What Gets a Panel: Panel Themes at AJLS

About This Visualization

Methodology

This visualization analyzes the thematic content of AJLS panel and paper titles across 26 years of conferences (1999–2025). Three views offer different perspectives on what drives panel formation.

The treemap shows the overall distribution of topics, with tile size reflecting frequency and color indicating category (theme, period, genre, or literary figure). The timeline tracks how specific topics rise and fall. The theme alignment view shows which topics co-occur on the same panels, revealing thematic clusters and interdisciplinary connections.

Data Source

Topic extraction uses four regex-based keyword dictionaries applied to both panel titles and individual paper titles from 26 years of AJLS conference programs (1999–2025).

Keyword Dictionaries

Themes: 15 categories (~180 keywords) covering gender, empire, modernity, war, etc.
Periods: 7 eras (~40 terms) from classical/Heian through contemporary
Genres/Media: 9 types (~50 terms) including novel, poetry, film, manga
Literary Figures: ~120 author name variants with romanization handling

Matching Method

Matching applies Unicode diacritics flattening and word-boundary enforcement for short terms. This is regex-based keyword matching, not NLP or machine learning. The treemap aggregates all assignments across the full 26-year span. The timeline uses 3-year rolling windows to smooth year-to-year variation.

Theme Alignment

Co-occurrence is computed from shared panel membership: if two topics both appear in titles within the same panel, they share an alignment edge, with weight proportional to frequency of co-occurrence.

Limitations

Title-only analysis underestimates topic coverage compared to abstract-based methods.
Substring matching can produce false positives (e.g., “Post-Edo” matches “Edo”).
Japanese-language titles rarely match English keywords.
Panel titles sometimes describe the session theme rather than individual paper topics.