JPH0785030A - Language processing system - Google Patents

Abstract

PURPOSE:To provide the exact morpheme analyzed result by decreasing the number of unregistered words without increasing the capacity of a dictionary for morpheme analysis even concerning a Japanese sentence composed of a lot of Japanese syllabary(KANA) character strings with high possibility to be conventionally processed as unregistered words. CONSTITUTION:This system is provided with an input part 1 for inputting the Japanese sentence mixing Chinese characters(KANJI) and KANA, dictionary 12 for morpheme analysis, morpheme analysis part 5 for performing morpheme analysis to the Japanese sentence inputted from the input part 1 while using the dictionary 12 for morpheme analysis and for extracting the KANA character string from the Japanese sentence, and KANA/KANJI conversion part 8 for performing the KANA/KANJI conversion of the extracted KANA character string corresponding to the request of a user, and the morpheme analysis part 5 performs the morpheme analysis to the character string as the result of KANA /KANJI conversion again.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a language processing system having a morphological analysis processing function for Japanese sentences.

[0002]

2. Description of the Related Art In general, Japanese sentence morphological analysis processing is
It is done by matching the input Japanese sentence, that is, the input character string with the word registered in advance in the morphological analysis dictionary, identifying the word, and dividing it into word units. If there is an unregistered word (unregistered word), the word cannot be identified for this word and it is processed as an unregistered word.

A problem in processing unregistered words is that the unregistered words are hiragana character strings.
When the unregistered word is a hiragana character string, it is difficult to distinguish it from the adjunct word, and it becomes very difficult to determine the range of the unregistered word, which is a factor that reduces the accuracy of morphological analysis.

In order to avoid such a problem in morphological analysis, conventionally, for example, a technique disclosed in Japanese Patent Laid-Open No. 63-095572 is known. In this technique, a Japanese sentence containing kanji and kana is mixed. In morphological analysis,
An unregistered word is detected as an unregistered word on the assumption that it is an independent word.

[0005]

With the above-mentioned conventional technique, it is possible to detect unregistered words (range identification) of "Hiragana character string" when performing morphological analysis of a sentence with many Hiragana character strings. Although it is possible, the detected unregistered word remains the unregistered word, and the correct reading or part-of-speech is not positively added to it. That is, for a sentence with many hiragana character strings, in general, many unregistered words are included in the morphological analysis result, and even if it can be detected as an unregistered word, it is conventionally not attached with a part of speech or the like. Therefore, there is a drawback that an accurate morphological analysis result cannot be obtained.

[0006] When performing morphological analysis on a sentence with many hiragana character strings, in order to reduce the number of unregistered words occurring, the same word may be registered in the morphological analysis dictionary in hiragana notation together with the kanji notation. It is conceivable, but in this case, the number of registered words increases and the capacity of the morphological analysis dictionary increases, which causes a problem.

The present invention reduces the number of unregistered words without increasing the capacity of the morphological analysis dictionary, even for Japanese sentences with many hiragana character strings that were previously likely to be processed as unregistered words. It is an object of the present invention to provide a language processing system that is capable of obtaining accurate morphological analysis results.

[0008]

In order to achieve the above object, the invention according to claim 1 uses an input means for inputting a Japanese sentence mixed with kanji and kana, a morphological analysis dictionary, and an input means. Morphological analysis is performed on the extracted Japanese sentence using a morphological analysis dictionary, and a morphological analysis unit that extracts a hiragana character string from the Japanese sentence, and the extracted hiragana character string is converted into kana-kanji characters as needed by the user. It has a kana-kanji conversion means, and the morpheme analysis means is characterized in that the character string resulting from the kana-kanji conversion is subjected to morpheme analysis again. As a result, even for Japanese sentences with many hiragana character strings that were likely to be processed as unregistered words, without increasing the ease of the morphological analysis dictionary, and without changing the morphological analysis dictionary at all, The number of unregistered words can be reduced and an accurate morphological analysis result can be obtained.

The inventions according to claims 2 and 3 are:
The above morphological analysis dictionary is configured as one combined with a kana-kanji conversion dictionary. In this case,
It is characterized in that a dictionary lookup control means for switching the dictionary between a search form for morphological analysis and a search form for kana-kanji conversion is further provided. As a result, the dictionary capacity of the entire system can be reduced and a very compact system can be realized.

