JPH0477857A - Improper expression detecting device - Google Patents

Abstract

PURPOSE:To automatically detect only the improper expressions for each type of written messages by providing a module which applies the warnings for the improper expressions based on the contents of an improper expression dictionary. CONSTITUTION:When a written message and its type are received from a user, the message is analyzed at a morpheme analyzing part 3 and the morpheme information is given to each word. Then an improper expression dictionary 7 is retrieved and an improper expression detecting part 4 detects automatically whether the message includes an improper expression or not in a field to which the message belongs. If so, a warning is produced based on the contents of the dictionary 7 and the message is outputted. Thus only the improper expressions can be surely detected out of the input written messages for each type of these messages.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention provides an inappropriate expression detection method that can automatically determine and detect inappropriate expressions contained in a communication based on the type of the communication. It is related to the device.

[Conventional technology]

The expressions that are inappropriate in a communication differ depending on the field (communication type) to which the communication belongs. For example, the expression ``congratulations'' is inappropriate in the case of a message type of ``condolences,'' but it is not inappropriate in the case of ``congratulations.'' Inappropriate expressions in sentences differ depending on the type of communication, so they cannot be detected automatically and are all detected and corrected manually.

In addition, in a device that detects prohibited words contained in newspapers, prohibited words are registered in advance in the dictionary used when dividing sentences into word units, and words are successfully divided using these prohibited words. At times, the word was considered a forbidden word. In addition,
In the case of this prohibited word, the field (corresponding to the message type)
A fixed number was registered regardless of the condition. Conventional literature of this type includes the 33rd Information Processing Society of Japan National Conference 45-9
It is described in ゜p,, p, 1.719 (1,986).

[Problem to be solved by the invention]

As described above, in the case of (a) prohibition processing using a word processor, conventional devices only automatically correct prohibited characters and symbols at the beginning and end of a line, and cannot correct inappropriate expressions.

Furthermore, (b) when manually detecting inappropriate expressions in correspondence, there is a risk that detections may be missed because the detection is done by humans, and the criteria for determining inappropriate expressions differs from person to person.

In addition, (c) a device that detects prohibited words contained in newspapers detects prohibited words regardless of the type of correspondence, so it may detect expressions that are not inappropriate for the type of correspondence. , Moreover, in order to detect inappropriate expressions as words, expressions that include inappropriate characters (for example, ``celebration'')
All characters ``death'') in j had to be registered.

An object of the present invention is to provide an inappropriate expression detection device that can solve these conventional problems and automatically detect only expressions that are inappropriate in the type of message to which a message belongs.

[Means for Solving the Problems] In order to achieve the above object, the inappropriate expression detection device of the present invention arranges inappropriate expressions for each type of communication, and identifies inappropriate expressions for each inappropriate expression. It is equipped with an inappropriate expression dictionary in which a warning level representing the level of the person's behavior and the contents of the warning are described as a set, and a means for inputting a message type including, for example, congratulations, condolence, congratulations, and regrets; a module that divides the input correspondence into words and adds morphological information to each word; and a module that divides the input correspondence into words and adds morphological information to each word; A module that searches an appropriate expression dictionary and determines whether a word is an inappropriate expression for the input communication type, and a word that is determined to include an inappropriate expression by executing the module. It is equipped with a module that assigns a warning level and warning content for inappropriate expressions according to the contents of an inappropriate expression dictionary, and a means for outputting the input message along with the warning level and warning content assigned by executing the module. There is a characteristic in doing.

[For production]

In the present invention, inappropriate expressions are collected for each type of communication, and an inappropriate expression dictionary is created that describes the warning level indicating the level of inappropriateness of each inappropriate expression and the content of the warning. I'll keep it. When the message and the message type to which the message belongs are received from the user as input, the message is first morphologically analyzed and morpheme information is added to each word.

Next, by searching the inappropriate expression dictionary, it is automatically detected whether or not there is an inappropriate expression in the field (communication type) to which the communication belongs in the communication. If it is determined that the message contains inappropriate expressions, a warning is given according to the contents of the inappropriate expression dictionary, and then
Output the message. This makes it possible to reliably detect only expressions that are inappropriate for the message type of the communication from the input message.

〔Example] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a basic block diagram of an inappropriate expression detection device showing an embodiment of the present invention.

