JPH0262656A - Back-up device for elaboration of japanese word document - Google Patents

Abstract

PURPOSE:To keep the uniform quality of documents by performing the elaborating working based on the warning messages which are produced and displayed according to the passages of the detected improper expressions and the types of these expressions. CONSTITUTION:An improper expression inspecting part 3 decides the activation state of an initiation node of an improper expression detecting rule network contained in an improper expression inspecting rule dictionary 4 based on the characteristic data collected by a characteristic data collecting part 2. Then the part 3 activates or inactivates successively the nodes included in the rule network based on the connecting relation among those nodes, the described activating conditions of nodes, and the node value calculating method. Each termination node in the network corresponds to an improper expression, and the termination node that is finally activated means that the improper expression corresponding to the relevant node is detected. Furthermore a position is held in a sentence of an improper expression as the value of the termination node. Thus it is possible to obtain even the positions of the improper expressions in the sentences.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention is a Japanese document processing system that inputs and processes Japanese documents. Regarding an elaboration support device.

[Industrial application field]

Technologies related to document editing in conventional Japanese document processing systems include detection of errors in okurikana included in one document,
Detecting inconsistencies in spelling, detecting the presence of homophones, etc.
The main technology is to detect errors at the word level, and there is no practical technology for detecting inappropriate expressions at the sentence level and presenting an appropriate message to the user based on the detected inappropriate expressions.

Also, for example, Nikkei Baito March 1988 issue [Will it be the deciding factor for the next generation word processor? Composition support/readability evaluation tool"
As described in , readability evaluation technology has been developed to evaluate the readability of sentences. [Problem to be solved by the invention] As described above, conventional document editing technology has been used to detect inappropriate expressions at the level of one sentence and to develop appropriate responses to them. Since there is no practical technique for presenting a message to a user, there is a problem in that the user performing the editing work is unable to comprehensively review inappropriate expressions in a document. Also,
Readability evaluation results are only presented as numerical values based on factors such as the proportion of kanji in a sentence and the average length of a sentence.

It has been impossible to implement a redaction support function that targets documents in a specific field.

An object of the present invention is to provide a Japanese document elaboration support device that supports a user's comprehensive elaboration of inappropriate expressions in a Japanese document processing system, and is adapted to a specific target field of documents. .

[Means to solve the problem] In the Japanese document elaboration support device of the present invention.

A characteristic data collection unit that collects characteristic data for elaboration support for each Japanese sentence by inputting morphological analysis results and syntactic analysis results for each sentence composing the target Japanese document, and a characteristic data collection unit that collects characteristic data for elaboration support for each Japanese sentence, An inappropriate expression testing unit that detects the type of inappropriate expression and the portion containing inappropriate expressions that require elaboration in each Japanese sentence based on registered inappropriate expression detection rules, and the inappropriate expression detection unit. an inappropriate expression detection rule dictionary that stores rules; a warning message generator that generates a warning message based on the detected inappropriate expression location and inappropriate expression type; and a warning message generator that stores the warning message generation rule. Equipped with a rule dictionary.

[For production]

A document that has already been subjected to morphological analysis and syntactic analysis on each Japanese sentence constituting the Japanese document to be revised is input to the Japanese document revision support device.

The characteristic data collection unit collects characteristic data indicating the possibility that inappropriate expressions exist in each Japanese sentence constituting the document to be processed. The inappropriate expression testing section first determines the activation state of the start node of the inappropriate expression detection rule network in the inappropriate expression detection rule dictionary according to the characteristic data collected by the characteristic data collection section. After that, the inappropriate expression testing unit determines the connection relationship of each node of the inappropriate expression detection rule network in the inappropriate expression detection rule dictionary,
Each terminal node in the network activates or deactivates each node in the network sequentially based on the node activation condition described in the node and the node value calculation method. A terminal node that corresponds to an appropriate expression and is finally activated means that an inappropriate expression corresponding to that node has been detected. Also,
By holding the position of the inappropriate expression in the sentence as the value of the terminal node, the position of the inappropriate expression in the sentence can also be obtained.

