Datasets:

Sub-tasks:
parsing
Languages:
Chinese
Multilinguality:
monolingual
Size Categories:
1K<n<10K
Language Creators:
found
Annotations Creators:
crowdsourced
Source Datasets:
original
Tags:
License:
albertvillanova HF staff commited on
Commit
53b5b42
1 Parent(s): 2c63d54

Convert dataset to Parquet

Browse files

Convert dataset to Parquet.

README.md CHANGED
@@ -19,6 +19,7 @@ task_ids:
19
  - parsing
20
  pretty_name: AMTTL
21
  dataset_info:
 
22
  features:
23
  - name: id
24
  dtype: string
@@ -32,19 +33,28 @@ dataset_info:
32
  '1': I
33
  '2': E
34
  '3': S
35
- config_name: amttl
36
  splits:
37
  - name: train
38
- num_bytes: 1132212
39
  num_examples: 3063
40
  - name: validation
41
- num_bytes: 324374
42
  num_examples: 822
43
  - name: test
44
- num_bytes: 328525
45
  num_examples: 908
46
- download_size: 685534
47
- dataset_size: 1785111
 
 
 
 
 
 
 
 
 
 
48
  ---
49
 
50
  # Dataset Card for AMTTL
 
19
  - parsing
20
  pretty_name: AMTTL
21
  dataset_info:
22
+ config_name: amttl
23
  features:
24
  - name: id
25
  dtype: string
 
33
  '1': I
34
  '2': E
35
  '3': S
 
36
  splits:
37
  - name: train
38
+ num_bytes: 1132196
39
  num_examples: 3063
40
  - name: validation
41
+ num_bytes: 324358
42
  num_examples: 822
43
  - name: test
44
+ num_bytes: 328509
45
  num_examples: 908
46
+ download_size: 274351
47
+ dataset_size: 1785063
48
+ configs:
49
+ - config_name: amttl
50
+ data_files:
51
+ - split: train
52
+ path: amttl/train-*
53
+ - split: validation
54
+ path: amttl/validation-*
55
+ - split: test
56
+ path: amttl/test-*
57
+ default: true
58
  ---
59
 
