DNorm: disease name normalization with pairwise learning to rank

Abstract

Motivation: Despite the central role of diseases in biomedical research, there have been much fewer attempts to automatically determine which diseases are mentioned in a text—the task of disease name normalization (DNorm)—compared with other normalization tasks in biomedical text mining research.Methods: In this article we introduce the first machine learning approach for DNorm, using the NCBI disease corpus and the MEDIC vocabulary, which combines MeSH® and OMIM. Our method is a high-performing and mathematically principled framework for learning similarities between mentions and concept names directly from training data. The technique is based on pairwise learning to rank, which has not previously been applied to the normalization task but has proven successful in large optimization problems for information retrieval.Results: We compare our method with several techniques based on lexical normalization and matching, MetaMap and Lucene. Our algorithm achieves 0.782 micro-averaged F-measure and 0.809 macro-averaged F-measure, an increase over the highest performing baseline method of 0.121 and 0.098, respectively.Availability: The source code for DNorm is available at http://www.ncbi.nlm.nih.gov/CBBresearch/Lu/Demo/DNorm, along with a web-based demonstration and links to the NCBI disease corpus. Results on PubMed abstracts are available in PubTator: http://www.ncbi.nlm.nih.gov/CBBresearch/Lu/Demo/PubTatorContact: zhiyong.lu@nih.gov

Joint Entity Recognition and Linking in Technical Domains Using Undirected Probabilistic Graphical Models

“…We have shown that our approach can be easily applied to different domains by merely exchanging the underlying ontology and training data. On the task of recognizing and linking disease names, we show that our approach outperforms the state-of-the-art systems DNorm [12] and TaggerOne [11], as well as two lexicon-based baselines. On the task of recognizing and linking chemical names, our system achieves comparable performance to the state-of-the-art.…”

“…We apply the same model to both problems, only exchanging the underlying reference knowledge base. With an F 1 score of 85.9 in disease linking, we outperform the state-of-the-art systems DNorm [12] and TaggerOne [11]; in chemical compounds linking, our system achieves an F 1 score of 86.6, which is comparable to the stateof-the-art. Thus, J-Link provides high performance on both domains without major need of manual adaptation or system tuning.…”

“…We focus our discussion on the biomedical domain and disease/chemical recognition and linking, as this is our application scenario in this paper. The DNorm system [12] relies on a learning-to-rank approach in order to induce similarities between disease mentions and concept names directly from training data. However, the system does not include any information about coherence between different entities within the same text.…”

“…We compare our approach to the two state-of-the-art systems DNorm [12] and TaggerOne [11], as well as against two simple baselines (LMB and LMB + ). The latter baselines are based on non-overlapping longest matches, using the dictionary as described in Section 3.5.…”

Identifying Scientific Project-generated Data Citation from Full-text Articles: An Investigation of TCGA Data Citation

“…For example, in the results section of one article (PMCID: PMC3910500), it described the genomic landscape of glioblastoma using the wholeexome (WES), whole-genome sequencing (WGS), and RNA-Sequencing (RNA) (Brennan et al, 2013). To identify the TCGA cancer type and high-throughput platform concept from the free texts, we developed a named entity recognition method that is based on a biomedical text mining tool (Leaman, Islamaj, & Lu, 2013).…”