JP2685808B2 - User-supported input sentence response processing device - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a response processing device in a computer system in which inquiry conditions for internal knowledge and a database prepared in advance are designated by a natural language.

[Conventional technology]

In a computer system in which query conditions are specified by natural language, morphological analysis, syntactic analysis, semantic analysis, intention analysis, etc. of input query sentences are specified in advance using dictionaries, grammatical rules, and field-dependent knowledge. According to the rules described above, the information is converted into an internal representation that can be understood by the computer system, and information that should respond to the user is generated from the internal knowledge and the database.

[Problems to be solved by the invention]

In a computer system designed to specify query conditions in natural language, no matter how much knowledge is stored, there are always sentences that cannot be parsed or cannot respond. However, since the user cannot know what sentence the computer system can understand and what sentence cannot be understood at the time of input, it was necessary to change the sentence again and again.

In order to solve such a problem, there is a device in which a computer system displays an example sentence that the computer system can respond to, and the user inputs the sentence while referring to it. Since the input sentence must be generated while considering the responsive sentence, the user's input burden was heavy. Also, if there are a large number of assumed example sentences, it is impossible to display these example sentences at once. Therefore, specify the field that is the target of the input sentence and select only the example sentences that belong to that field. A way to display is needed. However, in this case, there are problems that the user has to input the field to specify the field, and that the field that the user does not enter must be considered.

Furthermore, in a conventional computer system for inputting natural language, in order to understand various expressions, various grammatical knowledge, field-dependent knowledge, morphological analysis rules, syntactic analysis rules, semantic analysis rules, intention analysis rules, etc. Since a lot of knowledge and rules were needed, many changes such as rule changes and knowledge changes were required to increase the number of sentences that can be processed and the applicable fields, and a great deal of labor was required. Met.

[Object of the invention]

An object of the present invention is to exemplify a sentence to which a computer system can respond based on a sentence input by a user, thereby changing the information that the user wants to convey to the computer system to a sentence that can be responded to by the computer system and input it. It is possible to provide a user-supported input sentence response processing device that can easily change the system even when the application field is changed.

[Means for solving the problem]

The input-sentence response processing device according to the present invention is provided with an example-sentence storage unit that stores and retains an example sentence of an input sentence to which the response processing device can respond by pre-expressing it with a sequence of words that compose the example sentence, and extracting word meaning attributes. The means for extracting the words constituting the input sentence and the semantic attribute of the word from the input sentence input by the user, and the example sentence evaluation value calculation means for storing the input sentence and each example sentence stored in the example sentence storage unit. An evaluation value indicating the similarity with is calculated based on the semantic attributes of the words that form each sentence, and the display example sentence determination unit determines the evaluation value based on the size of the evaluation value.
The example sentence to be displayed is determined, the word of the example sentence to be displayed is replaced with the word of the input sentence by the user intention selecting means, the display example sentence is displayed, and the sentence intended by the user is selected from the display example sentence. By doing so, one of the example sentences held in the example sentence accumulating unit is determined as a processing target example sentence, and a response processing means performs a response process to the processing target example sentence. Is.

(Operation)

In the present invention, the computer system exemplifies the sentences that can be responded to by the computer system based on the example sentences held in the example sentence storage unit and the semantic attributes of the words of the sentences input by the user, and selects the sentence that matches the user's intention. Since the response processing is performed with the selected display example sentence as the processing target, the user can easily and correctly input the intention that the user wants to convey to the computer system without imposing a heavy burden on the user, and the user can accurately You can get a quick response.

In addition, it is possible to increase the number of sentences that can be responded only by adding the example sentence and the processing procedure information to be held in the example sentence storage unit. Furthermore, even when the application field is changed, it can be easily dealt with by simply changing the example sentence and the processing procedure information to be held in the example sentence storage unit and the semantic information to be given to the word.

〔Example〕

FIG. 1 is a block diagram showing a configuration of an embodiment of the present invention.

