Building a high-quality sense inventory for improved abbreviation disambiguation

Thalia: semantic search engine for biomedical abstracts

SUMMARY

Although the publication rate of the biomedical literature has been growing steadily during the last decades, the accessibility of pertinent research publications for biologist and medical practitioners remains a challenge. This article describes Thalia, which is a semantic search engine that can recognize eight different types of concepts occurring in biomedical abstracts. Thalia is available via a web-based interface or a RESTful API. A key aspect of our search engine is that it is updated from PubMed on a daily basis. We describe here the main building blocks of our tool as well as an evaluation of the retrieval capabilities of Thalia in the context of a precision medicine dataset.

AVAILABILITY AND IMPLEMENTATION

Thalia is available at http://nactem.ac.uk/Thalia_BI/.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

A Text Mining Pipeline Using Active and Deep Learning Aimed at Curating Information in Computational Neuroscience

The curation of neuroscience entities is crucial to ongoing efforts in neuroinformatics and computational neuroscience, such as those being deployed in the context of continuing large-scale brain modelling projects. However, manually sifting through thousands of articles for new information about modelled entities is a painstaking and low-reward task. Text mining can be used to help a curator extract relevant information from this literature in a systematic way. We propose the application of text mining methods for the neuroscience literature. Specifically, two computational neuroscientists annotated a corpus of entities pertinent to neuroscience using active learning techniques to enable swift, targeted annotation. We then trained machine learning models to recognise the entities that have been identified. The entities covered are Neuron Types, Brain Regions, Experimental Values, Units, Ion Currents, Channels, and Conductances and Model organisms. We tested a traditional rule-based approach, a conditional random field and a model using deep learning named entity recognition, finding that the deep learning model was superior. Our final results show that we can detect a range of named entities of interest to the neuroscientist with a macro average precision, recall and F1 score of 0.866, 0.817 and 0.837 respectively. The contributions of this work are as follows: 1) We provide a set of Named Entity Recognition (NER) tools that are capable of detecting neuroscience entities with performance above or similar to prior work. 2) We propose a methodology for training NER tools for neuroscience that requires very little training data to get strong performance. This can be adapted for any sub-domain within neuroscience. 3) We provide a small corpus with annotations for multiple entity types, as well as annotation guidelines to help others reproduce our experiments.

A corpus of potentially contradictory research claims from cardiovascular research abstracts

Background

Research literature in biomedicine and related fields contains a huge number of claims, such as the effectiveness of treatments. These claims are not always consistent and may even contradict each other. Being able to identify contradictory claims is important for those who rely on the biomedical literature. Automated methods to identify and resolve them are required to cope with the amount of information available. However, research in this area has been hampered by a lack of suitable resources. We describe a methodology to develop a corpus which addresses this gap by providing examples of potentially contradictory claims and demonstrate how it can be applied to identify these claims from Medline abstracts related to the topic of cardiovascular disease.

Methods

A set of systematic reviews concerned with four topics in cardiovascular disease were identified from Medline and analysed to determine whether the abstracts they reviewed contained contradictory research claims. For each review, annotators were asked to analyse these abstracts to identify claims within them that answered the question addressed in the review. The annotators were also asked to indicate how the claim related to that question and the type of the claim.

Results

A total of 259 abstracts associated with 24 systematic reviews were used to form the corpus. Agreement between the annotators was high, suggesting that the information they provided is reliable.

Conclusions

The paper describes a methodology for constructing a corpus containing contradictory research claims from the biomedical literature. The corpus is made available to enable further research into this area and support the development of automated approaches to contradiction identification.

Beyond accuracy: creating interoperable and scalable text-mining web services

The biomedical literature is a knowledge-rich resource and an important foundation for future research. With over 24 million articles in PubMed and an increasing growth rate, research in automated text processing is becoming increasingly important. We report here our recently developed web-based text mining services for biomedical concept recognition and normalization. Unlike most text-mining software tools, our web services integrate several state-of-the-art entity tagging systems (DNorm, GNormPlus, SR4GN, tmChem and tmVar) and offer a batch-processing mode able to process arbitrary text input (e.g. scholarly publications, patents and medical records) in multiple formats (e.g. BioC). We support multiple standards to make our service interoperable and allow simpler integration with other text-processing pipelines. To maximize scalability, we have preprocessed all PubMed articles, and use a computer cluster for processing large requests of arbitrary text.

