Datasets:
input_ids listlengths 8.19k 8.19k | attention_mask listlengths 8.19k 8.19k | labels listlengths 8.19k 8.19k | token_start int32 0 0 | token_end int32 56 5.92k | char_start int32 0 0 | char_end int32 77 7.14k | work_id stringlengths 18 50 | cid stringlengths 4 14 | source stringclasses 8
values | region stringclasses 2
values | region_start int32 0 0 | window_index int32 0 0 | n_tokens_region int32 56 5.92k |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
[
2,
235248,
245131,
49513,
244608,
244146,
246634,
238324,
109339,
237304,
241751,
237819,
242896,
238274,
237937,
242896,
237831,
241751,
237780,
238134,
241211,
238274,
241751,
238451,
241501,
239791,
241751,
237780,
237831,
238277,
243655,
238627,
23... | [
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1... | [
-100,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
... | 0 | 4,206 | 0 | 6,006 | IE00EGS1016703_VE00EGS1016703_004_1_tei#0 | IE00EGS1016703 | tei | both | 0 | 0 | 4,206 |
[
2,
235248,
245131,
49513,
244608,
244146,
246634,
238324,
109339,
237937,
237780,
237819,
242896,
237780,
237937,
242896,
237831,
241751,
237780,
238134,
241211,
238274,
241751,
238451,
241501,
239791,
241751,
237780,
237831,
238277,
243655,
238627,
23... | [
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1... | [
-100,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
... | 0 | 4,134 | 0 | 6,002 | IE00EGS1016703_VE00EGS1016703_004_1_tei#1 | IE00EGS1016703 | tei | both | 0 | 0 | 4,134 |
[2,235248,245131,49513,244608,244146,246634,238324,109339,238808,239791,241198,241751,240268,237780,(...TRUNCATED) | [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1(...TRUNCATED) | [-100,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,(...TRUNCATED) | 0 | 4,086 | 0 | 6,006 | IE00EGS1016703_VE00EGS1016703_004_1_tei#2 | IE00EGS1016703 | tei | both | 0 | 0 | 4,086 |
[2,235248,245131,49513,244608,244146,246634,238324,109339,237937,241317,238034,103158,238679,242386,(...TRUNCATED) | [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1(...TRUNCATED) | [-100,0,0,0,0,0,0,0,0,1,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,(...TRUNCATED) | 0 | 4,064 | 0 | 6,005 | IE00EGS1016703_VE00EGS1016703_004_1_tei#3 | IE00EGS1016703 | tei | both | 0 | 0 | 4,064 |
[2,235248,245131,49513,244608,244146,246634,238324,109339,238134,241198,242610,237304,238627,237304,(...TRUNCATED) | [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1(...TRUNCATED) | [-100,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,(...TRUNCATED) | 0 | 3,963 | 0 | 6,003 | IE00EGS1016703_VE00EGS1016703_004_1_tei#4 | IE00EGS1016703 | tei | both | 0 | 0 | 3,963 |
[2,235248,245131,49513,244608,244146,246634,238324,109339,238547,243555,103158,237819,238808,93673,2(...TRUNCATED) | [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1(...TRUNCATED) | [-100,0,0,0,0,0,0,0,0,1,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,(...TRUNCATED) | 0 | 4,060 | 0 | 6,008 | IE00EGS1016703_VE00EGS1016703_004_1_tei#5 | IE00EGS1016703 | tei | both | 0 | 0 | 4,060 |
[2,235248,245131,49513,244608,244146,246634,238324,109339,237937,237780,241501,238277,237819,242386,(...TRUNCATED) | [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1(...TRUNCATED) | [-100,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,(...TRUNCATED) | 0 | 3,975 | 0 | 6,006 | IE00EGS1016703_VE00EGS1016703_004_1_tei#6 | IE00EGS1016703 | tei | both | 0 | 0 | 3,975 |
[2,235248,245131,241317,244608,244146,246634,238324,109339,238451,243655,241751,240268,241751,238274(...TRUNCATED) | [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1(...TRUNCATED) | [-100,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,(...TRUNCATED) | 0 | 3,974 | 0 | 6,005 | IE00EGS1016703_VE00EGS1016703_005_1_tei#0 | IE00EGS1016703 | tei | both | 0 | 0 | 3,974 |
[2,235248,244146,246634,238324,109339,237304,241751,237819,242896,238277,237819,238034,238134,249765(...TRUNCATED) | [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1(...TRUNCATED) | [-100,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,(...TRUNCATED) | 0 | 3,939 | 0 | 6,006 | IE00EGS1016703_VE00EGS1016703_005_1_tei#1 | IE00EGS1016703 | tei | both | 0 | 0 | 3,939 |
[2,235248,245131,240801,244608,244146,246634,238324,109339,240246,241751,238679,237780,238451,238277(...TRUNCATED) | [1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1(...TRUNCATED) | [-100,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,(...TRUNCATED) | 0 | 1,544 | 0 | 2,188 | IE00EGS1016703_VE00EGS1016703_006_1_tei#0 | IE00EGS1016703 | tei | full | 0 | 0 | 1,544 |
title-author-training
Token-classification data for finding the title and the author of each work in Tibetan volumes, for fine-tuning jhu-clsp/mmBERT-base. One model learns both.
Labels (label_map.json): O=0, B-TITLE=1, I-TITLE=2, B-AUTHOR=3, I-AUTHOR=4. Special tokens and padding are -100. A token gets B if a span starts in it, I for the following tokens of the span.
Input: each work is cropped to its first and last ~3,000 characters (edges moved to a line break; whole work if 6,000 characters or fewer; a span is never cut). Windows are up to 8,192 tokens (in practice about 2,000). Tokenizer: jhu-clsp/mmBERT-base.
Columns: input_ids, attention_mask, labels, token_start, token_end, char_start, char_end, work_id, cid (collection id), source, region (full/head/tail), region_start, window_index, n_tokens_region.
Split (this version): all 34,774 annotated works were pooled; copies of the same work (same typed title+author, same first 300 characters, or a near-identical text start) were reduced to one (30,009 unique works; the higher-ranked source is kept: tei, google_books, google_vision, paddleocr_v2, ocrv1-ws-ldv1, ocr_qwen_v1, paddleocr_v1, then the longer text). The split is about 80/10/10, stratified on source x book type x length, spread over collection families; duplicate works never sit in different splits (a collection can have works in more than one split). Validation and test contain annotator spans only.
Annotator works by source
| source | train | validation | test | total |
|---|---|---|---|---|
| tei | 6,105 | 537 | 484 | 7,126 |
| google_books | 3,810 | 377 | 349 | 4,536 |
| google_vision | 2,994 | 304 | 225 | 3,523 |
| paddleocr_v2 | 5,075 | 637 | 640 | 6,352 |
| ocr_qwen_v1 | 3,206 | 308 | 341 | 3,855 |
| ocrv1-ws-ldv1 | 2,012 | 207 | 226 | 2,445 |
| paddleocr_v1 | 1,061 | 124 | 119 | 1,304 |
| v3_auto | 7,811 | 0 | 0 | 7,811 |
Counts
| split | works | windows |
|---|---|---|
| train | 32,074 | 32,074 |
| validation | 2,494 | 2,494 |
| test | 2,384 | 2,384 |
Train label counts: B-TITLE 32,074, I-TITLE 1,197,216, B-AUTHOR 32,074, I-AUTHOR 439,826.
- Downloads last month
- 89