The sentence input by the user is input to the word meaning attribute extraction means 1 and is executed by a generally known morphological analysis method or the like.
Word division is performed, and by referring to the word dictionary,
The semantic attribute of each word is obtained and sent to the example sentence evaluation value calculation means 2. In the example sentence evaluation value calculation means 2, the example sentence accumulating unit 6 holds each of the example sentence evaluation units as the evaluation value calculation criterion that gives a high evaluation value to a sentence including many words having the same semantic attribute as the word of the input sentence. An evaluation value for each example sentence is calculated, and the example sentence number and its evaluation value are sent to the display example sentence determining means 3. In addition,
In calculating the evaluation value, the weight according to the semantic attribute is taken into consideration. The display example sentence determination means 3 determines the output rank of the example sentence to be displayed according to the size of the evaluation value and sends it to the user intention selection means 4. The user intention selecting means 4 displays the display example sentence generated by replacing the word in the example sentence with the word of the input sentence having the same meaning attribute to the user. Next, the user waits for a character string to be input, and an example sentence to be sent to the response processing means 5 is determined by the user's instruction. When the character string input by the user is not an instruction to determine the example sentence, the input character string is sent to the word meaning attribute extraction means 1.

When the character string input by the user is an instruction to determine the example sentence, the determined example sentence is sent to the response processing means 5. The response processing means 5 searches the database 7 for the determined example sentence, determines the information to be responded to, generates the response sentence, and displays the response sentence to the user.

 Next, a detailed description will be given using a specific example.

It is assumed that the example sentence storage unit 6 holds the example sentences shown in FIG. At this time, as shown in FIG. 3, the operation of each processing means will be described when the user inputs "Which inn can golf?". FIG. 4 shows an example of the result obtained by the word meaning attribute extraction means 1. In the word meaning attribute extraction means 1, a word segmentation process such as "golf / wa / able / inn / ha /?" Is performed by a morphological analysis method or the like, and the semantic attribute of each word is extracted. Here, description will be made assuming that weights other than nouns are 0 when the evaluation value is calculated. At this time, the word having a weight of 0 does not participate in the evaluation value calculation and can be ignored. Therefore, the semantic attributes of “golf” and “inn” are extracted, and the semantic attributes of “entertainment” and “accommodation facility” are obtained.

Next, the example sentence evaluation value calculation means 2 calculates the evaluation value of the example sentence held in the example sentence accumulating unit 6. FIG. 5 shows example sentence numbers of example sentences having the words of the semantic attributes “entertainment” and “accommodation facility”. If it is specified that 10 words will be added as an evaluation value if one word with the same semantic attribute is included, an evaluation value of 20 points will be obtained for example sentence numbers 2, 8 and 10 as shown in FIG. For example sentence numbers 1,3,4,5,6, you get 10 points. The display example sentence determination means 3 determines the example sentences to be output in descending order of the evaluation value. When the evaluation values are the same, the order is determined according to another evaluation criterion. For example, a word in which a semantic attribute is defined is called a keyword word, and a sentence having the smallest difference in the number of keyword words between the input sentence and the example sentence is set as a high-rank candidate. In this case, since the number of keyword words is as shown in FIG. 7, the order of the displayed example sentences is determined in the order of example sentence 10, example sentence 2, and example sentence 8. Here, it is assumed that only the example sentence having the highest evaluation value is specified to be sent to the user intention selecting means 4. Example sentence numbers 10, 2, 8 are sent to the user intention selecting means 4.

The user intention selection means 4 replaces the word in the example sentence with the same semantic attribute as the semantic attribute of the word of the input sentence with the word of the input sentence and displays it. For example, “tennis” in the example sentence 10 is replaced with “golf” and displayed. FIG. 8 shows an example of a display example sentence. Here, in [], the keyword words that cannot be determined from the sentence input by the user, and the words held in the example sentence storage unit 6 are displayed. The user specifies a display example sentence that matches the content to be transmitted to the computer system, or inputs a word corresponding to an undetermined keyword word in the display example sentence.

For example, when the user specifies the first display example sentence, the sentence “I want you to tell me the inn where golf is possible” is sent to the response processing means 5. The response processing means 5 performs a response process on this sentence by performing a semantic analysis process generally performed in natural language processing. Further, if the processing procedure corresponding to the example sentence is defined in the example sentence storage unit 6, the response process can be performed by referring to the processing procedure defined in the example sentence storage unit 6. For example, for example sentence number 10, "substitute the word corresponding to'amusement 'in variable X, search for'accommodation' containing X in'amusement ', and display the retrieved'hotel name'. If the processing procedure is defined as ".
A search for "accommodation facilities" including "golf" in "entertainment" is displayed, for example, "There are A inn and B hotel." When many example sentences are stored in the example sentence storage unit 6, the above processing procedure is held corresponding to the example sentence only for frequently used example sentences, and for other input sentences. After performing the semantic analysis process that is generally performed, the response process may be performed according to the analysis result.

