JP5151412B2 - Notation fluctuation analyzer - Google Patents

Description

  The present invention relates to a notation fluctuation analysis apparatus and program for analyzing notation fluctuations having the same concept but different notations.

  In general, there are a plurality of notations (words) for the same concept, which is one cause of hindering communication. It is almost clear that this difference in notation (notation fluctuation) is due to differences in geography, time, and education. In other words, documents that use the same language are similar in geography, time, and author. The background of a set of documents that would have used the same language is called “site” in the following.

Here, various inventions have been proposed with respect to the notation.
For example, a method for extracting synonym pairs by utilizing the fact that synonyms are often shown in parentheses immediately after the notation has been proposed (see Patent Document 1).
For example, an invention has been proposed in which synonymous relations between words are estimated using statistical information of words in the vicinity of those words in a document in which they appear (see Patent Document 2).
For example, there has been proposed an invention for finding a notation fluctuation using documents having the same meaning (monolingual parallel corpus) (see Patent Document 3).

JP 11-328205 A Japanese Patent Application Laid-Open No. 11-212975 JP 2006-251843 A

  Here, in the medical field, the notation of technical terms used internally by hospitals is different, even if the department is different within the same hospital, the notation is different, and the name of the disease used in the past is now called differently It is happening that doctors in the same department have different notations because they are from different universities. Although medical care is an act of solving problems through cooperation between medical institutions such as multiple institutions and doctors and nurses, communication errors due to these differences in notation have become one of the causes of medical errors. Yes.

  Therefore, a glossary has been created to unify terms used in hospitals, and a terminology such as a standard disease name master has been created to unify terms widely. However, creating a glossary or terminology is time consuming, and these means must force medical personnel to use a predetermined vocabulary and are not necessarily used successfully.

  On the other hand, to create a correspondence table or alias dictionary between sites for the fluctuation of the notation, and to think about how to convey the contents accurately by converting the mechanical notation when communicating between different sites it can. However, even in this case, there is a problem that it takes time to create the correspondence table.

  The present invention has been made in view of the above-described conventional circumstances, and an object thereof is to efficiently extract a pair of words that are presumed to be synonyms that represent the same concept although they have different notations.

The present invention according to claim 1 is a storage means for storing a plurality of electronic documents describing medical results, a selecting means for selecting a plurality of documents having similar description contents, and a medical care appearing in the description in the document. Extracting means for extracting a word relating to, a specifying means for specifying a word appearing in common in each document among the words extracted from each selected document, and the specified among the words extracted from each selected document For a set of words other than words and extracted from different documents, determination means for determining whether or not a predetermined criterion related to the identity of the word concept is satisfied, and a word determined to satisfy the predetermined criterion An output means for outputting a set , wherein the determination means includes a disease name or its identification code described as a diagnosis result in a document from which each word of the target has been extracted, and a disease characterized by the occurrence of the disease. When to show There are a representation swing analyzer and wherein the determining and satisfies a predetermined criterion when the preparation time of the documents extracted each word of the subject has common properties.

  According to a second aspect of the present invention, in the notation fluctuation analysis apparatus according to the first aspect, the document is provided with creation subject information indicating the creation subject, and the determination means includes each of the objects. It is characterized in that it is determined that a predetermined standard is satisfied when the creation subject information added to the document from which the word is extracted is different.

  According to a third aspect of the present invention, in the notation fluctuation analysis device according to the first or second aspect, the determination unit determines that a predetermined criterion is satisfied when the notation of each word of the target is similar. It is characterized by doing.

  According to a fourth aspect of the present invention, in the notation fluctuation analysis apparatus according to any one of the first to third aspects, the determination unit is configured such that the syntax related to each target word has commonality. It is determined that a predetermined standard is satisfied.

  According to a fifth aspect of the present invention, in the notation fluctuation analysis device according to any one of the first to fourth aspects, the determination unit is configured to generate the appearance of the word in the document from which the target word is extracted. When the position has commonality, it is determined that a predetermined criterion is satisfied.

  According to a sixth aspect of the present invention, in the notation fluctuation analysis apparatus according to any one of the first to fifth aspects, the determination unit describes the diagnosis result in a document in which each word of the target is extracted. It is characterized in that it is determined that a predetermined standard is satisfied when the disease name or its identification code is common.

