Challenges In Natural Language Processing Research Articles

De-identification of clinical notes via recurrent neural network and conditional random field

De-identification, identifying information from data, such as protected health information (PHI) present in clinical data, is a critical step to enable data to be shared or published. The 2016 Centers of Excellence in Genomic Science (CEGS) Neuropsychiatric Genome-scale and RDOC Individualized Domains (N-GRID) clinical natural language processing (NLP) challenge contains a de-identification track in de-identifying electronic medical records (EMRs) (i.e., track 1). The challenge organizers provide 1000 annotated mental health records for this track, 600 out of which are used as a training set and 400 as a test set. We develop a hybrid system for the de-identification task on the training set. Firstly, four individual subsystems, that is, a subsystem based on bidirectional LSTM (long-short term memory, a variant of recurrent neural network), a subsystem-based on bidirectional LSTM with features, a subsystem based on conditional random field (CRF) and a rule-based subsystem, are used to identify PHI instances. Then, an ensemble learning-based classifiers is deployed to combine all PHI instances predicted by above three machine learning-based subsystems. Finally, the results of the ensemble learning-based classifier and the rule-based subsystem are merged together. Experiments conducted on the official test set show that our system achieves the highest micro F1-scores of 93.07%, 91.43% and 95.23% under the “token”, “strict” and “binary token” criteria respectively, ranking first in the 2016 CEGS N-GRID NLP challenge. In addition, on the dataset of 2014 i2b2 NLP challenge, our system achieves the highest micro F1-scores of 96.98%, 95.11% and 98.28% under the “token”, “strict” and “binary token” criteria respectively, outperforming other state-of-the-art systems. All these experiments prove the effectiveness of our proposed method.

Automated de-identification of clinical free-text

ABSTRACTObjectivesIncreasing interest to use unstructured electronic patient records for research has attracted attention to automated de-identification methods to conduct large scale removal of Personal Identifiable Information (PII). PII mainly include identifiable information such as person names, dates (e.g., date of birth), reference numbers (e.g., hospital number, NHS number), locations (e.g., hospital names, addresses), contacts (e.g., telephone, e-mail), occupation, age, and other identity information (ethnical, religion, sexual) mentioned in a private context. 
 De-identification of clinical free-text remains crucial to enable large-scale data access for health research while adhering to legal (Data Protection Act 1998) and ethical obligations. Here we present a computational method developed to automatically remove PII from clinical text.
 ApproachIn order to automatically identify PII in clinical text, we have developed and validated a Natural Language Processing (NLP) method which combine knowledge- (lexical dictionaries and rules) and data-driven (linear-chain conditional random fields) techniques. In addition, we have designed a novel two-pass recognition approach that uses the output of the initial pass to create patient-level and run-time dictionaries used to identify PII mentions that lack specific contextual clues considered by the initial entity extraction modules. The labelled data used to model and validate our techniques were generated using six human annotators and two distinct types of free-text from The Christie NHS Foundation Trust: (1) clinical correspondence (400 documents) and (2) clinical notes (1,300 documents).
 ResultsThe de-identification approach was developed and validated using a 60/40 percent split between the development and test datasets. The preliminary results show that our method achieves 97% and 93% token-level F1-measure on clinical correspondence and clinical notes respectively. In addition, the proposed two-pass recognition method was found particularly effective for longitudinal records. Notably, the performances are comparable to human benchmarks (using inter annotator agreements) of 97% and 90% F1 respectively.
 ConclusionsWe have developed and validated a state-of-the-art method that matches human benchmarks in identification and removal of PII from free-text clinical records. The method has been further validated across multiple institutions and countries (United States and United Kingdom), where we have identified a notable NLP challenge of cross-dataset adaption and have proposed using active learning methods to address this problem. The algorithm, including an active learning component, will be provided as open source to the healthcare community.

