sdyz5210 / ihp

Identification of Human Phenotype Entities

IHP

Framework for identifying Human Phenotype entities

Dependencies and other uses should follow the original ReadMe.

This is a fork created to accomodate an annotator for the Human Phenotype Ontology. It uses Gold Standard Corpora and Test Suites Created by Bio-Lark. Link Here

Usage

If a corpus is to be loaded into IHP, it's necessary to run Stanford CoreNLP.

cd bin/stanford-corenlp-full-2015-12-09/
java -mx4g -cp "*" edu.stanford.nlp.pipeline.StanfordCoreNLPServer -timeout 500000 &

Load Corpus (For both Gold Standard Corpora and Test Suite)

   python src/main.py load_corpus --goldstd hpo_train --log DEBUG
   python src/main.py load_corpus --goldstd hpo_test --log DEBUG
   python src/main.py load_corpus --goldstd tsuite --log DEBUG

Train, Test and Evaluate with StanfordNER

   python src/main.py train --goldstd hpo_train --models models/hpo_train --log DEBUG
   python src/main.py test --goldstd hpo_test -o pickle data/results_hpo_train --models models/hpo_train --log DEBUG
   python src/evaluate.py evaluate hpo_test --results data/results_hpo_train --models models/hpo_train --log DEBUG

Train, Test and Evaluate with CRFSuite

   python src/main.py train --goldstd hpo_train --models models/hpo_train --log DEBUG --entitytype hpo --crf crfsuite
   python src/main.py test --goldstd hpo_test -o pickle data/results_hpo_train --models models/hpo_train --log DEBUG --entitytype hpo --crf crfsuite
   python src/evaluate.py evaluate hpo_test --results data/results_hpo_train --models models/hpo_train --log DEBUG --entitytype hpo

Test and Evaluate for Test Suites

   python src/main.py test --goldstd tsuite -o pickle data/results_hpo_train --models models/hpo_train --log DEBUG --entitytype hpo --crf crfsuite
   python src/evaluate.py evaluate tsuite --results data/results_hpo_train --models models/hpo_train --log DEBUG --entitytype hpo 

Rules can be added to the evaluation parameters:

   --rules andor stopwords small_ent twice_validated stopwords gowords posgowords longterms small_len quotes defwords digits lastwords

FAQ

How to run IHP in new, unlabeled, unstructured text?

Replace the sample corpus in corpora/hpo/test_corpus/ by the new, unlabeled, unstructured text and delete the content of corpora/hpo/test_ann/. Then run:

    python src/main.py load_corpus --goldstd hpo_test --log DEBUG
    python src/main.py test --goldstd hpo_test -o pickle data/results_hpo_train --models models/hpo_train --log DEBUG
    python src/evaluate.py evaluate hpo_test --results data/results_hpo_train --models models/hpo_train --log DEBUG

The report file in data/results_hpo_train_report.txt will have the generated annotations marked as false positives (because no annotation file was provided).

References:

• M. Lobo, A. Lamurias, and F. Couto, “Identifying human phenotype terms by combining machine learning and validation rules,” BioMed Research International, vol. 2017, pp. 1--14, 2017 (https://doi.org/10.1155/2017/8565739)