a13team/shikimori-dataset

--- annotations_creators: - no-annotation language_creators: - found language: - en - ru license: unknown multimodal: - text - tabular pretty_name: Shikimori Dataset size_categories: - n_10K source_datasets: - original task_categories: - tabular-regression task_ids: - tabular-multi-class-classification --- # Shikimori Dataset Anime database from [Shikimori](https://shikimori.one) with user ratings, suitable for recommender systems and ML research. ## Dataset Structure The dataset contains **3 separate data files**: ### anime.parquet / anime.jsonl (9,950 entries) Anime titles with metadata. | Field | Type | Description | |-------|------|-------------| | id | int | Shikimori anime ID | | name | string | English name | | russian | string | Russian name | | description | string | Synopsis (BBCode) | | score | float | Average community score | | rating | string | Age rating (g, pg_13, r, etc.) | | episodes | int | Total episodes | | episodes_aired | int | Aired episodes (for ongoing) | | duration | int | Episode duration in minutes | | genres | dict | Genre IDs and names | | season | string | Season (e.g. fall_2023) | | aired_on | string | Premiere date | | released_on | string | Release date | | status | string | released/ongoing/anons | | studios | dict | Studio info | | is_censored | bool | Censored flag | ### users_rates.parquet / users_rates.jsonl (67,071 entries) User anime viewing history and ratings. | Field | Type | Description | |-------|------|-------------| | id | int | Rate ID | | user_id | int | Shikimori user ID | | anime_id | int | Anime ID | | score | int | User's score (0 = unset) | | episodes | int | Episodes watched | | rewatches | int | Rewatch count | | created_at | string | First rate timestamp | | updated_at | string | Last update timestamp | ### genres.parquet / genres.jsonl (80 entries) Genre reference table. | Field | Type | Description | |-------|------|-------------| | id | int | Genre ID | | name | string | English name | | russian | string | Russian name | ## Use Cases - **Recommender Systems** (Collaborative Filtering, Matrix Factorization) - **Anime Classification** by genre/description - **Score Prediction** models - **User Behavior Analysis** - **LLM fine-tuning** for anime recommendations ## Data Source Original scraped from Shikimori API. See: https://shikimori.one/api/doc ## Notes - Anime descriptions contain BBCode formatting (`[character=id]name[/character]`) - Scores are community averages (float, 1-10 scale) - Users are anonymized by ID (not by name) - Dataset snapshot date: April 2026