Challenges in adapting existing clinical natural language processing systems to multiple, diverse health care settings

Widespread application of clinical natural language processing (NLP) systems requires taking existing NLP systems and adapting them to diverse and heterogeneous settings. We describe the challenges faced and lessons learned in adapting an existing NLP system for measuring colonoscopy quality. Colonoscopy and pathology reports from 4 settings during 2013-2015, varying by geographic location, practice type, compensation structure, and electronic health record. Though successful, adaptation required considerably more time and effort than anticipated. Typical NLP challenges in assembling corpora, diverse report structures, and idiosyncratic linguistic content were greatly magnified. Strategies for addressing adaptation challenges include assessing site-specific diversity, setting realistic timelines, leveraging local electronic health record expertise, and undertaking extensive iterative development. More research is needed on how to make it easier to adapt NLP systems to new clinical settings. A key challenge in widespread application of NLP is adapting existing systems to new clinical settings.

Multilingual Metaphor Processing: Experiments with Semi-Supervised and Unsupervised Learning

Highly frequent in language and communication, metaphor represents a significant challenge for Natural Language Processing (NLP) applications. Computational work on metaphor has traditionally evolved around the use of hand-coded knowledge, making the systems hard to scale. Recent years have witnessed a rise in statistical approaches to metaphor processing. However, these approaches often require extensive human annotation effort and are predominantly evaluated within a limited domain. In contrast, we experiment with weakly supervised and unsupervised techniques—with little or no annotation—to generalize higher-level mechanisms of metaphor from distributional properties of concepts. We investigate different levels and types of supervision (learning from linguistic examples vs. learning from a given set of metaphorical mappings vs. learning without annotation) in flat and hierarchical, unconstrained and constrained clustering settings. Our aim is to identify the optimal type of supervision for a learning algorithm that discovers patterns of metaphorical association from text. In order to investigate the scalability and adaptability of our models, we applied them to data in three languages from different language groups—English, Spanish, and Russian—achieving state-of-the-art results with little supervision. Finally, we demonstrate that statistical methods can facilitate and scale up cross-linguistic research on metaphor.

Chinese Word Sense Disambiguation using a LSTM

Word sense disambiguation (WSD) is a challenging natural language processing (NLP) problem. We propose a new strategy for WSD, which at first replaces the interesting word in a sentence by the different synonyms corresponding to the different meanings, and then justify whether the transformed sentence is “legal”. A legal sentence is still legal after one or more word are replaced by other ones with the same meaning. A long short-term memory (LSTM) network-based model is proposed to perform the sentence/text classification. Furthermore, we build a Chinese WSD dataset based on HIT-CIR Tongyici Cilin (Extended) dataset. The model is evaluated on the new dataset and achieves better performance than the state-of-the-art.

Combination of Deep Recurrent Neural Networks and Conditional Random Fields for Extracting Adverse Drug Reactions from User Reviews.

Adverse drug reactions (ADRs) are an essential part of the analysis of drug use, measuring drug use benefits, and making policy decisions. Traditional channels for identifying ADRs are reliable but very slow and only produce a small amount of data. Text reviews, either on specialized web sites or in general-purpose social networks, may lead to a data source of unprecedented size, but identifying ADRs in free-form text is a challenging natural language processing problem. In this work, we propose a novel model for this problem, uniting recurrent neural architectures and conditional random fields. We evaluate our model with a comprehensive experimental study, showing improvements over state-of-the-art methods of ADR extraction.

Sortal anaphora resolution to enhance relation extraction from biomedical literature.

BackgroundEntity coreference is common in biomedical literature and it can affect text understanding systems that rely on accurate identification of named entities, such as relation extraction and automatic summarization. Coreference resolution is a foundational yet challenging natural language processing task which, if performed successfully, is likely to enhance such systems significantly. In this paper, we propose a semantically oriented, rule-based method to resolve sortal anaphora, a specific type of coreference that forms the majority of coreference instances in biomedical literature. The method addresses all entity types and relies on linguistic components of SemRep, a broad-coverage biomedical relation extraction system. It has been incorporated into SemRep, extending its core semantic interpretation capability from sentence level to discourse level.ResultsWe evaluated our sortal anaphora resolution method in several ways. The first evaluation specifically focused on sortal anaphora relations. Our methodology achieved a F1 score of 59.6 on the test portion of a manually annotated corpus of 320 Medline abstracts, a 4-fold improvement over the baseline method. Investigating the impact of sortal anaphora resolution on relation extraction, we found that the overall effect was positive, with 50 % of the changes involving uninformative relations being replaced by more specific and informative ones, while 35 % of the changes had no effect, and only 15 % were negative. We estimate that anaphora resolution results in changes in about 1.5 % of approximately 82 million semantic relations extracted from the entire PubMed.ConclusionsOur results demonstrate that a heavily semantic approach to sortal anaphora resolution is largely effective for biomedical literature. Our evaluation and error analysis highlight some areas for further improvements, such as coordination processing and intra-sentential antecedent selection.Electronic supplementary materialThe online version of this article (doi:10.1186/s12859-016-1009-6) contains supplementary material, which is available to authorized users.