Further, in the invention according to claim 4, the morpheme analysis means determines a difference in the character type of the input Japanese sentence, extracts a hiragana character string, and extracts the hiragana character string from the extracted characters. Then, the connection with the morphemes before and after it is checked to determine whether or not it is an adjunct word, and the character part that is determined to be an adjunct word is treated as a morpheme, and the character part that is not determined to be an adjunct word. Is characterized in that it is determined as a continuous hiragana character string as a kana-kanji conversion target. As a result, it is possible to clearly identify the portion to be converted into Kana-Kanji and correctly perform the conversion into Kana-Kanji.

Further, in the invention according to claim 5, when the morpheme analysis means performs a morpheme analysis again on the character string resulting from the kana-kanji conversion, the immediately preceding morpheme of the continuous hiragana character string and the kana-kanji converted continuous It is characterized by performing a connection check with the first morpheme of the hiragana character string and a connection check between the morpheme immediately after the continuous hiragana character string and the tail morpheme of the kana-kanji converted continuous hiragana character string. . As a result, when there are a plurality of candidates during Kana-Kanji conversion, the candidates can be narrowed down.

Further, in the invention according to claim 6, the morpheme analysis means performs morpheme analysis again on the character string resulting from the kana-kanji conversion, and then the character string resulting from the kana-kanji conversion is converted into the original hiragana character string. It is characterized by being restored to. As a result, it is possible to finally obtain the result as if the morphological analysis was performed on the original Japanese sentence.

Further, the invention according to claims 7 to 9 further comprises a display means and a display control means for switching the display form of the display means between morphological analysis processing and kana-kanji conversion. have. As a result, the user can easily confirm and understand the processing results and the like during both the morphological analysis processing and the kana-kanji conversion processing.

Particularly, in the invention described in claim 8, the display control means is adapted to switch the display for morphological analysis processing to the display for kana-kanji conversion processing according to the need of the user. As a result, the user can always know which process is currently being performed.

Further, in the invention according to claim 9, when the continuous hiragana character string is extracted in the morpheme analysis processing, the display control means can distinguish the part of the continuous hiragana character string from other parts. It is supposed to be displayed. As a result, it is possible to clearly indicate which part of the character string requires kana-kanji conversion.

[0016]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram of an embodiment of a language processing system according to the present invention.

Referring to FIG. 1, this language processing system performs morphological analysis on an input unit 1 for inputting a Japanese sentence (text character string) mixed with Kanji and Kana and a character string input from the input unit 1. It has a morphological analysis section 5 for performing, a kana-kanji conversion section 8 for performing kana-kanji conversion processing, and a display control section 10 for performing predetermined display control on the display section 11.

Here, the morphological analysis unit 5 registers the word notation and the part-of-speech information for the input character string as shown in FIG.
The morphological analysis is performed using the morphological analysis dictionary 12 and the connection table 4 in which the connection relationships between morphemes are described, and the morphological analysis result is held in the morphological holding buffer 6. ing.

Further, the morphological analysis unit 5 further determines the difference in the character type such as hiragana or numeral for the input character string, and when the hiragana character string is extracted, stores it in the hiragana character string holding buffer 7. It also has a function to do.
In addition, for a character determined to be hiragana, a connection check is performed by referring to the part-of-speech of the morpheme before and after that character, and the adjunct word (word that should be treated as one morpheme) of the previous morpheme is determined from the connection relation. It is also designed to identify whether there is any. In this case, the hiragana character string holding buffer 7 holds the hiragana character string identified as an adjunct word and the other continuous hiragana character strings so that they can be discriminated from each other.

The kana-kanji conversion unit 8 receives the hiragana character string (continuous kana character string) held in the hiragana character string holding buffer 7 as needed by the user (for example, the user selects a predetermined key). -Has a function of performing Kana-Kanji conversion processing (by operating, for example, a conversion key). When performing this Kana-Kanji conversion processing, word notation and reading are stored in association with each other. The kana-kanji conversion dictionary 3 as shown in FIG. 3 is used, and the kana-kanji conversion result is stored in the conversion character string buffer 9.

