Surveying biomedical relation extraction: a critical examination of current datasets and the proposal of a new resource

Natural language processing (NLP) has become an essential technique in various fields, offering a wide range of possibilities for analyzing data and developing diverse NLP tasks. In the biomedical domain, understanding the complex relationships between compounds and proteins is critical, especially in the context of signal transduction and biochemical pathways. Among these relationships, protein-protein interactions (PPIs) are of particular interest, given their potential to trigger a variety of biological reactions. To improve the ability to predict PPI events, we propose the protein event detection dataset (PEDD), which comprises 6823 abstracts, 39 488 sentences and 182 937 gene pairs. Our PEDD dataset has been utilized in the AI CUP Biomedical Paper Analysis competition, where systems are challenged to predict 12 different relation types. In this paper, we review the state-of-the-art relation extraction research and provide an overview of the PEDD's compilation process. Furthermore, we present the results of the PPI extraction competition and evaluate several language models' performances on the PEDD. This paper's outcomes will provide a valuable roadmap for future studies on protein event detection in NLP. By addressing this critical challenge, we hope to enable breakthroughs in drug discovery and enhance our understanding of the molecular mechanisms underlying various diseases.

Automatic Extraction of Medication Mentions from Tweets-Overview of the BioCreative VII Shared Task 3 Competition

This study presents the outcomes of the shared task competition BioCreative VII (Task 3) focusing on the extraction of medication names from a Twitter user's publicly available tweets (the user's 'timeline'). In general, detecting health-related tweets is notoriously challenging for natural language processing tools. The main challenge, aside from the informality of the language used, is that people tweet about any and all topics, and most of their tweets are not related to health. Thus, finding those tweets in a user's timeline that mention specific health-related concepts such as medications requires addressing extreme imbalance. Task 3 called for detecting tweets in a user's timeline that mentions a medication name and, for each detected mention, extracting its span. The organizers made available a corpus consisting of 182 049 tweets publicly posted by 212 Twitter users with all medication mentions manually annotated. The corpus exhibits the natural distribution of positive tweets, with only 442 tweets (0.2%) mentioning a medication. This task was an opportunity for participants to evaluate methods that are robust to class imbalance beyond the simple lexical match. A total of 65 teams registered, and 16 teams submitted a system run. This study summarizes the corpus created by the organizers and the approaches taken by the participating teams for this challenge. The corpus is freely available at https://biocreative.bioinformatics.udel.edu/tasks/biocreative-vii/track-3/. The methods and the results of the competing systems are analyzed with a focus on the approaches taken for learning from class-imbalanced data.

Semi-supervised learning for topographic map analysis over time: a study of bridge segmentation

Geographical research using historical maps has progressed considerably as the digitalization of topological maps across years provides valuable data and the advancement of AI machine learning models provides powerful analytic tools. Nevertheless, analysis of historical maps based on supervised learning can be limited by the laborious manual map annotations. In this work, we propose a semi-supervised learning method that can transfer the annotation of maps across years and allow map comparison and anthropogenic studies across time. Our novel two-stage framework first performs style transfer of topographic map across years and versions, and then supervised learning can be applied on the synthesized maps with annotations. We investigate the proposed semi-supervised training with the style-transferred maps and annotations on four widely-used deep neural networks (DNN), namely U-Net, fully-convolutional network (FCN), DeepLabV3, and MobileNetV3. The best performing network of U-Net achieves [Formula: see text] and [Formula: see text] trained on style-transfer synthesized maps, which indicates that the proposed framework is capable of detecting target features (bridges) on historical maps without annotations. In a comprehensive comparison, the [Formula: see text] of U-Net trained on Contrastive Unpaired Translation (CUT) generated dataset ([Formula: see text]) achieves 57.3 % than the comparative score ([Formula: see text]) of the least valid configuration (MobileNetV3 trained on CycleGAN synthesized dataset). We also discuss the remaining challenges and future research directions.

Task reformulation and data-centric approach for Twitter medication name extraction

Automatically extracting medication names from tweets is challenging in the real world. There are many tweets; however, only a small proportion mentions medications. Thus, datasets are usually highly imbalanced. Moreover, the length of tweets is very short, which makes it hard to recognize medication names from the limited context. This paper proposes a data-centric approach for extracting medications in the BioCreative VII Track 3 (Automatic Extraction of Medication Names in Tweets). Our approach formulates the sequence labeling problem as text entailment and question-answer tasks. As a result, without using the dictionary and ensemble method, our single model achieved a Strict F1 of 0.77 (the official baseline system is 0.758, and the average performance of participants is 0.696). Moreover, combining the dictionary filtering and ensemble method achieved a Strict F1 of 0.804 and had the highest performance for all participants. Furthermore, domain-specific and task-specific pretrained language models, as well as data-centric approaches, are proposed for further improvements. Database URL https://competitions.codalab.org/competitions/23925 and https://biocreative.bioinformatics.udel.edu/tasks/biocreative-vii/track-3/.

Increased Online Aggression During COVID-19 Lockdowns: Two-Stage Study of Deep Text Mining and Difference-in-Differences Analysis

BACKGROUND

The COVID-19 pandemic caused a critical public health crisis worldwide, and policymakers are using lockdowns to control the virus. However, there has been a noticeable increase in aggressive social behaviors that threaten social stability. Lockdown measures might negatively affect mental health and lead to an increase in aggressive emotions. Discovering the relationship between lockdown and increased aggression is crucial for formulating appropriate policies that address these adverse societal effects. We applied natural language processing (NLP) technology to internet data, so as to investigate the social and emotional impacts of lockdowns.

OBJECTIVE

This research aimed to understand the relationship between lockdown and increased aggression using NLP technology to analyze the following 3 kinds of aggressive emotions: anger, offensive language, and hate speech, in spatiotemporal ranges of tweets in the United States.

METHODS

We conducted a longitudinal internet study of 11,455 Twitter users by analyzing aggressive emotions in 1,281,362 tweets they posted from 2019 to 2020. We selected 3 common aggressive emotions (anger, offensive language, and hate speech) on the internet as the subject of analysis. To detect the emotions in the tweets, we trained a Bidirectional Encoder Representations from Transformers (BERT) model to analyze the percentage of aggressive tweets in every state and every week. Then, we used the difference-in-differences estimation to measure the impact of lockdown status on increasing aggressive tweets. Since most other independent factors that might affect the results, such as seasonal and regional factors, have been ruled out by time and state fixed effects, a significant result in this difference-in-differences analysis can not only indicate a concrete positive correlation but also point to a causal relationship.