Modified Random Walk Algorithm to Improve the Efficiency of Word Sense Disambiguation (WSD)

Natural language processing (NLP) is a field in computer science, artificial intelligence and the linguistics which mainly concentrates on the interactions between human languages (natural language) and the computer. One of the main challenges in NLP is ambiguity. Every language is ambiguous in nature, in the way that one word has multiple meaning and multiple words have same meaning. The ambiguities are generally categorized into two groups: lexical and structural ambiguities. Lexical ambiguity arises where there are two or more possible meaning for a single word. Structural ambiguities appear when a given sentence is interpreted in more than one way due to ambiguous sentence structure. Word Sense Disambiguation (WSD) is defined as the task of finding the correct sense of a word in a specific context. This paper presents our preliminary work towards building WSD system by constructing a corpus. We include a detailed analysis of the factor that affects the WSD algorithm and propose a modified algorithm based on random walk algorithm and compare the working of each of these algorithms

An Algorithm of Word Indexing Model for Document Summarization based on Perspective of Document

Natural language processing (NLP) is an area of computer science, artificial intelligence, and computational linguistics connected with the communications between computers and natural languages. There are many challenges in NLP involve natural language understanding, that is, enabling computers to derive meaning from human or natural language input, and others involve natural language generation. Document summarization is a part of it. Many different classes of such process based on machine learning are developed. In researches earlier document summarization mostly use the similarity between sentences in the document to extract the most significant sentences. The documents as well as the sentences are indexed using traditional term indexing measures, which do not take the context into consideration. The resulting indexing weights are used to compute the sentence similarity matrix. The proposed sentence similarity measure has been used with the baseline graph-based ranking models for sentence extraction.

Sentiment analysis and the complex natural language

AbstractThere is huge amount of content produced online by amateur authors, covering a large variety of topics. Sentiment analysis (SA) extracts and aggregates users’ sentiments towards a target entity. Machine learning (ML) techniques are frequently used as the natural language data is in abundance and has definite patterns. ML techniques adapt to domain specific solution at high accuracy depending upon the feature set used. The lexicon-based techniques, using external dictionary, are independent of data to prevent overfitting but they miss context too in specialized domains. Corpus-based statistical techniques require large data to stabilize. Complex network based techniques are highly resourceful, preserving order, proximity, context and relationships. Recent applications developed incorporate the platform specific structural information i.e. meta-data. New sub-domains are introduced as influence analysis, bias analysis, and data leakage analysis. The nature of data is also evolving where transcribed customer-agent phone conversation are also used for sentiment analysis. This paper reviews sentiment analysis techniques and highlight the need to address natural language processing (NLP) specific open challenges. Without resolving the complex NLP challenges, ML techniques cannot make considerable advancements. The open issues and challenges in the area are discussed, stressing on the need of standard datasets and evaluation methodology. It also emphasized on the need of better language models that could capture context and proximity.

A Coreference Resolution Approach using Morphological Features in Arabic

Coreference resolution is considered one of the challenges in natural language processing. It is an important task that includes determining which pronouns are referring to which entities. Most of the earlier approaches for coreference resolution are rule-based or machine learning approaches. However, these types of approaches have many limitations especially with Arabic language. In this paper, a different approach to coreference resolution is presented. The approach uses morphological features and dependency trees instead. It has fivestages, which overcomes the limitations of using annotated datasets for learning or a set of rules. The approach was evaluatedusing our own customized annotated dataset and “AnATAr” dataset. The evaluation show encouraging results with average F1 score of 89%.

