JPH09160929A - Document processor and method therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically generate a proper retrieval key from document data. SOLUTION: When document data to be stored in a data base 14 is inputted in a document input means 16, a document division means 17 divides document data into a prescribed unit and a character extraction means 19 extracts the partial characters from the divided unit. Because a key generation means 20 combines the extracted characters to generate a retrieval key, the retrieval key on which the contents of document data is reflected is automatically generated.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document processing apparatus and method for retrieving document data from a database.

[0002]

2. Description of the Related Art A document processing device such as a document management system is
It has a database in which a large number of document data are stored in advance, and desired document data can be searched and output from this database. In a general document processing device, a search key is added to the document data of a database by using a word that is a keyword. Therefore, when a user creates a search key from a desired word and inputs it into the document processing device, the search key matches. The document data to be retrieved is retrieved from the database.

At present, a translation support device for assisting translation of document data has also been developed, and it is also possible to use the document processing device as described above as such a translation support device. Has been proposed to. That is, when translating a sentence, the same sentence or a similar sentence may appear. Therefore, in the translation support device disclosed in Japanese Patent Laid-Open No. 4-160473, the translated document data is clustered and stored in the database together with the translated text, and the document data similar to the document data to be translated is stored in the database. Search for. When this is searched, only the different portion is automatically translated, and this is combined with the translated sentence of the retrieved document data to be the translated sentence of the document data to be translated.

[0004]

With the above-described document processing apparatus, it is possible to search the database for document data whose search keys match. However, when storing the document data in the database, it is necessary to add a search key reflecting the content to the document data, and this work is complicated. Similarly, when the document data is searched, the user needs to appropriately create a search key that reflects the content of the document data, and this work is also complicated.

Further, in the translation support device using the document processing device, since the translated document data can be used as a material, the accuracy of automatic translation can be improved. However, in the translation support device disclosed in Japanese Patent Laid-Open No. 4-160473, the similarity is calculated in order to search for similar document data, so the load of this process is large and the speed is low.

[0006]

According to another aspect of the present invention, there is provided a document processing apparatus, a database in which a large number of document data are stored, a document inputting unit for inputting document data stored in the database, and an input document. Document dividing means for dividing the data into predetermined units, character extracting means for extracting some characters from the divided units, key generating means for generating a search key by combining the extracted characters, and And a key registration means for registering the search key in association with the document data. Therefore, when the document data to be stored in the database is input, the document data is divided into predetermined units, and some characters are extracted from the divided units. A search key is generated by combining the extracted characters, and the generated search key is registered in association with the document data. Therefore, if the document data is input in natural language or the like, the search key that reflects the content will be displayed. It is automatically generated.

According to another aspect of the document processing apparatus of the present invention, a database storing a large number of document data and a search key obtained by dividing the document data into predetermined units, extracting some characters, and combining the search keys are the document data. Key registration means registered corresponding to, document input means for receiving the input of the document data of the search request, document dividing means for dividing the input document data into predetermined units, and a part of the divided units. Character extracting means for extracting the characters, key generating means for generating the search key by combining the extracted characters, and searching the generated search key from the key registration means to detect corresponding document data from the database. And a document search means for performing the search. Therefore, a large number of document data are stored in a database, the document data is divided into predetermined units, some characters are extracted, and the combined search key is registered in the key registration means in association with the document data. When the document data of the search request is input, the input document data is divided into predetermined units, and some characters are extracted from the divided units. A search key is generated by combining the extracted characters, the generated search key is searched from the key registration means, and corresponding document data is detected from the database. Therefore, the search request is made in natural language or the like. If you input as, the search key reflecting the content is automatically generated.

According to a third aspect of the present invention, in the document processing apparatus according to the first aspect, when the document search means searches the generated search key from the key registration means, the pattern matching is incomplete even if the pattern matching is incomplete. Search for the search key. Therefore, not only the document data whose search keys completely match, but also the document data whose search keys are similar are searched.

