JP2005174003A - Summary preparing method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a summary preparing method and its program for briefly and effectively noticing only a main text. <P>SOLUTION: When document information (for example, patient's electronic chart) is inputted, the morphemic analysis of such document information is performed, and whether a part of the text matches the other whole text is decided. When they matched, the partially matched character string is set as a simple sentence candidate, and in the other case, the text is set as a simplification candidate, as it is. Even in the case of the partially matched character string, when the number of characters of the matched character string is less than M, or when the number of morphemes is less than N, the partially matched character string is not set as a simple sentence candidate, and the text is set as the simplification candidate as it is. The simplifying candidate, including a keyword is extracted from the generated simplifying candidates, and set as a summary candidate. Then, the part pertinent to the summary candidate in the input document is marked, and the summary is generated. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

The present invention relates to a summary generation method for generating a summary from document information such as an electronic medical record and a program therefor.

  When a large number of pieces of document information are included in one file, a summary is often generated so that the contents of each piece of document information can be easily confirmed. For example, by extracting important parts from the document information and generating a separate summary document, or by adding underline or highlight to only the important parts, the contents of each document information can be simplified. To be able to grasp. Thereby, desired document information can be easily extracted from the file.

Here, in the case where a summary is generated from a document in which the same expression frequently appears, such as an electronic medical record, it is effective to generate a summary by extracting a sentence including a specific keyword. For example, in Patent Document 1 below, sentences including important expressions prepared in advance are extracted to create an e-mail summary.
Japanese Patent Laid-Open No. 11-316762

  However, when a sentence including a specific keyword is extracted in this way, sentences having a date, a conjunction and the like added before and after the sentence having the same content are extracted. However, when generating a summary, dates, conjunctions, and the like have no particular significance, and conversely, if dates, conjunctions, etc. are included, they are difficult to read. The summary is easier to read and understand if only the central sentence is omitted, with the date and conjunctions omitted.

Therefore, an object of the present invention is to provide a summary creation method and a program for displaying only main sentences in a concise and effective manner.

  In view of the above problems, the present invention has the following features.

  The invention of claim 1 is a summary generation method for generating a summary from document information, an important sentence extraction step for extracting a sentence including a keyword from sentences included in the document information, and one extracted important sentence A matching step for comparing a sentence with another important sentence to determine whether a part of the important sentence matches another important sentence, and a summary for setting a summary candidate according to the determination result of the matching determination step A candidate setting step; and a summary generation step for extracting a portion corresponding to the summary candidate set by the summary candidate setting step in the document information and generating a summary. When it is determined that part of the important sentence matches another important sentence in the matching determination step, the character string of the matching part is set as a summary candidate. When part of the key sentence in the matching determination step is determined not to match the other key sentences, and sets the key sentences as the summary candidates.

  According to a second aspect of the present invention, in the summary generation method according to the first aspect, the matching determination step determines whether a part of the important sentence matches the entire other important sentence.

  According to a third aspect of the present invention, in the summary generation method according to the first or second aspect, the summary candidate setting step determines that a part of the important sentence matches another important sentence in the matching determination step. The number of characters in the matching part is compared with a threshold value. When the number of characters is equal to or greater than the threshold value, the character string in the matching part is set as a summary candidate. When the number of characters is less than the threshold value, the important sentence is set as a summary candidate. It is characterized by setting.

  According to a fourth aspect of the present invention, in the summary generation method according to any one of the first to third aspects, the summary candidate setting step determines that a part of the important sentence matches another important sentence in the matching determination step. When the number of characters is less than the threshold, the morpheme number of the matching part is compared with a threshold, and when the number of the morpheme is equal to or greater than the threshold, the character string of the matching part is set as a summary candidate. The sentence is set as a summary candidate.

  According to a fifth aspect of the present invention, in the summary generation method according to any one of the first to fourth aspects, the summary generation step displays the full text of the document information, and the summary candidate set by the summary candidate setting step. The character string portion is marked.