A study of active learning methods for named entity recognition in clinical text

ObjectivesNamed entity recognition (NER), a sequential labeling task, is one of the fundamental tasks for building clinical natural language processing (NLP) systems. Machine learning (ML) based approaches can achieve good performance, but they often require large amounts of annotated samples, which are expensive to build due to the requirement of domain experts in annotation. Active learning (AL), a sample selection approach integrated with supervised ML, aims to minimize the annotation cost while maximizing the performance of ML-based models. In this study, our goal was to develop and evaluate both existing and new AL methods for a clinical NER task to identify concepts of medical problems, treatments, and lab tests from the clinical notes. MethodsUsing the annotated NER corpus from the 2010 i2b2/VA NLP challenge that contained 349 clinical documents with 20,423 unique sentences, we simulated AL experiments using a number of existing and novel algorithms in three different categories including uncertainty-based, diversity-based, and baseline sampling strategies. They were compared with the passive learning that uses random sampling. Learning curves that plot performance of the NER model against the estimated annotation cost (based on number of sentences or words in the training set) were generated to evaluate different active learning and the passive learning methods and the area under the learning curve (ALC) score was computed. ResultsBased on the learning curves of F-measure vs. number of sentences, uncertainty sampling algorithms outperformed all other methods in ALC. Most diversity-based methods also performed better than random sampling in ALC. To achieve an F-measure of 0.80, the best method based on uncertainty sampling could save 66% annotations in sentences, as compared to random sampling. For the learning curves of F-measure vs. number of words, uncertainty sampling methods again outperformed all other methods in ALC. To achieve 0.80 in F-measure, in comparison to random sampling, the best uncertainty based method saved 42% annotations in words. But the best diversity based method reduced only 7% annotation effort. ConclusionIn the simulated setting, AL methods, particularly uncertainty-sampling based approaches, seemed to significantly save annotation cost for the clinical NER task. The actual benefit of active learning in clinical NER should be further evaluated in a real-time setting.

An automatic system to identify heart disease risk factors in clinical texts over time

Despite recent progress in prediction and prevention, heart disease remains a leading cause of death. One preliminary step in heart disease prediction and prevention is risk factor identification. Many studies have been proposed to identify risk factors associated with heart disease; however, none have attempted to identify all risk factors. In 2014, the National Center of Informatics for Integrating Biology and Beside (i2b2) issued a clinical natural language processing (NLP) challenge that involved a track (track 2) for identifying heart disease risk factors in clinical texts over time. This track aimed to identify medically relevant information related to heart disease risk and track the progression over sets of longitudinal patient medical records. Identification of tags and attributes associated with disease presence and progression, risk factors, and medications in patient medical history were required. Our participation led to development of a hybrid pipeline system based on both machine learning-based and rule-based approaches. Evaluation using the challenge corpus revealed that our system achieved an F1-score of 92.68%, making it the top-ranked system (without additional annotations) of the 2014 i2b2 clinical NLP challenge.

Algorithm for Text to Graph Conversion and Summarizing Using NLP: A New Approach for Business Solutions

Text can be analysed by splitting the text and extracting the keywords .These may be represented as summaries, tabular representation, graphical forms, and images. In order to provide a solution to large amount of information present in textual format led to a research of extracting the text and transforming the unstructured form to a structured format. The paper presents the importance of Natural Language Processing (NLP) and its two interesting applications in Python Language: 1. Automatic text summarization [Domain: Newspaper Articles] 2. Text to Graph Conversion [Domain: Stock news]. The main challenge in NLP is natural language understanding i.e. deriving meaning from human or natural language input which is done using regular expressions, artificial intelligence and database concepts. Automatic Summarization tool converts the newspaper articles into summary on the basis of frequency of words in the text. Text to Graph Converter takes in the input as stock article, tokenize them on various index (points and percent) and time and then tokens are mapped to graph. This paper proposes a business solution for users for effective time management.

