A program to recognize self-acknowledged limitation sentences in biomedical articles

The repository contains the source code for the system described in the article Automatic recognition of self-acknowledged limitations in clinical research literature. The best performing rule-based system is presented (gov.nih.nlm.limitations.RuleBasedLimitationSentenceRecognizer), as well as the rule-based baseline (gov.nih.nlm.limitations.RuleBasedLimitationSentenceRecognizerBaseline).

To replicate the results, run gov.nih.nlm.limitations.RuleBasedLimitationSentenceRecognizer with three arguments:

• DATA/XML: directory that contains the parsed XML of the test set
• DATA/limitation_sentences_final.txt: gold annotations
• Output file name (after the run, this file should match DATA/rule_based_test.out.txt)

The parsed XML is generated from PubMed Central XML using gov.nih.nlm.limitations.CorpusParser.

Stanford CoreNLP model jar file that is needed for processing raw text for lexical and syntactic information (stanford-corenlp-3.3.1-models.jar) is not included with the distribution due to its size. It can be downloaded from https://stanfordnlp.github.io/CoreNLP/ and copied to lib directory.

• Halil Kilicoglu: [email protected]