AVAILABILITY AND IMPLEMENTATION

Our text-mining web service is freely available at http://www.ncbi.nlm.nih.gov/CBBresearch/Lu/Demo/tmTools/#curl

CONTACT

: Zhiyong.Lu@nih.gov.

Detection of sentence boundaries and abbreviations in clinical narratives

BACKGROUND

In Western languages the period character is highly ambiguous, due to its double role as sentence delimiter and abbreviation marker. This is particularly relevant in clinical free-texts characterized by numerous anomalies in spelling, punctuation, vocabulary and with a high frequency of short forms.

METHODS

The problem is addressed by two binary classifiers for abbreviation and sentence detection. A support vector machine exploiting a linear kernel is trained on different combinations of feature sets for each classification task. Feature relevance ranking is applied to investigate which features are important for the particular task. The methods are applied to German language texts from a medical record system, authored by specialized physicians.

RESULTS

Two collections of 3,024 text snippets were annotated regarding the role of period characters for training and testing. Cohen's kappa resulted in 0.98. For abbreviation and sentence boundary detection we can report an unweighted micro-averaged F-measure using a 10-fold cross validation of 0.97 for the training set. For test set based evaluation we obtained an unweighted micro-averaged F-measure of 0.95 for abbreviation detection and 0.94 for sentence delineation. Language-dependent resources and rules were found to have less impact on abbreviation detection than on sentence delineation.

CONCLUSIONS

Sentence detection is an important task, which should be performed at the beginning of a text processing pipeline. For the text genre under scrutiny we showed that support vector machines exploiting a linear kernel produce state of the art results for sentence boundary detection. The results are comparable with other sentence boundary detection methods applied to English clinical texts. We identified abbreviation detection as a supportive task for sentence delineation.

A cross-lingual similarity measure for detecting biomedical term translations

Bilingual dictionaries for technical terms such as biomedical terms are an important resource for machine translation systems as well as for humans who would like to understand a concept described in a foreign language. Often a biomedical term is first proposed in English and later it is manually translated to other languages. Despite the fact that there are large monolingual lexicons of biomedical terms, only a fraction of those term lexicons are translated to other languages. Manually compiling large-scale bilingual dictionaries for technical domains is a challenging task because it is difficult to find a sufficiently large number of bilingual experts. We propose a cross-lingual similarity measure for detecting most similar translation candidates for a biomedical term specified in one language (source) from another language (target). Specifically, a biomedical term in a language is represented using two types of features: (a) intrinsic features that consist of character n-grams extracted from the term under consideration, and (b) extrinsic features that consist of unigrams and bigrams extracted from the contextual windows surrounding the term under consideration. We propose a cross-lingual similarity measure using each of those feature types. First, to reduce the dimensionality of the feature space in each language, we propose prototype vector projection (PVP)—a non-negative lower-dimensional vector projection method. Second, we propose a method to learn a mapping between the feature spaces in the source and target language using partial least squares regression (PLSR). The proposed method requires only a small number of training instances to learn a cross-lingual similarity measure. The proposed PVP method outperforms popular dimensionality reduction methods such as the singular value decomposition (SVD) and non-negative matrix factorization (NMF) in a nearest neighbor prediction task. Moreover, our experimental results covering several language pairs such as English–French, English–Spanish, English–Greek, and English–Japanese show that the proposed method outperforms several other feature projection methods in biomedical term translation prediction tasks.

Text Mining on Big and Complex Biomedical Literature

Driven by the rapidly advancing techniques and increasing interests in biology and medicine, about 2,000 to 4,000 references are added daily to MEDLINE, the US national biomedical bibliographic database. Even for a specific research topic, extracting useful and comprehensive information out of the huge literature data pool is challenging. Text mining techniques become extremely useful when dealing with the abundant biomedical information and they have been applied to various areas in the realm of biomedical research. Instead of providing a brief overview of all text mining techniques and every major biomedical text mining application, this chapter explores in-depth the microRNA profiling area and related text mining tools. As an illustrative example, one rule-based text mining system developed by the authors is discussed in detail. This chapter also includes the discussion of the challenges and potential research areas in biomedical text mining.

