C06: Add Franz Kafka — Hebrew language exercises (1 page, PDM-1.0)

[shaypal5]

Summary

Add Franz Kafka (1883–1924) as a writer using his Hebrew language-exercise notebook, scanned from the Max Brod estate and available as a single PDM-1.0 page in the upstream repo. Kafka's Hebrew was that of an adult learner (he began seriously studying the language c. 1920); the resulting handwriting is careful and non-fluent — an underrepresented style that adds meaningful distributional breadth to the HTR training corpus.

Writer record

Field	Value
writer_id	franz_kafka
display_name	Franz Kafka
also_known_as	פרנץ קפקא, František Kafka
born / died	1883 / 1924 — PDM in all life+70 jurisdictions
scripts_written	Hebr, Latn
languages_written	he (learner), de, cs
period	1920–1924 (Hebrew study period)
VIAF	https://viaf.org/viaf/4927658/
Wikipedia	https://en.wikipedia.org/wiki/Franz_Kafka

Upstream source

entry_id	license	Notes
commons__kafka_hebrew_writings__p0001	PDM-1.0	Open-notebook spread of Hebrew language exercises, Max Brod estate scan

rights_basis: public_domain, attribution_required: false.

Known quality constraints

Kafka's Hebrew was learner-level, not fluent. Expect:

• Non-fluent letter forms: careful block-printing rather than cursive; some letters may be atypical or over-deliberate
• Low variety per page: exercises may repeat high-frequency letters and skip rare ones
• Mixed legibility: some exercises clear, others cramped or uncertain
• Many crops will likely be usable_for_syngen: false (non-fluent forms inappropriate as style exemplars for synthetic generation)
• quality.notes on each entry must document the learner-hand caveat explicitly

Despite these caveats, learner-hand samples are a valid and underrepresented data point for HTR generalisation. Ingest all crops that are at minimum legibility: medium; include legibility: low crops only if they are unambiguously identifiable as a specific letter.

Acceptance criteria

• New franz_kafka writer row in writers.jsonl with status: verified
• All crops with legibility >= medium ingested; borderline crops documented with usable_for_htr: false, usable_for_syngen: false and a quality.notes explanation
• letter.style field reflects learner/block-print hand (not standard cursive Ashkenazi)
• All entries pass python3 scripts/validate_indexes.py --upstream-path <upstream>
• python3 scripts/generate_release_artifacts.py --check passes
• python3 -m pytest passes