RESULTS

In the first 6 months of lockdown in 2020, aggression levels in all users increased compared to the same period in 2019. Notably, users under lockdown demonstrated greater levels of aggression than those not under lockdown. Our difference-in-differences estimation discovered a statistically significant positive correlation between lockdown and increased aggression (anger: P=.002, offensive language: P<.001, hate speech: P=.005). It can be inferred from such results that there exist causal relations.

CONCLUSIONS

Understanding the relationship between lockdown and aggression can help policymakers address the personal and societal impacts of lockdown. Applying NLP technology and using big data on social media can provide crucial and timely information for this effort.

Biomedical named entity recognition and linking datasets: survey and our recent development

Natural language processing (NLP) is widely applied in biological domains to retrieve information from publications. Systems to address numerous applications exist, such as biomedical named entity recognition (BNER), named entity normalization (NEN) and protein–protein interaction extraction (PPIE). High-quality datasets can assist the development of robust and reliable systems; however, due to the endless applications and evolving techniques, the annotations of benchmark datasets may become outdated and inappropriate. In this study, we first review commonlyused BNER datasets and their potential annotation problems such as inconsistency and low portability. Then, we introduce a revised version of the JNLPBA dataset that solves potential problems in the original and use state-of-the-art named entity recognition systems to evaluate its portability to different kinds of biomedical literature, including protein–protein interaction and biology events. Lastly, we introduce an ensembled biomedical entity dataset (EBED) by extending the revised JNLPBA dataset with PubMed Central full-text paragraphs, figure captions and patent abstracts. This EBED is a multi-task dataset that covers annotations including gene, disease and chemical entities. In total, it contains 85000 entity mentions, 25000 entity mentions with database identifiers and 5000 attribute tags. To demonstrate the usage of the EBED, we review the BNER track from the AI CUP Biomedical Paper Analysis challenge. Availability: The revised JNLPBA dataset is available at https://iasl-btm.iis.sinica.edu.tw/BNER/Content/Re vised_JNLPBA.zip. The EBED dataset is available at https://iasl-btm.iis.sinica.edu.tw/BNER/Content/AICUP _EBED_dataset.rar. Contact: Email: thtsai@g.ncu.edu.tw, Tel. 886-3-4227151 ext. 35203, Fax: 886-3-422-2681 Email: hsu@iis.sinica.edu.tw, Tel. 886-2-2788-3799 ext. 2211, Fax: 886-2-2782-4814 Supplementary information: Supplementary data are available at Briefings in Bioinformatics online.

The extraction of complex relationships and their conversion to biological expression language (BEL) overview of the BioCreative VI (2017) BEL track

Knowledge of the molecular interactions of biological and chemical entities and their involvement in biological processes or clinical phenotypes is important for data interpretation. Unfortunately, this knowledge is mostly embedded in the literature in such a way that it is unavailable for automated data analysis procedures. Biological expression language (BEL) is a syntax representation allowing for the structured representation of a broad range of biological relationships. It is used in various situations to extract such knowledge and transform it into BEL networks. To support the tedious and time-intensive extraction work of curators with automated methods, we developed the BEL track within the framework of BioCreative Challenges. Within the BEL track, we provide training data and an evaluation environment to encourage the text mining community to tackle the automatic extraction of complex BEL relationships. In 2017 BioCreative VI, the 2015 BEL track was repeated with new test data. Although only minor improvements in text snippet retrieval for given statements were achieved during this second BEL task iteration, a significant increase of BEL statement extraction performance from provided sentences could be seen. The best performing system reached a 32% F-score for the extraction of complete BEL statements and with the given named entities this increased to 49%. This time, besides rule-based systems, new methods involving hierarchical sequence labeling and neural networks were applied for BEL statement extraction.

Using a Large Margin Context-Aware Convolutional Neural Network to Automatically Extract Disease-Disease Association from Literature: Comparative Analytic Study

BACKGROUND

Research on disease-disease association (DDA), like comorbidity and complication, provides important insights into disease treatment and drug discovery, and a large body of the literature has been published in the field. However, using current search tools, it is not easy for researchers to retrieve information on the latest DDA findings. First, comorbidity and complication keywords pull up large numbers of PubMed studies. Second, disease is not highlighted in search results. Finally, DDA is not identified, as currently no disease-disease association extraction (DDAE) dataset or tools are available.

OBJECTIVE

As there are no available DDAE datasets or tools, this study aimed to develop (1) a DDAE dataset and (2) a neural network model for extracting DDA from the literature.

METHODS

In this study, we formulated DDAE as a supervised machine learning classification problem. To develop the system, we first built a DDAE dataset. We then employed two machine learning models, support vector machine and convolutional neural network, to extract DDA. Furthermore, we evaluated the effect of using the output layer as features of the support vector machine-based model. Finally, we implemented large margin context-aware convolutional neural network architecture to integrate context features and convolutional neural networks through the large margin function.

RESULTS

Our DDAE dataset consisted of 521 PubMed abstracts. Experiment results showed that the support vector machine-based approach achieved an F1 measure of 80.32%, which is higher than the convolutional neural network-based approach (73.32%). Using the output layer of convolutional neural network as a feature for the support vector machine does not further improve the performance of support vector machine. However, our large margin context-aware-convolutional neural network achieved the highest F1 measure of 84.18% and demonstrated that combining the hinge loss function of support vector machine with a convolutional neural network into a single neural network architecture outperforms other approaches.

CONCLUSIONS

To facilitate the development of text-mining research for DDAE, we developed the first publicly available DDAE dataset consisting of disease mentions, Medical Subject Heading IDs, and relation annotations. We developed different conventional machine learning models and neural network architectures and evaluated their effects on our DDAE dataset. To further improve DDAE performance, we propose an large margin context-aware-convolutional neural network model for DDAE that outperforms other approaches.

Statistical principle-based approach for gene and protein related object recognition

The large number of chemical and pharmaceutical patents has attracted researchers doing biomedical text mining to extract valuable information such as chemicals, genes and gene products. To facilitate gene and gene product annotations in patents, BioCreative V.5 organized a gene- and protein-related object (GPRO) recognition task, in which participants were assigned to identify GPRO mentions and determine whether they could be linked to their unique biological database records. In this paper, we describe the system constructed for this task. Our system is based on two different NER approaches: the statistical-principle-based approach (SPBA) and conditional random fields (CRF). Therefore, we call our system SPBA-CRF. SPBA is an interpretable machine-learning framework for gene mention recognition. The predictions of SPBA are used as features for our CRF-based GPRO recognizer. The recognizer was developed for identifying chemical mentions in patents, and we adapted it for GPRO recognition. In the BioCreative V.5 GPRO recognition task, SPBA-CRF obtained an F-score of 73.73% on the evaluation metric of GPRO type 1 and an F-score of 78.66% on the evaluation metric of combining GPRO types 1 and 2. Our results show that SPBA trained on an external NER dataset can perform reasonably well on the partial match evaluation metric. Furthermore, SPBA can significantly improve performance of the CRF-based recognizer trained on the GPRO dataset.