By the way, when the morpheme analysis dictionary 12 similar to that used in the conventional morpheme analysis is used, generally only the kanji notation of the word is registered in the morpheme analysis dictionary 12, and the hiragana of the word is registered. The notation is often not registered, and as described above, when a sentence with many hiragana character strings is input from the input unit 1, the morpheme analysis unit 5 processes the word in hiragana notation as an unregistered word. . Therefore, the continuous hiragana character string held in the hiragana character string holding buffer 6 is highly likely to be morphologically analyzed as an unregistered word.

The inventor of the present application has paid attention to the fact that a word processed as an unregistered word because it is in Hiragana notation is very often correctly morphologically analyzed if it is in Kanji notation. It has been found that the number of unregistered words can be reduced by converting the continuous hiragana character string held in the character string holding buffer 6 into a character string mixed with kanji kana and then performing morphological analysis again.
Therefore, in the present embodiment, the continuous hiragana character string stored in the hiragana character string holding buffer 7 is converted into kana characters by the kana-kanji conversion unit 8 as required by the user and stored in the converted character string buffer 9. Then, the converted character string stored in the converted character string buffer 9 is further given to the morpheme analysis unit 5 so that the morpheme analysis unit 5 performs the morpheme analysis again.

As described above, by converting a word processed as an unregistered word because it is in Hiragana notation into a Kanji character and again performing morphological analysis, the number of registered words in the morphological analysis dictionary 12 is not increased at all. Furthermore, it is possible to reduce the number of unregistered words without changing the contents of the morphological analysis dictionary 12 (that is, in the state where only the Kanji notation of the word is registered). Further, when converting the continuous hiragana character string to the kana-kanji character by the kana-kanji conversion unit 8, this kana-kanji conversion processing generally enables the man-machine interface to select the desired kanji character of the user. Can be converted into a character string containing the correct kanji and kana characters intended by the user, and thereby an accurate morphological analysis result can be obtained with certainty.

By providing the kana-kanji conversion unit 8,
This is not only used for morphological analysis processing,
It can also be used to further implement other functions such as, for example, a word processor. That is, the system of FIG.
It is possible to easily extend to a more general-purpose language processing system including a function such as a deprocessor.

In the configuration example of FIG. 1, the dictionary used for morpheme analysis processing, that is, the morpheme analysis dictionary 12 and the dictionary used for kana-kanji conversion processing, that is, kana-kanji conversion dictionary 3, are provided separately. In one dictionary,
It is also possible to share both processes.

FIG. 4 is a diagram showing a configuration example of a system in which a dictionary used for morpheme analysis processing and a dictionary used for kana-kanji conversion processing are integrated. Referring to FIG. 4, in this system, a dictionary 22 as shown in FIG. 5 in which word notation, reading, and part-of-speech information are associated, and this dictionary 22.
There is provided a dictionary lookup control unit 2 that can switch the search form between morphological analysis processing and kana-kanji conversion processing.

Here, the dictionary 22 is configured as a collection of the morphological analysis dictionary 12 and the kana-kanji conversion dictionary 3 shown in FIGS. 2 and 3, respectively.
Under the control of, the morpheme analysis process and the kana-kanji conversion process are commonly used. That is, the dictionary look-up control unit 2 makes the dictionary 2 when performing the morphological analysis.
2 is searched by notation (heading area), and when Kana-Kanji conversion processing is being performed, the dictionary 2
To perform a search by reading (reading area) for 2,
The search form is switched and controlled so that the dictionary 22 is used in both processes.

In the configuration of FIG. 4, the search form of the dictionary 22 can be switched as described above under the control of the dictionary lookup control unit 2, so that it is not necessary to provide a dictionary for each process.
Only one dictionary 22 can be used, the dictionary capacity of the entire system can be reduced, and a very compact system can be realized.

In the configuration example of FIGS. 1 and 4, after the continuous hiragana character string is converted by the kana-kanji conversion unit 8 into a character string mixed with kanji kana, the morpheme analysis is performed again to determine the morpheme, Checking the connection between the immediately preceding morpheme of the continuous hiragana character string and the leading morpheme of the converted hiragana-kanji character string, and the connection between the morpheme immediately after the continuous hiragana character string and the tail morpheme of the converted hiragana character string The morpheme analysis unit 5 can be provided with the function of checking, and the morpheme analysis unit 5 can also be provided with the function of restoring the character string converted from Kana-Kanji into the Hiragana character string of the original text.