  The invention according to claim 6 is a summary generation method for generating a summary from document information, wherein one sentence included in the document information is compared with another sentence, and a part of the sentence matches another sentence. Of the simple sentence candidates set by the matching determination step for determining whether to perform, the simple sentence candidate setting step for setting the simple sentence candidate according to the determination result in the matching determination step, and the simple sentence candidate setting step, the keyword A summary candidate setting step for extracting a simple sentence candidate including the summary candidate and setting it as a summary candidate; and a summary for generating a summary by extracting a portion of the document information corresponding to the summary candidate set by the summary candidate setting step The simplified sentence candidate setting step includes a generation step, and when the part of the sentence matches another sentence in the matching determination step When separated, the matching part character string is set as a simple sentence candidate, and when it is determined in the matching determination step that a part of the sentence does not match another sentence, the sentence is set as a simple sentence candidate It is characterized by doing.

  The invention according to claim 7 is the summary generation method according to claim 6, wherein the matching determination step determines whether a part of the sentence matches the whole of another sentence.

  The invention of claim 8 is the summary generation method according to claim 6 or 7, wherein the simplified sentence candidate setting step determines that a part of the sentence matches another sentence in the matching determination step, The number of characters in the matching part is compared with a threshold value.If the number of characters is equal to or greater than the threshold value, the character string in the matching part is set as a simple sentence candidate.If the number of characters is less than the threshold value, the sentence is simplified. It is set as a candidate.

  The invention according to claim 9 is the summary generation method according to any one of claims 6 to 8, wherein the simplified sentence candidate setting step is determined in the matching determination step that a part of the sentence matches another sentence. When the number of characters is less than the threshold, the number of matching morphemes is compared with a threshold value. Is set as a simple sentence candidate.

  A tenth aspect of the present invention is the summary generation method according to any one of the sixth to ninth aspects, wherein the summary generation step displays the full text of the document information, and the summary candidate set by the summary candidate setting step. The character string portion is marked.

  The invention of claim 11 is a program for giving a computer a summary generation function according to the summary generation method of any one of claims 1 to 10.

  In the above description, “sentence” includes not only a character string divided by a phrase “.” To the next phrase “.” But also a character string partitioned by a next line break mark from a line break mark. . “Marking” is a method of differentiating the display by changing the thickness, size color, etc. of the character in addition to highlighting the character string by adding an underline or highlight display. Is also widely included.

The features of the present invention will become more apparent from the following description of embodiments. However, the following embodiment is merely one embodiment of the present invention, and the meaning of the term of the present invention or each constituent element is not limited to that described in the following embodiment. Absent.

  According to the present invention, among sentences containing a keyword, sentences to which a date or conjunction such as “after” is added are simplified to sentences that exclude these and are regarded as summary candidates. Thus, it is possible to generate an effective summary (summary) without unnecessary expressions.

According to the third, fourth, eighth, and ninth aspects of the invention, when the number of characters of the part to be partially matched is less than the minimum number of characters, or when the number of morphemes is less than the minimum number of morphemes, partial matching is performed. Since the character string to be used is not a summary candidate and the sentence is used as a summary candidate as it is, the summary candidate is not oversimplified, and thus a moderately simplified summary (summary) is generated. Can do.

  Embodiments of the present invention will be described below with reference to the drawings.

  First, FIG. 1 shows a configuration of a summary creation device according to the embodiment.

  Note that the summary creation device in the present embodiment can be realized in hardware by a CPU, memory, or other LSI of an arbitrary computer. In terms of software, it is realized by a program having a recording control function loaded in a memory. FIG. 1 shows functional blocks of a summary creation device realized by hardware and software. However, it goes without saying that these functional blocks can be realized in various forms such as hardware only, software only, or a combination thereof.

  As shown in the figure, the summary creation device includes a text input unit 101, a morpheme analysis unit 102, a keyword setting unit 103, a keyword dictionary 104, an important sentence extraction unit 105, a summary candidate setting unit 106, and a summary output unit 107. Is included.

  The text input unit 101 receives document information such as an electronic medical record from an input port or a disk drive. The morpheme analysis unit 102 includes a database for morpheme analysis, parses the document information (single document information) input from the input unit 101 into morpheme units, divide information, and independent words of each unit Information indicating the distinction between attached words is output to the keyword setting unit 103 and the important sentence extracting unit 105 along with the document information.

