JP3908919B2 - Morphological analysis system and morphological analysis method - Google Patents

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a morpheme analysis system and a morpheme analysis method, and more particularly to a morpheme analysis of a character string including Japanese kanji and hiragana.
[0002]
[Prior art]
In Japanese orthography, words are not written separately, so in a morpheme analysis program that performs morphological analysis using a computer or the like, the main problem is to identify morphemes by obtaining word boundaries.
[0003]
As the morpheme analysis method, from the beginning of the input character string, the longest match method in which the longest match among the morphemes recorded in the dictionary is sequentially analyzed as candidates, and the minimum division number method, learning There are methods for performing morphological analysis using functions and statistical language models.
[0004]
[Problems to be solved by the invention]
Here, since many Japanese character strings contain ambiguity and ambiguity, there are the following problems in morpheme identification in a morpheme analysis program.
[0005]
(1) An incorrect analysis result may be obtained for an input sentence with a series of kanji characters.
For example, with respect to the input character string “used vehicle”, a result of “used | vehicle” (“|” represents a partitioning position between morphemes; the same applies hereinafter) is desirable. However, if a dictionary read by the program contains multiple morphemes “middle, old, car, both, used, vehicle, used car”, then analyzing this using the longest match method, “used car | both” The analysis result is obtained.
[0006]
(2) The analysis time of input sentences including hiragana is long.
[0007]
A program in order to analyze the input sentence including hiragana in consideration of the possibility of grammatical concatenation of these morphemes while obtaining morpheme candidates of all possibilities by sequentially decomposing the input character string. The amount of calculation becomes very large, and quick analysis is difficult.
[0008]
In view of the above problems, an object of the present invention is to provide a morpheme analysis system and a morpheme analysis method capable of performing morpheme analysis accurately and quickly.
[0009]
[Means for Solving the Problems and Effects of the Invention]
(1) A kanji character string morpheme analysis method according to the present invention includes:
A kanji character string morpheme analysis method for performing a morpheme analysis of a character string including kanji by referring to a morpheme recording unit in which a morpheme is recorded by a computer system,
In the morpheme recording unit,
If the morpheme is a morpheme that, when combined with any other character string, results in at least two morphemes separated in the character string of the morpheme, it indicates the delimited position in association with the morpheme Record the break position information together,
When the computer system performs morphological analysis of the input character string,
The first morpheme candidate is obtained by the longest match method from the first character of the character string by reading with reference to the morpheme from the morpheme recording unit,
If the position information is recorded in the first morpheme candidate, the second morpheme candidate including a character string subsequent to the character string of the first morpheme candidate by the longest match method with the subsequent character at the delimiter position as the first character To determine whether or not
If the second morpheme candidate can be obtained, the first morpheme candidate and the second morpheme candidate when the first morpheme candidate is divided at the delimiter position are used as the morpheme analysis result,
If the second morpheme candidate cannot be acquired, the first morpheme candidate is used as a morpheme analysis result.
[0010]
Thus, even if one morpheme obtained by the longest match method is a morpheme that, when combined with another subsequent character string, becomes at least two morphemes separated in the character string of the morpheme Therefore, it is determined as two morphemes that are separated at the separation position. Therefore, morphological analysis can be performed efficiently and accurately. Further, since it is only necessary to record the delimiter position information together, it is easy to create a dictionary that can obtain a certain analysis result and to maintain and manage the recorded contents.
[0013]
  (6)The morphological analysis system according to the present invention is:
  A character input part for inputting a character string;
  At least one of a kanji morpheme recording unit that records a morpheme of a character string that includes kanji and a hiragana morpheme recording unit that records a morpheme of a character string that includes hiragana;
  A morphological analysis unit that performs morphological analysis;
  A morphological analysis system comprising:
  The kanji morpheme recording unit is
  When a morpheme of a character string containing Kanji is combined with any other character string after at least one of the first half character string or the second half character string when the character string is separated, a character string that becomes at least two morphemes If included,
  Separation position information indicating the division position of the morpheme in association with the morpheme,
  Is also recorded,
  The delimiter position information is obtained by a computer,
  Information used to determine whether or not a morpheme including the subsequent character can be obtained by the longest match method, with the subsequent character recorded in the morpheme as the first character as the first character,
  It is characterized by
  The hiragana morpheme recording unit is
  Record at least a hiragana morpheme combination, which is a combination of morphemes that can be grammatically correct among a combination of two or more hiragana morphemes containing a particle or auxiliary verb or a formal noun or basic verb;
  In the hiragana morpheme junction,
  Separation position for each morpheme,
  Is recorded together.