The inappropriate expression information (type of inappropriate expression, position of inappropriate expression) detected as described above is as follows.

Sent to the warning message generator. The warning message generation section generates a warning message to be presented to the user by applying the warning message generation rules in the warning message generation rule dictionary based on the inappropriate expression information sent from the inappropriate expression verification section.

This generated warning message through the output section.

By displaying the warning message on a display device such as a display device, the user can be notified of the warning message.
According to this device, the type and position of inappropriate expressions that exist in a document are notified to the user as a warning message, so the user can follow the content of the warning message. Be able to refine aptitude expressions into appropriate expressions. In addition, when targeting documents in a specific field, inappropriate expression detection rules that correspond to expressions unique to that field or expressions that are prohibited to be used in that field may be changed to inappropriate expression detection rules. Just by registering it in a dictionary, it can be used as a Japanese document editing support device specialized for the field of application.

〔Example〕

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

FIG. 1 is a block diagram showing an embodiment of the Japanese document elaboration support device of the present invention. The Japanese document elaboration support device includes an input section 1. Characteristic data collection unit 2, inappropriateness expression testing unit 3. Inappropriate expression detection rule dictionary 4, warning message generation unit 5. It is composed of a warning message generation rule dictionary 6 and an output section 7.

The inappropriate expression detection rule dictionary 4 is a rule dictionary in which inappropriate expression detection rules configured in a network are registered in advance. The warning message generation rule dictionary 6 is a rule dictionary in which warning message generation rules corresponding one-to-one to each unsuitability expression item are registered in advance.

It is assumed that a document is input from the input unit 1, in which each Japanese sentence constituting the Japanese document to be revised has been subjected to morphological analysis and syntactic analysis.

Morphological analysis is the process of dividing Japanese sentences written in solid kanji and kana into word units.

During this process, phrases are also recognized at the same time. Syntactic analysis is the process of analyzing the syntactic structure of the entire sentence into a tree structure based on the word segmentation and clause information obtained from morphological analysis. Figures 2 (a) and (b) show the morphological analysis and syntactic analysis results for ``Read''.Normally, in a Japanese document processing system.

The processing functions for such morphological analysis and syntactic analysis are provided, and the results of the morphological analysis and syntactic analysis are inputted from the input unit 1.

The characteristic data collection unit 2 collects characteristic data indicating the possibility that inappropriate expressions exist in each Japanese sentence constituting the processing target document.

The inappropriate expression testing section 3 first determines the activation state of the start node of the inappropriate expression detection rule network in the inappropriate expression testing rule dictionary 4 according to the characteristic data collected by the characteristic data collecting section 2. Thereafter, the inappropriate expression testing unit 3 sequentially examines each node in the network based on the connection relationship of each node in the inappropriate expression detection rule network, the activation condition of the node described in the node, and the method of calculating the value of the node. Activate or deactivate each node,
Each end node in the network corresponds to one inadequacy expression, and a terminal node that is finally activated means that the inadequacy expression corresponding to that node has been detected. Furthermore, by holding the position of the inappropriate expression in the sentence as the value of the terminal node, the position of the inappropriate expression in one sentence can also be obtained.

The inappropriate expression information (type of inappropriate expression, position of inappropriate expression) detected as described above is as follows.

The message is sent to the warning message generating section 5. The warning message generation unit 5 applies the corresponding warning message generation rule stored in the warning message generation rule dictionary 6 based on the inappropriate expression information sent from the inappropriate expression verification unit 3.
This generates a warning message to present to the user.

The generated warning message is displayed on a display device such as a display device via the output unit 7.

This allows the user to be notified of a warning message.

A specific example will be explained below. Now, assume that the Japanese document to be processed contains the following two sentences.

■ Easily construct a distributed processing system.

■ Input/output devices can be operated in parallel.

The example sentence of
The noun (in this example, "distributed processing system") and the case particle "
A clause consisting of "no" is a noun (in this example, "construction")
This is an inappropriate expression that modifies a clause consisting of the case particle "ga" and the latter clause modifies the verb "dekiru". Furthermore, the example sentence ■ is an inappropriate expression in which an intransitive verb (in this example, ``act'') is used as if it were a transitive verb. These should be refined and rewritten as first-order.