  The keyword setting unit 103 detects the frequency of occurrence of each independent word included in the document information, and stores words having a frequency of occurrence equal to or higher than a predetermined threshold in a memory (not shown) as keyword candidates. At this time, a score corresponding to the occurrence frequency is set for each keyword candidate and stored in the memory.

  The keyword dictionary 104 stores keyword candidates set in advance by the user via input means such as a keyboard. When setting such keyword candidates, the user sets the importance of each keyword candidate. The keyword dictionary 104 stores a score corresponding to the degree of importance in association with each keyword.

  The keyword setting unit 103 generates an important word table from the keyword candidates stored in the memory and the keyword candidates registered in the keyword dictionary 104. The important word table is referred to in the important sentence extraction unit 105 when extracting the important sentence.

  Note that such an important word table is generated by, for example, all the keyword candidates registered in the keyword dictionary 104 and the keyword candidates with the highest score among the keyword candidates stored in the memory. Or you may make it produce | generate by the top few of the keyword candidate registered into the keyword 104 and the top few of the keyword candidate stored in the said memory. Here, it is preferable that the user can appropriately set up to what number is taken into the important word table.

  The important sentence extraction unit 105 extracts sentences including important words (keyword candidates) in the important word table set by the keyword setting unit 103 as morphemes from the sentences included in the input document as important sentence candidates. And output to the summary candidate setting unit 106. Here, for example, an important sentence candidate is extracted with one sentence from the phrase “.” To the next phrase “.”. Or you may make it make one sentence from a newline mark to the following newline mark.

  The summary candidate setting unit 106 compares the important sentence candidate input from the important sentence extracting unit 105 with other important sentence candidates, and when the important sentence candidate partially includes other important sentence candidates, matching is performed. The character string of the part to be set is set as the summary candidate, and when the other important sentence candidates are not partially included, the important sentence candidate is set as the summary candidate as it is. However, if the number of characters in the character string of the matching part is less than the preset minimum number of characters M, or the morpheme number is less than the preset minimum morpheme number N, the character string of the matching part is selected as the summary candidate. Instead, the important sentence candidate is set as a summary candidate as it is.

  The summary output unit 107 generates a summary (summary) from the document information and displays it on the monitor. For example, the input document information is displayed in full text, and markings such as underline and highlight are added to a character string that matches the summary candidate set by the summary candidate setting unit 106. Alternatively, a summary format may be prepared separately, and a character string that matches the summary candidate may be moved to this.

  FIG. 2 shows a processing flow of the summary creation device in the present embodiment.

  When document information is input from the text input unit 101 (S101), the input document information is morphologically analyzed by the morpheme analysis unit 102 (S102). Then, the frequency of independent words is counted by the keyword setting unit 103, and a score corresponding to the frequency is set for each independent word (S103). Among these, an important word table is generated from independent words (keyword candidates) having a score equal to or higher than the threshold K and independent words (keyword candidates) registered in the keyword dictionary 104 (S105). Then, a sentence including the important word in the generated important word table as a morpheme is extracted as an important sentence candidate in the important sentence extraction unit 105 (S105).

  When important sentence candidates are extracted from the input document in this way, the summary candidate setting unit 106 then executes the summary candidate setting process (S106 to S111). First, the important sentence candidate to be determined is compared with other important sentence candidates to determine whether the important sentence candidate partially includes another important sentence candidate (partial matching) (S106). Here, if there is no partial matching, the important sentence candidate to be determined is set as a summary candidate as it is (S109).

  On the other hand, if the partial matching is performed, it is determined whether the number of characters in the character string of the partial matching portion is less than the set value M (S107). If the number of characters is less than the set value M, the process proceeds to S109, and the important sentence candidate to be determined is set as a summary candidate as it is. On the other hand, if the number of characters is greater than or equal to the set value M, it is next determined whether or not the number of morphemes in the character string of the partially matched portion is less than the set value N (S107). If the morpheme number is less than the set value N, the process proceeds to S109, and the important sentence candidate to be determined is set as a summary candidate as it is. On the other hand, if the number of morphemes is N or more, the partially matched character string is set as a summary candidate (S110).