[0014]
As a result, the morphological analysis is performed accurately and quickly even when the character string to be analyzed is a mixture of characters such as kanji and hiragana.
[0015]
  (7)The network morphological analysis system according to the present invention is:
  Communication terminal
  A character string transmission means for transmitting a Japanese character string to a computer server;
  A morpheme analysis result receiving means for receiving a morpheme analysis result transmitted from a computer server;
  With
  A computer server connected to a communication terminal so that it can communicate with
  A receiving unit that receives information of the transmitted Japanese character string;
  At least one of a kanji morpheme recording unit that records a morpheme of a character string that includes kanji and a hiragana morpheme recording unit that records a morpheme of a character string that includes hiragana;
  A morphological analysis unit that performs morphological analysis;
  A morpheme analysis result transmitting means for transmitting a morpheme analysis result by the morpheme analysis unit to a communication terminal;
  A network morphological analysis system comprising:
  The kanji morpheme recording unit is
  When the morpheme of the character string including the kanji is a morpheme that becomes at least two morphemes separated in the character string of the morpheme when the other arbitrary character strings are subsequently combined,
  Separation position information indicating the division position in association with the morpheme,
  Is also recorded,
  The hiragana morpheme recording unit is at least
  A hiragana morpheme combination that is a combination of two or more hiragana morphemes that can be grammatically joined,
  Is recorded,
  In the hiragana morpheme junction,
  The separation position for each morpheme is recorded together,
  The morpheme analysis unit that performs morphological analysis of a character string including kanji is
  First morpheme candidate acquisition means for acquiring a first morpheme candidate from a first character of a character string by a longest match method by reading out the morpheme from the kanji morpheme recording unit;
  If the delimiter position information is recorded in the first morpheme candidate, the second morpheme candidate including a character string subsequent to the character string of the first morpheme candidate by the longest match method with the subsequent character at the delimiter position as the first character Second morpheme candidate acquisition determination means for determining whether or not can be acquired;
  If the second morpheme candidate can be acquired, the first morpheme candidate and the second morpheme candidate when the first morpheme candidate is divided at the dividing position are set as the morpheme analysis results. A kanji character string morpheme determining means which takes one morpheme candidate as a morpheme analysis result;
  With
  The morpheme analysis unit that performs morphological analysis of a character string including hiragana,
  If there is a character string that matches the hiragana morpheme junction by reading out the hiragana morpheme junction and the hiragana morpheme from the hiragana morpheme recording unit, the hiragana character string having the combination of the hiragana morpheme as a morpheme analysis result Morpheme determination means,
  It is characterized by having.
[0016]
  Thereby, it is not necessary to provide a kanji morpheme recording unit, a hiragana morpheme recording unit, and a morpheme analysis unit on the terminal device side. Therefore, a larger number of terminal device users can easily obtain the morphological analysis results.For morphological analysis of character strings containing Kanji characters, even if one morpheme is obtained by the longest match method, at least two morphemes that are separated in the character string of the morpheme when combined with other subsequent character strings If it is such a morpheme, it will be judged as two morphemes that are definitely separated at the separation position. Therefore, morphological analysis can be performed efficiently and accurately. Further, since it is only necessary to record the delimiter position information together, it is easy to create a dictionary that can obtain a certain analysis result and to maintain and manage the recorded contents. As for the morphological analysis of the character string including the hiragana, since the morphological analysis is performed by collating with the morpheme recording unit recorded in advance, the calculation amount of the program until the identification of the morpheme is greatly reduced. Therefore, morphological analysis is performed very quickly.
[0017]
  (8)The delimiter position information according to the present invention is:
  It is information indicating that it is located at the Nth character from the character behind the morpheme,
  The information indicating the position of the Nth character is as follows.
  Information used to determine whether or not a morpheme including the subsequent character can be obtained by the longest match method with the subsequent character at the position of the Nth character recorded in the morpheme as the first character;
  It is characterized by.
more than
[0018]
As a result, the process of moving back N characters from the position to be determined when the first morpheme candidate is acquired is performed, and therefore the process of returning to the beginning of the character string of the first morpheme candidate and moving N characters further is performed. Compared with, the amount of calculation of the program is reduced. Therefore, morphological analysis is performed more quickly.
[0019]
The correspondence between the definition of terms and the embodiments will be described.
[0020]
In this invention,
“Delimiter position information” is a morpheme that, when it is assumed that a certain morpheme is combined with another arbitrary character string later, it becomes at least two morphemes separated in the character string of the morpheme When it is determined, it indicates a delimiter position in the character string of the morpheme. In the embodiment, the N character backward data in FIG. 3A corresponds to this, and this data is the N (integer) of the case where such a delimiting position is the Nth character from the character after the morpheme. The value is shown.