Challenges in clinical natural language processing for automated disorder normalization

BackgroundIdentifying key variables such as disorders within the clinical narratives in electronic health records has wide-ranging applications within clinical practice and biomedical research. Previous research has demonstrated reduced performance of disorder named entity recognition (NER) and normalization (or grounding) in clinical narratives than in biomedical publications. In this work, we aim to identify the cause for this performance difference and introduce general solutions. MethodsWe use closure properties to compare the richness of the vocabulary in clinical narrative text to biomedical publications. We approach both disorder NER and normalization using machine learning methodologies. Our NER methodology is based on linear-chain conditional random fields with a rich feature approach, and we introduce several improvements to enhance the lexical knowledge of the NER system. Our normalization method – never previously applied to clinical data – uses pairwise learning to rank to automatically learn term variation directly from the training data. ResultsWe find that while the size of the overall vocabulary is similar between clinical narrative and biomedical publications, clinical narrative uses a richer terminology to describe disorders than publications. We apply our system, DNorm-C, to locate disorder mentions and in the clinical narratives from the recent ShARe/CLEF eHealth Task. For NER (strict span-only), our system achieves precision=0.797, recall=0.713, f-score=0.753. For the normalization task (strict span+concept) it achieves precision=0.712, recall=0.637, f-score=0.672. The improvements described in this article increase the NER f-score by 0.039 and the normalization f-score by 0.036. We also describe a high recall version of the NER, which increases the normalization recall to as high as 0.744, albeit with reduced precision. DiscussionWe perform an error analysis, demonstrating that NER errors outnumber normalization errors by more than 4-to-1. Abbreviations and acronyms are found to be frequent causes of error, in addition to the mentions the annotators were not able to identify within the scope of the controlled vocabulary. ConclusionDisorder mentions in text from clinical narratives use a rich vocabulary that results in high term variation, which we believe to be one of the primary causes of reduced performance in clinical narrative. We show that pairwise learning to rank offers high performance in this context, and introduce several lexical enhancements – generalizable to other clinical NER tasks – that improve the ability of the NER system to handle this variation. DNorm-C is a high performing, open source system for disorders in clinical text, and a promising step toward NER and normalization methods that are trainable to a wide variety of domains and entities. (DNorm-C is open source software, and is available with a trained model at the DNorm demonstration website: http://www.ncbi.nlm.nih.gov/CBBresearch/Lu/Demo/tmTools/#DNorm.)

Using local lexicalized rules to identify heart disease risk factors in clinical notes

Heart disease is the leading cause of death globally and a significant part of the human population lives with it. A number of risk factors have been recognized as contributing to the disease, including obesity, coronary artery disease (CAD), hypertension, hyperlipidemia, diabetes, smoking, and family history of premature CAD. This paper describes and evaluates a methodology to extract mentions of such risk factors from diabetic clinical notes, which was a task of the i2b2/UTHealth 2014 Challenge in Natural Language Processing for Clinical Data. The methodology is knowledge-driven and the system implements local lexicalized rules (based on syntactical patterns observed in notes) combined with manually constructed dictionaries that characterize the domain. A part of the task was also to detect the time interval in which the risk factors were present in a patient.The system was applied to an evaluation set of 514 unseen notes and achieved a micro-average F-score of 88% (with 86% precision and 90% recall). While the identification of CAD family history, medication and some of the related disease factors (e.g. hypertension, diabetes, hyperlipidemia) showed quite good results, the identification of CAD-specific indicators proved to be more challenging (F-score of 74%). Overall, the results are encouraging and suggested that automated text mining methods can be used to process clinical notes to identify risk factors and monitor progression of heart disease on a large-scale, providing necessary data for clinical and epidemiological studies.

Automatic de-identification of electronic medical records using token-level and character-level conditional random fields

De-identification, identifying and removing all protected health information (PHI) present in clinical data including electronic medical records (EMRs), is a critical step in making clinical data publicly available. The 2014 i2b2 (Center of Informatics for Integrating Biology and Bedside) clinical natural language processing (NLP) challenge sets up a track for de-identification (track 1). In this study, we propose a hybrid system based on both machine learning and rule approaches for the de-identification track. In our system, PHI instances are first identified by two (token-level and character-level) conditional random fields (CRFs) and a rule-based classifier, and then are merged by some rules. Experiments conducted on the i2b2 corpus show that our system submitted for the challenge achieves the highest micro F-scores of 94.64%, 91.24% and 91.63% under the “token”, “strict” and “relaxed” criteria respectively, which is among top-ranked systems of the 2014 i2b2 challenge. After integrating some refined localization dictionaries, our system is further improved with F-scores of 94.83%, 91.57% and 91.95% under the “token”, “strict” and “relaxed” criteria respectively.