In the user intention selecting means 4, when the user inputs a word designating a new condition instead of designating the display example sentence, the word meaning attribute extracting means 1, the example sentence evaluation value extracting means 2, and the display are made again. The same processing is performed in the displayed example sentences through the example sentence determination means 3, and the example sentences are displayed again. For example, when the user inputs "Hakone", an example sentence as shown in FIG. 9 is displayed. In this case, since the meaning attribute of "Gora" and "Hakone" is "hot spring place name", a display example sentence in which "Gora" is replaced with "Hakone" is displayed. In this way, the user can refer to the display example sentence and input a sentence to which a new condition is added. This corresponds to the realization of a kind of context processing function.

〔The invention's effect〕

As described above, according to the present invention, by exemplifying the sentences that the computer system can respond to based on the sentences input by the user, the user only has to select the sentence that suits his or her intention. The user's intention to be transmitted to the computer system can be input to the computer system without imposing a heavy burden on the user, and the user can quickly obtain an appropriate response. In addition, the user can know what sentence the computer system can respond to by referring to the displayed example sentence. Therefore, for example, by applying it as a user interface of a database search system, it is possible to realize a database search system that can be easily used even by an unfamiliar user.

Also, if the sentence is similar to the example sentence stored in the example sentence storage unit, the user can input the sentence in various expressions. For example,
If you keep "I want you to teach me an inn where you can play tennis" as an example sentence, "I want to stay at an inn where you can play golf.",
You can enter various expressions such as "I want to know a hotel that can play billiards." Therefore, it is possible to easily increase the number of responsive sentences only by adding the example sentences to be held in the example sentence storage unit and the processing procedure information. Furthermore, even when the application field is changed, it can be easily dealt with by simply changing the example sentence and the processing procedure information to be held in the example sentence storage unit and the semantic information to be given to the word.

[Brief description of the drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention.
The figure shows an example of an example sentence stored in the example sentence storage unit. Fig. 3 shows an example of an input sentence. Fig. 4 shows an example of word meaning attribute extraction. Fig. 5 shows an input sentence. Showing a word having the same semantic attribute as the word in FIG. 6, FIG. 6 showing an example of the evaluation result of example sentence evaluation values, FIG. 7 showing a keyword word, and FIG.
FIG. 9 and FIG. 9 are diagrams showing examples of display example sentences. In the figure, 1 is a word meaning attribute extracting means, 2 is an example sentence evaluation value calculating means, 3 is a display example sentence determining means, 4 is a user intention selecting means, 5 is a response processing means, 6 is an example sentence accumulating section, and 7 is a database. Show.

Continuation of the front page (56) References JP 61-118834 (JP, A) Yu Nakagawa, Tsuneaki Kato, "Methods of Understanding Semantics in Japanese Database Retrieval Systems" IPSJ Transactions, Vol. 27, No. 11 P. 1069-1076 (Sho 61-11-15)

Claims (2)

(57) [Claims] 1. A response processing device which understands an input sentence input by a user and responds based on internal knowledge and a database prepared in advance. From the example sentence storage unit that holds the word string of the example sentence that is expressed in advance by the string of constituent words, and the words that make up the input sentence and the semantic attributes of the words, from the input sentence that the user inputs Word meaning attribute extraction means, and example sentence evaluation value calculation means for calculating an evaluation value indicating the similarity between the input sentence and each example sentence held in the example sentence accumulating unit based on the meaning attributes of the words forming each sentence A display example sentence determining means for determining an example sentence to be displayed based on the size of the evaluation value; and a display example sentence generated by replacing the word of the example sentence to be displayed with the word of the input sentence, Select the sentence intended by the user from the displayed example sentence The user support includes a user intention selection unit that uses one of the example sentences stored in the example sentence storage unit as a processing target example sentence, and a response processing unit that performs a response process to the processing target example sentence. Type input sentence response processing device. 2. The example sentence accumulating section holds processing sequence information that defines a processing procedure to respond to each of the example sentences together with the word string of the example sentence, and the response processing means stores the processing target example sentence. 6. The user-supported input sentence response processing device according to claim 1, wherein the response processing is performed in accordance with the processing procedure information defined by the example sentence storage unit.