[0021]
DETAILED DESCRIPTION OF THE INVENTION
(1) Configuration of a morphological analysis system according to an embodiment of the present invention
An embodiment of a morphological analysis system according to the present invention will be described with reference to the drawings. A Japanese morpheme analysis system as a morpheme analysis system according to the present invention is provided in a personal computer having a configuration represented by the block diagram shown in FIG. This system includes a memory 100, a keyboard 103, a CPU 101, a CRT 102, and a hard disk 104. In the hard disk 104, the character type distinguishing program 10 shown in FIG. 1, the reading imparting program 20 for morphemes, the kanji character string morpheme analysis program 14 for performing the kanji character string morpheme analysis method of the present invention, and the present invention. A hiragana character string morpheme analysis program 16 for performing a hiragana character string morpheme analysis method, a kanji character string morpheme N character registration dictionary 12 as a kanji morpheme recording unit, and a hiragana morpheme junction list dictionary 18 as a hiragana morpheme recording unit It is recorded. The Kanji character string morpheme N character registration dictionary 12 and the hiragana morpheme junction list dictionary 18 are recorded in the floppy disk (registered trademark) 105 or the CD-ROM 106, and are recorded in the hard disk 104 by reading them. The kanji character string morpheme N character registration dictionary 12 and the hiragana morpheme junction list dictionary 18 are not limited to the present embodiment, and may be installed from a computer-readable recording medium such as an IC card, and further using a communication line. You can also download it.
[0022]
The program for the operation of the CPU 101 is not limited to this embodiment, and may be installed from a computer-readable recording medium such as an IC card. Further, the program can be downloaded using a communication line. Further, by installing the program from the CD-ROM 106 or the like, the program stored in the CD-ROM 106 or the like is not indirectly executed by the computer but the program stored in the CD-ROM 106 or the like is directly executed. You may make it perform to.
[0023]
As a program that can be executed by a computer, a program that can be directly executed by simply installing it as well as a program that needs to be converted into another form or the like (for example, decompressing a data-compressed program) Furthermore, what is executable in combination with another module part is also included.
[0024]
The flow of morphological analysis of the entire system includes hiragana by the character type discrimination program 10 while moving the position to be judged from the beginning of the sentence toward the end of the sentence with respect to the character string input by the keyboard 103 or the like as the character input unit. A character string and a character string including kanji are distinguished and analyzed by a corresponding program, and analysis data is obtained and recorded in the memory 100 by repeatedly moving the determination target to the next character string.
[0025]
Further, the reading imparting program 20 for morphemes is not limited to this, and the program can be changed according to the purpose of using the analysis result, for example, giving an accent for speech synthesis, There may be a case where a program for converting hiragana to kanji, performing part-of-speech analysis, or simply performing only morphological analysis is provided.
[0026]
(2) Configuration of Kanji character string morpheme N character registration dictionary 12
FIG. 3A shows a part of an example of the Kanji character string morpheme N character registration dictionary 12. This dictionary is provided with a column of a Chinese character dictionary as a Chinese character morpheme recording unit and a column of N character backward data as delimiter position information.
[0027]
In the column of the kanji dictionary, kanji character strings as morphemes, mixed character strings of kanji and katakana or hiragana, and the like are recorded.
[0028]
In the column of N character backward data, the value (integer) of N when there is a position delimited at the Nth character from the character after the morpheme is recorded in association with each morpheme. In other words, the value of N is a morpheme whose morpheme becomes at least two morphemes that are separated by positions in the character string of the morpheme when subsequent arbitrary character strings are combined. This indicates that the delimiter position in the character string is the Nth character from the character after the morpheme.
[0029]
For example, “used car” in the dictionary 12 is one morpheme, but by combining another character string at the back, “used | garage”, “used | wheels”, “used | vehicles”, etc. Two morphemes can be obtained for each. Therefore, in this case, it can be determined that there is a possibility that the first character from the end of “used car” may be separated, and therefore, “1” is recorded in the N character backward data. Similarly, “energy saving” is one morpheme, but two morphemes “energy saving | energy” can be obtained, and it can be determined that there is a possibility that the second character after “energy saving” may be separated. Therefore, “2” is recorded in the N character backward data.
[0030]
In other words, the morpheme is recorded in the dictionary 12 as one morpheme even though it is a morpheme that becomes two or more morphemes that are separated in the character string of the morpheme when the other character string is combined later. If there is a single morpheme, it is assumed that there may be two or more such morphemes, and the division position is recorded in advance in association with the morpheme.