NERChem: adapting NERBio to chemical patents via full-token features and named entity feature with chemical sub-class composition

Chemical patents contain detailed information on novel chemical compounds that is valuable to the chemical and pharmaceutical industries. In this paper, we introduce a system, NERChem that can recognize chemical named entity mentions in chemical patents. NERChem is based on the conditional random fields model (CRF). Our approach incorporates (1) class composition, which is used for combining chemical classes whose naming conventions are similar; (2) BioNE features, which are used for distinguishing chemical mentions from other biomedical NE mentions in the patents; and (3) full-token word features, which are used to resolve the tokenization granularity problem. We evaluated our approach on the BioCreative V CHEMDNER-patent corpus, and achieved an F-score of 87.17% in the Chemical Entity Mention in Patents (CEMP) task and a sensitivity of 98.58% in the Chemical Passage Detection (CPD) task, ranking alongside the top systems.

Database URL: Our NERChem web-based system is publicly available at iisrserv.csie.n cu.edu.tw/nerchem.

BelSmile: a biomedical semantic role labeling approach for extracting biological expression language from text

Biological expression language (BEL) is one of the most popular languages to represent the causal and correlative relationships among biological events. Automatically extracting and representing biomedical events using BEL can help biologists quickly survey and understand relevant literature. Recently, many researchers have shown interest in biomedical event extraction. However, the task is still a challenge for current systems because of the complexity of integrating different information extraction tasks such as named entity recognition (NER), named entity normalization (NEN) and relation extraction into a single system. In this study, we introduce our BelSmile system, which uses a semantic-role-labeling (SRL)-based approach to extract the NEs and events for BEL statements. BelSmile combines our previous NER, NEN and SRL systems. We evaluate BelSmile using the BioCreative V BEL task dataset. Our system achieved an F-score of 27.8%, ∼7% higher than the top BioCreative V system. The three main contributions of this study are (i) an effective pipeline approach to extract BEL statements, and (ii) a syntactic-based labeler to extract subject-verb-object tuples. We also implement a web-based version of BelSmile (iii) that is publicly available at iisrserv.csie.ncu.edu.tw/belsmile.

Collective Web-Based Parenthetical Translation Extraction Using Markov Logic Networks

Parenthetical translations are translations of terms in otherwise monolingual text that appear inside parentheses. Parenthetical translations extraction (PTE) is the task of extracting parenthetical translations from natural language documents. One of the main difficulties in PTE is to detect the left boundary of the translated term in preparenthetical text. In this article, we propose a collective approach that employs Markov logic to model multiple constraints used in the PTE task. We show how various constraints can be formulated and combined in a Markov logic network (MLN). Our experimental results show that the proposed collective PTE approach significantly outperforms a current state-of-the-art method, improving the average F-measure up to 27.11% compared to the previous word alignment approach. It also outperforms an individual MLN-based system by 8.2% and a system based on conditional random fields by 5.9%.

A context-aware approach for progression tracking of medical concepts in electronic medical records

Electronic medical records (EMRs) for diabetic patients contain information about heart disease risk factors such as high blood pressure, cholesterol levels, and smoking status. Discovering the described risk factors and tracking their progression over time may support medical personnel in making clinical decisions, as well as facilitate data modeling and biomedical research. Such highly patient-specific knowledge is essential to driving the advancement of evidence-based practice, and can also help improve personalized medicine and care. One general approach for tracking the progression of diseases and their risk factors described in EMRs is to first recognize all temporal expressions, and then assign each of them to the nearest target medical concept. However, this method may not always provide the correct associations. In light of this, this work introduces a context-aware approach to assign the time attributes of the recognized risk factors by reconstructing contexts that contain more reliable temporal expressions. The evaluation results on the i2b2 test set demonstrate the efficacy of the proposed approach, which achieved an F-score of 0.897. To boost the approach's ability to process unstructured clinical text and to allow for the reproduction of the demonstrated results, a set of developed .NET libraries used to develop the system is available at https://sites.google.com/site/hongjiedai/projects/nttmuclinicalnet.

Enhancing of chemical compound and drug name recognition using representative tag scheme and fine-grained tokenization

Background

The functions of chemical compounds and drugs that affect biological processes and their particular effect on the onset and treatment of diseases have attracted increasing interest with the advancement of research in the life sciences. To extract knowledge from the extensive literatures on such compounds and drugs, the organizers of BioCreative IV administered the CHEMical Compound and Drug Named Entity Recognition (CHEMDNER) task to establish a standard dataset for evaluating state-of-the-art chemical entity recognition methods.

Methods

This study introduces the approach of our CHEMDNER system. Instead of emphasizing the development of novel feature sets for machine learning, this study investigates the effect of various tag schemes on the recognition of the names of chemicals and drugs by using conditional random fields. Experiments were conducted using combinations of different tokenization strategies and tag schemes to investigate the effects of tag set selection and tokenization method on the CHEMDNER task.

Results

This study presents the performance of CHEMDNER of three more representative tag schemes-IOBE, IOBES, and IOB₁₂E-when applied to a widely utilized IOB tag set and combined with the coarse-/fine-grained tokenization methods. The experimental results thus reveal that the fine-grained tokenization strategy performance best in terms of precision, recall and F-scores when the IOBES tag set was utilized. The IOBES model with fine-grained tokenization yielded the best-F-scores in the six chemical entity categories other than the "Multiple" entity category. Nonetheless, no significant improvement was observed when a more representative tag schemes was used with the coarse or fine-grained tokenization rules. The best F-scores that were achieved using the developed system on the test dataset of the CHEMDNER task were 0.833 and 0.815 for the chemical documents indexing and the chemical entity mention recognition tasks, respectively.

Conclusions

The results herein highlight the importance of tag set selection and the use of different tokenization strategies. Fine-grained tokenization combined with the tag set IOBES most effectively recognizes chemical and drug names. To the best of the authors' knowledge, this investigation is the first comprehensive investigation use of various tag set schemes combined with different tokenization strategies for the recognition of chemical entities.