■′Distributed processing systems can be easily constructed.

■'I/O devices can be operated in parallel.

Fig. 3 shows the morphological analysis results and syntactic analysis results for the example sentences of ■, and Fig. 4 shows the morphological analysis results and syntactic analysis results for the example sentences of ■.

It is assumed that each Japanese sentence constituting the document to be revised has been identified in clause and word units through morphological analysis, and the syntactic structure of the entire sentence has been analyzed into a tree structure through syntactic analysis.

The input unit 1 inputs the example sentences of ■ and ■ in Figures 3 and 4.
Morphological analysis results and syntactic analysis results as shown in the figure are input and transferred to the characteristic data collection section 2.

The characteristic data collection unit 2 analyzes the results of each sentence in the target document based on the morphological analysis and syntactic analysis results for the sentence.
Characteristic data indicating the possibility of the existence of inappropriate expressions in a sentence is extracted. The characteristic data collection section 2 consists of a plurality of independent parts.

Each part corresponds to one or more inappropriate expression items. That is, the characteristic data collected by one part indicates the possible existence of an associated expression of inadequacy.

It is assumed that one first-order characteristic data can be obtained for the example sentence (2).

■-1: Number of clauses containing the verb "dekiru" It is assumed that one characteristic data of the fifth order can be obtained for the box-sentence of ■.

-1 = Number of clauses containing intransitive verbs Each characteristic data collected by the characteristic data collection unit 2 is collected from one or more starting nodes of the inappropriate expression detection rule network stored in the inappropriate expression detection rule dictionary 4. is associated with.

FIG. 5 shows a conceptual diagram of the inappropriate expression detection rule network stored in the inappropriate expression detection rule dictionary 4. As seen in FIG. 5, the inappropriate expression detection rule network consists of nodes and nodes. Consists of links that indicate connections between There are three types of nodes: starting nodes, intermediate nodes, and terminal nodes.

All these nodes have the same structure.

Note that the start node corresponds to the characteristic data collection point, and the end node corresponds to the unsuitability expression item.

FIG. 6 shows a conceptual diagram of the structure of each node in the inappropriate expression detection rule network. As seen in FIG. 6, each node consists of five data storage areas. The node ID is
This is a symbol for uniquely identifying a node. A node value is a value that a node has. The input tool list is a list of links from nodes that have a connection relationship to the node. The output tourist list is a list of links from this node to nodes that have a connection relationship. The previous node activation condition is a precondition for activating the node, which indicates in what state the node having a connection relationship with the node is likely to be activated. The node value calculation description describes a procedure for calculating the node value of the node when the previous node activation condition, which is a prerequisite for node activation, is satisfied. This calculation procedure generally includes condition determination for activation. Therefore, in such a case, just because the previous node activation condition is satisfied does not necessarily mean that the node will be activated.By activating this calculation procedure, the calculated node value of the node will be , is stored in the node value storage area mentioned above.

FIG. 7 shows an inappropriate expression detection rule network corresponding to example sentences (■) and (■).

The inappropriate expression testing section 3 first determines the activation state of the starting node in the inappropriate expression detection rule network based on the characteristic data collected by the characteristic data collecting section 2. Thereafter, the inappropriate expression testing unit 3 calculates the connection relationship of each node in the inappropriate expression detection rule network in the inappropriate expression testing rule dictionary 4, the activation condition of the node described in the node, and the value of the node. Based on the method, each node in the network is sequentially activated or deactivated. Each terminal node in the network corresponds to one unsuitability expression, and finally the activated terminal node means that an inappropriate expression corresponding to that node has been detected. Furthermore, as in the inappropriate expression detection rule network shown in Figure 7, by holding the position of the inappropriate expression in the sentence as the value of the terminal node, the position of the inappropriate expression in the sentence can also be determined. Obtainable.

