ccdv/patent-classification

1---2language: en3size_categories: 10K<n<100K4task_categories:5- text-classification6task_ids:7- multi-class-classification8- topic-classification9tags:10- long context11dataset_info:12- config_name: abstract13  features:14  - name: text15    dtype: string16  - name: label17    dtype:18      class_label:19        names:20          '0': Human Necessities21          '1': Performing Operations; Transporting22          '2': Chemistry; Metallurgy23          '3': Textiles; Paper24          '4': Fixed Constructions25          '5': Mechanical Engineering; Lightning; Heating; Weapons; Blasting26          '6': Physics27          '7': Electricity28          '8': General tagging of new or cross-sectional technology29  splits:30  - name: train31    num_bytes: 1722510132    num_examples: 2500033  - name: validation34    num_bytes: 347285435    num_examples: 500036  - name: test37    num_bytes: 345673338    num_examples: 500039  download_size: 1206795340  dataset_size: 2415468841- config_name: patent42  features:43  - name: text44    dtype: string45  - name: label46    dtype:47      class_label:48        names:49          '0': Human Necessities50          '1': Performing Operations; Transporting51          '2': Chemistry; Metallurgy52          '3': Textiles; Paper53          '4': Fixed Constructions54          '5': Mechanical Engineering; Lightning; Heating; Weapons; Blasting55          '6': Physics56          '7': Electricity57          '8': General tagging of new or cross-sectional technology58  splits:59  - name: train60    num_bytes: 46678862561    num_examples: 2500062  - name: validation63    num_bytes: 9531510764    num_examples: 500065  - name: test66    num_bytes: 9384486967    num_examples: 500068  download_size: 27296625169  dataset_size: 65594860170configs:71- config_name: abstract72  data_files:73  - split: train74    path: abstract/train-*75  - split: validation76    path: abstract/validation-*77  - split: test78    path: abstract/test-*79- config_name: patent80  data_files:81  - split: train82    path: patent/train-*83  - split: validation84    path: patent/validation-*85  - split: test86    path: patent/test-*87  default: true88---89 90**Patent Classification: a classification of Patents and abstracts (9 classes).** 91 92This dataset is intended for long context classification (non abstract documents are longer that 512 tokens). \93Data are sampled from "BIGPATENT: A Large-Scale Dataset for Abstractive and Coherent Summarization." by Eva Sharma, Chen Li and Lu Wang 94 * See: https://aclanthology.org/P19-1212.pdf 95 * See: https://evasharma.github.io/bigpatent/96 97It contains 9 unbalanced classes, 35k Patents and abstracts divided into 3 splits: train (25k), val (5k) and test (5k). 98 99**Note that documents are uncased and space separated (by authors)**100 101Compatible with [run_glue.py](https://github.com/huggingface/transformers/tree/master/examples/pytorch/text-classification) script:102```103export MODEL_NAME=roberta-base104export MAX_SEQ_LENGTH=512105 106python run_glue.py \107  --model_name_or_path $MODEL_NAME \108  --dataset_name ccdv/patent-classification  \109  --do_train \110  --do_eval \111  --max_seq_length $MAX_SEQ_LENGTH \112  --per_device_train_batch_size 8 \113  --gradient_accumulation_steps 4 \114  --learning_rate 2e-5 \115  --num_train_epochs 1 \116  --max_eval_samples 500 \117  --output_dir tmp/patent118```