According to a fourth aspect of the present invention, in the document processing apparatus according to the first, second or third aspect, the character extracting means extracts a predetermined number of characters from the head of the divided unit of the document data. Therefore, the search key is formed by a small number of characters.

In the document processing apparatus according to the fifth aspect, in the document processing apparatus according to the first, second or third aspect, the character extracting means extracts every other character from the beginning of the divided unit of the document data. Therefore, the search key is formed by a small number of characters.

According to a sixth aspect of the present invention, in the document processing apparatus according to the first, second, third, fourth or fifth aspect, the syntactic analysis means for syntactically analyzing the document data to determine the role of the divided words. The document dividing means divides the document data into words, the syntactic analysis means parses the document data to determine the role of the divided words, and the key generation means determines the characters extracted from the words. A search key is generated by combining the roles of and. Therefore, not only the character of the word of the document data but also the role of the word is included in the search key.

According to a seventh aspect of the present invention, there is provided the document processing apparatus according to the sixth aspect, wherein the character extracting means is a target for extracting characters of only words having a specific role. Therefore,
In the syntactic structure of document data, a search key can be generated only from words having a main role or an essential role.

The document processing apparatus according to claim 8 is the document processing apparatus according to any one of claims 1, 2, 3, 4 and 5, further comprising a part-of-speech estimating means for estimating the part-of-speech of the divided word, and the document dividing means is , The document data is divided into words, the part-of-speech estimation means estimates the part-of-speech of the divided words, and the key generation means combines the characters extracted from the words with the part-of-speech to generate a search key. Therefore, not only the character of the word of the document data but also the part of speech of the word is included in the search key.

According to a ninth aspect of the present invention, there is provided the document processing apparatus according to the eighth aspect, wherein the character extraction means targets only the words of a specific part of speech. Therefore,
The search key can be generated only from the main part of speech of the document data and the word of the required part of speech.

The document processing apparatus according to claim 10 is the document processing apparatus according to any one of claims 1, 2, 3, 4 and 5,
Providing attribute storage means for storing attribute data of words in advance,
Attribute detecting means for detecting attribute data of the divided words is provided, the document dividing means divides the document data into words, the attribute detecting means detects the attribute data of the divided words, and the key generating means , The search key is generated by combining the character extracted from the word with the attribute data. Therefore, not only the character of the word of the document data but also the attribute data of the word is included in the search key.

In the document processing method according to the eleventh aspect, when the document data to be stored in the database is input, the document data is divided into predetermined units, and some characters are extracted from the divided units, A search key is generated by combining the extracted characters, and the generated search key is registered in association with the document data. Therefore, if the document data is input in natural language or the like, a search key that reflects the content is automatically generated.

In the document processing method according to the twelfth aspect, a large number of document data are stored in a database, the document data is divided into predetermined units, some characters are extracted, and a search key obtained by combining them is used as a document. When the document data of the search request is input in advance by registering in the key registration means corresponding to the data, the input document data is divided into predetermined units,
Some characters are extracted from the divided units, the extracted characters are combined to generate a search key, the generated search key is searched from the key registration means, and the corresponding document data is searched for in the database. It was detected from. Therefore, when a search request is input as document data in natural language or the like, a search key reflecting the content is automatically generated.

[0018]

BEST MODE FOR CARRYING OUT THE INVENTION An embodiment of a document processing apparatus of the present invention will be described below with reference to the drawings. First, the document processing device 1 illustrated here is provided as a part of a translation support device (not shown), and has a data processing device 2 and an electronic file system 3 as shown in FIG. There is.
The data processing device 2 is an existing microcomputer or workstation, and the electronic file system 3 is an HD (Hard Disk) or MO (Magneto Optical D).
It consists of a large-capacity storage device such as isk).