[0031]
It should be noted that, in addition to the case where a character string is combined after the morpheme as described above, the determination of the presence of such a delimited position in the character string of one morpheme It is also conceivable to assume a case in which any other character string is subsequently joined to the first half character string when the columns are separated. For example, in the above example of “energy saving”, assuming that the character string in the first half of the character string is divided into “savings” and another arbitrary character string is combined, “saving” | resources ” Therefore, it can be determined that there is a delimited position after “saving”.
[0032]
Here, the basis for the determination of the separation position shown in the above example is that “car” in “used car” and “saving” in “energy saving” are so-called coined elements in each morpheme. Is.
[0033]
(3) Morphological analysis method for character strings including kanji
FIG. 4 shows processing when the kanji character string morphological analysis program 14 performs morphological analysis of a character string including kanji according to the present embodiment. Note that this morpheme analysis method complements the shortcomings while using the conventional longest match method. Hereinafter, this method is referred to as “longest match-N-character backward method”.
[0034]
First, the kanji character string morpheme analysis program 14 refers to the kanji character string morpheme N character registration dictionary 12 to acquire a first morpheme candidate from the first character by the longest match method (step S1). Then, the N character backward data of the acquired character string is determined (step S2). If N = 0, the first morpheme candidate is determined as the morpheme analysis result (step S3). On the other hand, if N is not 0, the determination target position is moved backward from the character string of the first morpheme candidate to a position backward by N characters, and the second morpheme candidate by the longest match method for the subsequent character string from that position again Is started (step S7).
[0035]
Next, it is determined whether or not a character string morpheme of N + 1 or more can be acquired as a second morpheme candidate (step S8). If not acquired, the first morpheme candidate acquired first is determined as a morpheme analysis result. (Step S3). On the other hand, if it is acquired, the first half character string when the first morpheme candidate is separated at a position backward by N characters and the second morpheme candidate are determined as the morpheme analysis result (step S9).
[0036]
Then, the morpheme determined in step S3 or step S9 is recorded (step S4). If not all the morphemes of the character string data are recorded, the first morpheme candidate by the longest match method is applied to the remaining character strings. Acquisition is started (step S10), and thereafter the same flow is repeated to advance morpheme analysis. When all data morphemes have been recorded, the analysis ends (step S6).
[0037]
As a specific example of the morphological analysis method as described above, FIG. 5 shows an example in which the input character string “used vehicle” is given. Since not only “used” but also “used cars” are recorded in the dictionary 12 as morphemes, “used cars” is selected as the first morpheme candidate according to the longest match method (step S50). Since “1” is recorded in “used car” (step S51), the position to be judged is moved backward by one character, and the second morpheme candidate is determined by the longest match method for the character string following “car”. It is determined whether or not can be acquired (step S52). As illustrated in FIG. 3A, since “vehicle” is registered in the dictionary (step S53), “used” (the first half character string in which “used car” is separated by the position of “car”) and “vehicle” ”As morphemes, and“ used | vehicles ”as analysis results (steps S54, S55, S56, S57).
[0038]
(4) Effects of the morphological analysis method for character strings containing kanji
According to the above method, even one morpheme acquired by the longest match method becomes at least two morphemes that are separated in the character string of the morpheme when combined with other subsequent character strings. In the case of a morpheme, it is determined as two morphemes that are definitely separated at the separation position. Therefore, morphological analysis can be performed efficiently and accurately. Further, since it is only necessary to record the delimiter position information together, it is easy to create a dictionary that can obtain a certain analysis result and to maintain and manage the recorded contents.
[0039]
Hereinafter, the effects of the present embodiment will be described using the example shown in FIG. 5 while comparing with the conventional morphological analysis method.
[0040]
First, according to the conventional longest match method, if the morpheme “used car” is recorded in the dictionary, an analysis result undesirable for the user “used car | both” is obtained. On the other hand, if “used vehicle” is recorded, “used vehicle” is taken as the analysis result, and a more preferable result than the result “used car | both” is obtained. However, even if such a record is made, it can only be effective for "used vehicles", and it examines all possibilities such as "used wheels" and "used garages" that can cause similar problems. Recording a dictionary requires a great deal of effort. If “used car” is not recorded, the analysis result “used | vehicle” can be obtained as in the present embodiment. However, it is difficult to predict at the dictionary recording stage that the desired result will not be obtained if the "used car" as a morpheme is recorded in the dictionary. If it is changed, it will affect other recorded contents and programs, and it will take much effort to maintain and manage the recorded contents of the dictionary so that a certain analysis result can be obtained.