In addition, the display control unit 10 has a function of switching display between the morphological analysis process and the kana-kanji conversion process.
You can also give it. For example, during the morphological analysis process, the display control unit 10 causes the display unit 11 to display the input character string,
A morpheme analysis result held in the morpheme holding buffer 6,
The hiragana character string held in the hiragana character string holding buffer 7 is displayed, and when a continuous hiragana character string is extracted from the hiragana character string, only the part of the continuous hiragana character string of the hiragana character string is changed to black and white. It is designed to be inverted. Also, when the kana-kanji conversion process is selected according to the user's need by operating a predetermined key such as a conversion key, the display unit 11 is switched to the kana-kanji conversion process screen. It has become.

Next, the operation of the language processing system having such a configuration will be described with reference to the flowcharts of FIGS. In the following, for convenience, the language processing system is shown in FIG.
It is assumed that the dictionary lookup control unit 2 is initially configured.
Suppose that the search form of the dictionary 22 is switched for morphological analysis, that is, a search by notation is performed.

When a text character string is input from the input unit 1 (step S1), the morphological analysis unit 5 reads this character string (step S2), the dictionary 22 and the connection table 4.
Using the input character type determination process (steps S3, S4),
The morphological analysis process (step S5) is performed, and the display unit 11
The input character string, the morphological analysis result and the hiragana character string are displayed at (step S6). Here, steps S3 and S4
In the input character type determination process (1), it is determined whether the input character is a hiragana or a number, and when it is determined that the input character is a hiragana, the character is held in the hiragana character string holding buffer 7. The morpheme analysis result is held in the morpheme holding buffer 6. The contents of the morpheme holding buffer 6 and the hiragana character string holding buffer 7 are displayed on the display unit 11.
Are displayed as a morphological analysis result and a hiragana character string, respectively.

In this way, after performing a series of processes as described above on the input text character string, the morphological analysis unit 5 determines whether or not a hiragana character string exists in this text character string. Judgment is made based on whether or not a character string is held in the hiragana character string holding buffer 7 (step S
7). As a result, when the hiragana character string holding buffer 7 does not hold the character string, the hiragana character string does not exist in the text character string, so it is determined that all the morphological analysis has been performed in the process of step S4, and the process ends. To do.

On the other hand, the hiragana character string holding buffer 7
If the hiragana character string is stored in the hiragana character string, the morpheme analysis unit 5 further determines that the leading character of the hiragana character string held in the hiragana character string holding buffer 7 is the previous morpheme of the hiragana character string. A connection check is performed (step S8), and it is identified whether or not this is an adjunct word of the immediately preceding morpheme (step S9). When it is an adjunct word, this character shifted by one character is regarded as the first character (step S9). S10), the process returns to step S7, and it is checked whether or not the character shifted by one character is also an adjunct word.

When the first character is no longer an adjunct word after repeating the processing of steps S7 to S10, in this hiragana character string, the hiragana character string portion identified as an adjunct word and the hiragana character string other than the adjunct word And the hiragana character string portion identified as an adjunct word is treated as a morpheme, and the hiragana character string portion other than the adjunct word is detected as a continuous hiragana character string (unregistered word at this stage). can do. At this time, the hiragana character string portion detected as the continuous hiragana character string is held in the hiragana character string holding buffer 7 in a form distinguishable from other character string portions (for example, a flag or the like is attached). In addition, the display unit 11 displays the hiragana character string portion detected as the continuous hiragana character string in a form according to the user's need (for example, by black and white reverse display) and distinguishes it from other character strings. (Steps S12 to S14).

