forked from NVIDIA/DeepLearningExamples
-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Merge pull request NVIDIA#644 from swethmandava/master
Bert tf update (triton v2, fixes)
- Loading branch information
Showing
17 changed files
with
619 additions
and
149 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
169 changes: 169 additions & 0 deletions
169
TensorFlow/LanguageModeling/BERT/data/ChemProtTextFormatting.py
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,169 @@ | ||
| # Copyright (c) 2019 NVIDIA CORPORATION. All rights reserved. | ||
| # Licensed under the Apache License, Version 2.0 (the "License"); | ||
| # you may not use this file except in compliance with the License. | ||
| # You may obtain a copy of the License at | ||
| # | ||
| # http://www.apache.org/licenses/LICENSE-2.0 | ||
| # | ||
| # Unless required by applicable law or agreed to in writing, software | ||
| # distributed under the License is distributed on an "AS IS" BASIS, | ||
| # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| # See the License for the specific language governing permissions and | ||
| # limitations under the License. | ||
|
|
||
| import os | ||
| import csv | ||
| import zipfile | ||
| import argparse | ||
| import re | ||
|
|
||
| class ChemProtTextFormatting: | ||
| """A basic formatter to preprocess the chemprot dataset. | ||
| """ | ||
|
|
||
| def __init__(self, input_folder, output_folder): | ||
|
|
||
| chemprot_folder = input_folder | ||
| with zipfile.ZipFile(os.path.join(chemprot_folder, "ChemProt_Corpus.zip"), "r") as zip: | ||
| zip.extractall(chemprot_folder) | ||
|
|
||
| chemprot_folder = os.path.join(input_folder, "ChemProt_Corpus") | ||
|
|
||
| with zipfile.ZipFile(os.path.join(chemprot_folder, "chemprot_development.zip")) as zip: | ||
| zip.extractall(chemprot_folder) | ||
|
|
||
| if not os.path.exists(output_folder): | ||
| os.makedirs(output_folder) | ||
|
|
||
| self.format(os.path.join(chemprot_folder, "chemprot_development"), | ||
| "chemprot_development_entities.tsv", "chemprot_development_relations.tsv", | ||
| "chemprot_development_abstracts.tsv", os.path.join(output_folder, "dev.tsv")) | ||
|
|
||
| with zipfile.ZipFile(os.path.join(chemprot_folder, "chemprot_test_gs.zip")) as zip: | ||
| zip.extractall(chemprot_folder) | ||
| self.format(os.path.join(chemprot_folder, "chemprot_test_gs"), | ||
| "chemprot_test_entities_gs.tsv", "chemprot_test_relations_gs.tsv", | ||
| "chemprot_test_abstracts_gs.tsv", os.path.join(output_folder, "test.tsv")) | ||
|
|
||
| with zipfile.ZipFile(os.path.join(chemprot_folder, "chemprot_training.zip")) as zip: | ||
| zip.extractall(chemprot_folder) | ||
| self.format(os.path.join(chemprot_folder, "chemprot_training"), | ||
| "chemprot_training_entities.tsv", "chemprot_training_relations.tsv", | ||
| "chemprot_training_abstracts.tsv", os.path.join(output_folder, "train.tsv")) | ||
|
|
||
|
|
||
|
|
||
| def format(self, chemprot_path, entity_filename, relations_filename, abstracts_filename, output_filename): | ||
| """ | ||
| Constructs ChemProt dataset for Relation Extraction. | ||
| Args: | ||
| chemprot_path: Path to files | ||
| entity_filename: Contains labelled mention annotations of chemical compounds and genes/proteins. | ||
| <PMID> <EntityNumber> <Type of Entity> <Start Character offset> <End Character Offset> <Text String> | ||
| relations_filename: Contains a subset of chemical-protein relations annotations for the Chemprot dataset | ||
| <PMID> <CPR Group> <EntityNumber1> <EntityNumber2> | ||
| abstracts_filename: Contains plain text CHEMPROT PubMed Data | ||
| <PMID> <Title of the Article> <Abstract of the Article> | ||
| output_filename: Path to output file that will contain preprocessed data | ||
| <PMID.EntityNumber1.EntityNumber2> <Preprocessed Sentence> <CPR Group> | ||
| """ | ||
|
|
||
| data = {} | ||
| train_entities = csv.reader(open(os.path.join(chemprot_path, entity_filename), | ||
| mode="r"), delimiter="\t") | ||
| for entity in train_entities: | ||
| id = entity[0] | ||
