Naomi Sager

Natural Language Processing and the Representation of Clinical Data

OBJECTIVE Develop a representation of clinical observations and actions and a method of processing free-text patient documents to facilitate applications such as quality assurance. DESIGN The Linguistic String Project (LSP) system of New York University utilizes syntactic analysis, augmented by a sublanguage grammar and an information structure that are specific to the clinical narrative, to map free-text documents into a database for querying. MEASUREMENTS Information precision (I-P) and information recall (I-R) were measured for queries for the presence of 13 asthma-health-care quality assurance criteria in a database generated from 59 discharge letters. RESULTS I-P, using counts of major errors only, was 95.7% for the 28-letter training set and 98.6% for the 31-letter test set. I-R, using counts of major omissions only, was 93.9% for the training set and 92.5% for the test set.

Grammatically-based automatic word class formation

Abstract Most previous attempts at producing word classes (thesauri) by statistical analysis have used very limited distributional information such as word co-occurrence in a document or a sentence. This paper describes an automatic procedure which uses the syntactic relations as the basis for grouping words into classes. It forms classes by grouping together nouns that occur as subject (or object) of the same verbs, and similarly by grouping together verbs occurring with the same subject or object. The program was applied to a small corpus of sentences in a subfield of pharmacology. This procedure yielded the word classes for the subfield, in good agreement with the word classes recognized by pharmacologists. The word classes can be used to describe the informational patterns that occur in texts of the subfield, to disambiguate parses of a sentence, and perhaps to improve the performance of current information retrieval systems.

Natural Language Information Formatting: The Automatic Conversion of Texts to a Structured Data Base

Publisher Summary The field of computerized language processing encompasses a wide range of goals and methodologies, ranging from such theoretical objectives as the modeling of human linguistic behavior and human language acquisition, to such goals as machine translation, natural language systems for man–machine communication, and speech recognition. What unites these varied endeavors is the need to come to grips with the special features of natural language as a communication system. The chapter approaches the problems from the viewpoint of information science, with applications envisioned primarily in science information retrieval and data base management. The chapter also attempts to provide a general solution to the following problem: Given a collection of documents on a particular subject written in English or another natural language, how can computer programs arrange the information contained in the documents so that it can be accessed from different points of view for a variety of informational tasks. This assumes that there are underlying common features in the texts that can be made explicit by formal procedures. For the field of information management the solution to this problem would extend the data bases on which data processing functions that have already been automated can operate. The chapter discusses the principles and methods of analysis, and computer programs for information formatting. If indeed it is possible to obtain structured data bases from textual material, then the potential for new computer applications is large. In this section a few applications are sketched with reference to how they would be realized on a data base of computer-formatted medical records.

Syntactic Analysis of Natural Language

Publisher Summary The methods of syntactic analysis, segment a sentence into parts and these in turn into smaller parts. The chapter discusses the relevance of string analysis for computational purposes. It overcomes the problem of discontinuity in a natural way and provides a framework for introducing further linguistic refinements without adding appreciably to the bulk or complexity of the grammar. These features are both because the linguistic string is the least segment of a sentence with respect to which grammatical restrictions can be stated. The chapter describes the left-to-right procedure for string analysis of sentences and the computer program based on it, which emphasizes the means of handling such essential features of language structure as detailed subclass restrictions, coordinate and comparative conjunctions, and syntactic ambiguity. The left-to-right procedure for string analysis of sentences is based on an axiomatic formulation of linguistic string theory, which presents, in terms of particular syntactic categories for words of the language, a set of elementary strings of word categories and rules for combining the elementary strings to form sentences.

Analysis and processing of compact text

This paper describes the characteristics of compact text as revealed in computer analysis of a set of physician notes. Computer processing of the documents was performed using the LSP system for natural language analysis. A numerical breakdown of syntactic and semantic patterns found in the texts is presented. It is found that four major properties of compact text make it possible to process the content of the documents with syntactic procedures that operate on full free text.

An experiment in automated health care evaluation from narrative medical records.

Abstract This paper describes an experiment in the automatic application of health care evaluation criteria to narrative hospital discharge summaries. The processing of the documents was done in two steps. First a three-stage system for natural language analysis accepted the sentences of each discharge summary as input and produced as output a tabular representation of the information in the document. Second, the health care evaluation criteria were implemented as a set of retrieval routines on this tabular representation. A comparison of the computer-generated evaluation results with those obtained by a physician reviewer showed good agreement, indicating that it is possible to process medical narrative automatically and to perform complex, medically significant fact retrieval on narrative input.

Automatic encoding of clinical narrative

An experiment in the automatic encoding of English-language medical data is described. The encoding program has two stages. First, the free-text input is parsed and the information is arranged in a tabular format by a general-purpose natural language processor developed at New York University. Then a simple code-dependent subprogram assigns numerical values to the entries on the basis of the positions the input words occupy in the information format. Results of a blind test of the encoding program using the code employed at Roswell Park Memorial Institute for earliest symptoms of head-neck cancer are presented.

Computerized Language Processing for Multiple Use of Narrative Discharge Summaries

At New York University, computer programs have been developed that convert natural language medical records into a structured data base, i.e. into a table containing the same information as the stored documents. In this form specific information can be quickly retrieved, and summaries of the different kinds of information in the documents can be automatically generated. The automatic conversion of the information from its free-text form to a tabular form is called information formatting. This paper describes the application of the information formatting programs to a small set of pediatric discharge summaries for hospitalizations due to sickle cell disease. The programs created a table of approximately 50 columns in which each different type of information in the documents appeared under a separate heading. From this, a retrieval program extracted instances where symptoms of possible infection preceded symptoms of painful crisis, as suggested by the literature on sickle cell disease. In answer to more detailed queries the program checked the time-order of findings within one document. The potential use of such tables in continuing medical education and other applications in the hospital setting are discussed.

Sublanguage grammers in science information processing

This paper presents the results of an investigation into information structures in natural language science texts. A novel hypothesis was tested; namely, that the literature of a science subfield has characteristic restrictions on lanugage usage which can be used to develop information formats for text sentences in the subfield. The formats provide a standard representation of the specific types of information found in sentences of subfield articles, though a priori semantic categories are not used. The method of sublanguage grammars for obtaining information formats is described. Illustrations are drawn from a sublanguage grammar written for a subfield of pharmacology. Parts of the procedure are computerized or are being implemented.

Processing free-text input to obtain a database of medical information

The Linguistic String Project of New York University has developed computer programs that convert the information in free-text documents of a technical specialty into a structured form suitable for mapping into a relational database. The processing is based upon the restrictions on the use of language that are characteristic of the subject matter and the document type. These restrictions are summarized in a “sublanguage grammar” that provides a set of word classes and formulas corresponding to the objects and relations of interest in the domain. The programs are independent of the particular sublanguage grammar employed. The application to narrative patient records will be described and the applicability of the methods to other domains discussed.