[0041]
Note that “used car | both” and “used | vehicles” are also 2 by the minimum number of division method (a technique that gives priority to the result of minimizing the total number of morphemes constituting the input character string). Since it is a morpheme, it is not determined which is the analysis result.
[0042]
In this regard, if the longest match-N character retraction method according to the present embodiment is adopted, the morpheme requested by the user such as “used | vehicle” or “used | wheel” is reliably identified.
[0043]
In addition, as other analysis methods, there are existing analysis programs in which a desired combination of morphemes for various input sentences is learned, or an analysis result that gives priority to the statistically most probable analysis result has already existed. Recording all examples of what is expected to cause problems is labor intensive and the cost of learning / statistical data creation is a major problem. In addition, it is not guaranteed whether the learned analysis result or the statistically most likely analysis result is the desired morpheme identification result for the analysis of an arbitrary input sentence.
[0044]
In this regard, according to the kanji character string morpheme analysis program 14 of the present embodiment, since it is not necessary to provide the learning / statistical program as described above, a very simple program can be adopted as compared with the conventional program. And the program capacity can be reduced.
[0045]
Although the Kanji character string morpheme N character registration dictionary 12 in this embodiment records the delimiter position information as data indicating the position of the Nth character from the character behind the morpheme, it is not limited to this. . For example, it may be recorded as the Nth character from the character before the morpheme. However, since the position of the Nth character from the back character is indicated as in the present embodiment, a process of moving back N characters from the determination target position when the first morpheme candidate is acquired is performed. Compared to the case where the process of returning to the beginning of the character string of the first morpheme candidate and moving N characters further is performed, the calculation amount of the program is reduced. Therefore, morphological analysis is performed more quickly.
[0046]
As another embodiment of the Kanji character string morpheme N character registration dictionary 12, FIG. 3B shows a part of an example of a Kanji character string morpheme partition position pointer registration dictionary. In this dictionary, not the N-character backward data as shown in FIG. 3A but the delimiter position pointer as delimiter position information is recorded in the character string of the morpheme. In this case, the determination when the first morpheme candidate is acquired The process of acquiring the second morpheme candidate by moving the target position to the position of the pointer is performed.
[0047]
(5) Configuration of the hiragana morpheme junction list dictionary 18
FIG. 6 shows a part of an example of the hiragana morpheme junction list dictionary 18 as a hiragana morpheme recording unit. In the dictionary column, a hiragana dictionary as a hiragana morpheme recording unit in which hiragana morphemes and hiragana morpheme junctions are recorded is recorded.
[0048]
Here, the hiragana morpheme joint is a character string in which a plurality of morphemes represented by hiragana or the like are joined in consideration of the correctness of the grammatical joint and their break positions are also recorded. In the present embodiment, hiragana morpheme conjunctions are recorded for particles, auxiliary verbs, basic verbs, formal nouns, etc., which generally occupy most of character strings represented by hiragana and the like.
[0049]
For example, “No” and “Place” can be judged as separate morphemes, but it is grammatically correct that “Place” is joined to the case particle “No” to express the relationship of existence. Therefore, a morpheme junction, which is a combination of two morphemes “no | place”, is recorded in the dictionary.
[0050]
(6) Morphological analysis method for character string including hiragana
FIG. 7 shows processing when the hiragana character string morphological analysis program 16 performs morphological analysis of a character string including hiragana according to the present embodiment.
[0051]
First, the hiragana character string morpheme analysis program 16 refers to the hiragana character string morpheme junction list dictionary 18 and collates the data (step S20). This collation is performed by repeating so-called brute force search processing until all characters in the series of character string data match any morpheme or morpheme junction recorded in the dictionary 18 (step S21). . Next, it is determined whether or not the grammatical joining rule between each morpheme of the acquired morpheme candidates is appropriate (step S22). Will get. The determination of the joining rule is the same as that generally used for morphological analysis programs. If the joining rule is appropriate, the acquired one is determined and recorded as a fixed morpheme (step S23), and the same matching process is performed for the remaining hiragana character strings (steps S24 and S25). ) If the analysis results of all data are recorded, the analysis is terminated (step S26).