FIG. 8 shows the process of processing example sentence (■) using the inappropriate expression detection rule network shown in FIG. 7. First, due to the presence of the corresponding characteristic data, node X-0002 becomes inactive. As the node X-0002 becomes inactive, the nodes that can be traced from X-0002 are the unprocessed nodes A-0001 and A- which are the nodes connected to the node X-0001.
0002 satisfy the previous node activation condition, so
Node value calculation is performed according to the node value calculation method described in the node, and each node has a node value as shown in the figure, and is in an activated state. In the next stage, nodes Y-0001 and Y-0002 are connected to node A-0001.
Y-0003 is processed as a node connected to 002. Node Y-0002 is not activated because the previous node activation condition is satisfied, but the node value calculation condition is not satisfied.

Since both the previous node activation condition and the node value calculation condition are satisfied, node Y-0001 is activated and has a node value N as shown in the figure. Node Y-
0003 also satisfies both the previous node activation condition and the node value calculation condition, so it is activated and has the node value as shown in the figure.

As explained above, the following two types of inappropriate expressions and the position of the inappropriate expressions in the sentence are detected for the example sentence (3).

■-A: Type; Noun + clause of "no" modifies the clause of pseudonym + "ga".

and the latter clause is the verb “de Modify "kiru".

Position; (Phrase 12 Clause 22 Clause 4) [Node Y-0
[By activation of node Y-0003] ■-B: Type; position where adverb and verb "dekiru" are consecutive; (clause 31 clause 4) [By activation of node Y-0003] Next, the inappropriateness shown in Figure 7 FIG. 9 shows the process of processing example sentence (2) using the expression detection rule network. First, node X-0002 becomes activated due to the presence of corresponding characteristic data.

Node X-0001 becomes inactive. Node X-
As 0001 becomes inactive, X-0001
Since node A-0003, which is a node connected to node X-0002 and is connected to node X-0002, satisfies the previous node activation condition, the node value calculation method described in this node is Accordingly, the node value is calculated and has the node value as shown in Figure 5, and in the first stage of activation state, the node A-0
Node A-000 as the node connected to 003
4 and A-0005 are processed. Node A-0005
is not activated because its previous node activation condition is not satisfied, and node A-〇〇〇4 is activated because both the previous node activation condition and the condition during node value calculation are satisfied, as shown in the figure. 5. It has a node value such as 5. Node y-oo connected to node A-0004.

4 is activated by satisfying the previous node activation condition, and has the node value shown in the figure.

As explained above, for the example sentence (3), the following type of inappropriate expression and the position of the inappropriate expression in the sentence are detected.

■-A: Type; position where intransitive verbs are used as transitive verbs;
(Phrase 19, Clause 3) The inappropriate expression information (unsuitable expression type, inappropriate expression position) detected as described above [by activation of node Y-0004] is sent to the warning message generation unit 5. . In the warning message generation rule dictionary 6 referred to by the warning message generation unit 5, message generation rules are described in one-to-one correspondence to each inappropriateness expression item. Each message generation rule consists of three parts: a list of inappropriate expressions, a message generation pattern, and a variable conversion rule. The inappropriate expression location list is a list of variables indicating positional information in a sentence related to the inappropriate expression. That is, each element of the list corresponds to each element of the node value at the terminal node in the inappropriate expression detection rule, which has a one-to-one correspondence with the warning message generation rule. The message generation pattern is a template for the message that is finally generated. The message generation pattern has a variable part, and there is a correspondence between the variables described in the list of inappropriate expressions and the variables described in the message generation pattern. That is, the variables in the template are the results of applying the variable conversion rules to the variables in the inappropriate expression location. The warning message generation unit 5 first associates the position information of the detected inappropriate expression with the inappropriate expression information list in the warning message rule, and then performs variable conversion on each variable in the inappropriate expression information list. Apply rules. Finally, a warning message is generated by replacing the variable part in the message generation pattern with the result converted by the conversion rule.

The warning message rules for inappropriate expressions detected by the process shown in FIGS. 8 and 9 and the warning messages generated based on the rules are shown below. Here, Xi
, X2, etc. are the variables in the unsuitable part list,
X2' etc. indicate variables in the message generation pattern.

(a) Message generation rules and generated messages related to Y-0001 $ List of inappropriate expressions for message generation rules: (Xi X2 X3) Message generation pattern: Rewrite to the form 'rX1'X2' J.