The CHEMDNER corpus of chemicals and drugs and its annotation principles

The automatic extraction of chemical information from text requires the recognition of chemical entity mentions as one of its key steps. When developing supervised named entity recognition (NER) systems, the availability of a large, manually annotated text corpus is desirable. Furthermore, large corpora permit the robust evaluation and comparison of different approaches that detect chemicals in documents. We present the CHEMDNER corpus, a collection of 10,000 PubMed abstracts that contain a total of 84,355 chemical entity mentions labeled manually by expert chemistry literature curators, following annotation guidelines specifically defined for this task. The abstracts of the CHEMDNER corpus were selected to be representative for all major chemical disciplines. Each of the chemical entity mentions was manually labeled according to its structure-associated chemical entity mention (SACEM) class: abbreviation, family, formula, identifier, multiple, systematic and trivial. The difficulty and consistency of tagging chemicals in text was measured using an agreement study between annotators, obtaining a percentage agreement of 91. For a subset of the CHEMDNER corpus (the test set of 3,000 abstracts) we provide not only the Gold Standard manual annotations, but also mentions automatically detected by the 26 teams that participated in the BioCreative IV CHEMDNER chemical mention recognition task. In addition, we release the CHEMDNER silver standard corpus of automatically extracted mentions from 17,000 randomly selected PubMed abstracts. A version of the CHEMDNER corpus in the BioC format has been generated as well. We propose a standard for required minimum information about entity annotations for the construction of domain specific corpora on chemical and drug entities. The CHEMDNER corpus and annotation guidelines are available at: http://www.biocreative.org/resources/biocreative-iv/chemdner-corpus/.

LiverCancerMarkerRIF: a liver cancer biomarker interactive curation system combining text mining and expert annotations

Biomarkers are biomolecules in the human body that can indicate disease states and abnormal biological processes. Biomarkers are often used during clinical trials to identify patients with cancers. Although biomedical research related to biomarkers has increased over the years and substantial effort has been expended to obtain results in these studies, the specific results obtained often contain ambiguities, and the results might contradict each other. Therefore, the information gathered from these studies must be appropriately integrated and organized to facilitate experimentation on biomarkers. In this study, we used liver cancer as the target and developed a text-mining–based curation system named LiverCancerMarkerRIF, which allows users to retrieve biomarker-related narrations and curators to curate supporting evidence on liver cancer biomarkers directly while browsing PubMed. In contrast to most of the other curation tools that require curators to navigate away from PubMed and accommodate distinct user interfaces or Web sites to complete the curation process, our system provides a user-friendly method for accessing text-mining–aided information and a concise interface to assist curators while they remain at the PubMed Web site. Biomedical text-mining techniques are applied to automatically recognize biomedical concepts such as genes, microRNA, diseases and investigative technologies, which can be used to evaluate the potential of a certain gene as a biomarker. Through the participation in the BioCreative IV user-interactive task, we examined the feasibility of using this novel type of augmented browsing-based curation method, and collaborated with curators to curate biomarker evidential sentences related to liver cancer. The positive feedback received from curators indicates that the proposed method can be effectively used for curation. A publicly available online database containing all the aforementioned information has been constructed at http://btm.tmu.edu.tw/livercancermarkerrif in an attempt to facilitate biomarker-related studies.

Database URL:http://btm.tmu.edu.tw/LiverCancerMarkerRIF/

BioC interoperability track overview

BioC is a new simple XML format for sharing biomedical text and annotations and libraries to read and write that format. This promotes the development of interoperable tools for natural language processing (NLP) of biomedical text. The interoperability track at the BioCreative IV workshop featured contributions using or highlighting the BioC format. These contributions included additional implementations of BioC, many new corpora in the format, biomedical NLP tools consuming and producing the format and online services using the format. The ease of use, broad support and rapidly growing number of tools demonstrate the need for and value of the BioC format.

Database URL:http://bioc.sourceforge.net/

A resource-saving collective approach to biomedical semantic role labeling

Background

Biomedical semantic role labeling (BioSRL) is a natural language processing technique that identifies the semantic roles of the words or phrases in sentences describing biological processes and expresses them as predicate-argument structures (PAS’s). Currently, a major problem of BioSRL is that most systems label every node in a full parse tree independently; however, some nodes always exhibit dependency. In general SRL, collective approaches based on the Markov logic network (MLN) model have been successful in dealing with this problem. However, in BioSRL such an approach has not been attempted because it would require more training data to recognize the more specialized and diverse terms found in biomedical literature, increasing training time and computational complexity.

Results

We first constructed a collective BioSRL system based on MLN. This system, called collective BIOSMILE (CBIOSMILE), is trained on the BioProp corpus. To reduce the resources used in BioSRL training, we employ a tree-pruning filter to remove unlikely nodes from the parse tree and four argument candidate identifiers to retain candidate nodes in the tree. Nodes not recognized by any candidate identifier are discarded. The pruned annotated parse trees are used to train a resource-saving MLN-based system, which is referred to as resource-saving collective BIOSMILE (RCBIOSMILE). Our experimental results show that our proposed CBIOSMILE system outperforms BIOSMILE, which is the top BioSRL system. Furthermore, our proposed RCBIOSMILE maintains the same level of accuracy as CBIOSMILE using 92% less memory and 57% less training time.

Conclusions

This greatly improved efficiency makes RCBIOSMILE potentially suitable for training on much larger BioSRL corpora over more biomedical domains. Compared to real-world biomedical corpora, BioProp is relatively small, containing only 445 MEDLINE abstracts and 30 event triggers. It is not large enough for practical applications, such as pathway construction. We consider it of primary importance to pursue SRL training on large corpora in the future.

Collective instance-level gene normalization on the IGN corpus

A high proportion of life science researches are gene-oriented, in which scientists aim to investigate the roles that genes play in biological processes, and their involvement in biological mechanisms. As a result, gene names and their related information turn out to be one of the main objects of interest in biomedical literatures. While the capability of recognizing gene mentions has made significant progress, the results of recognition are still insufficient for direct use due to the ambiguity of gene names. Gene normalization (GN) goes beyond the recognition task by linking a gene mention to a database ID. Unlike most previous works, we approach GN on the instance-level and evaluate its overall performance on the recognition and normalization steps in abstracts and full texts. We release the first instance-level gene normalization (IGN) corpus in the BioC format, which includes annotations for the boundaries of all gene mentions and the corresponding IDs for human gene mentions. Species information, along with existing co-reference chains and full name/abbreviation pairs are also provided for each gene mention. Using the released corpus, we have designed a collective instance-level GN approach using not only the contextual information of each individual instance, but also the relations among instances and the inherent characteristics of full-text sections. Our experimental results show that our collective approach can achieve an F-score of 0.743. The proposed approach that exploits section characteristics in full-text articles can improve the F-scores of information lacking sections by up to 1.8%. In addition, using the proposed refinement process improved the F-score of gene mention recognition by 0.125 and that of GN by 0.03. Whereas current experimental results are limited to the human species, we seek to continue updating the annotations of the IGN corpus and observe how the proposed approach can be extended to other species.

