Biomedical subjects

Serguei V Pakhomov

Publications and source records attributed to Serguei V Pakhomov.

8 recordsLinked to original sources

PubMed2005

Developing a corpus of clinical notes manually annotated for part-of-speech.

PURPOSE: This paper presents a project whose main goal is to construct a corpus of clinical text manually annotated for part-of-speech (POS) information. We describe and discuss the process of training three domain experts to perform linguistic annotation. METHODS: Three domain experts were trained to perform manual annotation of a corpus of clinical notes. A part of this corpus was combined with the Penn Treebank corpus of general purpose English text and another part was set aside for testing. The corpora were then used for training and testing statistical part-of-speech taggers. We list some of the challenges as well as encouraging results pertaining to inter-rater agreement and consistency of annotation. RESULTS: We used the Trigrams'n'Tags (TnT) [T. Brants, TnT-a statistical part-of-speech tagger, In: Proceedings of NAACL/ANLP-2000 Symposium, 2000] tagger trained on general English data to achieve 89.79% correctness. The same tagger trained on a portion of the medical data annotated for this project improved the performance to 94.69%. Furthermore, we find that discriminating between different types of discourse represented by different sections of clinical text may be very beneficial to improve correctness of POS tagging. CONCLUSION: Our preliminary experimental results indicate the necessity for adapting state-of-the-art POS taggers to the sublanguage domain of clinical text.

Abstracting and Indexing↗

PubMed2005

Domain-specific language models and lexicons for tagging.

Accurate and reliable part-of-speech tagging is useful for many Natural Language Processing (NLP) tasks that form the foundation of NLP-based approaches to information retrieval and data mining. In general, large annotated corpora are necessary to achieve desired part-of-speech tagger accuracy. We show that a large annotated general-English corpus is not sufficient for building a part-of-speech tagger model adequate for tagging documents from the medical domain. However, adding a quite small domain-specific corpus to a large general-English one boosts performance to over 92% accuracy from 87% in our studies. We also suggest a number of characteristics to quantify the similarities between a training corpus and the test data. These results give guidance for creating an appropriate corpus for building a part-of-speech tagger model that gives satisfactory accuracy results on a new domain at a relatively small cost.

Humans↗

PubMed2005

Prospective recruitment of patients with congestive heart failure using an ad-hoc binary classifier.

This paper addresses a very specific problem of identifying patients diagnosed with a specific condition for potential recruitment in a clinical trial or an epidemiological study. We present a simple machine learning method for identifying patients diagnosed with congestive heart failure and other related conditions by automatically classifying clinical notes dictated at Mayo Clinic. This method relies on an automatic classifier trained on comparable amounts of positive and negative samples of clinical notes previously categorized by human experts. The documents are represented as feature vectors, where features are a mix of demographic information as well as single words and concept mappings to MeSH and HICDA classification systems. We compare two simple and efficient classification algorithms (Naïve Bayes and Perceptron) and a baseline term spotting method with respect to their accuracy and recall on positive samples. Depending on the test set, we find that Naïve Bayes yields better recall on positive samples (95 vs. 86%) but worse accuracy than Perceptron (57 vs. 65%). Both algorithms perform better than the baseline with recall on positive samples of 71% and accuracy of 54%.

Artificial Intelligence↗

PubMed2005

Medical facts to support inferencing in natural language processing.

We report on the use of medical facts to support the enhancement of natural language processing of biomedical text. Inferencing in semantic interpretation depends on a fact repository as well as an ontology. We used statistical methods to construct a repository of drug-disorder co-occurrences from a large collection of clinical notes, and this resource is used to validate inferences automatically drawn during semantic interpretation of Medline citations about pharmacologic interventions for disease. We evaluated the results against a published reference standard for treatment of diseases.

Disease↗

PubMed2004

Using compound codes for automatic classification of clinical diagnoses.

Classification of diagnoses (a.k.a. coding) is the central part of current concept based medical IR systems. Some classification systems contain over 30,000 distinct codes which makes classifying clinical documents a time consuming labor intensive and error prone process. This paper presents a simple methodology for cleaning up and reusing existing manually coded diagnostic statements mainly extracted from clinical notes to build predictive models using a sparse-feature implementation of a Naïve Bayes classifier. One of the problems addressed is that diagnostic statements often contain several diagnoses and are assigned several codes resulting in a multi-class classification problem. We investigate one possible way of addressing this problem by introducing compound (multiple code) categories. We present experimental results of classifying >16,000 randomly selected diagnostic strings into 19 top level categories. A small improvement (3%) with using compound categories over simple categories indicates that using multiple code categories is a promising solution, although clearly in need of further research and refinement.

Abstracting and Indexing↗

PubMed2004

A corpus driven approach applying the "frame semantic" method for modeling functional status terminology.

In an effort to unearth semantic models that could prove fruitful to functional-status terminology development we applied the "frame semantic" method, derived from the linguistic theory of thematic roles currently exemplified in the Berkeley "FrameNet" Project. Full descriptive sentences with functional-status conceptual meaning were derived from structured content within a corpus of questionnaire assessment instruments commonly used in clinical practice for functional-status assessment. Syntactic components in those sentences were delineated through manual annotation and mark-up. The annotated syntactic constituents were tagged as frame elements according to their semantic role within the context of the derived functional-status expression. Through this process generalizable "semantic frames" were elaborated with recurring "frame elements". The "frame semantic" method as an approach to rendering semantic models for functional-status terminology development and its use as a basis for machine recognition of functional status data in clinical narratives are discussed.

Information Science↗

PubMed2003

A data-driven approach for extracting "the most specific term" for ontology development.

We present a data-driven approach to extract the "most specific" terms relevant to an ontology of functioning, disability and health. The algorithm is a combination of statistical and linguistic approaches. The statistical filter is based on the frequency of the content words in a given text string; the linguistic heuristic is an extension of existing algorithms but goes beyond noun phrases and is formulated as a "complete syntactic node". Thus, it can be applied to any syntactic node of interest in the particular domain. Two test sets were marked by three experts. Test set 1 is a well-constructed text from pain abstracts; test set 2 is actual medical reports. Results are reported as recall, precision, F-score and rate of valid terms in false positives. A limitation of the current research is the relatively small test set.

Algorithms↗

PubMed2002

Maximum entropy modeling for mining patient medication status from free text.

Using a classification scheme of patient medication status we sought to recognize and categorize medications mentioned in the unrestricted text of clinical documents generated in clinical practice. The categories refer to the patient's status with respect to the medication such as discontinuation, start or initiation, and continuation of a given medication. This categorization is performed with a machine learning technique, Maximum Entropy (ME), that is well suited to incorporating heterogeneous sources of information necessary for classifying patient's medication status. We use hand labeled training data to generate ME models and test 5 different training feature sets. Our results show that the most optimal feature set includes a combination of the following: two words preceding and following the mention of the drug, the subject of the sentence in which the drug mention occurs, the 2 words following the subject, and a binary feature vector of lexicalized semantic cues indicative of medication status or its change. The average predictive power of a model trained on these features is approximately 89%.

Artificial Intelligence↗