This paper reports on a shared task involving the assignment of emotions to suicide notes. Two features distinguished this task from previous shared tasks in the biomedical domain. One is that it resulted in the corpus of fully anonymized clinical text and annotated suicide notes. This resource is permanently available and will (we hope) facilitate future research. The other key feature of the task is that it required categorization with respect to a large set of labels. The number of participants was larger than in any previous biomedical challenge task. We describe the data production process and the evaluation measures, and give a preliminary analysis of the results. Many systems performed at levels approaching the inter-coder agreement, suggesting that human-like performance on this task is within the reach of currently available technologies.
Sentiment analysis; suicide; suicide notes; natural language processing; computational linguistics; shared task; challenge 2011
We describe the Open University team’s submission to the 2011 i2b2/VA/Cincinnati Medical Natural Language Processing Challenge, Track 2 Shared Task for sentiment analysis in suicide notes. This Shared Task focused on the development of automatic systems that identify, at the sentence level, affective text of 15 specific emotions from suicide notes. We propose a hybrid model that incorporates a number of natural language processing techniques, including lexicon-based keyword spotting, CRF-based emotion cue identification, and machine learning-based emotion classification. The results generated by different techniques are integrated using different vote-based merging strategies. The automated system performed well against the manually-annotated gold standard, and achieved encouraging results with a micro-averaged F-measure score of 61.39% in textual emotion recognition, which was ranked 1st place out of 24 participant teams in this challenge. The results demonstrate that effective emotion recognition by an automated system is possible when a large annotated corpus is available.
emotion recognition; keyword-based model; machine-learning-based model; hybrid model; result integration
The reasons that drive someone to commit suicide are complex and their study has attracted the attention of scientists in different domains. Analyzing this phenomenon could significantly improve the preventive efforts. In this paper we present a method for sentiment analysis of suicide notes submitted to the i2b2/VA/Cincinnati Shared Task 2011. In this task the sentences of 900 suicide notes were labeled with the possible emotions that they reflect. In order to label the sentence with emotions, we propose a hybrid approach which utilizes both rule based and machine learning techniques. To solve the multi class problem a rule-based engine and an SVM model is used for each category. A set of syntactic and semantic features are selected for each sentence to build the rules and train the classifier. The rules are generated manually based on a set of lexical and emotional clues. We propose a new approach to extract the sentence’s clauses and constitutive grammatical elements and to use them in syntactic and semantic feature generation. The method utilizes a novel method to measure the polarity of the sentence based on the extracted grammatical elements, reaching precision of 41.79 with recall of 55.03 for an f-measure of 47.50. The overall mean f-measure of all submissions was 48.75% with a standard deviation of 7%.
NLP; sentiment analysis; emotion classification; polarity measurement; machine learning
In 2007, suicide was the tenth leading cause of death in the U.S. Given the significance of this problem, suicide was the focus of the 2011 Informatics for Integrating Biology and the Bedside (i2b2) Natural Language Processing (NLP) shared task competition (track two). Specifically, the challenge concentrated on sentiment analysis, predicting the presence or absence of 15 emotions (labels) simultaneously in a collection of suicide notes spanning over 70 years. Our team explored multiple approaches combining regular expression-based rules, statistical text mining (STM), and an approach that applies weights to text while accounting for multiple labels. Our best submission used an ensemble of both rules and STM models to achieve a micro-averaged F1 score of 0.5023, slightly above the mean from the 26 teams that competed (0.4875).
sentiment analysis; machine learning; text analysis; i2b2 competition
An ensemble of supervised maximum entropy classifiers can accurately detect and identify sentiments expressed in suicide notes. Using lexical and syntactic features extracted from a training set of externally annotated suicide notes, we trained separate classifiers for each of fifteen pre-specified emotions. This formed part of the 2011 i2b2 NLP Shared Task, Track 2. The precision and recall of these classifiers related strongly with the number of occurrences of each emotion in the training data. Evaluating on previously unseen test data, our best system achieved an F1 score of 0.534.