TEMPTING system: a hybrid method of rule and machine learning for temporal relation extraction in patient discharge summaries

Patient discharge summaries provide detailed medical information about individuals who have been hospitalized. To make a precise and legitimate assessment of the abundant data, a proper time layout of the sequence of relevant events should be compiled and used to drive a patient-specific timeline, which could further assist medical personnel in making clinical decisions. The process of identifying the chronological order of entities is called temporal relation extraction. In this paper, we propose a hybrid method to identify appropriate temporal links between a pair of entities. The method combines two approaches: one is rule-based and the other is based on the maximum entropy model. We develop an integration algorithm to fuse the results of the two approaches. All rules and the integration algorithm are formally stated so that one can easily reproduce the system and results. To optimize the system's configuration, we used the 2012 i2b2 challenge TLINK track dataset and applied threefold cross validation to the training set. Then, we evaluated its performance on the training and test datasets. The experiment results show that the proposed TEMPTING (TEMPoral relaTion extractING) system (ranked seventh) achieved an F-score of 0.563, which was at least 30% better than that of the baseline system, which randomly selects TLINK candidates from all pairs and assigns the TLINK types. The TEMPTING system using the hybrid method also outperformed the stage-based TEMPTING system. Its F-scores were 3.51% and 0.97% better than those of the stage-based system on the training set and test set, respectively.

T-HOD: a literature-based candidate gene database for hypertension, obesity and diabetes

Researchers are finding it more and more difficult to follow the changing status of disease candidate genes due to the exponential increase in gene mapping studies. The Text-mined Hypertension, Obesity and Diabetes candidate gene database (T-HOD) is developed to help trace existing research on three kinds of cardiovascular diseases: hypertension, obesity and diabetes, with the last disease categorized into Type 1 and Type 2, by regularly and semiautomatically extracting HOD-related genes from newly published literature. Currently, there are 837, 835 and 821 candidate genes recorded in T-HOD for hypertension, obesity and diabetes, respectively. T-HOD employed the state-of-art text-mining technologies, including a gene/disease identification system and a disease–gene relation extraction system, which can be used to affirm the association of genes with three diseases and provide more evidence for further studies. The primary inputs of T-HOD are the three kinds of diseases, and the output is a list of disease-related genes that can be ranked based on their number of appearance, protein–protein interactions and single-nucleotide polymorphisms. Unlike manually constructed disease gene databases, the content of T-HOD is regularly updated by our text-mining system and verified by domain experts. The interface of T-HOD facilitates easy browsing for users and allows T-HOD curators to verify data efficiently. We believe that T-HOD can help life scientists in search for more disease candidate genes in a less time- and effort-consuming manner.

Database URL:http://bws.iis.sinica.edu.tw/THOD

Improving protein-protein interaction pair ranking with an integrated global association score

Protein-protein interaction (PPI) database curation requires text-mining systems that can recognize and normalize interactor genes and return a ranked list of PPI pairs for each article. The order of PPI pairs in this list is essential for ease of curation. Most of the current PPI pair ranking approaches rely on association analysis between the two genes in the pair. However, we propose that ranking an extracted PPI pair by considering both the association between the paired genes and each of those genes’ global associations with all other genes mentioned in the paper can provide a more reliable ranked list. In this work, we present a composite interaction score that considers not only the association score between two interactors (pair association score) but also their global association scores. We test three representative data fusion algorithms to estimate this global association score—two Borda-Fuse models and one linear combination model (LCM). The three estimation methods are evaluated using the data set of the BioCreative II.5 Interaction Pair Task (IPT) in terms of area under the interpolated precision/recall curve (AUC iP/R). Our experimental results indicate that using LCM to estimate the global association score can boost the AUC iP/R score from 0.0175 to 0.2396, outperforming the best BioCreative II.5 IPT system.

Coreference resolution of medical concepts in discharge summaries by exploiting contextual information

OBJECTIVE

Patient discharge summaries provide detailed medical information about hospitalized patients and are a rich resource of data for clinical record text mining. The textual expressions of this information are highly variable. In order to acquire a precise understanding of the patient, it is important to uncover the relationship between all instances in the text. In natural language processing (NLP), this task falls under the category of coreference resolution.

DESIGN

A key contribution of this paper is the application of contextual-dependent rules that describe relationships between coreference pairs. To resolve phrases that refer to the same entity, the authors use these rules in three representative NLP systems: one rule-based, another based on the maximum entropy model, and the last a system built on the Markov logic network (MLN) model.

RESULTS

The experimental results show that the proposed MLN-based system outperforms the baseline system (exact match) by average F-scores of 4.3% and 5.7% on the Beth and Partners datasets, respectively. Finally, the three systems were integrated into an ensemble system, further improving performance to 87.21%, which is 4.5% more than the official i2b2 Track 1C average (82.7%).

CONCLUSION

In this paper, the main challenges in the resolution of coreference relations in patient discharge summaries are described. Several rules are proposed to exploit contextual information, and three approaches presented. While single systems provided promising results, an ensemble approach combining the three systems produced a better performance than even the best single system.

Multi-stage gene normalization for full-text articles with context-based species filtering for dynamic dictionary entry selection

Background

Gene normalization (GN) is the task of identifying the unique database IDs of genes and proteins in literature. The best-known public competition of GN systems is the GN task of the BioCreative challenge, which has been held four times since 2003. The last two BioCreatives, II.5 & III, had two significant differences from earlier tasks: firstly, they provided full-length articles in addition to abstracts; and secondly, they included multiple species without providing species ID information. Full papers introduce more complex targets for GN processing, while the inclusion of multiple species vastly increases the potential size of dictionaries needed for GN. BioCreative III GN uses Threshold Average Precision at a median of k errors per query (TAP-k), a new measure closely related to the well-known average precision, but also reflecting the reliability of the score provided by each GN system.