[0052]
As a specific example of the hiragana morphological analysis method as described above, FIG. 8 shows an example in which an input character string “up to the limit” is given. First, the hiragana character string morpheme analysis program 16 refers to the hiragana character string morpheme junction list dictionary 18 and collates the data (step S70). Then, if the morpheme junction “no | place” and the morpheme “barely” “up to” recorded in the dictionary 18 are acquired, the morpheme is acquired for all characters in the series of character strings. Therefore, these are acquired as morpheme candidates (step S71), it is determined that the joining rules are appropriate (step S72), they are determined and recorded as confirmed morphemes (steps S73, S74), and the morpheme analysis is performed. The process ends, and “up to | where | of last minute |” is set as the analysis result (step S75).
[0053]
(7) Effects of the morphological analysis method for character strings including hiragana
According to the above method, since the morpheme analysis is performed by collating with the dictionary 18 recorded in advance, the calculation amount of the program until the morpheme is determined is greatly reduced. Therefore, morphological analysis is performed very quickly.
[0054]
Hereinafter, such an effect of the present embodiment will be described using the example shown in FIG. 8 while comparing with the conventional morphological analysis method.
[0055]
A conventional morpheme analysis program obtains all possible morpheme candidates by sequentially decomposing an input character string, and performs analysis in consideration of the possibility of joining these morphemes. For example, in the case of the conventional program, in the example of FIG. 8, the morpheme candidates are determined and acquired while sequentially decomposing the input character string. ("Noto" includes the possibility of an erroneous analysis result such as "no" is joined to the case particle "no" to indicate the relationship of equality). Processing time until acquisition will be long.
[0056]
In this regard, if the collation processing with the junction list dictionary according to the present embodiment is employed, the morpheme requested by the user “up to | where |” is quickly and reliably identified.
[0057]
Note that the dictionary 18 records hiragana morpheme junctions for particles, auxiliary verbs, basic verbs, formal nouns, etc., which occupy most of the character strings represented by hiragana or the like in ordinary Japanese sentences. According to this, it is general that the capacity of the dictionary becomes larger than the conventional method in which only the morpheme of the character string including the hiragana is recorded in the dictionary. However, according to the present embodiment, since the amount of calculation of the program is greatly reduced, the program capacity can be reduced, and as a result, quick analysis is performed, thereby realizing a morphological analysis system preferable for the user. Will be.
[0058]
As described above, the morphological analysis method for character strings including kanji and the morphological analysis method for character strings including hiragana have been described, but the Japanese morphological analysis system in this embodiment includes a program for performing both of these, Morphological analysis can be performed accurately and quickly even when the character string to be analyzed is a mixture of characters such as kanji and hiragana. In particular, for Japanese character strings, the ratio of hiragana and the ratio of kanji is about 50%, respectively. In other words, the analysis is performed by the program corresponding to each distinction, so that the analysis is performed more quickly.
[0059]
Note that the criteria for determining the quality of morphological analysis results may differ depending on the purpose of use of the results, so the recorded contents of the dictionary in FIG. 3A and FIG. 6 are not limited to this, and their usage The dictionary is recorded so that a desired analysis result corresponding to the purpose can be obtained. For this, not only the creator of the dictionary but also the user himself / herself may be able to change the contents of the dictionary according to its purpose. Specifically, N character data can be changed or a new hiragana morpheme connection can be made. It can be recorded or corrected if there is an error.
[0060]
(8) Other embodiments of the morphological analysis system according to the present invention
In the above embodiment, a system in which a personal computer program analyzes an input character string is shown as a Japanese morpheme analysis system, but the present invention is not limited to this, and other general use of Japanese morpheme analysis, for example, It can also be used for word processor kanji-kana conversion programs and dictionary recording.
[0061]
Further, as shown in FIG. 9, it can be employed in a morphological analysis service using a network. In this case, the user of the terminal device 50 as a communication terminal inputs a character string to be analyzed into the device, and the character string transmitting means sends the data to the morphological analysis service server 56 as a computer server via the Internet 52. Send. The morpheme analysis service server 56 including the receiving unit, at least one of the kanji morpheme recording unit and the hiragana morpheme recording unit, the morpheme analysis unit, and the morpheme analysis result transmission unit performs the morpheme analysis of the data. The analysis result is transmitted to the terminal device 50.
[0062]
Thereby, it is not necessary to provide a kanji morpheme recording unit, a hiragana morpheme recording unit, and a morpheme analysis unit on the terminal device side. Therefore, a larger number of terminal device users can easily obtain the morphological analysis results.
[Brief description of the drawings]
FIG. 1 is a diagram showing a program configuration of a Japanese morphological analysis system according to an embodiment of the present invention.
FIG. 2 is a block diagram of a personal computer equipped with a Japanese morphological analysis system.