In this state, when the user operates a conversion key or the like to issue a kana-kanji conversion request (when the kana-kanji conversion mode is selected) (step S1).
5) The display control unit 10 switches the display screen of the display unit 11 from the display for morphological analysis to the display for Kana-Kanji conversion (step S16). Further, the dictionary lookup control unit 2 switches the search mode of the dictionary 22 to the search using the kana character string (reading) (step S17). At this stage, the kana-kanji conversion unit 8 performs kana-kanji conversion of the continuous hiragana character string held in the hiragana character string holding buffer 7 by the man-machine interface with the user (step S18), and the result is obtained. The display unit 1 stores the converted character string in the buffer 9.
No. 1 is displayed (step S19). If the conversion result is intended by the user, the conversion result is confirmed by the user pressing a predetermined key (for example, a confirmation key (not shown)) (step S20), and the kana-kanji conversion is performed. The mode is exited (step S21), and control is transferred to the morpheme analysis unit 5. Also, at this stage, the dictionary lookup control unit 2
Switches the search form of the dictionary 22 to the search by notation.

In this way, when control is transferred to the morpheme analysis unit 5, the morpheme analysis unit 5 causes the converted character string buffer 9
Morphological analysis is performed on the kana-kanji conversion result stored in, that is, the kana-kanji converted character string (step S22). In addition, at this time, the connection between the immediately preceding morpheme of the continuous hiragana character string and the leading morpheme of the kana-kanji converted continuous hiragana character string is checked, the morpheme immediately after the continuous hiragana character string and the tail of the kana-kanji converted continuous hiragana string Also check the connection with morphemes.

The morpheme analysis result in step S22 is held in the morpheme analysis buffer 6 and displayed on the display unit 11 (step S23). When the morpheme is determined by the user based on this display (step S24), the morpheme analysis unit 5 converts the kana-kanji converted portion into the original hiragana character string, that is, restores the original text character string (step S24). S25), in this way, the morphological analysis of the input sentence is completed.

Next, a specific example will be described. When the character string “A large number of people are walking around” as shown in FIG. 9A is input from the input unit 1, the morphological analysis unit 5 first performs a morphological analysis on this character string. . This morphological analysis result is displayed on the display unit 11 as shown in FIG.

At this time, as shown in FIG. 9 (c),
“Ozei no” and “go gakuro” are determined as hiragana character strings and are held in the hiragana character string holding buffer 7. In this morphological analysis process, the dictionary lookup control unit 2 is designed to search the dictionary 22 by notation. Therefore, in the above Hiragana character string, "Ozei", "Tsuro", and "Toru" Is treated as an unregistered word at this stage. Of the hiragana character strings held in the hiragana character string holding buffer 7 as described above, the morphological analysis unit 5 first pays attention to the hiragana character string “Ozei no” and immediately before the leading character “o”. Check whether there is a morpheme. In this case, since the leading character "O" has no immediately preceding morpheme, the hiragana character string "Ozei no" having this character "O" as the leading character is determined as a continuous hiragana character string. Further, the morpheme analysis unit 5 next pays attention to the hiragana character string “go through” and checks whether or not the leading character “ga” has the immediately preceding morpheme. In the present case, the first character "ga" has "person" as the immediately preceding morpheme, and the character "ga" is determined to be an adjunct word to "person", and is determined as one morpheme of the case particle "ga". To be done. Next, the character is shifted by one character, and the character "tsu" is similarly examined. However, the character "tsu" is the case particle "ga" as the immediately preceding morpheme, so the hiragana character string "tsuuro "Toru" is determined as a continuous hiragana character string.

Thus, from the character string of FIG. 9 (a) to FIG.
Two consecutive hiragana character strings "Ozei no" like (d),
When “Thurs through the strings” are extracted, these are held in the hiragana character string holding buffer 7 so as to be distinguishable by, for example, attaching a flag or the like, and are also displayed on the display unit 11 in a required form (eg, black and white). It is displayed (inverted).

At this stage, when the user selects the kana-kanji conversion mode, the dictionary search form is replaced by the kana character string search form. If the kanji “massive”, “passage”, “pass” is registered in the dictionary 22 in correspondence with the kana character strings “Ozei”, “tsuuro”, and “toru”, kana-kanji conversion The section 8 converts the two consecutive hiragana character strings in FIG. 9 (d) into the character strings “large number” and “passage” as shown in FIG. 9 (e) and displays the kana-kanji conversion result. It is displayed on the part 11. The user looks at this display and confirms whether or not it has been converted to the intended one. If it is the intended one, for example, the confirmation key (not shown) is selected, and kana kanji End the conversion process.