60
  # Dataset Card for AMTTL
amttl/test-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e162219c3d2e9a4b234407072169e58475c70f69a1118c4c92c1cc8bdb7fddcf
3
+ size 51311
amttl/train-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:93ff0e728fa5bf6cf4c32805ac01529c1b022f29b39f28406a5e7fd28b9b6342
3
+ size 172615
amttl/validation-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e7992b50bd6d87521937260ed7ebce5a986b8eb52ad0905373fe94d6b155c53e
3
+ size 50425
dataset_infos.json CHANGED
@@ -1 +1,65 @@
1
- {"amttl": {"description": "Chinese word segmentation (CWS) trained from open source corpus faces dramatic performance drop\nwhen dealing with domain text, especially for a domain with lots of special terms and diverse\nwriting styles, such as the biomedical domain. However, building domain-specific CWS requires\nextremely high annotation cost. In this paper, we propose an approach by exploiting domain-invariant\nknowledge from high resource to low resource domains. Extensive experiments show that our mode\nachieves consistently higher accuracy than the single-task CWS and other transfer learning\nbaselines, especially when there is a large disparity between source and target domains.\n\nThis dataset is the accompanied medical Chinese word segmentation (CWS) dataset.\nThe tags are in BIES scheme.\n\nFor more details see https://www.aclweb.org/anthology/C18-1307/\n", "citation": "@inproceedings{xing2018adaptive,\n title={Adaptive multi-task transfer learning for Chinese word segmentation in medical text},\n author={Xing, Junjie and Zhu, Kenny and Zhang, Shaodian},\n booktitle={Proceedings of the 27th International Conference on Computational Linguistics},\n pages={3619--3630},\n year={2018}\n}\n", "homepage": "https://www.aclweb.org/anthology/C18-1307/", "license": "", "features": {"id": {"dtype": "string", "id": null, "_type": "Value"}, "tokens": {"feature": {"dtype": "string", "id": null, "_type": "Value"}, "length": -1, "id": null, "_type": "Sequence"}, "tags": {"feature": {"num_classes": 4, "names": ["B", "I", "E", "S"], "names_file": null, "id": null, "_type": "ClassLabel"}, "length": -1, "id": null, "_type": "Sequence"}}, "post_processed": null, "supervised_keys": null, "builder_name": "amttl", "config_name": "amttl", "version": {"version_str": "1.0.0", "description": null, "major": 1, "minor": 0, "patch": 0}, "splits": {"train": {"name": "train", "num_bytes": 1132212, "num_examples": 3063, "dataset_name": "amttl"}, "validation": {"name": "validation", "num_bytes": 324374, "num_examples": 822, "dataset_name": "amttl"}, "test": {"name": "test", "num_bytes": 328525, "num_examples": 908, "dataset_name": "amttl"}}, "download_checksums": {"https://raw.githubusercontent.com/adapt-sjtu/AMTTL/master/medical_data/forum_train.txt": {"num_bytes": 434357, "checksum": "9819373963ea04d1d28844d5bc83b6b0332fad8b5f2e73092bcfc58dc6d6292a"}, "https://raw.githubusercontent.com/adapt-sjtu/AMTTL/master/medical_data/forum_dev.txt": {"num_bytes": 124973, "checksum": "1a2eb461b98d2a9160baad7f76d003cc0917b998e8283bcffa52b71224dd9d17"}, "https://raw.githubusercontent.com/adapt-sjtu/AMTTL/master/medical_data/forum_test.txt": {"num_bytes": 126204, "checksum": "aea1a8cf244cd565e94bd193a1eef7a10b16eeb0b6fbb6ed1d2fefbd55360dd6"}}, "download_size": 685534, "post_processing_size": null, "dataset_size": 1785111, "size_in_bytes": 2470645}}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "amttl": {
3
+ "description": "Chinese word segmentation (CWS) trained from open source corpus faces dramatic performance drop\nwhen dealing with domain text, especially for a domain with lots of special terms and diverse\nwriting styles, such as the biomedical domain. However, building domain-specific CWS requires\nextremely high annotation cost. In this paper, we propose an approach by exploiting domain-invariant\nknowledge from high resource to low resource domains. Extensive experiments show that our mode\nachieves consistently higher accuracy than the single-task CWS and other transfer learning\nbaselines, especially when there is a large disparity between source and target domains.\n\nThis dataset is the accompanied medical Chinese word segmentation (CWS) dataset.\nThe tags are in BIES scheme.\n\nFor more details see https://www.aclweb.org/anthology/C18-1307/\n",
4
+ "citation": "@inproceedings{xing2018adaptive,\n title={Adaptive multi-task transfer learning for Chinese word segmentation in medical text},\n author={Xing, Junjie and Zhu, Kenny and Zhang, Shaodian},\n booktitle={Proceedings of the 27th International Conference on Computational Linguistics},\n pages={3619--3630},\n year={2018}\n}\n",
5
+ "homepage": "https://www.aclweb.org/anthology/C18-1307/",
6
+ "license": "",
7
+ "features": {
8
+ "id": {
9
+ "dtype": "string",
10
+ "_type": "Value"
11
+ },
12
+ "tokens": {
13
+ "feature": {
14
+ "dtype": "string",
15
+ "_type": "Value"
16
+ },
17
+ "_type": "Sequence"
18
+ },
19
+ "tags": {
20
+ "feature": {
21
+ "names": [
22
+ "B",
23
+ "I",
24
+ "E",
25
+ "S"
26
+ ],
27
+ "_type": "ClassLabel"
28
+ },
29
+ "_type": "Sequence"
30
+ }
31
+ },
32
+ "builder_name": "parquet",
33
+ "dataset_name": "amttl",
34
+ "config_name": "amttl",
35
+ "version": {
36
+ "version_str": "1.0.0",
37
+ "major": 1,
38
+ "minor": 0,
39
+ "patch": 0
40
+ },
41
+ "splits": {
42
+ "train": {
43
+ "name": "train",
44
+ "num_bytes": 1132196,
45
+ "num_examples": 3063,
46
+ "dataset_name": null
47
+ },
48
+ "validation": {
49
+ "name": "validation",
50
+ "num_bytes": 324358,
51
+ "num_examples": 822,
52
+ "dataset_name": null
53
+ },
54
+ "test": {
55
+ "name": "test",
56
+ "num_bytes": 328509,
57
+ "num_examples": 908,
58
+ "dataset_name": null
59
+ }
60
+ },
61
+ "download_size": 274351,
62
+ "dataset_size": 1785063,
63
+ "size_in_bytes": 2059414
64
+ }
65
+ }