FIG. 3A is a diagram showing a part of an example of a Kanji character string morpheme N character backward dictionary according to an embodiment of the present invention; FIG. 3B is a diagram illustrating a part of an example of a kanji character string morpheme break position pointer registration dictionary according to another embodiment of the present invention.
FIG. 4 is a diagram illustrating a morphological analysis method for a character string including kanji according to an embodiment of the present invention.
FIG. 5 is a diagram illustrating a specific example of a morphological analysis method for a character string including kanji according to an embodiment of the present invention.
FIG. 6 is a diagram illustrating a part of an example of a hiragana morpheme junction list dictionary according to an embodiment of the present invention.
FIG. 7 is a diagram illustrating a morphological analysis method for a character string including hiragana according to an embodiment of the present invention.
FIG. 8 is a diagram illustrating a specific example of a morphological analysis method for a character string including hiragana according to an embodiment of the present invention.
FIG. 9 is a diagram illustrating a network Japanese morphological analysis system according to another embodiment of the present invention.
[Explanation of symbols]
10 ... Character type distinction program
12 ... Kanji character string morpheme N character registration dictionary
14 ... Kanji character string morphological analysis program
18 ... Hiragana Morphological Junction List Dictionary
16 ... Hiragana character string morpheme analysis program
20 ... Reading program for morphemes

Claims (8)

A kanji character string morpheme analysis method for performing a morpheme analysis of a character string including kanji by referring to a morpheme recording unit in which a morpheme is recorded by a computer system,
In the morpheme recording unit,
If the morpheme is a morpheme that, when combined with any other character string, results in at least two morphemes separated in the character string of the morpheme, it indicates the delimited position in association with the morpheme Record the break position information together,
When the computer system performs morphological analysis of the input character string,
The first morpheme candidate is obtained by the longest match method from the first character of the character string by reading with reference to the morpheme from the morpheme recording unit,
If the position information is recorded in the first morpheme candidate, the second morpheme candidate including a character string subsequent to the character string of the first morpheme candidate by the longest match method with the subsequent character at the delimiter position as the first character To determine whether or not
If the second morpheme candidate can be obtained, the first morpheme candidate and the second morpheme candidate when the first morpheme candidate is divided at the delimiter position are used as the morpheme analysis result,
If the second morpheme candidate cannot be acquired, the kanji character string morpheme analysis method using the first morpheme candidate as the morpheme analysis result.   In the Chinese character string morphological analysis method of Claim 1,
  The delimiter position information is
  It is information indicating that it is located at the Nth character from the character behind the morpheme,
  The information indicating the position of the Nth character is as follows.
  Information used to determine whether or not a morpheme including the subsequent character can be obtained by the longest match method with the subsequent character at the position of the Nth character recorded in the morpheme as the first character;
  Kanji character string morphological analysis method characterized by the above. A kanji character string morpheme analysis system that performs a morpheme analysis of a character string including kanji by referring to a morpheme recording unit in which a morpheme is recorded by a computer system,
In the morpheme recording unit,
If the morpheme is a morpheme that, when combined with any other character string, results in at least two morphemes separated in the character string of the morpheme, it indicates the delimited position in association with the morpheme Record the break position information together,
When the computer system performs morphological analysis of the input character string,
The first morpheme candidate is obtained by the longest match method from the first character of the character string by reading with reference to the morpheme from the morpheme recording unit,
If the position information is recorded in the first morpheme candidate, the second morpheme candidate including a character string subsequent to the character string of the first morpheme candidate by the longest match method with the subsequent character at the delimiter position as the first character To determine whether or not
If the second morpheme candidate can be obtained, the first morpheme candidate and the second morpheme candidate when the first morpheme candidate is divided at the delimiter position are used as the morpheme analysis result,
If a 2nd morpheme candidate cannot be acquired, a 1st morpheme candidate will be made into a morpheme analysis result, The kanji character string morpheme analysis system characterized by the above-mentioned. A computer-readable program for causing a computer system to function as a kanji character string morphological analysis system for performing morphological analysis of a character string including kanji, or a recording medium on which the program is recorded,
In the morpheme recording unit,
If the morpheme is a morpheme that, when combined with any other character string, results in at least two morphemes separated in the character string of the morpheme, it indicates the delimited position in association with the morpheme Record the break position information together,
When the computer system performs morphological analysis of the input character string,
The first morpheme candidate is obtained by the longest match method from the first character of the character string by reading with reference to the morpheme from the morpheme recording unit,
If the position information is recorded in the first morpheme candidate, the second morpheme candidate including a character string subsequent to the character string of the first morpheme candidate by the longest match method with the subsequent character at the delimiter position as the first character To determine whether or not
If the second morpheme candidate can be obtained, the first morpheme candidate and the second morpheme candidate when the first morpheme candidate is divided at the delimiter position are used as the morpheme analysis result,
If the second morpheme candidate cannot be acquired, a program for causing a computer system to execute a process of using the first morpheme candidate as a morpheme analysis result, or a recording medium on which the program is recorded.   In the program of Claim 4, or the recording medium which recorded the program,
  The delimiter position information is
  It is information indicating that it is located at the Nth character from the character behind the morpheme,
  The information indicating the position of the Nth character is as follows.