Link-topic model for biomedical abbreviation disambiguation

INTRODUCTION

The ambiguity of biomedical abbreviations is one of the challenges in biomedical text mining systems. In particular, the handling of term variants and abbreviations without nearby definitions is a critical issue. In this study, we adopt the concepts of topic of document and word link to disambiguate biomedical abbreviations.

METHODS

We newly suggest the link topic model inspired by the latent Dirichlet allocation model, in which each document is perceived as a random mixture of topics, where each topic is characterized by a distribution over words. Thus, the most probable expansions with respect to abbreviations of a given abstract are determined by word-topic, document-topic, and word-link distributions estimated from a document collection through the link topic model. The model allows two distinct modes of word generation to incorporate semantic dependencies among words, particularly long form words of abbreviations and their sentential co-occurring words; a word can be generated either dependently on the long form of the abbreviation or independently. The semantic dependency between two words is defined as a link and a new random parameter for the link is assigned to each word as well as a topic parameter. Because the link status indicates whether the word constitutes a link with a given specific long form, it has the effect of determining whether a word forms a unigram or a skipping/consecutive bigram with respect to the long form. Furthermore, we place a constraint on the model so that a word has the same topic as a specific long form if it is generated in reference to the long form. Consequently, documents are generated from the two hidden parameters, i.e. topic and link, and the most probable expansion of a specific abbreviation is estimated from the parameters.

RESULTS

Our model relaxes the bag-of-words assumption of the standard topic model in which the word order is neglected, and it captures a richer structure of text than does the standard topic model by considering unigrams and semantically associated bigrams simultaneously. The addition of semantic links improves the disambiguation accuracy without removing irrelevant contextual words and reduces the parameter space of massive skipping or consecutive bigrams. The link topic model achieves 98.42% disambiguation accuracy on 73,505 MEDLINE abstracts with respect to 21 three letter abbreviations and their 139 distinct long forms.

A method for integrating and ranking the evidence for biochemical pathways by mining reactions from text

Motivation: To create, verify and maintain pathway models, curators must discover and assess knowledge distributed over the vast body of biological literature. Methods supporting these tasks must understand both the pathway model representations and the natural language in the literature. These methods should identify and order documents by relevance to any given pathway reaction. No existing system has addressed all aspects of this challenge.

Method: We present novel methods for associating pathway model reactions with relevant publications. Our approach extracts the reactions directly from the models and then turns them into queries for three text mining-based MEDLINE literature search systems. These queries are executed, and the resulting documents are combined and ranked according to their relevance to the reactions of interest. We manually annotate document-reaction pairs with the relevance of the document to the reaction and use this annotation to study several ranking methods, using various heuristic and machine-learning approaches.

Results: Our evaluation shows that the annotated document-reaction pairs can be used to create a rule-based document ranking system, and that machine learning can be used to rank documents by their relevance to pathway reactions. We find that a Support Vector Machine-based system outperforms several baselines and matches the performance of the rule-based system. The success of the query extraction and ranking methods are used to update our existing pathway search system, PathText.

Availability: An online demonstration of PathText 2 and the annotated corpus are available for research purposes at http://www.nactem.ac.uk/pathtext2/.

Contact:makoto.miwa@manchester.ac.uk

Supplementary information:Supplementary data are available at Bioinformatics online.

An easily implemented method for abbreviation expansion for the medical domain in Japanese text. A preliminary study

BACKGROUND

One of the barriers for the effective use of computerized health-care related text is the ambiguity of abbreviations. To date, the task of disambiguating abbreviations has been treated as a classification task based on surrounding words. Application of this framework for languages that have no word boundaries requires pre-processing to segment a sentence into separate word sequences. While the segmentation processing is often a source of problem, it is unknown whether word information is really requisite for abbreviation expansion.

OBJECTIVES

The present study examined and compared abbreviation expansion methods with and without the incorporation of word information as a preliminary study.

METHODS

We implemented two abbreviation expansion methods: 1) a morpheme-based method that relied on word information and therefore required pre-processing, and 2) a character-based method that relied on simple character information. We compared the expansion accuracies for these two methods using eight medical abbreviations. Experimental data were automatically built as a pseudo-annotated corpus using the Internet.