natural language processing; text analysis; emotion classification; suicide notes
In this paper we report on the approaches that we developed for the 2011 i2b2 Shared Task on Sentiment Analysis of Suicide Notes. We have cast the problem of detecting emotions in suicide notes as a supervised multi-label classification problem. Our classifiers use a variety of features based on (a) lexical indicators, (b) topic scores, and (c) similarity measures. Our best submission has a precision of 0.551, a recall of 0.485, and a F-measure of 0.516.
similarity method; statistical method; sentiment classification; suicide notes
The authors present a system developed for the 2011 i2b2 Challenge on Sentiment Classification, whose aim was to automatically classify sentences in suicide notes using a scheme of 15 topics, mostly emotions. The system combines machine learning with a rule-based methodology. The features used to represent a problem were based on lexico–semantic properties of individual words in addition to regular expressions used to represent patterns of word usage across different topics. A naïve Bayes classifier was trained using the features extracted from the training data consisting of 600 manually annotated suicide notes. Classification was then performed using the naïve Bayes classifier as well as a set of pattern–matching rules. The classification performance was evaluated against a manually prepared gold standard consisting of 300 suicide notes, in which 1,091 out of a total of 2,037 sentences were associated with a total of 1,272 annotations. The competing systems were ranked using the micro-averaged F-measure as the primary evaluation metric. Our system achieved the F-measure of 53% (with 55% precision and 52% recall), which was significantly better than the average performance of 48.75% achieved by the 26 participating teams.
natural language processing; sentiment analysis; topic classification; naïve Bayes classifier
This paper describes the sentiment classification system developed by the Mayo Clinic team for the 2011 I2B2/VA/Cincinnati Natural Language Processing (NLP) Challenge. The sentiment classification task is to assign any pertinent emotion to each sentence in suicide notes. We have implemented three systems that have been trained on suicide notes provided by the I2B2 challenge organizer—a machine learning system, a rule-based system, and a system consisting of a combination of both. Our machine learning system was trained on re-annotated data in which apparently inconsistent emotion assignment was adjusted. Then, the machine learning methods by RIPPER and multinomial Naïve Bayes classifiers, manual pattern matching rules, and the combination of the two systems were tested to determine the emotions within sentences. The combination of the machine learning and rule-based system performed best and produced a micro-average F-score of 0.5640.
sentiment classification; suicidal emotion; natural language processing; machine learning
We present a system to automatically identify emotion-carrying sentences in suicide notes and to detect the specific fine-grained emotion conveyed. With this system, we competed in Track 2 of the 2011 Medical NLP Challenge,14 where the task was to distinguish between fifteen emotion labels, from guilt, sorrow, and hopelessness to hopefulness and happiness.
Since a sentence can be annotated with multiple emotions, we designed a thresholding approach that enables assigning multiple labels to a single instance. We rely on the probability estimates returned by an SVM classifier and experimentally set thresholds on these probabilities. Emotion labels are assigned only if their probability exceeds a certain threshold and if the probability of the sentence being emotion-free is low enough. We show the advantages of this thresholding approach by comparing it to a naïve system that assigns only the most probable label to each test sentence, and to a system trained on emotion-carrying sentences only.
emotion detection; multi-label classification; thresholds; probability estimates
This paper describes a system for automatic emotion classification, developed for the 2011 i2b2 Natural Language Processing Challenge, Track 2. The objective of the shared task was to label suicide notes with 15 relevant emotions on the sentence level. Our system uses 15 SVM models (one for each emotion) using the combination of features that was found to perform best on a given emotion. Features included lemmas and trigram bag of words, and information from semantic resources such as WordNet, SentiWordNet and subjectivity clues. The best-performing system labeled 7 of the 15 emotions and achieved an F-score of 53.31% on the test data.
emotion classification; topic classification; suicide; suicide notes; machine learning
This paper describes the Duluth systems that participated in the Sentiment Analysis track of the i2b2/VA/Cincinnati Children’s 2011 Challenge. The top Duluth system was a rule-based approach derived through manual corpus analysis and the use of measures of association to identify significant ngrams. This performed in the median range of systems, attaining an F-measure of 0.45. The second system was automatically derived from the most frequent bigrams unique to one or two emotions. It achieved an F-measure of 0.36. The third system was the union of the first two, and reached an F-measure of 0.44.