  The processes of S106 to S110 are repeated until the summary candidate setting process is completed for all important sentence candidates (S111). When the summary candidate setting process for all important sentence candidates is completed, the summary output unit 107 then executes a summary (summary) output process based on the summary candidates (S112). For example, the input document information is displayed in full text, and markings such as underline and highlight are added to the character string that matches the summary candidate set in S106 to S111.

  FIG. 3 shows a specific processing example when setting summary candidates.

  When one unit of document information (for example, electronic medical record) is input to the input unit, such document information is morphologically analyzed as shown in FIG. In the figure, “/” indicates a morpheme break. At this time, if the keywords “progress” and “especially” are set in the important word table, only sentences including “progress” and “particularly” as morphemes are extracted from the text included in the document. The important sentence candidate is generated as shown in FIG.

  Next, it is determined whether or not a part of the generated important sentence candidate matches another important sentence candidate (partial matching). When they match, the character string to be partially matched is set as a summary candidate. For example, in the figure (b), “the progress is then smooth” means that the part “the progress is good” partially matches other important sentence candidates as shown in the figure (d). “The progress is smooth” is set as the summary candidate.

  On the other hand, when partial matching is not performed, the important sentence candidate is set as a summary candidate as it is. For example, in the figure (b), “No symptoms in particular” overlaps with “No pain in particular” and “No particular pain”, but includes “No pain in particular” as part of it. Because it is not, it is assumed that partial matching is not performed. Therefore, as shown in FIG. 5C, “no particular symptom” is set as a summary candidate as it is. The same applies to “no pain”.

  In this way, in this embodiment, among sentences (important sentence candidates) including keywords, sentences to which a date or a conjunction such as “after” is added are simplified to sentences that exclude these, and summary candidates are added. Therefore, it is possible to generate and output a summary (summary) without unnecessary expressions such as dates and conjunctions.

  Although not illustrated in FIG. 3, if the number of characters of the part to be partially matched is less than the minimum number of characters M, or if the number of morphemes is less than the minimum number of morphemes N, the character string to be partially matched is selected as a summary candidate. The important sentence candidate is set as a summary candidate instead, and it is possible to prevent the important sentence candidate from being oversimplified, thereby generating a summary (summary) that is simplified enough to grasp the contents. Can be output.

  Note that the minimum number of characters M and the minimum number of morphemes N are set to values that the designer recognizes that the most effective summary is output, for example, by trying to generate a summary while changing them. . Alternatively, these values may be appropriately set by the user.

  In the first embodiment, after extracting important sentence candidates based on keywords (important words), summary candidates are set by simplifying them, but in this embodiment, first, they are included in the input document. After a simplified sentence is extracted, a sentence including a keyword is extracted and set as a summary candidate.

  FIG. 4 shows the configuration of the summary generation device according to the present embodiment.

  In the figure, functions of a text input unit 101, a morpheme analysis unit 102, a keyword setting unit 103, a keyword dictionary 104, and a summary output unit 107 are the same as those shown in FIG. In this embodiment, a simple sentence extraction unit 110 and a summary candidate setting unit 111 are employed instead of the important sentence extraction unit 105 and the summary candidate setting unit 106 of the first embodiment.

  The simplified sentence extraction unit 110 compares one sentence among the sentences included in the input document with another sentence, and when the sentence partially matches with another sentence, the simplified sentence extraction unit 110 calculates the character string of the matching part. When a simple sentence candidate is used and partial matching is not performed, the sentence is directly used as a simple sentence candidate. However, if the number of characters in the character string of the matching part is less than the preset minimum number of characters M, or the morpheme number is less than the preset minimum morpheme number N, the character string of the matching part is simplified Instead of a candidate, the sentence is used as a simple sentence candidate.

  The summary candidate setting unit 111 extracts sentences including important words (keyword candidates) in the important word table set by the keyword setting unit 103 as morphemes from the generated simplified sentence candidates and sets them as summary candidates. To do.