The present invention described in claim 7, a computer, a storage means for storing a plurality of electronic documents that describes the clinical results, and selection means for selecting a plurality of documents describing the contents are similar, the description in the document Extraction means for extracting words related to medical care that appears, specification means for specifying words that appear in common among the documents extracted from the selected documents, and among words extracted from the selected documents A determination means for determining whether or not a predetermined criterion related to the identity of the word concept is satisfied for a set of words extracted from different documents other than the specified word, and determined to satisfy the predetermined criterion a program for functioning as an output means for outputting a set of words with the determination means, disease name or identification code is written as a diagnosis result in a document to extract each word of the object A case showing the disease with time characteristics in the development of the disease, and judging to satisfy the predetermined criterion when the preparation time of the documents extracted each word of the subject has a commonality This is a notation fluctuation analysis program.

According to the notation fluctuation analysis apparatus according to claim 1, a medical document (for example, an electronic document such as an electronic medical record or a medical report) having a similar description is selected, and a concept is set for different word pairs in the similar document. Therefore, it is possible to efficiently extract a pair of words estimated as synonyms.
In addition, for diseases that are characterized by the occurrence of the disease (such as influenza), medical documents related to the diagnosis results are created at the same time, and almost the same words are used in each medical document, while for different words Focusing on the property that they often represent the same concept, it is possible to evaluate synonyms for different sets of words in medical documents with similar descriptions when the timing of medical document creation is common. Based on the criterion of affirmatively, it is possible to extract a set of words estimated to be synonyms.

  According to the notation fluctuation analysis apparatus according to claim 2, there is little notation fluctuation (that is, the notation of a certain concept is unified) when the medical document creation subject (doctor or department or hospital to which the doctor belongs) is the same. On the other hand, by focusing on the property that different notations are often used for certain concepts when the authors are different, it is possible to identify different word pairs in medical documents with similar descriptions. Based on the criterion that the synonym is positively evaluated when the preparation subject is different, a set of words estimated to be synonyms can be extracted.

  According to the notation fluctuation analysis device according to claim 3, when words having similar notations represent the same concept, such as when the constituent characters of each word are partially different or when the constituent characters are the same and the arrangement thereof is different By focusing on the property that there are many words, synonyms are created based on the criteria of positive evaluation as synonyms when different word pairs are similar in medical documents with similar descriptions. A set of words estimated as words can be extracted.

  According to the notation fluctuation analysis apparatus according to claim 4, words having syntactic commonality in the appearance pattern such as morpheme information such as part of speech and syntax information when the periphery of the word is processed in natural language are the same. By focusing on the nature of often representing concepts, different word pairs in medical documents with similar descriptions are evaluated as synonyms when the syntax related to each word has commonality. Based on the above criteria, a set of words estimated to be synonyms can be extracted.

  According to the notation fluctuation analysis apparatus according to claim 5, the appearance in the document, such as an item name in which a word appears in the document, an absolute appearance position, or a relative appearance position in consideration of words commonly appearing in the vicinity. By paying attention to the property that words with common position often represent the same concept, the appearance position of each word in the medical document is common to different sets of words in medical documents with similar descriptions. Based on the criterion that the evaluation as a synonym is positive when it has, a set of words estimated as a synonym can be extracted.

  According to the notation fluctuation analysis apparatus according to claim 6, the same word is used in the medical document described with respect to the disease name having the same diagnosis result, while the different words often represent the same concept. By focusing on, the combination of different words in a medical document with similar description content is evaluated positively as a synonym when the disease name or its identification code described as a diagnosis result in the medical document is common. Based on the criteria, a set of words estimated to be synonyms can be extracted.

According to the notation fluctuation analysis program of the seventh aspect , the notation fluctuation analysis apparatus that exhibits the above-described effects can be realized using a computer.

The present invention will be specifically described based on an embodiment exemplified below.
In this example, the two sites, Hospital A and Hospital B, each have an electronic medical record, which is used by each electronic medical record to facilitate communication between the two hospitals for reasons such as the integration of the two hospitals. An example will be described in which it is necessary to create a correspondence table for the words that have been created.

FIG. 1 shows a functional block diagram of the notation fluctuation analysis apparatus according to this example.
The notation fluctuation analysis apparatus of this example includes a document collection means 1, a similar document finding means 2, a technical term extraction means 3, an identical notation word extraction means 4, a notation difference word correspondence extraction means 5, and a correspondence table construction means 6. Yes.