  Information used to determine whether or not a morpheme including the subsequent character can be obtained by the longest match method with the subsequent character at the position of the Nth character recorded in the morpheme as the first character;
  Or a recording medium on which the program is recorded. A character input part for inputting a character string;
At least one of a kanji morpheme recording unit that records a morpheme of a character string that includes kanji and a hiragana morpheme recording unit that records a morpheme of a character string that includes hiragana;
A morphological analysis unit that performs morphological analysis;
A morphological analysis system comprising:
The kanji morpheme recording unit is
When a morpheme of a character string containing Kanji is combined with any other character string after at least one of the first half character string or the second half character string when the character string is separated, a character string that becomes at least two morphemes If included,
Separation position information indicating the division position of the morpheme in association with the morpheme,
Is also recorded,
The delimiter position information is obtained by a computer,
Information used to determine whether or not a morpheme including the subsequent character can be obtained by the longest match method, with the subsequent character recorded in the morpheme as the first character as the first character,
It is characterized by
The hiragana morpheme recording unit is
Record at least a hiragana morpheme combination, which is a combination of morphemes that can be grammatically correct among a combination of two or more hiragana morphemes containing a particle or auxiliary verb or a formal noun or basic verb;
In the hiragana morpheme junction,
Separation position for each morpheme,
A morphological analysis system characterized by the fact that is recorded together. Communication terminal
A character string transmission means for transmitting a Japanese character string to a computer server;
A morpheme analysis result receiving means for receiving a morpheme analysis result transmitted from a computer server;
With
A computer server connected to a communication terminal so that it can communicate with
A receiving unit that receives information of the transmitted Japanese character string;
At least one of a kanji morpheme recording unit that records a morpheme of a character string that includes kanji and a hiragana morpheme recording unit that records a morpheme of a character string that includes hiragana;
A morphological analysis unit that performs morphological analysis;
A morpheme analysis result transmitting means for transmitting a morpheme analysis result by the morpheme analysis unit to a communication terminal;
A network morphological analysis system comprising:
The kanji morpheme recording unit is
When the morpheme of the character string including the kanji is a morpheme that becomes at least two morphemes separated in the character string of the morpheme when the other arbitrary character strings are subsequently combined,
Separation position information indicating the division position in association with the morpheme,
Is also recorded,
The hiragana morpheme recording unit is a combination of two or more hiragana morphemes that can be joined at least grammatically,
Is recorded,
In the hiragana morpheme junction,
The separation position for each morpheme is recorded together,
The morpheme analysis unit that performs morphological analysis of a character string including kanji is
First morpheme candidate acquisition means for acquiring a first morpheme candidate from a first character of a character string by a longest match method by reading out the morpheme from the kanji morpheme recording unit;
If the position information is recorded in the first morpheme candidate, the second morpheme candidate including a character string subsequent to the character string of the first morpheme candidate by the longest match method with the subsequent character at the delimiter position as the first character Second morpheme candidate acquisition determination means for determining whether or not can be acquired;
If the second morpheme candidate can be acquired, the first morpheme candidate and the second morpheme candidate when the first morpheme candidate is divided at the dividing position are set as the morpheme analysis results. A kanji character string morpheme determining means which takes one morpheme candidate as a morpheme analysis result;
With
The morpheme analysis unit that performs morphological analysis of a character string including hiragana,
If there is a character string that matches the hiragana morpheme junction by reading out the hiragana morpheme junction and the hiragana morpheme from the hiragana morpheme recording unit, the hiragana character string having the combination of the hiragana morpheme as a morpheme analysis result Morpheme determination means,
A network morphological analysis system characterized by comprising: The morphological analysis system according to claim 3, 6 or 7,
The delimiter position information is
It is information indicating that it is located at the Nth character from the character behind the morpheme,
The information indicating the position of the Nth character is as follows.
Information used to determine whether or not a morpheme including the subsequent character can be obtained by the longest match method with the subsequent character at the position of the Nth character recorded in the morpheme as the first character;
A morphological analysis system characterized by

Images