The data processing device 2 has a CPU (Central
Processing Unit) 4, which has a ROM (Read Only Memory) 5 and a RAM (Random Acce
An ss memory (6), a keyboard 8 having a mouse 7, a display 9, and communication I / Fs (Interfaces) 10 and 11 are connected. The communication line 1 is provided in the first communication I / F 10.
2 is connected to the communication line 12. An FDD (Floppy Disk Drive) for outputting document data and a host computer (both not shown) are connected to the communication line 12. The electronic file system 3 is connected to the second communication I / F 11 by a connector 13, and a database 14 and a key index 15 as a key registration means are built in the electronic file system 3. There is.

As described above, the document processing apparatus 1 according to the present embodiment is provided as a part of the translation support apparatus, and this translation support apparatus converts the whole sentence of English long document data into Japanese one by one. translate. Therefore, the document processing apparatus 1 searches for a sentence obtained by dividing the whole sentence of one piece of document data to be translated into sentences, and retrieves a similar sentence using a sentence to be translated as a search request. Therefore, the whole sentence of one large document data to be translated is divided into sentence units and stored as a plurality of document data items to be searched in the database 14, and the key index 15 stores a plurality of document data items to be searched. Search keys generated one by one from each of the document data are registered.

The document processing apparatus 1 according to the present embodiment executes the storage and retrieval of the document data as described above, and therefore, FIG.
As schematically shown in FIG. 1, it has a document inputting unit 16, a document dividing unit 17, a syntax analyzing unit 18, a character extracting unit 19, a key generating unit 20, and a document searching unit 21.

The document input means 16 receives, as a search target, full-text document data stored in the database 14 and one-sentence document data converted into a search key as a search request by, for example, the communication I / F 10. It is input from the translation support device. When the document data input in this way is the whole sentence to be searched, this document data is stored in the database 14 by the communication I / F 11. When the translation support device translates the document data, this translated sentence is also sequentially added to the document data stored in the database 14.

The document dividing means 17 performs morphological analysis on the input document data and divides the document data into predetermined units. That is, when the input document data is a whole sentence to be searched, this is divided into sentence units to generate a plurality of document data to be searched, and each of the plurality of document data is divided into words. When the input document data is one sentence of the search request, this document data is divided into word units. Here, since the document data is written in English, the sentence-based division is performed in response to the detection of a period, and the word-based division is performed in response to the detection of spaces or commas.

The syntactic analysis means 18 syntactically analyzes the document data input as described above, and determines the role of the divided words. The role of this word is a subject, a verb, an object, a complement, etc., which is determined by existing syntactic analysis using a word dictionary. The character extracting means 19 extracts some characters from the words divided as described above, but only the words having a specific role in the syntactic structure are extracted from the characters.

More specifically, as shown in FIG. 4, since the result of the syntax analysis is generated in a tree structure, here, the leading parent node and the child node directly related to this are detected, and these are detected. From the node of "S (subject), V (verb), O (object),
Only the word "C (complement)" is detected and the first three letters are extracted from each of these words.
"V, SVO, SVC, SVOO, SVOC" is "S,
Since it is formed by V, O, and C ″, by performing the above-described processing, only the portion corresponding to the skeleton of each sentence pattern is processed.

The key generation means 20 generates a search key by combining each of the three characters extracted from a plurality of words with the role of the word as described above. When this search key is generated from the document data to be searched, this search key is registered in the key index 15 corresponding to the document data. Here, since one search key is generated from each of the plurality of document data to be searched, as shown in FIG. 3, the address and length of one document data are set for each search key. Since the document data to be searched is one full-text document data divided into each sentence, its address is set as the number of characters from the beginning of each sentence to the beginning of each sentence, and its length is the number of characters. It is set.

On the other hand, when the search key is generated from the document data of the search request, this search key is output to the document search means 21. The document search means 21 searches the key index 15 for the search key generated from the document data of the search request, and detects the document data corresponding to the search result from the database 14. Such search processing of the search key is executed by general pattern matching of the character string, and the detected document data is displayed on the display 9 together with the translated text.