The document collection means 1 collects and stores the electronic medical records of both sites.
The electronic medical record handled in this example is, as illustrated in FIG. 2, actual data including text data describing medical results such as a document ID for identifying each electronic medical record, findings, chief complaints, and disease names, Attribute data such as creator and creation date / time are included. Note that the document collection means 1 adds the identification code of the collection source site (in this example, the hospital identification code) to the attribute data when collecting the electronic medical record, and identifies from which site the electronic medical record is collected. It is possible to grasp after the fact.

  The similar document finding means 2 selects the electronic medical records stored in the document collecting means 1 that have similar description contents (actual data). In this example, electronic medical records having similar description contents are selected based on the similarity of a word that appears in the description of the electronic medical record and the keyword vector based on the appearance frequency.

Note that electronic documents with similar description contents may be selected by a method other than the statistical evaluation method.
For example, the name of the disease described as the diagnosis result is based on the fact that the same words are used in electronic medical records that describe the disease name with the same diagnosis result, but the different words often represent the same concept. Alternatively, electronic medical records having a common identification code are determined to be similar. For example, for diseases that are characterized by the occurrence of the disease (for example, influenza), the electronic medical records related to the diagnosis results are created at the same time, and almost the same words are used in each electronic medical record, while those for different words Based on the property that often represents the same concept, electronic medical records having a common creation time are determined to be similar.
Further, a plurality of the above-described methods may be combined, thereby improving the accuracy of selecting similar electronic medical records.

The technical term extraction means 3 extracts words related to medical care (for example, technical terms in the medical field) that appear in the description of the electronic medical record.
In this example, word extraction is performed using a known proper name extraction technique, but it is also possible to extract words using other methods such as a method of removing stop words and general words.
Each extracted word may be stored as a list for each electronic medical record, or a tag may be attached to a corresponding portion of the electronic medical record from which the word is extracted.

Among the words extracted by the technical term extraction means 3 from the similar electronic medical records selected by the similar document finding means 2, the same notation word extraction means 4 is a word (common word) that appears in common in each electronic medical record. ).
In this example, the commonality of each word is determined by simply looking at the identity of the character string, but the process is simplified. However, the commonality of each word may be determined in consideration of notation changes such as the use of words. . In addition, when there are three or more similar electronic medical records, in this example, those appearing in all the electronic medical records are common words, for example, those appearing in at least two electronic medical records may be common words, What appears in an electronic medical record of a predetermined ratio or more may be a common language.
Each identified common word may be held as a list, each word in the word list for each electronic medical record may be flagged with a common word, and common to the corresponding part of the extraction electronic medical record You may attach the tag which shows a word.

  The notation different word correspondence extraction means 5 is determined as a common word by the same notation word extraction means 4 among the words extracted by the technical term extraction means 3 from the similar electronic medical records selected by the similar document discovery means 2. Whether or not the concept of each word is the same is determined for a set of words extracted from different electronic medical records.

Here, for example, the following can be considered as an evaluation standard for determining the identity of the concept of each word (determination as to whether it is a synonym).
(1) Affirmatively evaluate as a synonym when the subject that created the electronic chart of the extraction source is different.
This evaluation standard is created when the subject of the electronic medical record (the doctor or the department or hospital to which the doctor belongs) is the same (ie, the notation of a certain concept is unified) This is based on the property that when a subject is different, a different notation is often used for a certain concept. In order to adopt this standard, it is necessary that the name of the creator or its identification code is added (or recorded as attribute data) to the electronic medical record.

(2) When the notation of each word is similar, it is positively evaluated as a synonym.
This evaluation criterion is based on the property that words with similar notation often represent the same concept, such as when the constituent characters of each word are partially different or when the constituent characters are the same and the arrangement is different. .

(3) When the syntax concerning each word has commonality, it evaluates positively as a synonym.
This evaluation criterion is the same for words that have syntactic commonality in their appearance patterns, such as morpheme information such as parts of speech and syntactic information (for example, appearing as “ga”) when the surroundings of the word are processed in natural language. It is based on the nature that it often represents the concept of. In order to adopt this standard, natural language analysis functions such as morphological analysis and syntax analysis are required.