In FIG. 1, reference numeral 9 denotes the main body of the inappropriate expression detection device consisting of a CPU, memory, etc., 1 denotes a message type input unit for inputting the type of message, and 2 denotes a communication from which inappropriate expressions are to be detected. 3 is a morphological analysis unit that performs morphological analysis; 4 is an inappropriate expression detection unit that detects inappropriate expressions; and 5 is a unit for expressions detected by the inappropriate expression detection unit 4. , a warning adding section that gives a warning level and warning content, 6 an output part that outputs a message, a warning level, and a warning content, 7 an inappropriate expression dictionary that registers inappropriate expressions, 8
is an output device such as a display or printer. Here, the inappropriate expression dictionary 7 is a memory device for registering expressions, the output device 8 is a hardware device such as a display, printer, or disk, and the message type input section 1 is an input device such as a keyboard, numeric keypad, etc. The human power section 2 is a character input device such as a word processor, and the output section 6 is a driver, register, etc., both of which are hardware. On the other hand, the morphological analysis unit 3, inappropriate expression detection unit 4
, and the warning adding unit 5 are program modules for executing their respective functions with a processor, and are stored in the memory of the inappropriate expression detection device 9.

The message type input unit 1 receives input from the user about the type of message (congratulations, condolence, congratulations, regrets, etc.), and the communication human resources unit 2 receives input from the user about the message itself. . Thereby, the morphological analysis section 3 divides the correspondence received by the correspondence human resources section 2 into word units, and gives each word morphological information such as prototype information, part of speech, and conjugated form. Next, the inappropriate expression detection unit 4 searches the inappropriate expression dictionary 7, and if the input message contains an expression that is inappropriate in the received message type, this expression Detect.

Next, the warning giving unit 5 determines the level of the warning (fatal, caution, etc.) and the content of the warning (forbidden phrase, hateful word, cautionary word, etc.) for the inappropriate expression detected by the inappropriate expression detecting unit 4. Grant. Next, the output unit 6 outputs the input message to the output device 8 along with the type of the message, the warning level of inappropriate expressions included in the message, and the content of the warning.

Figure 2 shows the records of the inappropriate expression dictionary in Figure 1.
It is a format diagram.

As shown in FIG. 2, the inappropriate expression dictionary 7 has prototype information 10 of words that can be inappropriate expressions or characters that can be inappropriate expressions as keys, and for each prototype information 10, this is a message that is inappropriate. Type 11 and its warning level (fatal level or cautionary level, etc.)1
This dictionary is made up of records 14 that describe warning information 2 and warning content 13. In other words, the inappropriate expression dictionary 7 is 1
A plurality of records 14 consisting of fields 0 to 13 are registered.

The prototype information field 10 is a key part when searching the inappropriate expression dictionary 7, and the message type field 11 is a part that describes the type of communication for which the prototype information of this record is inappropriate. . Warning level field 1
2 is a part that describes the warning level, such as whether the word with the prototype information of this record is fatally inappropriate or should be used with caution. This is a part that describes warning contents such as whether a word with prototype information is a disliked character, a discriminatory word, a cautionary word, etc.

FIG. 3 is a processing flowchart of the inappropriate expression detection device of the present invention.

First, the message type input unit 1 receives an input of the type of message (congratulations, condolence, congratulations, condolences, etc.) from the user (step Sl). The input message type is transferred to the inappropriate expression detection section 4.

Next, the correspondence human resources department 2 receives input of a correspondence from the user (step S2). The input message is transferred to the morphological analysis section 3.

The morphological analysis unit 3 divides the input correspondence into words,
Further, prototype information, parts of speech, etc. are added to each word, and the results are transferred to the inappropriate expression detection section 4. Here, the number of words in the message is n, and in order to process the first word,
Set i=1 (step S3).

The inappropriate expression detection unit 4 searches the inappropriate expression dictionary 7 using the prototype information of the first word as a key (step S4).

If the search is successful (YES in step S5), the information in the inappropriate expression dictionary 7 corresponding to the key prototype information is acquired, and then the process moves to step S6. If the search fails (No in step S5), the process moves to step S8.

If the search is successful, the message type in the record for the prototype information is detected, and it is determined whether the message type matches the message type input by the user (input in step S1) (step S6).

If they match, the process moves to step S7, and if they do not match, the process moves to step S8.

The warning adding unit 6 issues a message to the jth word, that is, a word detected as an inappropriate expression and whose message type matches.
The content of the warning is given to the word along with the warning level (step S7).

Since the processing for that word has been completed, in order to move on to the next word, word number 1 is incremented by 1, so that j=i+1 (step 38).

To determine whether all words have been processed, i>n?
) (step S9), if it has been completed, the process moves to step S10, and if there are still unprocessed words,
Return to step S4.

The output unit 6 outputs the input message to the output device 8 together with the message type, the warning level and warning content of inappropriate expressions included in the record (step 510
).

FIG. 4 is a diagram showing an example of the contents of a dictionary used in the morphological analysis section in FIG. 1.