  FIG. 5 shows a processing flow of the summary creation device in the present embodiment.

  Of the processing flow shown in the figure, S101 to S104 are the same as the processing flow of FIG.

  When the important word table is generated in S104, next, among the sentences included in the input document, one sentence (sentence candidate) is compared with another sentence, and the sentence candidate partially substitutes another sentence. Whether it is included (partial matching) is determined (S121). Here, if there is no partial matching, the sentence candidate is used as a simple sentence candidate as it is (S124).

  On the other hand, if the partial matching is performed, it is determined whether the number of characters in the character string of the partial matching portion is less than the set value M (S122). If the number of characters is less than the set value M, the process proceeds to S124, and the sentence candidate is directly used as a simple sentence candidate. On the other hand, if the number of characters is greater than or equal to the set value M, it is next determined whether or not the number of morphemes in the character string of the part to be partially matched is less than the set value N (S123). Here, if the morpheme number is less than the set value N, the process proceeds to S124, and the sentence candidate is directly used as a simple sentence candidate. On the other hand, if the number of morphemes is N or more, the partially matched character string is set as a simple sentence candidate (S125).

  The processes of S121 to S125 are repeated until the generation process of the simplified sentence candidate is completed for all sentences (S126). When the generation process of the simplified sentence candidates for all sentences is completed, a simplified sentence candidate including the important words in the important word table generated in S104 as morphemes is extracted from the simplified sentence candidates, and the summary candidates are extracted. It is set (S127). Based on the set summary candidates, the summary output unit 107 executes summary (summary) output processing (S128). For example, the input document information is displayed in full text, and markings such as underline and highlight display are added to the character string that matches the summary candidates set in S121 to S127.

  FIG. 6 shows a specific processing example when setting summary candidates.

  When one unit of document information (for example, electronic medical record) is input to the input unit, such document information is morphologically analyzed as shown in FIG. Next, each sentence subjected to morphological analysis is discriminated whether or not a part thereof matches another sentence (partial matching). When they match, a character string to be partially matched is set as a simple sentence candidate. On the other hand, when they do not match, the sentence is directly used as a simplification candidate.

  For example, in the figure (a), “the progress is smooth afterwards” means that “the progress is smooth” because the portion of “the progress is smooth” partially matches with other sentences. Set as a candidate.

  Note that “no abnormality”, “no symptom in particular”, and “no pain in particular” in FIG. 9A partially match with “none”, but the number of characters of “none” to be partially matched is the minimum value M ( For example, since it is less than M = 4), as shown in FIG. 4D, the simplified sentence candidates for “no abnormality”, “no particular symptom”, and “no special pain” are set to “none”. There is nothing. These are “no abnormality”, “no symptom in particular”, and “no pain in particular”, respectively, as simplification candidates.

  Next, among the generated simplification candidates, simplification candidates including keywords are extracted and set as summary candidates. For example, if the keywords “progress” and “especially” are set in the keyword table, only simplification candidates in FIG. 5B that include “progress” and “particularly” as morphemes are extracted. Then, it is set as a summary candidate as shown in FIG.

  As described above, in this embodiment, similar to the first embodiment, a summary (summary) having no unnecessary expressions such as dates and conjunctions can be generated and output. Further, by setting the minimum number of characters M and the minimum number of morphemes N, excessive simplification can be prevented, and a simplified summary can be generated and output effectively.

It goes without saying that the present invention is not limited to the above-described embodiment, and various other modifications are possible. The embodiment of the present invention can be appropriately changed within the scope of the technical idea shown in the claims.