(4) When the appearance position of each word in the extraction source electronic medical record has commonality, it is positively evaluated as a synonym.
This evaluation criterion is based on the appearance position in the electronic medical record, such as the item name (field name) in which the word appears in the electronic medical record, the absolute appearance position, or the relative appearance position in consideration of the word appearing in the vicinity in common. Words with commonality are based on the property that they often represent the same concept.

(5) When a disease name or its identification code (ICD10 code, medical reward code, etc.) described as a diagnosis result is common in the extraction source electronic medical record, it is positively evaluated as a synonym.
This evaluation criterion is based on the property that almost the same word is used in the electronic medical record in which the diagnosis results are described for the same disease name, while the different words often represent the same concept. The disease name and identification code can be specified by various methods such as specifying a word described in the disease name column as a disease name and specifying a character string according to a predetermined rule as an identification code, for example. it can.

(6) Affirmatively evaluate as a synonym when the extraction time of the electronic medical record of the extraction source has commonality.
This evaluation standard is different for diseases with temporal characteristics in the occurrence of disease (for example, influenza), while electronic medical records related to the diagnosis results are created at the same time and the same words are used in each electronic medical record. Words are based on the nature that they often represent the same concept. In order to adopt this standard, it is necessary that the creation date is added to the electronic medical record (or recorded as attribute data).

Although it may be determined whether or not it is a synonym using any of these evaluation criteria, it is desirable to determine by combining a plurality of these.
Here, the determination as to whether or not the evaluation criterion is satisfied may be performed by calculating the degree of conformity with the evaluation criterion and comparing it with a predetermined threshold value. When multiple evaluation criteria are used, the degree of conformity for each evaluation criterion may be compared with the threshold value for each evaluation criterion, and synonyms may be determined. A synonym may be determined by comparing the calculated overall fitness with a threshold. In addition, when evaluating the degree of conformity with respect to each evaluation criterion by individually comparing with the threshold value for each evaluation criterion, an evaluation criterion is evaluated, and only when the evaluation criterion is satisfied, the next evaluation criterion is evaluated. In this way, the processing load may be reduced by performing processing while narrowing down evaluation targets.

The correspondence table construction unit 6 outputs a set of words determined as synonyms by the notation difference word correspondence extraction unit 5 in a correspondence table format. That is, for example, as shown in FIG. 3, a correspondence table in which “liver cancer” and “liver cancer” determined to be synonyms are arranged in parallel is output. In the figure, the data is output in a format in which words are associated with each hospital related to the creation of the electronic medical record of the extraction source.
The output of the correspondence table may be displayed by a display means such as a liquid crystal display and presented to the user, or may be printed by a printing means on a medium such as paper and presented to the user. You may make it output to the memory | storage means to hold | maintain and make it usable later.

FIG. 4 shows a processing flow for creating a correspondence table by the notation fluctuation analysis apparatus of this example.
The similar document finding means 2 selects the electronic medical records stored in the document collecting means 1 and has similar description contents (step S1), and the technical term extraction means 3 selects the selected similar electronic medical records. Words relating to medical care appearing in the description are extracted (step S2), and the same notation word extracting means 4 identifies a common word that appears in common in each electronic medical record among the extracted words (step S3). . Then, the notation different word correspondence extracting means 5 determines whether the concept of each word is the same for a set of words other than the common word among the extracted words and extracted from different electronic medical records (synonymous) The correspondence table forming means 6 outputs a set of words determined to be synonyms in a correspondence table format (step S5).

  When the similar document finding means 2 selects similar electronic medical records using the keyword vector method, the technical term extraction means 3 first performs word extraction from each electronic medical record, and based on the extraction result Thus, the similar document finding means 2 may determine the similarity of the electronic medical record by obtaining the keyword vector of each electronic document.

FIG. 5 shows a main hardware configuration of the notation fluctuation analysis apparatus according to this example.
That is, the notation fluctuation analysis apparatus of this example includes a CPU that performs various arithmetic processes, a RAM that is a work area of the CPU, a ROM that stores a basic control program, a program for realizing each function according to the present invention, and the like. Communication between the HDD to be stored, the liquid crystal display that displays and outputs information to the user, the input / output I / F that is an interface with devices such as a mouse / keyboard that accepts input of information from the user, and other devices It is comprised with the computer which has hardware resources, such as communication I / F which is an interface to perform.