As a result, the control is transferred to the morpheme analysis unit 5 again, and the morpheme analysis unit 5 analyzes the input character string in the Kana-Kanji converted state again and checks the connection. For example, for the continuous hiragana character string "Touro Toru", the morpheme is analyzed into "passage,""o," and "Toru," and the morpheme immediately preceding the continuous Hiragana character string "Touro Toru" (that is, "ga" Morpheme) and the leading morpheme of the kana-kanji converted continuous hiragana character string (that is, the morpheme of the "passage"), and also the morpheme immediately after the continuous hiragana character string "Tsuurou Toru" (that is, , ".") And the trailing morpheme (that is, "pass" morpheme) of the kana-kanji converted continuous hiragana character string. In this way, the morpheme is identified after the analysis and connection check are performed again. As a result, in the Hiragana notation, a word that does not exist in the dictionary and becomes an unregistered word can be converted into kanji to be morphologically analyzed, and the input character string in FIG. ) Can be obtained. This morpheme analysis result is displayed on the display unit 11, and if this is what the user intended, the user selects the confirmation key. When the confirmation key is selected, each morpheme of FIG. 9 (f) is restored to the original text notation as shown in FIG. 9 (g),
The morpheme analysis on the character string in FIG. 9A is completed.

[0044]

As described above, according to the invention of claim 1, the input means for inputting a Japanese sentence mixed with kanji and kana, the morphological analysis dictionary, and the Japanese sentence input by the input means are stored. Morphological analysis using a morphological analysis dictionary, and morphological analysis means for extracting hiragana character strings from the Japanese sentence, and kana-kanji conversion means for converting the extracted hiragana character strings into kana-kanji characters as required by the user. Since the morpheme analysis means is configured to perform morpheme analysis again on the character string resulting from Kana-Kanji conversion, Japanese sentences with many Hiragana character strings that were likely to be processed as unregistered words Also, the number of unregistered words can be reduced without increasing the ease of the morphological analysis dictionary and without changing the morphological analysis dictionary, and accurate morphological analysis results can be obtained. That.

According to the second and third aspects of the invention, the morphological analysis dictionary is a kana-kanji conversion dictionary and a morphological analysis dictionary.
In this case, dictionary lookup control means for switching the dictionary between a search form for morphological analysis and a search form for kana-kanji conversion is further provided. Therefore, this dictionary can be shared by both processes, and the dictionary capacity of the entire system can be reduced.
A very compact system can be realized.

According to the invention described in claim 4, the morpheme analysis means determines the difference in the character type of the input Japanese sentence, extracts the hiragana character string, and extracts the hiragana character string into the extracted characters. On the other hand, it is determined whether or not it is an adjunct word by performing a connection check with the morphemes before and after it, and the character part that is determined to be an adjunct word is treated as one morpheme, and the character that is not determined to be an adjunct word. Since the part is determined as a continuous hiragana character string to be converted to kana-kanji, the part to be converted to kana-kanji can be clearly identified and kana-kanji conversion can be performed correctly.

Further, according to the invention described in claim 5, when the morpheme analysis means again performs morpheme analysis on the character string resulting from the kana-kanji conversion, the morpheme immediately before the continuous hiragana character string and kana-kanji conversion are performed. It is designed to perform a connection check with the leading morpheme of the character string and a connection between the morpheme immediately after the continuous hiragana character string and the tail morpheme of the converted kana-kanji character. If there are candidates, they can be narrowed down.

Further, according to the invention described in claim 6, the morpheme analyzing means performs the morpheme analysis again on the character string resulting from the kana-kanji conversion, and then returns the character string resulting from the kana-kanji conversion to the original hiragana character. Since it is supposed to be restored to the column,
It is possible to finally obtain the result as if the morphological analysis was performed on the original Japanese sentence.

Further, according to the inventions of claims 7 to 9, a display control for switching between the display means and the display form of the display means for morphological analysis processing or kana-kanji conversion. Since the means is provided, the user can easily confirm and grasp the processing results and the like at both the morphological analysis processing and the kana-kanji conversion processing.

In particular, according to the invention described in claim 8, the display control means is adapted to switch the display for morphological analysis processing to the display for kana-kanji conversion processing according to the need of the user. The user can always keep track of which process is currently being performed.

