Add instructions for GENIA.

conf/genia.json

@@ -0,0 +1,39 @@
+{
+    "run_name": "base-run",
+    "dataset_name": "GENIA",
+    "model_name_or_path": "microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract",
+    "train_file": "data/genia/train.json",
+    "validation_file": "data/genia/dev.json",
+    "test_file": "data/genia/test.json",
+    "entity_type_file": "entity_types.json",
+    "dataset_entity_types": ["DNA", "RNA", "protein", "cell_line", "cell_type"],
+    "report_to": "none",
+    "max_seq_length": 128,
+    "use_span_width_embedding": true,
+    "doc_stride": 16,
+    "preprocessing_num_workers": 6,
+    "dataloader_num_workers": 6,
+    "overwrite_cache": false,
+    "overwrite_output_dir": true,
+    "do_train": true,
+    "fp16": true,
+    "gradient_checkpointing": true,
+    "optim": "adamw_torch",
+    "label_names": ["ner"],
+    "num_train_epochs": 20,
+    "per_device_train_batch_size": 8,
+    "gradient_accumulation_steps": 2,
+    "learning_rate": 3e-5,
+    "logging_steps": 50,
+    "load_best_model_at_end": true,
+    "metric_for_best_model": "f1",
+    "greater_is_better": true,
+    "save_strategy": "steps",
+    "save_steps": 100,
+    "save_total_limit": 1,
+    "do_eval": true,
+    "evaluation_strategy": "steps",
+    "eval_steps": 100,
+    "do_predict": true,
+    "output_dir": "/tmp/genia"
+}

data_preproc/README.md

@@ -15,6 +15,7 @@ The data preprocessing code is adapted from [DyGIE](https://github.com/luanyi/Dy
 * ACE 2004 (https://catalog.ldc.upenn.edu/LDC2005T09)
 * ACE 2005 (https://catalog.ldc.upenn.edu/LDC2006T06)
 * CoNLL 2003 (We use the preprocessed version from [Yu et al., 2020](https://github.com/juntaoy/biaffine-ner/issues/16))
+* GENIA (We use the preprocessed version from [Yu et al., 2020](https://github.com/juntaoy/biaffine-ner/issues/17))
 
 # Usage
 
@@ -74,4 +75,16 @@ python convert_to_hf_ds_format.py conll2003/dev.json ${CoNLL2003}/dev.json --tas
 python convert_to_hf_ds_format.py conll2003/test.json ${CoNLL2003}/test.json --task conll2003
 ```
 
+### GENIA
+```bash
+cd genia
+split -l 1599 train_dev.genia.jsonlines train_dev.split.
+cd ..
+mkdir -p ../data/genia
+python convert_to_hf_ds_format.py genia/train_dev.split.aa ../data/genia/train.json --task conll2003
+python convert_to_hf_ds_format.py genia/train_dev.split.ab ../data/genia/dev.json --task conll2003
+python convert_to_hf_ds_format.py genia/test.genia.jsonlines ../data/genia/test.json --task conll2003
+```
+
+
 If you want to use other datasets, please convert them into the same format as above.

entity_types.json

@@ -8,4 +8,9 @@
 {"dataset": "CoNLL2003", "name": "PER", "description": "a person entity is limited to human including a single individual or a group.", "description_source": "ACE2005 annotation guidelines"}
 {"dataset": "CoNLL2003", "name": "ORG", "description": "organization entities are limited to companies, corporations, agencies, institutions and other groups of people.", "description_source": "ACE2005 annotation guidelines"}
 {"dataset": "CoNLL2003", "name": "LOC", "description": "location entities are limited to geographical entities such as geographical areas and landmasses, mountains, bodies of water, and geological formations.", "description_source": "ACE2005 annotation guidelines"}
-{"dataset": "CoNLL2003", "name": "MISC", "description": "examples of miscellaneous entities include events, nationalities, products and works of art.", "description_source": "MRC4NER Github repo"}
+{"dataset": "CoNLL2003", "name": "MISC", "description": "examples of miscellaneous entities include events, nationalities, products and works of art.", "description_source": "MRC4NER Github repo"}
+{"dataset": "GENIA", "name": "DNA", "description": "DNA, which consists of a polysugar-phosphate backbone possessing projections of purines and pyrimidines, forms a double helix that is held together by hydrogen bonds between these purines and pyrimidines.", "description_source": "UMLS definition"}
+{"dataset": "GENIA", "name": "RNA", "description": "RNA is a polynucleotide consisting essentially of chains with a repeating backbone of phosphate and ribose units to which nitrogenous bases are attached.", "description_source": "UMLS definition"}
+{"dataset": "GENIA", "name": "protein", "description": "Protein is Linear POLYPEPTIDES that are synthesized on RIBOSOMES and may be further modified, crosslinked, cleaved, or assembled into complex proteins with several subunits. The specific sequence of AMINO ACIDS determines the shape the polypeptide will take, during PROTEIN FOLDING, and the function of the protein.", "description_source": "UMLS definition"}
+{"dataset": "GENIA", "name": "cell_line", "description": "Cell line are cells propagated in vitro in special media conducive to their growth. Cultured cells are used to study developmental, morphologic, metabolic, physiologic, and genetic processes, among others.", "description_source": "UMLS definition"}
+{"dataset": "GENIA", "name": "cell_type", "description": "Cells are the fundamental, structural, and functional units or subunits of living organisms. They are composed of CYTOPLASM containing various ORGANELLES and a CELL MEMBRANE boundary.", "description_source": "UMLS definition"}