Results

To use full-paper text, we employed a multi-stage GN algorithm and a ranking method which exploit information in different sections and parts of a paper. To handle the inclusion of multiple unknown species, we developed two context-based dynamic strategies to select dictionary entries related to the species that appear in the paper—section-wide and article-wide context. Our originally submitted BioCreative III system uses a static dictionary containing only the most common species entries. It already exceeds the BioCreative III average team performance by at least 24% in every evaluation. However, using our proposed dynamic dictionary strategies, we were able to further improve TAP-5, TAP-10, and TAP-20 by 16.47%, 13.57% and 6.01%, respectively in the Gold 50 test set. Our best dynamic strategy outperforms the best BioCreative III systems in TAP-10 on the Silver 50 test set and in TAP-5 on the Silver 507 set.

Conclusions

Our experimental results demonstrate the superiority of our proposed dynamic dictionary selection strategies over our original static strategy and most BioCreative III participant systems. Section-wide dynamic strategy is preferred because it achieves very similar TAP-k scores to article-wide dynamic strategy but it is more efficient.

The gene normalization task in BioCreative III

BACKGROUND

We report the Gene Normalization (GN) challenge in BioCreative III where participating teams were asked to return a ranked list of identifiers of the genes detected in full-text articles. For training, 32 fully and 500 partially annotated articles were prepared. A total of 507 articles were selected as the test set. Due to the high annotation cost, it was not feasible to obtain gold-standard human annotations for all test articles. Instead, we developed an Expectation Maximization (EM) algorithm approach for choosing a small number of test articles for manual annotation that were most capable of differentiating team performance. Moreover, the same algorithm was subsequently used for inferring ground truth based solely on team submissions. We report team performance on both gold standard and inferred ground truth using a newly proposed metric called Threshold Average Precision (TAP-k).

RESULTS

We received a total of 37 runs from 14 different teams for the task. When evaluated using the gold-standard annotations of the 50 articles, the highest TAP-k scores were 0.3297 (k=5), 0.3538 (k=10), and 0.3535 (k=20), respectively. Higher TAP-k scores of 0.4916 (k=5, 10, 20) were observed when evaluated using the inferred ground truth over the full test set. When combining team results using machine learning, the best composite system achieved TAP-k scores of 0.3707 (k=5), 0.4311 (k=10), and 0.4477 (k=20) on the gold standard, representing improvements of 12.4%, 21.8%, and 26.6% over the best team results, respectively.

CONCLUSIONS

By using full text and being species non-specific, the GN task in BioCreative III has moved closer to a real literature curation task than similar tasks in the past and presents additional challenges for the text mining community, as revealed in the overall team results. By evaluating teams using the gold standard, we show that the EM algorithm allows team submissions to be differentiated while keeping the manual annotation effort feasible. Using the inferred ground truth we show measures of comparative performance between teams. Finally, by comparing team rankings on gold standard vs. inferred ground truth, we further demonstrate that the inferred ground truth is as effective as the gold standard for detecting good team performance.

Integration of gene normalization stages and co-reference resolution using a Markov logic network

MOTIVATION

Gene normalization (GN) is the task of normalizing a textual gene mention to a unique gene database ID. Traditional top performing GN systems usually need to consider several constraints to make decisions in the normalization process, including filtering out false positives, or disambiguating an ambiguous gene mention, to improve system performance. However, these constraints are usually executed in several separate stages and cannot use each other's input/output interactively. In this article, we propose a novel approach that employs a Markov logic network (MLN) to model the constraints used in the GN task. Firstly, we show how various constraints can be formulated and combined in an MLN. Secondly, we are the first to apply the two main concepts of co-reference resolution-discourse salience in centering theory and transitivity-to GN models. Furthermore, to make our results more relevant to developers of information extraction applications, we adopt the instance-based precision/recall/F-measure (PRF) in addition to the article-wide PRF to assess system performance.

RESULTS

Experimental results show that our system outperforms baseline and state-of-the-art systems under two evaluation schemes. Through further analysis, we have found several unexplored challenges in the GN task.

CONTACT

hongjie@iis.sinica.edu.tw

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Dynamic programming re-ranking for PPI interactor and pair extraction in full-text articles

Background

Experimentally verified protein-protein interactions (PPIs) cannot be easily retrieved by researchers unless they are stored in PPI databases. The curation of such databases can be facilitated by employing text-mining systems to identify genes which play the interactor role in PPIs and to map these genes to unique database identifiers (interactor normalization task or INT) and then to return a list of interaction pairs for each article (interaction pair task or IPT). These two tasks are evaluated in terms of the area under curve of the interpolated precision/recall (AUC iP/R) score because the order of identifiers in the output list is important for ease of curation.

Results

Our INT system developed for the BioCreAtIvE II.5 INT challenge achieved a promising AUC iP/R of 43.5% by using a support vector machine (SVM)-based ranking procedure. Using our new re-ranking algorithm, we have been able to improve system performance (AUC iP/R) by 1.84%. Our experimental results also show that with the re-ranked INT results, our unsupervised IPT system can achieve a competitive AUC iP/R of 23.86%, which outperforms the best BC II.5 INT system by 1.64%. Compared to using only SVM ranked INT results, using re-ranked INT results boosts AUC iP/R by 7.84%. Statistical significance t-test results show that our INT/IPT system with re-ranking outperforms that without re-ranking by a statistically significant difference.

Conclusions

In this paper, we present a new re-ranking algorithm that considers co-occurrence among identifiers in an article to improve INT and IPT ranking results. Combining the re-ranked INT results with an unsupervised approach to find associations among interactors, the proposed method can boost the IPT performance. We also implement score computation using dynamic programming, which is faster and more efficient than traditional approaches.

Multistage gene normalization and SVM-based ranking for protein interactor extraction in full-text articles

The interactor normalization task (INT) is to identify genes that play the interactor role in protein-protein interactions (PPIs), to map these genes to unique IDs, and to rank them according to their normalized confidence. INT has two subtasks: gene normalization (GN) and interactor ranking. The main difficulties of INT GN are identifying genes across species and using full papers instead of abstracts. To tackle these problems, we developed a multistage GN algorithm and a ranking method, which exploit information in different parts of a paper. Our system achieved a promising AUC of 0.43471. Using the multistage GN algorithm, we have been able to improve system performance (AUC) by 1.719 percent compared to a one-stage GN algorithm. Our experimental results also show that with full text, versus abstract only, INT AUC performance was 22.6 percent higher.

HypertenGene: extracting key hypertension genes from biomedical literature with position and automatically-generated template features

Background

The genetic factors leading to hypertension have been extensively studied, and large numbers of research papers have been published on the subject. One of hypertension researchers' primary research tasks is to locate key hypertension-related genes in abstracts. However, gathering such information with existing tools is not easy: (1) Searching for articles often returns far too many hits to browse through. (2) The search results do not highlight the hypertension-related genes discovered in the abstract. (3) Even though some text mining services mark up gene names in the abstract, the key genes investigated in a paper are still not distinguished from other genes. To facilitate the information gathering process for hypertension researchers, one solution would be to extract the key hypertension-related genes in each abstract. Three major tasks are involved in the construction of this system: (1) gene and hypertension named entity recognition, (2) section categorization, and (3) gene-hypertension relation extraction.