Query-oriented unsupervised multi-document summarization via deep learning model

Capturing the compositional process from words to documents is a key challenge in natural language processing and information retrieval. Extractive style query-oriented multi-document summarization generates a summary by extracting a proper set of sentences from multiple documents based on pre-given query. This paper proposes a novel document summarization framework based on deep learning model, which has been shown outstanding extraction ability in many real-world applications. The framework consists of three parts: concepts extraction, summary generation, and reconstruction validation. A new query-oriented extraction technique is proposed to extract information distributed in multiple documents. Then, the whole deep architecture is fine-tuned by minimizing the information loss in reconstruction validation. According to the concepts extracted from deep architecture layer by layer, dynamic programming is used to seek most informative set of sentences for the summary. Experiment on three benchmark datasets (DUC 2005, 2006, and 2007) assess and confirm the effectiveness of the proposed framework and algorithms. Experiment results show that the proposed method outperforms state-of-the-art extractive summarization approaches. Moreover, we also provide the statistical analysis of query words based on Amazon’s Mechanical Turk (MTurk) crowdsourcing platform. There exists underlying relationships from topic words to the content which can contribute to summarization task.

Parsing clinical text: how good are the state-of-the-art parsers?

BackgroundParsing, which generates a syntactic structure of a sentence (a parse tree), is a critical component of natural language processing (NLP) research in any domain including medicine. Although parsers developed in the general English domain, such as the Stanford parser, have been applied to clinical text, there are no formal evaluations and comparisons of their performance in the medical domain.MethodsIn this study, we investigated the performance of three state-of-the-art parsers: the Stanford parser, the Bikel parser, and the Charniak parser, using following two datasets: (1) A Treebank containing 1,100 sentences that were randomly selected from progress notes used in the 2010 i2b2 NLP challenge and manually annotated according to a Penn Treebank based guideline; and (2) the MiPACQ Treebank, which is developed based on pathology notes and clinical notes, containing 13,091 sentences. We conducted three experiments on both datasets. First, we measured the performance of the three state-of-the-art parsers on the clinical Treebanks with their default settings. Then we re-trained the parsers using the clinical Treebanks and evaluated their performance using the 10-fold cross validation method. Finally we re-trained the parsers by combining the clinical Treebanks with the Penn Treebank.ResultsOur results showed that the original parsers achieved lower performance in clinical text (Bracketing F-measure in the range of 66.6%-70.3%) compared to general English text. After retraining on the clinical Treebank, all parsers achieved better performance, with the best performance from the Stanford parser that reached the highest Bracketing F-measure of 73.68% on progress notes and 83.72% on the MiPACQ corpus using 10-fold cross validation. When the combined clinical Treebanks and Penn Treebank was used, of the three parsers, the Charniak parser achieved the highest Bracketing F-measure of 73.53% on progress notes and the Stanford parser reached the highest F-measure of 84.15% on the MiPACQ corpus.ConclusionsOur study demonstrates that re-training using clinical Treebanks is critical for improving general English parsers' performance on clinical text, and combining clinical and open domain corpora might achieve optimal performance for parsing clinical text.

A Hybrid Approach to Pronominal Anaphora Resolution in Arabic

One of the challenges in natural language processing is to determine which pronouns to be referred to their intended referents in the discourse. Performing anaphora resolution is considered as an important task for a number of natural language processing applications such as information extraction, question answering and text summarization. Most of the earlier works of anaphora resolution have been applied to English and other languages. However, the work done in Arabic is not sufficiently studied. In this study, a hybrid approach that combines different architectures for resolving pronominal anaphora in Arabic language is presented. The hybrid model adopted the strategy based on the combination of a rule-based and machine learning approach. The collection of anaphora and respective possible antecedents was identified in a rule-based manner with morphological information taken into account. In addition, the selection of the most probable candidate as the antecedent of the anaphor was done by machine learning based on a k-Nearest Neighbor (k-NN) approach. In this study, the appropriate features to be used in this task were determined and their effect on the performance of anaphora resolution was investigated. Experiments of the proposed method were performed using the corpus of the Quran annotated with pronominal anaphora. The experimental results indicate that the proposed hybrid approach is completely reasonable and feasible for Arabic pronominal anaphora resolution.