The morphological analysis unit 3 divides the input communication into words, and assigns prototype information, part of speech, conjugation, etc. to each word. In that case, the above process is performed while referring to a normal word dictionary.

For example, as shown in Figure 4, a normal word dictionary contains ``word notation J'' and its corresponding ``prototype information,'' which are divided sentences, and ``part of speech'' and word notation for the prototype information. The ``conjugated form'' of is described. Note that the contents of the dictionary shown in FIG. 4 will be used in a specific example described later.

FIG. 5 is a diagram showing an example of the contents of the inappropriate expression dictionary in FIG. 1.

As shown in FIG. 5, the inappropriate expression dictionary 7 includes "prototype information J, the message type J belonging to the prototype information, the warning level if it is inappropriate, and the warning content". ” is written as a roux code. In FIG. 5, the prototype information F*death*J indicates that all words with prototype information including the character "death" can be inappropriate expressions. The contents of the inappropriate expression dictionary 7 shown in FIG. 5 will be used in specific examples described later.

FIG. 6 is a diagram illustrating the processing process of specific example 1 of the present invention.

In the message type input section 1 and the message human power section 2, it is assumed that (message type) ``Congratulations on your marriage'' and the above message type have been received (steps St and S2 in Fig. 3). ).

The morphological analysis unit 3 divides the correspondence into word units and assigns morphological information to each word.If the dictionary used by the morphological analysis unit 3 has the contents shown in Figure 4, the results of the morphological analysis are as follows: 11 (1). That is, since there are four words: "marriage", ",", "congratulations", and ".", n4.1 is set to 1 (step S3 in Figure 3).
).

The inappropriate expression detection unit 4 first searches the inappropriate expression dictionary 7 using the prototype information of the first word "marriage J" as a key. Assuming that the inappropriate expression dictionary 7 has the contents shown in FIG. Search fails. Second word prototype information In,
The same failure occurs for the prototype information "." of JJ and the fourth word (steps 4 to 8 in FIG. 3). For the third word "Congratulations", when the inappropriate expression dictionary 7 is searched using the prototype information of the third word "Congratulations" as a key, the inappropriate expression dictionary 7 has the content shown in FIG. (Steps S4 and S5 in FIG. 3).

However, while the message type input in the message type human horn section 1 is "Kei", in the record of "Congratulations" written in the inappropriate expression dictionary 7, the message type is "Kei". Condolences J. Therefore, the message type of the input message and the message type in the inappropriate expression dictionary 7 do not match, and r Congratulations J is not inappropriate in the message type r Kei J. (Step S6 in FIG. 3).

This completes the process of the inappropriate expression detection unit 4 (step S9 in FIG. 3).

As shown in FIG. 6(2), the warning adding section 5 outputs a warning that there is no inappropriate expression. In the output unit 6, since the input message does not contain inappropriate expressions,
As shown in FIG. 6(3), only the message and the type of message are output to the output device 8.

In other words, even if the same message is included in the inappropriate expression dictionary 7, if the types of the messages are different, the messages will not be determined to be inappropriate.

FIG. 7 is an explanatory diagram showing the processing process of the second specific example of the present invention.

The following messages and types of messages are input to input units 1 and 2.

Message type: Condolences <Message> Congratulations on your marriage.

That is, in specific example 2, the message is the same as in specific example 1, but the message type is different. In other words, specific example 2
This corresponds to a case where the message type is input incorrectly.

In this case as well, processing is performed in the same manner as in Example 1, but unlike Example 1, ``Congratulations!'' is detected as an inappropriate expression (see FIG. 7 (2)). This is because the content of the inappropriate expression dictionary 7 indicates that "congratulations" is inappropriate in the message type T condolence. In this case, the warning adding section 5 adds a message indicating that it is a fatal or taboo phrase, and transfers it to the output section 6. The output result is shown in Figure 7 (3)
It becomes as shown in . The user who obtains this result knows that the message and the message type do not match. In Example 2, the user notices an error in inputting the message type.

In this way, when the message type does not match the content of the message, it is possible to detect an error in the message type by detecting inappropriate expressions included in the message.

Figure 8 shows 1. FIG. 7 is an explanatory diagram showing a processing process in Example 3 of the present invention.

The message type input section 1 and the message human resource section 2 receive the following message types and messages (step SL, 2 in FIG. 3).

(Message type) Kei (Correspondence) When the wind blows, we will block it together, and even when we die, we will be together, so please be the two of us.