RESULTS

As a result of the experiment, accuracies for the character-based method were from 0.890 to 0.942 while accuracies for the morpheme-based method were from 0.796 to 0.932. The character-based method significantly outperformed the morpheme-based method for three of the eight abbreviations (p < 0.05). For the remaining five abbreviations, no significant differences were found between the two methods.

CONCLUSIONS

Character information may be a good alternative in terms of simplicity to morphological information for abbreviation expansion in English medical abbreviations appeared in Japanese texts on the Internet.

Using rule-based natural language processing to improve disease normalization in biomedical text

BACKGROUND AND OBJECTIVE

In order for computers to extract useful information from unstructured text, a concept normalization system is needed to link relevant concepts in a text to sources that contain further information about the concept. Popular concept normalization tools in the biomedical field are dictionary-based. In this study we investigate the usefulness of natural language processing (NLP) as an adjunct to dictionary-based concept normalization.

METHODS

We compared the performance of two biomedical concept normalization systems, MetaMap and Peregrine, on the Arizona Disease Corpus, with and without the use of a rule-based NLP module. Performance was assessed for exact and inexact boundary matching of the system annotations with those of the gold standard and for concept identifier matching.

RESULTS

Without the NLP module, MetaMap and Peregrine attained F-scores of 61.0% and 63.9%, respectively, for exact boundary matching, and 55.1% and 56.9% for concept identifier matching. With the aid of the NLP module, the F-scores of MetaMap and Peregrine improved to 73.3% and 78.0% for boundary matching, and to 66.2% and 69.8% for concept identifier matching. For inexact boundary matching, performances further increased to 85.5% and 85.4%, and to 73.6% and 73.3% for concept identifier matching.

CONCLUSIONS

We have shown the added value of NLP for the recognition and normalization of diseases with MetaMap and Peregrine. The NLP module is general and can be applied in combination with any concept normalization system. Whether its use for concept types other than disease is equally advantageous remains to be investigated.

Discriminative application of string similarity methods to chemical and non-chemical names for biomedical abbreviation clustering

Background

Term clustering, by measuring the string similarities between terms, is known within the natural language processing community to be an effective method for improving the quality of texts and dictionaries. However, we have observed that chemical names are difficult to cluster using string similarity measures. In order to clearly demonstrate this difficulty, we compared the string similarities determined using the edit distance, the Monge-Elkan score, SoftTFIDF, and the bigram Dice coefficient for chemical names with those for non-chemical names.

Results

Our experimental results revealed the following: (1) The edit distance had the best performance in the matching of full forms, whereas Cohen et al. reported that SoftTFIDF with the Jaro-Winkler distance would yield the best measure for matching pairs of terms for their experiments. (2) For each of the string similarity measures above, the best threshold for term matching differs for chemical names and for non-chemical names; the difference is especially large for the edit distance. (3) Although the matching results obtained for chemical names using the edit distance, Monge-Elkan scores, or the bigram Dice coefficients are better than the result obtained for non-chemical names, the results were contrary when using SoftTFIDF. (4) A suitable weight for chemical names varies substantially from one for non-chemical names. In particular, a weight vector that has been optimized for non-chemical names is not suitable for chemical names. (5) The matching results using the edit distances improve further by dividing a set of full forms into two subsets, according to whether a full form is a chemical name or not. These results show that our hypothesis is acceptable, and that we can significantly improve the performance of abbreviation-full form clustering by computing chemical names and non-chemical names separately.

Conclusions

In conclusion, the discriminative application of string similarity methods to chemical and non-chemical names may be a simple yet effective way to improve the performance of term clustering.

Semantic text mining support for lignocellulose research

Background

Biofuels produced from biomass are considered to be promising sustainable alternatives to fossil fuels. The conversion of lignocellulose into fermentable sugars for biofuels production requires the use of enzyme cocktails that can efficiently and economically hydrolyze lignocellulosic biomass. As many fungi naturally break down lignocellulose, the identification and characterization of the enzymes involved is a key challenge in the research and development of biomass-derived products and fuels. One approach to meeting this challenge is to mine the rapidly-expanding repertoire of microbial genomes for enzymes with the appropriate catalytic properties.