Results

We first compare the retrieval performance achieved by individually adding template features and position features to the baseline system. Then, the combination of both is examined. We found that using position features can almost double the original AUC score (0.8140vs.0.4936) of the baseline system. However, adding template features only results in marginal improvement (0.0197). Including both improves AUC to 0.8184, indicating that these two sets of features are complementary, and do not have overlapping effects. We then examine the performance in a different domain--diabetes, and the result shows a satisfactory AUC of 0.83.

Conclusion

Our approach successfully exploits template features to recognize true hypertension-related gene mentions and position features to distinguish key genes from other related genes. Templates are automatically generated and checked by biologists to minimize labor costs. Our approach integrates the advantages of machine learning models and pattern matching. To the best of our knowledge, this the first systematic study of extracting hypertension-related genes and the first attempt to create a hypertension-gene relation corpus based on the GAD database. Furthermore, our paper proposes and tests novel features for extracting key hypertension genes, such as relative position, section, and template features, which could also be applied to key-gene extraction for other diseases.

PubMed-EX: a web browser extension to enhance PubMed search with text mining features

PubMed-EX is a browser extension that marks up PubMed search results with additional text-mining information. PubMed-EX's page mark-up, which includes section categorization and gene/disease and relation mark-up, can help researchers to quickly focus on key terms and provide additional information on them. All text processing is performed server-side, freeing up user resources.

AVAILABILITY

PubMed-EX is freely available at http://bws.iis.sinica.edu.tw/PubMed-EX and http://iisr.cse.yzu.edu.tw:8000/PubMed-EX/.

Semi-automatic conversion of BioProp semantic annotation to PASBio annotation

BACKGROUND

Semantic role labeling (SRL) is an important text analysis technique. In SRL, sentences are represented by one or more predicate-argument structures (PAS). Each PAS is composed of a predicate (verb) and several arguments (noun phrases, adverbial phrases, etc.) with different semantic roles, including main arguments (agent or patient) as well as adjunct arguments (time, manner, or location). PropBank is the most widely used PAS corpus and annotation format in the newswire domain. In the biomedical field, however, more detailed and restrictive PAS annotation formats such as PASBio are popular. Unfortunately, due to the lack of an annotated PASBio corpus, no publicly available machine-learning (ML) based SRL systems based on PASBio have been developed. In previous work, we constructed a biomedical corpus based on the PropBank standard called BioProp, on which we developed an ML-based SRL system, BIOSMILE. In this paper, we aim to build a system to convert BIOSMILE's BioProp annotation output to PASBio annotation. Our system consists of BIOSMILE in combination with a BioProp-PASBio rule-based converter, and an additional semi-automatic rule generator.

RESULTS

Our first experiment evaluated our rule-based converter's performance independently from BIOSMILE performance. The converter achieved an F-score of 85.29%. The second experiment evaluated combined system (BIOSMILE + rule-based converter). The system achieved an F-score of 69.08% for PASBio's 29 verbs.

CONCLUSION

Our approach allows PAS conversion between BioProp and PASBio annotation using BIOSMILE alongside our newly developed semi-automatic rule generator and rule-based converter. Our system can match the performance of other state-of-the-art domain-specific ML-based SRL systems and can be easily customized for PASBio application development.

Overview of BioCreative II gene mention recognition

Nineteen teams presented results for the Gene Mention Task at the BioCreative II Workshop. In this task participants designed systems to identify substrings in sentences corresponding to gene name mentions. A variety of different methods were used and the results varied with a highest achieved F₁ score of 0.8721. Here we present brief descriptions of all the methods used and a statistical analysis of the results. We also demonstrate that, by combining the results from all submissions, an F score of 0.9066 is feasible, and furthermore that the best result makes use of the lowest scoring submissions.

BIOSMILE web search: a web application for annotating biomedical entities and relations

BIOSMILE web search (BWS), a web-based NCBI-PubMed search application, which can analyze articles for selected biomedical verbs and give users relational information, such as subject, object, location, manner, time, etc. After receiving keyword query input, BWS retrieves matching PubMed abstracts and lists them along with snippets by order of relevancy to protein–protein interaction. Users can then select articles for further analysis, and BWS will find and mark up biomedical relations in the text. The analysis results can be viewed in the abstract text or in table form. To date, BWS has been field tested by over 30 biologists and questionnaires have shown that subjects are highly satisfied with its capabilities and usability. BWS is accessible free of charge at http://bioservices.cse.yzu.edu.tw/BWS.

Exploiting likely-positive and unlabeled data to improve the identification of protein-protein interaction articles

Background

Experimentally verified protein-protein interactions (PPI) cannot be easily retrieved by researchers unless they are stored in PPI databases. The curation of such databases can be made faster by ranking newly-published articles' relevance to PPI, a task which we approach here by designing a machine-learning-based PPI classifier. All classifiers require labeled data, and the more labeled data available, the more reliable they become. Although many PPI databases with large numbers of labeled articles are available, incorporating these databases into the base training data may actually reduce classification performance since the supplementary databases may not annotate exactly the same PPI types as the base training data. Our first goal in this paper is to find a method of selecting likely positive data from such supplementary databases. Only extracting likely positive data, however, will bias the classification model unless sufficient negative data is also added. Unfortunately, negative data is very hard to obtain because there are no resources that compile such information. Therefore, our second aim is to select such negative data from unlabeled PubMed data. Thirdly, we explore how to exploit these likely positive and negative data. And lastly, we look at the somewhat unrelated question of which term-weighting scheme is most effective for identifying PPI-related articles.

Results

To evaluate the performance of our PPI text classifier, we conducted experiments based on the BioCreAtIvE-II IAS dataset. Our results show that adding likely-labeled data generally increases AUC by 3~6%, indicating better ranking ability. Our experiments also show that our newly-proposed term-weighting scheme has the highest AUC among all common weighting schemes. Our final model achieves an F-measure and AUC 2.9% and 5.0% higher than those of the top-ranking system in the IAS challenge.

Conclusion

Our experiments demonstrate the effectiveness of integrating unlabeled and likely labeled data to augment a PPI text classification system. Our mixed model is suitable for ranking purposes whereas our hierarchical model is better for filtering. In addition, our results indicate that supervised weighting schemes outperform unsupervised ones. Our newly-proposed weighting scheme, TFBRF, which considers documents that do not contain the target word, avoids some of the biases found in traditional weighting schemes. Our experiment results show TFBRF to be the most effective among several other top weighting schemes.

