tkon3 commited on
Commit
99efcf8
1 Parent(s): bf96976

add readme

Browse files
Files changed (2) hide show
  1. README.md +40 -0
  2. patent-classification.py +1 -1
README.md ADDED
@@ -0,0 +1,40 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ languages: en
3
+ task_categories: text-classification
4
+ tags:
5
+ - long context
6
+ task_ids:
7
+ - multi-class-classification
8
+ - topic-classification
9
+ size_categories: 10K<n<100K
10
+ ---
11
+
12
+ **Patent Classification: a classification of Patents and abstracts (9 classes).**
13
+
14
+ This dataset is intended for long context classification (non abstract documents are longer that 512 tokens). \
15
+ Data are sampled from "BIGPATENT: A Large-Scale Dataset for Abstractive and Coherent Summarization." by Eva Sharma, Chen Li and Lu Wang
16
+ * See: https://aclanthology.org/P19-1212.pdf
17
+ * See: https://evasharma.github.io/bigpatent/
18
+
19
+ It contains 11 slightly unbalanced classes, 35k Patents and abstracts divided into 3 splits: train (25k), val (5k) and test (5k).
20
+
21
+ **Note that documents are uncased and space separated (by authors)**
22
+
23
+ Compatible with [run_glue.py](https://github.com/huggingface/transformers/tree/master/examples/pytorch/text-classification) script:
24
+ ```
25
+ export MODEL_NAME=roberta-base
26
+ export MAX_SEQ_LENGTH=512
27
+
28
+ python run_glue.py \
29
+ --model_name_or_path $MODEL_NAME \
30
+ --dataset_name ccdv/patent-classification \
31
+ --do_train \
32
+ --do_eval \
33
+ --max_seq_length $MAX_SEQ_LENGTH \
34
+ --per_device_train_batch_size 8 \
35
+ --gradient_accumulation_steps 4 \
36
+ --learning_rate 2e-5 \
37
+ --num_train_epochs 1 \
38
+ --max_eval_samples 500 \
39
+ --output_dir tmp/patent
40
+ ```
patent-classification.py CHANGED
@@ -9,7 +9,7 @@ _CITATION = None
9
 
10
  _DESCRIPTION = """
11
  Patent Classification Dataset: a classification of Patents (9 classes).
12
- It contains 11 slightly unbalanced classes, 25k Patents and summaries divided into 3 splits: train (25k), val (5k) and test (5k).
13
  Copied from "Long Document Classification From Local Word Glimpses via Recurrent Attention Learning" by JUN HE LIQUN WANG LIU LIU, JIAO FENG AND HAO WU
14
  See: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8675939
15
  See: https://github.com/LiqunW/Long-document-dataset
 
9
 
10
  _DESCRIPTION = """
11
  Patent Classification Dataset: a classification of Patents (9 classes).
12
+ It contains 9 unbalanced classes, 25k Patents and summaries divided into 3 splits: train (25k), val (5k) and test (5k).
13
  Copied from "Long Document Classification From Local Word Glimpses via Recurrent Attention Learning" by JUN HE LIQUN WANG LIU LIU, JIAO FENG AND HAO WU
14
  See: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8675939
15
  See: https://github.com/LiqunW/Long-document-dataset