| if data.get(id, None) is None: | ||
| data[id] = {"relations":{}, "entities":{"CHEMICAL":{}, "GENE":{}}} | ||
| data[id]["entities"]["CHEMICAL" if entity[2] == "CHEMICAL" else "GENE"][entity[1]] = (int(entity[3]), int(entity[4]), entity[2]) | ||
|
|
||
| train_relations=csv.reader(open(os.path.join(chemprot_path, relations_filename), | ||
| mode="r"), delimiter="\t") | ||
| for relation in train_relations: | ||
| try: | ||
| id = relation[0] | ||
| data[id]["relations"][(relation[4].split("Arg1:")[-1], relation[5].split("Arg2:")[-1])] = relation[1] if relation[2] == "Y " else "false" | ||
| except: | ||
| print("invalid id") | ||
| raise ValueError | ||
| # print(data[list(data.keys())[0]]) | ||
|
|
||
| with open(output_filename, 'w') as ofile: | ||
| train_abstracts = csv.reader(open(os.path.join(chemprot_path, abstracts_filename), | ||
| mode="r"), delimiter="\t") | ||
| owriter = csv.writer(ofile, delimiter='\t', lineterminator=os.linesep) | ||
| owriter.writerow(["index", "sentence", "label"]) | ||
|
|
||
| num_sentences = 0 | ||
| rejected = 0 | ||
| for abstract in train_abstracts: | ||
| id = abstract[0] | ||
| line = abstract[1] + "\n" + abstract[2] | ||
|
|
||
| for tag1 in data[id]["entities"]["CHEMICAL"].keys(): | ||
| for tag2 in data[id]["entities"]["GENE"].keys(): | ||
| tag1_details = data[id]["entities"]["CHEMICAL"][tag1] | ||
| tag2_details = data[id]["entities"]["GENE"][tag2] | ||
| if ((tag1_details[0] <= tag2_details[0] and tag2_details[0] <= tag1_details[1]) # x1 <= y1 <= x2 | ||
| or (tag1_details[0] <= tag2_details[1] and tag2_details[0] <= tag1_details[1])): # x1 <= y2 <= x2 | ||
| continue | ||
|
|
||
| relation = data[id]["relations"].get((tag2, tag1), None) | ||
| relation = data[id]["relations"].get((tag1, tag2), None) if relation is None else relation | ||
| if relation is None: | ||
| relation = "false" | ||
|
|
||
| start = 0 | ||
| line_protected = re.sub(r"(.)\.(?=[\d])", r"\1[PROTECTED_DOT]", line) | ||
| for sentence in re.split(r'\.|\?', line_protected): | ||
| sentence = sentence.replace("[PROTECTED_DOT]", ".") | ||
| original_sentence = sentence | ||
| end = start + len(sentence) | ||
|
|
||
| if (tag1_details[0] >= start and tag1_details[1] <= end) and \ | ||
| (tag2_details[0] >= start and tag2_details[1] <= end): | ||
| for offset_start, offset_end, value in sorted(list(data[id]["entities"]["CHEMICAL"].values()) + list(data[id]["entities"]["GENE"].values()), | ||
| reverse=True): | ||
| if (offset_start, offset_end) == (tag1_details[0], tag1_details[1]) or (offset_start, offset_end) == (tag2_details[0], tag2_details[1]): | ||
| if sentence[offset_start - start] == "@": | ||
| offset_end = start + sentence.find('$',offset_start - start) + 1 | ||
| word = value | ||
| elif offset_start < start or offset_end > end or sentence[offset_start - start] == "@": | ||
| continue | ||
| else: | ||
| word = "OTHER" | ||
| sentence = sentence[:offset_start-start] + "@" + word + "$" + sentence[offset_end-start:] | ||
| sentence = sentence.strip() | ||
| owriter.writerow([id+"."+tag1+"."+tag2, sentence, relation]) | ||
| num_sentences += 1 | ||
| if id == "23538201" and start == 1048: | ||
| print("Accepted", tag1, tag2) | ||
|
|
||
| else: | ||
| rejected += 1 | ||
|
|
||
| start = end + 1 | ||
| print("Succesfully written {} samples to {}".format(num_sentences, output_filename)) | ||
| print("Rejected are", rejected) | ||
|
|
||
|
|
||
| if __name__=="__main__": | ||
| parser = argparse.ArgumentParser( | ||
| description='Preprocessing Application for ChemProt' | ||
| ) | ||
|
|
||
| parser.add_argument( | ||
| '--input_folder', | ||
| type=str, | ||
| help='Specify the input files in a comma-separated list (no spaces)' | ||
| ) | ||
| parser.add_argument( | ||
| '--output_folder', | ||
| type=str, | ||
| help='Specify the input files in a comma-separated list (no spaces)' | ||
| ) | ||
|
|
||
|
|
||
| args = parser.parse_args() | ||
| preprocess_chemprot = ChemProtTextFormatting(args.input_folder, args.output_folder) |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.