” Variable conversion rules: x1→x1': Change the particle from "no" to "wo".

x2→x2′: Change to s-odd noun + [can do].

Skewer generation message (warning message) Please rewrite it as ``[Distributed processing system can be constructed.''

(b) Message generation rules for Y-0003 and list of inappropriate expressions for message generation rules: (Xi Rule: X14X1':X1' =
Message generation rules and message generation rules for Y-0004 List of inappropriate expressions: (Xi please.

Variable conversion rules: X1 → X1': XL' = XI X2→X2′: Add causative expression (“to cause”).

1-Generated message (alarm message) ``Rewrite the message using a causative expression such as ``The input/output device can be operated,'' or reconsider the verb used.

(a) is for an inappropriate expression detected by the process shown in FIG. In this case, variable X1 is clause 1
("j of distributed processing system), variable X2 is clause 2 ("
construction ga''), variable X3 is associated with clause 4 ([can be j). Next, Xl.

Variable conversion rules are applied to the variables of X2.

As a result, X1' has the character string J representing r distributed processing system as its value, and X2' has the character string "can be constructed" as its value.

This result is replaced with each corresponding variable in the message generation pattern, and a warning message "A distributed processing system can be constructed" is generated.

Warning messages are generated for (b) and (c) by a similar process.

The warning message generated as above is:

Via the output unit 7, the information is displayed on a display device such as a display device, and the user is notified. The user can efficiently perform editing work by referring to this warning message.

〔Effect of the invention〕

As explained above, according to the present invention, a warning is generated and presented based on the location of inappropriate expressions detected and the type of inappropriate expressions when revising a Japanese document consisting of multiple sentences. This can be done based on the message.

Since the detection of inappropriate expressions and the presentation of warning messages are carried out comprehensively by the device implementing the present invention, the user performing the editing work can comprehensively revise inappropriate expressions in one document. can maintain uniform quality. In addition, if you are targeting a document in a field that requires specific expressions, you can simply register inappropriate expression detection rules that are appropriate for the document in the inappropriate expression detection rule dictionary, and you will receive revision support that is appropriate for the target field of the document. functions can be realized.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing an embodiment of the Japanese document elaboration support device of the present invention, Fig. 2 is a diagram explaining morphological analysis and syntactic analysis, and Fig. 3 is the morphological analysis result and syntactic analysis result for example sentence ■. Figure 4 is a diagram showing the morphological analysis results and syntactic analysis results for example sentence ■. Figure 5 is a conceptual diagram of the inappropriate expression detection rule network. Figure 6 is a diagram showing the nodes in the inappropriate expression detection rule network. Conceptual diagram of structure, 7th
The figure shows the inappropriate expression detection rule network related to example sentences, Figure 8 is a diagram explaining the process of inappropriate expression testing for example sentence ■, and Figure 9 explains the process of inappropriate expression testing for example sentence ■. It is a diagram. 1... Input section, 2... Characteristic data collection section, 3...
- Inappropriate expression testing unit, 4... Inappropriate expression detection rule dictionary, 5... Warning message generation unit. 6... Warning message generation rule dictionary, 7... Output section. Figure 1: Funeral map: Figure 1;

Claims (1)

[Claims] (1) In a Japanese document processing system that inputs and processes a Japanese document consisting of multiple Japanese sentences, the morphological analysis results and syntactic analysis results for each Japanese sentence that make up the document are input, and the A characteristic data collection unit that collects characteristic data for elaboration support, and a characteristic data collection unit that collects inappropriate expressions that require elaboration work included in each Japanese sentence based on the collected characteristic data and pre-registered inappropriate expression detection rules. an inappropriate expression testing unit that detects the inappropriate expressions including the inappropriate expressions, an inappropriate expression rule dictionary that stores the inappropriate expression detection rules, and an inappropriate expression detection section that detects the inappropriate expressions and inappropriate expression types; A Japanese document elaboration support device comprising: a warning message generation unit that generates a warning message; and a warning message generation rule dictionary that stores the warning message generation rules.