The figure which shows the structure of the summary preparation apparatus which concerns on Example 1. FIG. 6 is a flowchart illustrating processing operations of the summary creation device according to the first embodiment. The figure which shows the specific example of the summary preparation operation | movement which concerns on Example 1. FIG. The figure which shows the structure of the summary preparation apparatus which concerns on Example 2. FIG. 10 is a flowchart illustrating processing operations of the summary creation device according to the second embodiment. The figure which shows the specific example of the summary preparation operation | movement which concerns on Example 2. FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 103 Keyword setting part 104 Keyword dictionary 105 Important sentence extraction part 106 Summary candidate setting part 107 Summary output part 110 Simplified sentence extraction part 111 Summary candidate setting part

Claims (11)

A summary generation method for generating a summary from document information,
An important sentence extraction step for extracting sentences including keywords from the sentences included in the document information;
A matching determination step for comparing one extracted important sentence with another important sentence and determining whether a part of the important sentence matches another important sentence;
A summary candidate setting step for setting a summary candidate according to the determination result of the matching determination step;
A summary generation step of generating a summary by extracting a portion corresponding to the summary candidate set by the summary candidate setting step in the document information,
The summary candidate setting step includes:
When it is determined in the matching determination step that a part of the important sentence matches another important sentence, a character string of the matching part is set as a summary candidate,
When it is determined in the matching determination step that a part of the important sentence does not match another important sentence, the important sentence is set as a summary candidate.
A summary generation method characterized by the above. In claim 1,
In the matching determination step, it is determined whether a part of the important sentence matches the entire other important sentence.
A summary generation method characterized by the above. In claim 1 or 2,
In the summary candidate setting step, when it is determined in the matching determination step that a part of the important sentence matches another important sentence, the number of characters in the matching part is compared with a threshold value, and the number of characters is equal to or more than the threshold value. Sets the character string of the matching part as a summary candidate, and when the number of characters is less than the threshold, sets the important sentence as a summary candidate.
A summary generation method characterized by the above. In any one of Claims 1 thru | or 3,
In the summary candidate setting step, when it is determined in the matching determination step that a part of the important sentence matches another important sentence, the number of matching morphemes is compared with a threshold value, and the morpheme number is equal to or greater than the threshold value. When the number of characters is less than the threshold, the important sentence is set as a summary candidate.
A summary generation method characterized by the above. In any of claims 1 to 4,
The summary generation step displays the full text of the document information and marks the character string portion of the summary candidate set by the summary candidate setting step.
A summary generation method characterized by the above. A summary generation method for generating a summary from document information,
A matching determination step of comparing one sentence included in the document information with another sentence, and determining whether a part of the sentence matches the other sentence;
A simple sentence candidate setting step for setting a simple sentence candidate according to the determination result of the matching determination step;
A summary candidate setting step of extracting a simplified sentence candidate including a keyword from the simplified sentence candidates set by the simplified sentence candidate setting step and setting it as a summary candidate;
A summary generation step of generating a summary by extracting a portion corresponding to the summary candidate set by the summary candidate setting step in the document information,
The simplified sentence candidate setting step includes:
When it is determined in the matching determination step that a part of the sentence matches another sentence, the matching part character string is set as a simple sentence candidate,
When it is determined in the matching determination step that a part of the sentence does not match another sentence, the sentence is set as a simple sentence candidate;
A summary generation method characterized by the above. In claim 6,
The matching determination step determines whether a part of the sentence matches the entire other sentence,
A summary generation method characterized by the above. In claim 6 or 7,
In the simplified sentence candidate setting step, when it is determined in the matching determination step that a part of the sentence matches another sentence, the number of matching parts is compared with a threshold, and when the number of characters is equal to or more than the threshold Set the matching part character string as a simple sentence candidate, and when the number of characters is less than the threshold, set the sentence as a simple sentence candidate,
A summary generation method characterized by the above. In any of claims 6 to 8,
In the simplified sentence candidate setting step, when it is determined in the matching determination step that a part of the sentence matches another sentence, the morpheme number of the matching part is compared with a threshold value, and the morpheme number is equal to or greater than the threshold value. When setting the matching part character string as a simple sentence candidate, when the number of characters is less than the threshold, set the sentence as a simple sentence candidate,
A summary generation method characterized by the above. In any of claims 6 to 9,
The summary generation step displays the full text of the document information and marks the character string portion of the summary candidate set by the summary candidate setting step.
A summary generation method characterized by the above. A program for giving a computer a summary generation function according to the summary generation method according to claim 1.

Images