According to the ninth aspect of the invention, the display control means can distinguish the part of the continuous hiragana character string from other parts when the continuous hiragana character string is extracted in the morphological analysis process. Since it is designed to be displayed in the form,
It is possible to clearly indicate which part of the character string is necessary for kana-kanji conversion.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of a language processing system according to the present invention.

FIG. 2 is a diagram showing an example of a morphological analysis dictionary.

FIG. 3 is a diagram showing an example of a kana-kanji conversion dictionary.

FIG. 4 is a diagram showing a modification of the language processing system of FIG.

5 is a diagram showing an example of a dictionary of the language processing system of FIG.

6 is a flowchart showing a processing operation of the language processing system of FIG.

7 is a flowchart showing a processing operation of the language processing system of FIG.

8 is a flowchart showing a processing operation of the language processing system of FIG.

FIG. 9 is a diagram showing a specific processing example of the language processing system of the present invention.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 input unit 2 dictionary lookup control unit 4 connection table 5 morpheme analysis unit 6 morpheme holding buffer 7 Hiragana character string holding buffer 8 Kana-Kanji conversion unit 9 conversion character string buffer 10 display control unit 11 display unit 12 morpheme analysis dictionary 22 dictionary

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Office reference number FI technical display location 8125-5L 15/38 E

Claims (9)

[Claims] 1. An input unit for inputting a Japanese sentence mixed with Kanji and Kana, a morphological analysis dictionary, and a Japanese sentence input by the input unit are morphologically analyzed using the morphological analysis dictionary, and the Japanese sentence is also analyzed. It has a morpheme analysis means for extracting a hiragana character string from a word sentence, and a kana-kanji conversion means for converting the extracted hiragana character string into kana-kanji characters as needed by the user, and the morpheme analysis means is kana-kanji converted. A language processing system characterized in that the resulting character string is again subjected to morphological analysis. 2. The language processing system according to claim 1, wherein the morphological analysis dictionary is configured as one combined with a kana-kanji conversion dictionary. In this case, the morphological analysis dictionary is used. A language processing system further comprising dictionary lookup control means for switching between a search form for Kana and Kana-Kanji conversion. 3. The language processing system according to claim 1, wherein the dictionary has at least word notation, word reading, and part-of-speech information, and the dictionary lookup control means sets a dictionary search form to: When morphological analysis is done,
A language processing system characterized by being switched to a search form based on notation, and being controlled to be switched to a search form based on reading when kana-kanji conversion is performed. 4. The language processing system according to claim 1, wherein the morpheme analysis unit determines a difference in the character type of the input Japanese sentence, extracts a hiragana character string, and extracts the hiragana character string as a character. On the other hand, it is determined whether or not it is an adjunct word by performing a connection check with the morphemes before and after it, and the character part that is determined to be an adjunct word is treated as one morpheme, and the character that is not determined to be an adjunct word. The language processing system is characterized in that the part is determined as a continuous hiragana character string to be converted into kana-kanji. 5. The language processing system according to claim 1, wherein the morpheme analysis means performs kana-kanji conversion with the morpheme immediately preceding the continuous hiragana kana character string when the character string resulting from the kana-kanji conversion is subjected to morpheme analysis again. It is characterized by performing a connection check with the leading morpheme of the continuous hiragana character string and a connection check between the morpheme immediately after the continuous hiragana character string and the tail morpheme of the kana-kanji converted continuous hiragana character string. And language processing system. 6. The language processing system according to claim 1, wherein the morpheme analysis unit performs morpheme analysis again on the character string resulting from the kana-kanji conversion, and then returns the character string resulting from the kana-kanji conversion to the original hiragana. A language processing system characterized by being restored to a character string. 7. The language processing system according to claim 1, further comprising display means and display control means for switching the display form of the display means between morphological analysis processing and kana-kanji conversion. A language processing system characterized by being. 8. The language processing system according to claim 7, wherein the display control means displays a display for morphological analysis processing.
A language processing system characterized by switching to a display for Kana-Kanji conversion processing according to the needs of the user. 9. The language processing system according to claim 7, wherein the display control means can distinguish a part of the continuous hiragana character string from another part when the continuous hiragana character string is extracted in the morphological analysis process. A language processing system characterized by being displayed in various forms.