BIOSMILE: a semantic role labeling system for biomedical verbs using a maximum-entropy model with automatically generated template features

Background

Bioinformatics tools for automatic processing of biomedical literature are invaluable for both the design and interpretation of large-scale experiments. Many information extraction (IE) systems that incorporate natural language processing (NLP) techniques have thus been developed for use in the biomedical field. A key IE task in this field is the extraction of biomedical relations, such as protein-protein and gene-disease interactions. However, most biomedical relation extraction systems usually ignore adverbial and prepositional phrases and words identifying location, manner, timing, and condition, which are essential for describing biomedical relations. Semantic role labeling (SRL) is a natural language processing technique that identifies the semantic roles of these words or phrases in sentences and expresses them as predicate-argument structures. We construct a biomedical SRL system called BIOSMILE that uses a maximum entropy (ME) machine-learning model to extract biomedical relations. BIOSMILE is trained on BioProp, our semi-automatic, annotated biomedical proposition bank. Currently, we are focusing on 30 biomedical verbs that are frequently used or considered important for describing molecular events.

Results

To evaluate the performance of BIOSMILE, we conducted two experiments to (1) compare the performance of SRL systems trained on newswire and biomedical corpora; and (2) examine the effects of using biomedical-specific features. The experimental results show that using BioProp improves the F-score of the SRL system by 21.45% over an SRL system that uses a newswire corpus. It is noteworthy that adding automatically generated template features improves the overall F-score by a further 0.52%. Specifically, ArgM-LOC, ArgM-MNR, and Arg2 achieve statistically significant performance improvements of 3.33%, 2.27%, and 1.44%, respectively.

Conclusion

We demonstrate the necessity of using a biomedical proposition bank for training SRL systems in the biomedical domain. Besides the different characteristics of biomedical and newswire sentences, factors such as cross-domain framesets and verb usage variations also influence the performance of SRL systems. For argument classification, we find that NE (named entity) features indicating if the target node matches with NEs are not effective, since NEs may match with a node of the parsing tree that does not have semantic role labels in the training set. We therefore incorporate templates composed of specific words, NE types, and POS tags into the SRL system. As a result, the classification accuracy for adjunct arguments, which is especially important for biomedical SRL, is improved significantly.

NERBio: using selected word conjunctions, term normalization, and global patterns to improve biomedical named entity recognition

Background

Biomedical named entity recognition (Bio-NER) is a challenging problem because, in general, biomedical named entities of the same category (e.g., proteins and genes) do not follow one standard nomenclature. They have many irregularities and sometimes appear in ambiguous contexts. In recent years, machine-learning (ML) approaches have become increasingly common and now represent the cutting edge of Bio-NER technology. This paper addresses three problems faced by ML-based Bio-NER systems. First, most ML approaches usually employ singleton features that comprise one linguistic property (e.g., the current word is capitalized) and at least one class tag (e.g., B-protein, the beginning of a protein name). However, such features may be insufficient in cases where multiple properties must be considered. Adding conjunction features that contain multiple properties can be beneficial, but it would be infeasible to include all conjunction features in an NER model since memory resources are limited and some features are ineffective. To resolve the problem, we use a sequential forward search algorithm to select an effective set of features. Second, variations in the numerical parts of biomedical terms (e.g., "2" in the biomedical term IL2) cause data sparseness and generate many redundant features. In this case, we apply numerical normalization, which solves the problem by replacing all numerals in a term with one representative numeral to help classify named entities. Third, the assignment of NE tags does not depend solely on the target word's closest neighbors, but may depend on words outside the context window (e.g., a context window of five consists of the current word plus two preceding and two subsequent words). We use global patterns generated by the Smith-Waterman local alignment algorithm to identify such structures and modify the results of our ML-based tagger. This is called pattern-based post-processing.

Results

To develop our ML-based Bio-NER system, we employ conditional random fields, which have performed effectively in several well-known tasks, as our underlying ML model. Adding selected conjunction features, applying numerical normalization, and employing pattern-based post-processing improve the F-scores by 1.67%, 1.04%, and 0.57%, respectively. The combined increase of 3.28% yields a total score of 72.98%, which is better than the baseline system that only uses singleton features.

Conclusion

We demonstrate the benefits of using the sequential forward search algorithm to select effective conjunction feature groups. In addition, we show that numerical normalization can effectively reduce the number of redundant and unseen features. Furthermore, the Smith-Waterman local alignment algorithm can help ML-based Bio-NER deal with difficult cases that need longer context windows.

Various criteria in the evaluation of biomedical named entity recognition

Background

Text mining in the biomedical domain is receiving increasing attention. A key component of this process is named entity recognition (NER). Generally speaking, two annotated corpora, GENIA and GENETAG, are most frequently used for training and testing biomedical named entity recognition (Bio-NER) systems. JNLPBA and BioCreAtIvE are two major Bio-NER tasks using these corpora. Both tasks take different approaches to corpus annotation and use different matching criteria to evaluate system performance. This paper details these differences and describes alternative criteria. We then examine the impact of different criteria and annotation schemes on system performance by retesting systems participated in the above two tasks.

Results

To analyze the difference between JNLPBA's and BioCreAtIvE's evaluation, we conduct Experiment 1 to evaluate the top four JNLPBA systems using BioCreAtIvE's classification scheme. We then compare them with the top four BioCreAtIvE systems. Among them, three systems participated in both tasks, and each has an F-score lower on JNLPBA than on BioCreAtIvE. In Experiment 2, we apply hypothesis testing and correlation coefficient to find alternatives to BioCreAtIvE's evaluation scheme. It shows that right-match and left-match criteria have no significant difference with BioCreAtIvE. In Experiment 3, we propose a customized relaxed-match criterion that uses right match and merges JNLPBA's five NE classes into two, which achieves an F-score of 81.5%. In Experiment 4, we evaluate a range of five matching criteria from loose to strict on the top JNLPBA system and examine the percentage of false negatives. Our experiment gives the relative change in precision, recall and F-score as matching criteria are relaxed.

Conclusion

In many applications, biomedical NEs could have several acceptable tags, which might just differ in their left or right boundaries. However, most corpora annotate only one of them. In our experiment, we found that right match and left match can be appropriate alternatives to JNLPBA and BioCreAtIvE's matching criteria. In addition, our relaxed-match criterion demonstrates that users can define their own relaxed criteria that correspond more realistically to their application requirements.