Results

Semantic technologies, including natural language processing, ontologies, semantic Web services and Web-based collaboration tools, promise to support users in handling complex data, thereby facilitating knowledge-intensive tasks. An ongoing challenge is to select the appropriate technologies and combine them in a coherent system that brings measurable improvements to the users. We present our ongoing development of a semantic infrastructure in support of genomics-based lignocellulose research. Part of this effort is the automated curation of knowledge from information on fungal enzymes that is available in the literature and genome resources.

Conclusions

Working closely with fungal biology researchers who manually curate the existing literature, we developed ontological natural language processing pipelines integrated in a Web-based interface to assist them in two main tasks: mining the literature for relevant knowledge, and at the same time providing rich and semantically linked information.

Allie: a database and a search service of abbreviations and long forms

Many abbreviations are used in the literature especially in the life sciences, and polysemous abbreviations appear frequently, making it difficult to read and understand scientific papers that are outside of a reader's expertise. Thus, we have developed Allie, a database and a search service of abbreviations and their long forms (a.k.a. full forms or definitions). Allie searches for abbreviations and their corresponding long forms in a database that we have generated based on all titles and abstracts in MEDLINE. When a user query matches an abbreviation, Allie returns all potential long forms of the query along with their bibliographic data (i.e. title and publication year). In addition, for each candidate, co-occurring abbreviations and a research field in which it frequently appears in the MEDLINE data are displayed. This function helps users learn about the context in which an abbreviation appears. To deal with synonymous long forms, we use a dictionary called GENA that contains domain-specific terms such as gene, protein or disease names along with their synonymic information. Conceptually identical domain-specific terms are regarded as one term, and then conceptually identical abbreviation-long form pairs are grouped taking into account their appearance in MEDLINE. To keep up with new abbreviations that are continuously introduced, Allie has an automatic update system. In addition, the database of abbreviations and their long forms with their corresponding PubMed IDs is constructed and updated weekly. Database URL: The Allie service is available at http://allie.dbcls.jp/.

Mining metabolites: extracting the yeast metabolome from the literature

Text mining methods have added considerably to our capacity to extract biological knowledge from the literature. Recently the field of systems biology has begun to model and simulate metabolic networks, requiring knowledge of the set of molecules involved. While genomics and proteomics technologies are able to supply the macromolecular parts list, the metabolites are less easily assembled. Most metabolites are known and reported through the scientific literature, rather than through large-scale experimental surveys. Thus it is important to recover them from the literature. Here we present a novel tool to automatically identify metabolite names in the literature, and associate structures where possible, to define the reported yeast metabolome. With ten-fold cross validation on a manually annotated corpus, our recognition tool generates an f-score of 78.49 (precision of 83.02) and demonstrates greater suitability in identifying metabolite names than other existing recognition tools for general chemical molecules. The metabolite recognition tool has been applied to the literature covering an important model organism, the yeast Saccharomyces cerevisiae, to define its reported metabolome. By coupling to ChemSpider, a major chemical database, we have identified structures for much of the reported metabolome and, where structure identification fails, been able to suggest extensions to ChemSpider. Our manually annotated gold-standard data on 296 abstracts are available as supplementary materials. Metabolite names and, where appropriate, structures are also available as supplementary materials. ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (doi:10.1007/s11306-010-0251-6) contains supplementary material, which is available to authorized users.

PathText: a text mining integrator for biological pathway visualizations

Motivation: Metabolic and signaling pathways are an increasingly important part of organizing knowledge in systems biology. They serve to integrate collective interpretations of facts scattered throughout literature. Biologists construct a pathway by reading a large number of articles and interpreting them as a consistent network, but most of the models constructed currently lack direct links to those articles. Biologists who want to check the original articles have to spend substantial amounts of time to collect relevant articles and identify the sections relevant to the pathway. Furthermore, with the scientific literature expanding by several thousand papers per week, keeping a model relevant requires a continuous curation effort. In this article, we present a system designed to integrate a pathway visualizer, text mining systems and annotation tools into a seamless environment. This will enable biologists to freely move between parts of a pathway and relevant sections of articles, as well as identify relevant papers from large text bases. The system, PathText, is developed by Systems Biology Institute, Okinawa Institute of Science and Technology, National Centre for Text Mining (University of Manchester) and the University of Tokyo, and is being used by groups of biologists from these locations.

Contact:brian@monrovian.com.