rule-based; sentiment classification; suicide notes
We describe our approach for creating a system able to detect emotions in suicide notes. Motivated by the sparse and imbalanced data as well as the complex annotation scheme, we have considered three hybrid approaches for distinguishing between the different categories. Each of the three approaches combines machine learning with manually derived rules, where the latter target very sparse emotion categories. The first approach considers the task as single label multi-class classification, where an SVM and a CRF classifier are trained to recognise fifteen different categories and their results are combined. Our second approach trains individual binary classifiers (SVM and CRF) for each of the fifteen sentence categories and returns the union of the classifiers as the final result. Finally, our third approach is a combination of binary and multi-class classifiers (SVM and CRF) trained on different subsets of the training data. We considered a number of different feature configurations. All three systems were tested on 300 unseen messages. Our second system had the best performance of the three, yielding an F1 score of 45.6% and a Precision of 60.1% whereas our best Recall (43.6%) was obtained using the third system.
emotion classification; hybrid; suicide; sentence classification
This paper describes the National Research Council of Canada’s submission to the 2011 i2b2 NLP challenge on the detection of emotions in suicide notes. In this task, each sentence of a suicide note is annotated with zero or more emotions, making it a multi-label sentence classification task. We employ two distinct large-margin models capable of handling multiple labels. The first uses one classifier per emotion, and is built to simplify label balance issues and to allow extremely fast development. This approach is very effective, scoring an F-measure of 55.22 and placing fourth in the competition, making it the best system that does not use web-derived statistics or re-annotated training data. Second, we present a latent sequence model, which learns to segment the sentence into a number of emotion regions. This model is intended to gracefully handle sentences that convey multiple thoughts and emotions. Preliminary work with the latent sequence model shows promise, resulting in comparable performance using fewer features.
natural language processing; text analysis; emotion classification; suicide notes; support vector machines; latent variable modeling
This paper reports on the results of an initiative to create and annotate a corpus of suicide notes that can be used for machine learning. Ultimately, the corpus included 1,278 notes that were written by someone who died by suicide. Each note was reviewed by at least three annotators who mapped words or sentences to a schema of emotions. This corpus has already been used for extensive scientific research.
natural language processing; computational linguistics; corpus; suicide
Acquiring and representing biomedical knowledge is an increasingly important component of contemporary bioinformatics. A critical step of the process is to identify and retrieve relevant documents among the vast volume of modern biomedical literature efficiently. In the real world, many information retrieval tasks are difficult because of high data dimensionality and the lack of annotated examples to train a retrieval algorithm. Under such a scenario, the performance of information retrieval algorithms is often unsatisfactory, therefore improvements are needed.
We studied two approaches that enhance the text categorization performance on sparse and high data dimensionality: (1) semantic-preserving dimension reduction by representing text with semantic-enriched features; and (2) augmenting training data with semi-supervised learning. A probabilistic topic model was applied to extract major semantic topics from a corpus of text of interest. The representation of documents was projected from the high-dimensional vocabulary space onto a semantic topic space with reduced dimensionality. A semi-supervised learning algorithm based on graph theory was applied to identify potential positive training cases, which were further used to augment training data. The effects of data transformation and augmentation on text categorization by support vector machine (SVM) were evaluated.
Results and Conclusion
Semantic-enriched data transformation and the pseudo-positive-cases augmented training data enhance the efficiency and performance of text categorization by SVM.
Extracting protein-protein interactions from biomedical literature is an important task in biomedical text mining. Supervised machine learning methods have been used with great success in this task but they tend to suffer from data sparseness because of their restriction to obtain knowledge from limited amount of labelled data. In this work, we study the use of unlabeled biomedical texts to enhance the performance of supervised learning for this task. We use feature coupling generalization (FCG) – a recently proposed semi-supervised learning strategy – to learn an enriched representation of local contexts in sentences from 47 million unlabeled examples and investigate the performance of the new features on AIMED corpus.