In such a configuration, the document processing device 1 of this embodiment functions as a part of the translation support device. This translation support device translates the whole sentence of a long English document data into Japanese one sentence at a time. For example, the translation is automatically executed and the result is presented to the user by the display on the display 9. When the user inputs the approval or correction of the translated text by manually operating the keyboard 8, the translation of the document data is completed in response to this input.

When the translation support device starts such a processing operation, the document processing device 1 divides the document data of the entire English sentence into a plurality of document data in units of sentence and stores them in the database 14, A search key individually corresponding to is generated and registered in the key index 15. In this state, when the translation support device performs sentence-by-sentence translation, the translated sentence is also stored in the database 14 by the document processing device 1 in correspondence with the document data of the original sentence.

In such a processing step, the document processing apparatus 1 searches the database 14 for a sentence similar to one sentence to be translated by the translation support device, and if a translated sentence is set in the retrieved document data, The translated text is provided to the translation support device. For this reason, the user who performs the translation work by the translation support device is provided with the previous translated text for the sentence similar to the previously translated text, so that the translation work can be performed easily and accurately. You can

The document processing method of the document processing apparatus 1 as described above will be sequentially described below with reference to the drawings. First, a document processing method of dividing the document data of the whole sentence to be translated into a plurality of document data of one sentence and storing the divided document data together with the search key will be described first.

When one piece of document data to be translated is set in the translation support device, this document data is transmitted to the document input means 16 of the document processing device 1 via the communication I / F 10 as shown in FIG.
Is entered from One full-text document data input as described above is used as a plurality of document data to be searched for by the communication I.
/ F11 to the database 14. Separately from this, the input document data is morphologically analyzed by the document dividing unit 17, divided into sentence units and converted into a plurality of document data to be searched, and then each of the plurality of document data is divided into word units. It is divided in order.

For example, the input full-text document data is "He certainly didn't want a wife who was fickl.
e as Ann. He had married her. ”, this sentence is“ Hecertainly didn't want a wife who was fickle as
Ann ”“ He had married her ”divided into two document data, then“ He / certainly / didn't / want / a
/ Wife / who / was / fickle / as / Ann ”“ He / had / mar
Each is divided into words "ried / her".

The search target document data divided as described above are parsed in order by the syntax analysis means 18, and the role of each of the divided words is determined as shown in FIG. Note that, here, only the syntactic structure of the first document data is shown in (a) in the figure, and the roles of words in the first document data are shown in order in (b) in the figure. When the syntactic analysis of this document data is completed, the specific words corresponding to this are extracted by the character extracting means 19, and the first three letters are extracted from these words.

Here, the result of the syntactic analysis is generated in a tree structure, and "S (subject), V (verb), O (object), C is calculated from the leading parent node and the child node directly related to it. (complement)"
Are detected, and the first three letters are extracted from each of these words. For example, in the above-described first document data, as shown in FIG. 4, the leading parent node is “V = wa
nt ”, and the child nodes of“ S, O, C ”directly connected to this are“ S = He, O = wife ”.

The characters thus extracted are combined by the key generation means 20 with the word role identifiers "S, V, O, C", which are registered in the key index 15 in alphabetical order as search keys. . As shown in FIG. 3, the search key of the above-mentioned first document data is “SHe
VwanOwif ”, and the search key for the second document data becomes“ SHeVmarOher. ”At this time, the address and length of the document data are also counted, and this is also registered in the key index 15 corresponding to the search key. The search key will be registered in a form corresponding to the document data.

When the generation and registration of the search key by the document processing apparatus 1 as described above is completed for all the plurality of divided document data, the translation support apparatus starts the process of translating the document data. At this time, when the translation support device completes the translation of one sentence, the document processing device 1 stores the translated sentence in the database 14 corresponding to the document data. at the same time,
The document processing apparatus 1 searches for a sentence similar to the sentence to be translated by the translation support device, and if a translation is set in the retrieved document data, provides the translation support device with the translation.