  Then, the program according to the present invention is read from the HDD, expanded in the RAM, and executed by the CPU, so that the storage means (document collecting means 1), the selecting means (similar document finding means 2), and the extraction according to the present invention are extracted. Means (technical term extraction means 3), identification means (same notation word extraction means 4), determination means (notation difference word correspondence extraction means 5), and output means (correspondence table construction means 6) are added to the computer of the notation fluctuation analyzer. Realized.

The program according to the present invention is provided to the practitioner of the present invention, for example, in a format for distributing an external storage medium such as a CD-ROM storing the program or a format for distributing via a network.
Each functional unit of the notation fluctuation analysis apparatus according to the present invention is not limited to a mode realized by the software configuration as in this example, and may be configured by a dedicated hardware module.
In addition, each functional unit of the notation fluctuation analysis apparatus according to the present invention is not limited to an aspect provided in one computer as in this example, and may be provided in a distributed manner in a plurality of computers.

It is a functional block diagram of the notation fluctuation analyzer which concerns on one Embodiment of this invention. It is a figure which illustrates the data structure of the electronic medical record which concerns on one Embodiment of this invention. It is a figure which illustrates the correspondence table concerning one embodiment of the present invention. It is a processing flow figure of correspondence table creation concerning one embodiment of the present invention. It is a hardware block diagram of the notation fluctuation analyzer which concerns on one Embodiment of this invention.

Explanation of symbols

1: Document collection means,
2: Similar document discovery means,
3: Technical term extraction means,
4: Same notation extraction means,
5: Mistaken word correspondence extraction means,
6: Correspondence table construction means

Claims (7)

Storage means for storing a plurality of electronic documents describing medical results;
A selection means for selecting a plurality of documents having similar description contents;
Extraction means for extracting words related to medical care appearing in the description in the document;
A specifying means for specifying a word that appears commonly in each document among the words extracted from each selected document;
Judgment to determine whether or not a predetermined criterion related to the identity of a word is satisfied for a set of words extracted from different documents other than the specified word among the words extracted from the selected documents Means,
Output means for outputting a set of words determined to satisfy a predetermined criterion;
Equipped with a,
The determination means is a case where a disease name or an identification code described as a diagnosis result in a document in which each word of the target is extracted indicates a disease having a temporal characteristic in the occurrence of the disease, and each word of the target A notation fluctuation analysis apparatus characterized by determining that a predetermined standard is satisfied when the creation time of a document from which a word is extracted has commonality . In the document, creation subject information indicating the creation subject is added,
The notation fluctuation analysis apparatus according to claim 1, wherein the determination unit determines that a predetermined criterion is satisfied when the creation subject information added to the document from which each target word is extracted is different.   The notation fluctuation analysis apparatus according to claim 1, wherein the determination unit determines that a predetermined criterion is satisfied when the notation of each word of the target is similar.   The notation fluctuation according to any one of claims 1 to 3, wherein the determination unit determines that a predetermined criterion is satisfied when the syntax related to each of the target words has commonality. Analysis device.   5. The determination unit according to claim 1, wherein the determination unit determines that a predetermined criterion is satisfied when the appearance position of the word in the document from which the target word is extracted has commonality. The notation fluctuation analyzer according to item 1.   The determination unit determines that a predetermined criterion is satisfied when a disease name described as a diagnosis result or an identification code thereof is common to a document from which each word of the target is extracted. 6. The notation fluctuation analyzer according to any one of 5 above. Computer
Storage means for storing a plurality of electronic documents describing medical results;
A selection means for selecting a plurality of documents having similar description contents;
Extraction means for extracting words related to medical care appearing in the description in the document;
A specifying means for specifying a word that appears commonly in each document among the words extracted from each selected document;
Judgment to determine whether or not a predetermined criterion related to the identity of a word is satisfied for a set of words extracted from different documents other than the specified word among the words extracted from the selected documents Means,
A program for functioning as output means for outputting a set of words determined to satisfy a predetermined criterion ,
The determination means is a case where a disease name or an identification code described as a diagnosis result in a document in which each word of the target is extracted indicates a disease having a temporal characteristic in the occurrence of the disease, and each word of the target A notation fluctuation analysis program that determines that a predetermined standard is satisfied when the creation time of the document from which the document is extracted has commonality .

Images