The new features generated by FCG achieve a 60.1 F-score and produce significant improvement over supervised baselines. The experimental analysis shows that FCG can utilize well the sparse features which have little effect in supervised learning. The new features perform better in non-linear classifiers than linear ones. We combine the new features with local lexical features, obtaining an F-score of 63.5 on AIMED corpus, which is comparable with the current state-of-the-art results. We also find that simple Boolean lexical features derived only from local contexts are able to achieve competitive results against most syntactic feature/kernel based methods.
FCG creates a lot of opportunities for designing new features, since a lot of sparse features ignored by supervised learning can be utilized well. Interestingly, our results also demonstrate that the state-of-the art performance can be achieved without using any syntactic information in this task.
To create a sentiment classification system for the Fifth i2b2/VA Challenge Track 2, which can identify thirteen subjective categories and two objective categories.
We developed a hybrid system using Support Vector Machine (SVM) classifiers with augmented training data from the Internet. Our system consists of three types of classification-based systems: the first system uses spanning n-gram features for subjective categories, the second one uses bag-of-n-gram features for objective categories, and the third one uses pattern matching for infrequent or subtle emotion categories. The spanning n-gram features are selected by a feature selection algorithm that leverages emotional corpus from weblogs. Special normalization of objective sentences is generalized with shallow parsing and external web knowledge. We utilize three sources of web data: the weblog of LiveJournal which helps to improve the feature selection, the eBay List which assists in special normalization of information and instructions categories, and the suicide project web which provides unlabeled data with similar properties as suicide notes.
The performance is evaluated by the overall micro-averaged precision, recall and F-measure.
Our system achieved an overall micro-averaged F-measure of 0.59. Happiness_peacefulness had the highest F-measure of 0.81. We were ranked as the second best out of 26 competing teams.
Our results indicated that classifying fine-grained sentiments at sentence level is a non-trivial task. It is effective to divide categories into different groups according to their semantic properties. In addition, our system performance benefits from external knowledge extracted from publically available web data of other purposes; performance can be further enhanced when more training data is available.
sentiment analysis; suicide note; spanning n-gram; web data; supervised approach
We report on the original integration of an automatic text categorization pipeline, so-called ToxiCat (Toxicogenomic Categorizer), that we developed to perform biomedical documents classification and prioritization in order to speed up the curation of the Comparative Toxicogenomics Database (CTD). The task can be basically described as a binary classification task, where a scoring function is used to rank a selected set of articles. Then components of a question-answering system are used to extract CTD-specific annotations from the ranked list of articles. The ranking function is generated using a Support Vector Machine, which combines three main modules: an information retrieval engine for MEDLINE (EAGLi), a gene normalization service (NormaGene) developed for a previous BioCreative campaign and finally, a set of answering components and entity recognizer for diseases and chemicals. The main components of the pipeline are publicly available both as web application and web services. The specific integration performed for the BioCreative competition is available via a web user interface at http://pingu.unige.ch:8080/Toxicat.
Detecting uncertain and negative assertions is essential in most BioMedical Text Mining tasks where, in general, the aim is to derive factual knowledge from textual data. This article reports on a corpus annotation project that has produced a freely available resource for research on handling negation and uncertainty in biomedical texts (we call this corpus the BioScope corpus).
The corpus consists of three parts, namely medical free texts, biological full papers and biological scientific abstracts. The dataset contains annotations at the token level for negative and speculative keywords and at the sentence level for their linguistic scope. The annotation process was carried out by two independent linguist annotators and a chief linguist – also responsible for setting up the annotation guidelines – who resolved cases where the annotators disagreed. The resulting corpus consists of more than 20.000 sentences that were considered for annotation and over 10% of them actually contain one (or more) linguistic annotation suggesting negation or uncertainty.
Statistics are reported on corpus size, ambiguity levels and the consistency of annotations. The corpus is accessible for academic purposes and is free of charge. Apart from the intended goal of serving as a common resource for the training, testing and comparing of biomedical Natural Language Processing systems, the corpus is also a good resource for the linguistic analysis of scientific and clinical texts.