Next, the document processing method of the document processing apparatus 1 will be sequentially described below. First, since the translation support device processes all the document data one sentence at a time, as shown in FIG. 6, this one sentence is input from the communication I / F 10 to the document input means 16 of the document processing device 1 as the document data of the search request. It As described above, the document data of one sentence input as the search request is divided into word units and then parsed, and the search key is composed of the word role identifier and the first three characters. Is generated.

When the search key of the search request is generated in this way, the document search means 21 searches the key index 15 by pattern matching such as binary search, so that the search key matching the pattern is searched. The address and the length of the document data are read, and the document data is detected from the database 14 according to the address and the length.

For example, when the document data of the search request to be translated is "He wants a wifeas Barbra.", The search key is "SHeVwanOwif", which is the "He certainly didn't want a". wife who
Matches with the search key of the first document data "was fickle as Ann."

When similar document data are searched in this way, it is determined here whether or not there is a translated sentence. When it is determined that this translated sentence exists, the translated sentence and the document data are transferred from the document processing device 1 to the translation support device. Provided to. Therefore, this translation support device displays the provided document data and translated text on the display 9, translates the current document data using the provided document data and translated text, and also displays this on the display 9. For this reason, the translation support device can easily and accurately perform the translation of the document data, and can present the user with the previously similar sentence and the translated sentence. Therefore, the user should confirm the accuracy of this translation well. You can

If the document data to be searched is input in natural language or the like as described above, the document processing apparatus 1 of the present embodiment can automatically generate a search key reflecting the contents of the document data. Does not need to create a search key, and the database 14 can be easily constructed.
Similarly, since an appropriate search key can be automatically generated from the document data of the search request, the user does not need to create a search key, and the document data can be searched easily.

Moreover, since all the characters of the word of the document data are not used for generating the search key, but only the first three characters are used, the number of characters forming the search key is limited. Therefore, the key index 1 for registering the search key
The storage capacity of 5 can be saved, and the document search means 21
It is possible to reduce the processing load of pattern matching due to. Moreover, because the document data is parsed to detect a specific word and only that word is targeted for extraction of characters, a search key is generated only from words that play a major role or an essential role in the syntax structure of the document data. Therefore, the content of the document data can be well reflected while the characters forming the search key are well reduced. Further, since not only the characters of the words but also the role of syntactic analysis is incorporated in the search key, the content of the document data can be better reflected in the search key, and the search accuracy can be improved.

The present invention is not limited to the document processing device 1 described above, and allows various modifications. For example, in the above-described embodiment, the document processing device 1 is used as a part of the translation support device, and one sentence obtained by dividing the whole sentence to be translated is targeted for processing. It is also possible to process general document data.

Further, in order to reduce the number of characters of the search key, only the first three letters from the word of the document data are used. However, in this case, the prefix is set in the search key, and there is a concern that the search accuracy is lowered. is there. If this causes a problem, for example, it is preferable to extract every other character from the word of the document data and use this for generating the search key.

Furthermore, in order to improve the retrieval accuracy, it has been illustrated that the document data is syntactically analyzed and the role of the word is also set as the retrieval key. However, the syntactic analysis means 18 for performing such syntactic analysis is replaced by the part of speech. It is also possible to provide an estimation means (not shown) and estimate the part of speech of the word by this part of speech estimation means. Such part-of-speech estimation can be performed by an existing process that uses morphological analysis and concatenation probability, and such a process can reduce the load more than parsing. It should be noted that the process of estimating the part of speech using the morphological analysis and the connection probability in this way is disclosed in Japanese Patent Laid-Open No. 2-254565.
It is described in detail in Japanese Patent Laid-Open Publication No. 4-112364.

When estimating the part-of-speech of a word in this way, it is also possible to extract characters of only a specific part-of-speech. In this case, the search key is obtained only from the main part-of-speech of the document data or the word of the required part-of-speech. Can be generated, so that the content of the document data can be well reflected while reducing the characters of the search key.

