Lexical patterns, features and knowledge resources for coreference resolution in clinical notes

Abstract

Generation of entity coreference chains provides a means to extract linked narrative events from clinical notes, but despite being a well-researched topic in natural language processing, general-purpose coreference tools perform poorly on clinical texts. This paper presents a knowledge-centric and pattern-based approach to resolving coreference across a wide variety of clinical records from two corpora (Ontology Development and Information Extraction (ODIE) and i2b2/VA), and describes a method for generating coreference chains using progressively pruned linked lists that reduces the search space and facilitates evaluation by a number of metrics. Independent evaluation results give an F-measure for each corpus of 79.2% and 87.5%, respectively. A baseline of blind coreference of mentions of the same class gives F-measures of 65.3% and 51.9% respectively. For the ODIE corpus, recall is significantly improved over the baseline (p<0.05) but overall there was no statistically significant improvement in F-measure (p>0.05). For the i2b2/VA corpus, recall, precision, and F-measure are significantly improved over the baseline (p<0.05). Overall, our approach offers performance at least as good as human annotators and greatly increased performance over general-purpose tools. The system uses a number of open-source components that are available to download.

Natural Language Analysis of Online Health Forums

“…Most of the existing work concentrates on processing a single sentence. However, our method is a formal approach that uses information from the whole post, which may contain several sentences, by making use of anaphoric relations [6] present in the sentences. At this stage we do not identify dosages or temporal information present in the text; these will be tackled in future work.…”

“…Disambiguation of semantic relations between two expressions (sentences in our case) is known as anaphora resolution, where a later expression (the anaphor) has some semantic relation to an earlier expression (the antecedent). The rule-based system described in [6] is a knowledge-centric and pattern-based approach for disambiguating anaphoric references in clinical records. For our work, we took a slightly different approach by considering protagonist theory [6], which suggests that narrative events are centered on one or more key actors.…”

“…The rule-based system described in [6] is a knowledge-centric and pattern-based approach for disambiguating anaphoric references in clinical records. For our work, we took a slightly different approach by considering protagonist theory [6], which suggests that narrative events are centered on one or more key actors. Our analysis of the posts reveals that a drug mention found in a sentence is often referred to in subsequent sentences.…”

A comparative study of evolving fuzzy grammar and machine learning techniques for text categorization

“…A major difficulty with the automated segmentation of unstructured text is that the information is designed to be read by humans that are able to extract the relevant attributes, even when it is not presented in a complete uniform fashion. This problem was approached by a content analysis study for text expression representation, such as syntactic and semantic representation (Gooch and Roudsari 2012) and lexical text pattern (Zamil et al 2011). This was used to identify text expressions, such as resembling a cognitive situation (Guo et al 2010), suicidal classification based on suicide notes (Pestian et al 2010), and disease outbreak, based on tweet analysis (Chew and Eysenbach 2009;Lampos and Cristianini 2010).…”