We describe the submission entered by SRI International and UC Davis for the I2B2 NLP Challenge Track 2. Our system is based on a machine learning approach and employs a combination of lexical, syntactic, and psycholinguistic features. In addition, we model the sequence and locations of occurrence of emotions found in the notes. We discuss the effect of these features on the emotion annotation task, as well as the nature of the notes themselves. We also explore the use of bootstrapping to help account for what appeared to be annotator fatigue in the data. We conclude a discussion of future avenues for improving the approach for this task, and also discuss how annotations at the word span level may be more appropriate for this task than annotations at the sentence level.
emotion detection; natural language processing; suicide note; psycholinguistic resources
We report on the design and evaluation of an original system to help assignment ICD (International Classification of Disease) codes to clinical narratives. The task is defined as a multi-class multi-document classification task. We combine a set of machine learning and data-poor methods to generate a single automatic text categorizer, which returns a ranked list of ICD codes. The combined ranking system currently obtains a precision of 75% at high ranks and a recall of about 63% for the top twenty returned codes for a theoretical upper bound of about 79% (inter-coder agreement). The performance of the data-poor classifier is weak, whereas the use of temporal features such as anamnesis and prescription contents results in a statistically significant improvement.
We introduce the linguistic annotation of a corpus of 97 full-text biomedical publications, known as the Colorado Richly Annotated Full Text (CRAFT) corpus. We further assess the performance of existing tools for performing sentence splitting, tokenization, syntactic parsing, and named entity recognition on this corpus.
Many biomedical natural language processing systems demonstrated large differences between their previously published results and their performance on the CRAFT corpus when tested with the publicly available models or rule sets. Trainable systems differed widely with respect to their ability to build high-performing models based on this data.
The finding that some systems were able to train high-performing models based on this corpus is additional evidence, beyond high inter-annotator agreement, that the quality of the CRAFT corpus is high. The overall poor performance of various systems indicates that considerable work needs to be done to enable natural language processing systems to work well when the input is full-text journal articles. The CRAFT corpus provides a valuable resource to the biomedical natural language processing community for evaluation and training of new models for biomedical full text publications.
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.
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.
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.
The extraction of complex events from biomedical text is a challenging task and requires in-depth semantic analysis. Previous approaches associate lexical and syntactic resources with ontologies for the semantic analysis, but fall short in testing the benefits from the use of domain knowledge.
We developed a system that deduces implicit events from explicitly expressed events by using inference rules that encode domain knowledge. We evaluated the system with the inference module on three tasks: First, when tested against a corpus with manually annotated events, the inference module of our system contributes 53.2% of correct extractions, but does not cause any incorrect results. Second, the system overall reproduces 33.1% of the transcription regulatory events contained in RegulonDB (up to 85.0% precision) and the inference module is required for 93.8% of the reproduced events. Third, we applied the system with minimum adaptations to the identification of cell activity regulation events, confirming that the inference improves the performance of the system also on this task.
Our research shows that the inference based on domain knowledge plays a significant role in extracting complex events from text. This approach has great potential in recognizing the complex concepts of such biomedical ontologies as Gene Ontology in the literature.
Event extraction following the GENIA Event corpus and BioNLP shared task models has been a considerable focus of recent work in biomedical information extraction. This work includes efforts applying event extraction methods to the entire PubMed literature database, far beyond the narrow subdomains of biomedicine for which annotated resources for extraction method development are available.
In the present study, our aim is to estimate the coverage of all statements of gene/protein associations in PubMed that existing resources for event extraction can provide. We base our analysis on a recently released corpus automatically annotated for gene/protein entities and syntactic analyses covering the entire PubMed, and use named entity co-occurrence, shortest dependency paths and an unlexicalized classifier to identify likely statements of gene/protein associations. A set of high-frequency/high-likelihood association statements are then manually analyzed with reference to the GENIA ontology.
We present a first estimate of the overall coverage of gene/protein associations provided by existing resources for event extraction. Our results suggest that for event-type associations this coverage may be over 90%. We also identify several biologically significant associations of genes and proteins that are not addressed by these resources, suggesting directions for further extension of extraction coverage.