日本語
DHSJR Collaboration

DHSJR Collaboration #

This page is an English-language summary of a working record completed on June 19, 2025 (analysis begun February 16, 2025), documenting the process of preparing KRM data for collaboration with DHSJR (Database of Historical Sino-Japanese Readings). The full account, including detailed worked examples, is kept in the Japanese version, DHSJRとの連携; this page presents only a summary. For KRM’s general treatment of Phonetic Glosses, see Types of Phonetic Glosses and Decipherment Issues.

This summary reflects the state of the collaboration as of June 2025, when the underlying analysis was completed. The mapped data was subsequently published on September 10, 2025; see krm_pronunciations for the current status.

1. Overview of DHSJR and the Approach to Mapping KRM Data #

  • Purpose of the DHSJR project: DHSJR (Database of Historical Sino-Japanese Readings) is a project to make Sino-Japanese character and word readings, from the Heian–Kamakura periods through the present, cross-searchable based on annotations such as Kana glosses, Tone marks, and Fanqie spellings. It is led by Professor Katō Daikaku of Waseda University.
  • Including the Myōgishō and resource IDs: The Myōgishō is not yet included in DHSJR. Work is underway to extract Phonetic Glosses from KRM’s full-text data (KRM.tsv and KRM_definitions.tsv) and adapt them to DHSJR’s format. Dedicated resource IDs have been reserved and organized for the Myōgishō’s several manuscripts (e.g., the Zushoryō manuscript: 30-048-01; the Kanchi-in manuscript: 30-048-02).
  • DHSJR’s 23-column structure: To load the data, each of DHSJR’s 23 defined data columns (e.g., character headword, tone marks, kana glosses, fanqie, similar-sound notes, material location) was individually mapped to the corresponding KRM data items.

2. Designing Phonetic Gloss IDs and the Row-Splitting Rule for One-to-Many Relationships (the core of krm_pronunciations) #

  • Basic ID design: In principle, the Definition Sequence ID (definition_seq_id, formerly KRID_no), which identifies a component of a dictionary Entry, and the Phonetic Gloss ID (pronunciation_id, formerly KRID_pron_no) are mapped one to one.
  • Where one-to-many records arise: When Kana glosses, Fanqie spellings, Similar sound notes, and Tone marks occur in combination within a single Definition, keeping everything on one row would obscure which character each Tone mark actually belongs to.
  • Adding suffix letters (b, c, …) and splitting rows: Where a one-to-many relationship arises, suffixes b, c, d, … n are appended to the Definition Sequence ID to generate multiple **Phonetic Gloss ID**s, and the data is split across multiple rows (affecting roughly 500-690 cases).
  • Priority order for row splitting:
    1. The Headword’s own direct Kana gloss and Tone mark (marked with a double circle “◎”) take priority and are recorded on the same row.
    2. A Phonetic Gloss stated as part of the Definition is recorded as the Phonetic Gloss for the Headword, on a separate row.
    3. Kana glosses or Tone marks applied to the individual characters used within a Similar sound note or Fanqie spelling (**Phonetic Gloss Character**s) are recorded on yet another row, with the target-character field changed accordingly.

3. Data Conversion and Normalization Rules for Tone Marks and Kana Glosses #

  • Replacing tone-mark symbols: The romanized Tone mark symbols KRM uses internally (L/F/H/R/T/S/V) are converted in bulk to DHSJR’s Japanese-language labels (平/平軽/上/去/入/入軽/濁). The underscore _ marking the absence of a Tone mark (formerly the at-sign @, before the March 2025 specification change) is replaced with a full-width asterisk (＊).
  • Concatenating multiple kana glosses: Where multiple Kana glosses exist, as in “シム／ニム”, they are joined with a full-width slash (／).
  • Handling tone marks and nasal marks attached to a kana gloss: Where a Tone mark or nasal-sound symbol (✓) is applied to the Kana gloss itself, that information is recorded inside full-width parentheses （） within the kana-gloss field — e.g., “和シヨ（＊平）” or “リヤウ（N＊＊）” — kept distinct from the Tone mark field used for Hanzi (Chinese characters).
  • Preserving representation via special characters: Specific characters inserted into a Kana gloss to mark a contracted or voiced sound — 火 for the contracted sound kwa, 所 for the contracted sound sho, 土 for a voiced sound — are retained as-is within the Kana gloss.

4. Detailed Processing Rules for Similar Sound Notes, Fanqie, and Reading-Type Markers #

  • Separating tone marks within a fanqie spelling: Where the final character of a Fanqie spelling (e.g., “曽” in “蘇曽反”) itself carries a Tone mark, the Headword’s own row (e.g., “僧”) records the fanqie with the tone-mark field left blank, and a separate row is created for the Phonetic Gloss Character record (e.g., “曽”) giving its Tone mark (平軽).
  • Compound notations such as “二音” and “二反”: Forms such as “二音”, “二反”, or “二切”, where multiple Phonetic Glosses are listed together under one Headword, are for now kept in their original notation rather than mechanically decomposed, to avoid losing nuance or internal consistency.
  • Protecting reading-type markers and source citations: Reading-type markers — the initial characters 和, 呉, 俗, 今, which indicate the nature of a reading — and source citations enclosed in double angle brackets (e.g., 《玉篇》《説文》) are essential scholarly information for interpreting a Phonetic Gloss and are retained in full in the data.