In the third specific example, a message different from those in the first and second specific examples is input. As in Example 1, as a result of analysis by the morphological analysis unit 3, prototype information, parts of speech, and conjugations as shown in FIG. 8(1) are obtained. In this case, since there are 19 words, the number is n-19 (step 33 in FIG. 3). Since the contents of the inappropriate expression dictionary 7 are as shown in Fig. 5,
The search in the inappropriate expression dictionary 7 fails for 17 words other than the two words ``obstruct J'' and ``die''. The processing for these failures is the same as for "marriage" in the first specific example. Here, the processing of ``object j'' and ``death'' described in the inappropriate expression dictionary 7 will be explained.

“Saeguru” is the sixth word and is written in the inappropriate expression dictionary 7. The message type input in the message type input section 1 is ``Keio J,'' and the message type in which ``object j'' is inappropriate in the inappropriate expression dictionary 7 is also ``Keio J.'' Therefore, the two match. Therefore, "Interrupting J" is detected as an inappropriate expression, and the processing moves to the processing of the warning adding unit 5.

The warning assigning unit 5 assigns the contents of the inappropriate expression dictionary 7 (warning level - caution, warning content - caution word) for "object" to "object" (step 87 in FIG. 3).

Next, regarding the eighth word "death", it does not exist as prototype information in the inappropriate expression dictionary 7, but "*death*
J, that is, all sentences containing the character "death J" indicate that they can be inappropriate expressions, so they are detected as inappropriate expressions in this case as well. Add the content (warning level matching fatal, warning content = abominable characters) to the town that will die (step S7 in Figure 3).

This state is shown in FIG. 8(2).

The last output unit 6 outputs information regarding the two words "Saeguru" and "Death Town" that were detected as inappropriate expressions to the output device 8 along with the message and the type of message (Step 510 in FIG. 3). The state of the output is as shown in FIG. 8(3).

In this way, if the inappropriate expression is included in the inappropriate expression dictionary 7 and the message type matches the input one, the inappropriate expression is detected and a corresponding warning is output. Furthermore, depending on the degree of inappropriateness, it is also output whether the level is fatal or cautionary. Further, if the content of the inappropriate expression dictionary 7 is a character such as ``Water death water J'', all expressions containing this character are detected as inappropriate.

〔Effect of the invention〕

As explained above, according to the present invention, it is determined whether or not the inputted message is inappropriate depending on the type of message.
Only expressions that are inappropriate in the corresponding message type can be detected. In this case, since an inappropriate expression dictionary is used, no inappropriate expressions will be missed as long as they are written there. Additionally, since inappropriate expressions are detected mechanically, there is no variation among users. Furthermore, when detecting inappropriate expressions, character-by-character detection is possible, so there is no need to write all words containing that character in the inappropriate expression dictionary.

[Brief explanation of drawings]

Fig. 1 is an overall configuration diagram of an inappropriate expression detection device showing an embodiment of the present invention, Fig. 2 is an explanatory diagram of the contents of the inappropriate expression dictionary in Fig. 1, and Fig. 3 is an inappropriate expression detection according to the present invention. 4 is a diagram showing the contents of the dictionary (word dictionary) used in the morphological analysis section of FIG. 1; FIG. 5 is a diagram showing an example of the contents of the inappropriate expression dictionary of the present invention; FIG. 6 is an explanatory diagram showing the processing process of specific example 1 of the present invention, FIG. 7 is an explanatory diagram showing the processing process of specific example 2 of the present invention, and FIG. 8 is an explanatory diagram showing the processing process of specific example 3 of the present invention. It is an explanatory diagram. 1: Correspondence type input section, 2: Correspondence writer section, 3 Morphological analysis section, 4: Inappropriate expression detection section, 5: Warning section, 6:
Output unit, 7: Inappropriate expression dictionary, 8: Output device, 14: Code of inappropriate expression dictionary.

Claims (1)

[Claims] (1) An inappropriate expression dictionary in which inappropriate expressions are arranged for each type of message, and for each inappropriate expression, a warning level indicating the level of inappropriateness of the expression and the content of the warning are described as a set. For example, a means for inputting the type of message including congratulations, condolence, congratulations, and regrets, a means for inputting the message, and a means for dividing the inputted message into word units and dividing the input message into words. A module that adds morphological information, and for each word divided by the execution of this module, searches the above-mentioned inappropriate expression dictionary using the prototype information of the word as a key, and searches the above-mentioned inappropriate expression dictionary to find out whether the word is inappropriate in the above-mentioned communication type into which the word is input. A module that determines whether an expression is inappropriate or not, and a warning level and warning content for inappropriate expressions are assigned to words that are determined to include inappropriate expressions by execution of the module, according to the contents of the inappropriate expression dictionary. What is claimed is: 1. An inappropriate expression detection device comprising: a module for assigning warnings; and means for outputting an input message along with a warning level and content of the warning assigned by executing the module.