Further, attribute storage means for storing attribute data of words in advance is provided, and attribute detection means for detecting attribute data of divided words is provided (neither is shown), and such attribute data is used as a search key. It is also possible to incorporate it.
For example, the words "apple" and "orange" have "F (= Frui
ts) ”attribute data is added. In this case, it is possible to add a desired attribute of a predetermined word, so that it is possible to abstract multiple nouns into a single symbol, and also search key characters. The content of the document data can be reflected well while reducing the.

Further, in the above-described document processing apparatus 1, since the document data is searched by the pattern matching of the search key, if one character of the search key is different, similar document data is not searched. If this causes a problem, it is preferable to use existing DP (Dynamic-Programming) matching in voice recognition so that a similar search key can be searched even if pattern matching is incomplete. In this case, not only the document data whose search key exactly matches,
Document data having a similar search key is also searched, so that the document data can be searched better.

A search key search process using such DP matching will be described below. First, DP matching is a typical processing method of voice recognition, and the similarity of character strings is detected by absorbing variations in pattern length and aligning. If this similarity is defined as the distance between the two search keys, the two character strings can be recognized as substantially the same if this distance is smaller than a predetermined threshold value. Here, when detecting the distance D of the character string "A, B" of the search key between the search request and the search target, A: character string A1, A2, ..., Am (m is the number of characters of A) B: character string B1, B2, ..., Bn (n is the number of characters of B) D (i, j): Distance between A and B between characters Ai and Bj d (i, j): Between characters Ai and Bj Distance INSERT: Penalty associated with insertion error DELETE: Penalty associated with dropout error, as shown in Equation 1 below, character strings "A, B"
The search key distance D (m, n) is

[0051]

[Equation 1]

It is calculated by the following equation. It is assumed here that “INSERT = 1, DELETE = 1”, but this can be set to various numerical values as desired. Since the distance between the two search keys is calculated by DP matching in this way, if the distance is smaller than a predetermined threshold, the two search keys can be determined to be similar.

The means for detecting similar search keys that do not completely match is not limited to the DP matching described above, and various existing processes can be applied. For example, by providing a confusion matrix as shown in Table 1 below,

[0054]

[Table 1]

By utilizing this, it is also possible to judge the similarity of characters individually, and to judge the similarity of the search key which is a character string based on this.

The confusion matrix described above is
At least one of the search key of the search target and the search request is O
It is assumed that the data is input as a CR (Optical Character Reader) reading result, and the character similarity is detected based on the shape, so that an optical reading error can also be compensated. When the input of the search key is a manual operation of the keyboard 8, erroneous input of a character by an adjacent key can be expected. Therefore, if a confusion matrix is created by using this as a character similarity, an operation error can be compensated. it can.

Although various modified examples have been described here, such modified examples can be freely combined within a range in which the contents do not conflict with each other.

[0058]

According to the document processing apparatus of the present invention, a database for storing a large number of document data, a document input means for inputting the document data to be stored in the database, and a predetermined document data for the input document data. A document dividing unit that divides into units, a character extracting unit that extracts some characters from the divided units, a key generating unit that combines the extracted characters to generate a search key, and a document that stores the generated search keys. By having the key registration means for registering in correspondence with the data, when the document data to be stored in the database is input, the search key reflecting the content is automatically generated, so that the user can create the search key. You do n’t have to
It is possible to easily build a database that can search document data satisfactorily.

According to another aspect of the document processing apparatus of the present invention, a database storing a large number of document data and a search key obtained by combining the document data by dividing the document data into predetermined units and extracting some characters Key registration means registered corresponding to, document input means for receiving the input of the document data of the search request, document dividing means for dividing the input document data into predetermined units, and a part of the divided units. A character extracting means for extracting the characters of the above, a key generating means for generating the search key by combining the extracted characters, a document for searching the generated search key from the key registration means and detecting the corresponding document data from the database. By having a search means, if a search request is input as document data in natural language or the like, a search key reflecting the content is automatically generated, and this search is performed. Since document data corresponding to the key is retrieved from the database, there is no need to create a user search key, it is possible to find the proper document data easily.

