Title Accuracy

The presentation demonstrates scientific rigor through a clear methodology, reproducible steps, and reference to prior studies (e.g., Evert et al., Kocher & Savoy, Alieva, Pronin & Kazartsev). The sources cited in the description are links to Telegram and VK channels, which are not academic references but are provided for following the institute’s news. The title accurately reflects the content, focusing on alternative stylometric tools. The discussion includes a formal discussant, adding critical perspective. No comments were provided for analysis.

The title accurately reflects the content, which focuses on alternative stylometric tools beyond Burrows' Delta.

8/10

The presentation is based on a systematic comparative study of stylometric metrics, with clear methodology, multiple corpora, and reference to prior research. The authors are affiliated with reputable academic institutions. However, the video is a seminar presentation, not a peer-reviewed publication, and some details (e.g., exact parameter ranges) are presented verbally without full reproducibility.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:00Introduction by seminar host Andrey Volodin, welcoming speakers from Kazan Federal University.
• 03:00Anton Afanasyev begins the presentation, outlining the problem of choosing distance metrics in stylometry.
• 10:00Review of prior comparative studies on stylometric metrics, including Evert et al. and Kocher & Savoy.
• 20:00Introduction of the six metrics and their classification into frequency-based, probabilistic, and rank-based groups.
• 30:00Description of the four corpora used in the study, including the hypercorpus of science fiction and village prose.
• 40:00Explanation of the experimental setup: fragment lengths, parameter ranges, and the k-nearest neighbors method for attribution.
• 50:00Presentation of results: Spearman distance achieves perfect accuracy across all corpora, while Jensen-Shannon divergence shows promise on the Strugatsky corpus.
• 1:00:00Discussion of the implications and limitations of the study, including corpus imbalance and the need for parameter tuning.
• 1:10:00Discussant Boris Kovalev provides critical commentary, followed by Q&A session.

• Telegram channel 'Борис и Борхес' — Mentioned as the discussant's Telegram channel.
• VK page 'Институт цифровых гуманитарных исследований' — Provided for following the institute's news.

• Evert, S., et al. (2017). Understanding and explaining Delta measures for authorship attribution. — Referenced in the presentation as a key comparative study of distance measures.
• Kocher, M., & Savoy, J. (2019). Distance measures in authorship attribution. — Referenced for comparing Tanimoto, Motosita, Clark, and cosine Delta.
• Alieva, O. V. (2024). Comparative analysis of distance measures in stylometry. — Referenced for testing various metrics on Ancient Greek texts.
• Pronin, D. D., & Kazartsev, E. V. (2026). New metrics for stylometric analysis. — Referenced for parallel work on Jensen-Shannon divergence.

• Evert, S., et al. (2017). Understanding and explaining Delta measures for authorship attribution. — This study found cosine Delta to be superior, whereas the present study suggests Spearman distance performs better. The discrepancy may be due to differences in corpora and experimental design.