sanjay73 commited on
Commit
8b0668c
·
verified ·
1 Parent(s): 7bdb657

Update README.md

Browse files

We present **Pralekha**, a document-level alignment benchmark designed to evaluate various long-context parallel data mining strategies, comprising over 2 million documents, with a 1:2 ratio of unaligned to aligned document pairs across 11 Indic languages and English.
Each instance in the dataset is represented as a tuple (*n_id*, *doc_id*, *lang*, *text*).

In this structure:
- *n_id* serves as a unique identifier for each parallel document pair,
- *doc_id* uniquely identifies individual documents,
- *lang* indicates the document’s language using its ISO-3 code, and
- *text* contains the corresponding textual content in that language.

Files changed (1) hide show
  1. README.md +49 -29
README.md CHANGED
@@ -1,29 +1,49 @@
1
- ---
2
- license: cc-by-4.0
3
- dataset_info:
4
- features:
5
- - name: n_id
6
- dtype: string
7
- - name: doc_id
8
- dtype: string
9
- - name: lang
10
- dtype: string
11
- - name: text
12
- dtype: string
13
- splits:
14
- - name: aligned
15
- num_bytes: 10276115509
16
- num_examples: 1566404
17
- - name: unaligned
18
- num_bytes: 4047766580
19
- num_examples: 634142
20
- download_size: 5548830328
21
- dataset_size: 14323882089
22
- configs:
23
- - config_name: default
24
- data_files:
25
- - split: aligned
26
- path: data/aligned-*
27
- - split: unaligned
28
- path: data/unaligned-*
29
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-4.0
3
+ dataset_info:
4
+ features:
5
+ - name: n_id
6
+ dtype: string
7
+ - name: doc_id
8
+ dtype: string
9
+ - name: lang
10
+ dtype: string
11
+ - name: text
12
+ dtype: string
13
+ splits:
14
+ - name: aligned
15
+ num_bytes: 10276115509
16
+ num_examples: 1566404
17
+ - name: unaligned
18
+ num_bytes: 4047766580
19
+ num_examples: 634142
20
+ download_size: 5548830328
21
+ dataset_size: 14323882089
22
+ configs:
23
+ - config_name: default
24
+ data_files:
25
+ - split: aligned
26
+ path: data/aligned-*
27
+ - split: unaligned
28
+ path: data/unaligned-*
29
+ language:
30
+ - bn
31
+ - en
32
+ - gu
33
+ - hi
34
+ - kn
35
+ - ml
36
+ - mr
37
+ - or
38
+ - pa
39
+ - ta
40
+ - te
41
+ - ur
42
+ tags:
43
+ - data-mining
44
+ - parallel-corpus
45
+ - document-alignment
46
+ pretty_name: Pralekha
47
+ size_categories:
48
+ - 1M<n<10M
49
+ ---