In the document processing apparatus according to the third aspect, the document search means searches the generated search key from the key registration means by searching a similar search key even if the pattern matching is incomplete. It is possible to search not only document data whose keys are exactly the same, but also document data whose search keys are similar.

In the document processing apparatus according to the fourth aspect, the character extraction means reduces the number of characters forming the search key to a small number by extracting a predetermined number of characters from the beginning of the divided unit of the document data. Therefore, the storage capacity of the key registration means can be saved, and the processing load of the document search means can be reduced.

In the document processing apparatus according to the present invention, the character extraction means reduces the number of characters forming the search key to a small number by extracting every other character from the beginning of the divided unit of the document data. Therefore, the storage capacity of the key registration unit can be saved, the processing load of the document search unit can be reduced, and erroneous search due to the prefix can be prevented.

In the document processing apparatus according to the sixth aspect, the document dividing unit divides the document data into words, and the syntactic analyzing unit syntactically analyzes the document data to determine the role of the divided words and generate a key. By generating a search key by combining the character extracted from the word with the role, the means can incorporate not only the character of the word of the document data but also the role of the word in the search key, so that the document data The content can be reflected in the search key.

In the document processing apparatus according to the seventh aspect, by extracting only the characters of the words having a specific role,
Since the search key can be generated only from the words of the main role and the essential role in the syntax structure of the document data,
The content of the document data can be well reflected in the search key while the number of characters of the search key is properly reduced.

In the document processing apparatus according to the eighth aspect, the document dividing unit divides the document data into words, the part-of-speech estimating unit estimates the part-of-speech of the divided word, and the key generating unit,
By generating the search key by combining the character extracted from the word with the part of speech, it is possible to incorporate not only the character of the word of the document data but also the part of speech of the word into the search key, so that the content of the document data can be searched better Can be reflected in the key.

In the document processing apparatus according to the ninth aspect, the character extracting means targets only the words of a specific part of speech to extract characters, and accordingly, only the main part of speech of the document data and the words of the essential part of speech are extracted. Since the search key can be generated, the content of the document data can be well reflected in the search key while the number of characters of the search key is favorably reduced.

In the document processing apparatus according to the tenth aspect, attribute storage means for storing attribute data of words in advance is provided, the document dividing means divides the document data into words, and the attribute detecting means detects the divided words. Detecting the attribute data, the key generating means incorporates not only the character of the word of the document data but also the attribute data of the word into the search key by combining the character extracted from the word with the attribute data to generate the search key. Therefore, the content of the document data can be reflected in the search key more favorably.

In the document processing method according to the eleventh aspect, when the document data to be stored in the database is input, the document data is divided into predetermined units, and some characters are extracted from the divided units, By generating a search key by combining the extracted characters and registering the generated search key in association with the document data, when the document data to be stored in the database is input, the content is changed. Since the reflected search key is automatically generated, it is not necessary for the user to create the search key, and it is possible to easily construct a database capable of satisfactorily searching document data.

In the document processing method according to the twelfth aspect of the present invention, a large number of document data are stored in a database, the document data is divided into predetermined units, some characters are extracted, and a search key obtained by combining them is used as the document When the document data of the search request is input in advance by registering in the key registration means corresponding to the data, the input document data is divided into predetermined units,
Some characters are extracted from the divided units, the extracted characters are combined to generate a search key, and the generated search key is searched from the key registration means to detect the corresponding document data from the database. By doing so, if a search request is input as document data in natural language or the like, a search key reflecting the content is automatically generated, and the document data corresponding to this search key is searched from the database. Since it is not necessary for the user to create a search key, it is possible to easily search for appropriate document data.

[Brief description of the drawings]

FIG. 1 is a schematic diagram functionally showing a document processing apparatus according to an embodiment of the present invention.

FIG. 2 is a block diagram showing an actual circuit structure of the document processing apparatus.

FIG. 3 is a schematic diagram showing stored contents of a key index which is a key registration means.

FIG. 4 is a schematic diagram showing a result of parsing document data.

FIG. 5 is a flowchart showing a document processing method for generating and registering a search key from document data.

FIG. 6 is a flowchart showing a document processing method for generating a search key from document data and searching the document data.

[Explanation of symbols]

 1 Document Processing Device 14 Database 15 Key Registration Means 16 Document Input Means 17 Document Dividing Means 18 Parsing Means 19 Character Extracting Means 20 Key Generating Means 21 Document Searching Means

Claims (12)

[Claims] 1. A database for storing a large number of document data, a document input means for inputting document data to be stored in this database, a document dividing means for dividing the input document data into predetermined units, and a division. Character extracting means for extracting a part of characters from the generated unit, key generating means for generating a search key by combining the extracted characters, and key registration means for registering the generated search key in correspondence with the document data. And a document processing device. 2. A database in which a large number of document data is stored, and a search key in which the document data is divided into predetermined units, some characters are extracted and combined, and the search keys are registered corresponding to the document data. Registration means, document input means for receiving input of document data of a search request, document dividing means for dividing the input document data into predetermined units, and character extracting means for extracting some characters from the divided units And key generation means for generating a search key by combining the extracted characters, and document search means for searching the generated search key from the key registration means and detecting corresponding document data from the database. A document processing device characterized by the above. 3. The document processing according to claim 1, wherein the document search means searches for a similar search key even if pattern matching is incomplete, when searching the generated search key from the key registration means. apparatus. 4. The document processing apparatus according to claim 1, wherein the character extracting means extracts a predetermined number of characters from the beginning of the divided unit of the document data. 5. The document processing apparatus according to claim 1, wherein the character extracting unit extracts every other character from the beginning of the divided unit of the document data. 6. A syntactic analysis unit for syntactically analyzing document data to determine a role of a word divided, the document dividing unit divides the document data into words, and the syntactic analysis unit includes:
4. The document data is syntactically analyzed to determine the role of the divided word, and the key generation means generates a search key by combining the character extracted from the word with the role. , 4 or 5 document processing device. 7. The document processing apparatus according to claim 6, wherein the character extracting means extracts characters only from words having a specific role. 8. A part-of-speech estimating unit for estimating a part-of-speech of a divided word is provided, the document dividing unit divides document data into words, and the part-of-speech estimating unit estimates a part of speech of the divided word, and a key 6. The document processing apparatus according to claim 1, wherein the generation unit combines the character extracted from the word with a part of speech to generate a search key. 9. The document processing apparatus according to claim 8, wherein the character extracting means extracts characters only from words having a specific part of speech. 10. An attribute storage unit for storing attribute data of a word in advance is provided, and an attribute detection unit for detecting attribute data of the divided word is provided. The document dividing unit divides the document data into words, and the attribute 4. The detecting means detects the attribute data of the divided words, and the key generating means combines the characters extracted from the words with the attribute data to generate the search key. 4. The document processing device according to 4 or 5. 11. When the document data to be stored in the database is input, the document data is divided into predetermined units,
It is characterized in that some characters are extracted from this divided unit, a search key is generated by combining the extracted characters, and the generated search key is registered in association with the document data. How to process documents. 12. A key registration means for storing a large number of document data in a database, dividing the document data into predetermined units, extracting some characters, and combining the retrieval keys in correspondence with the document data. When the search request document data is input, the input document data is divided into predetermined units, some characters are extracted from the divided units, and the extracted characters are extracted. A document processing method is characterized in that a search key is generated by combining the above, and the generated search key is searched from the key registration means to detect corresponding document data from the database.