JPH0793364A - Sentence form converting device - Google Patents

Abstract

PURPOSE:To provide the sentence form converting device which converts an input sentence in a natural language accompanied with omissions of words into forms that various information processing systems can handle. CONSTITUTION:This device is provided with a word dividing process part 1 which divides the input sentence into words, a request kind decision part 2 which decides the types of the indication and question that the input sentence shows, a concept dictionary 4 wherein information showing what system's concept the words belong to is described, an unnecessary word dictionary 5 wherein unnecessary words such as particles and auxiliary verbs which are required to constitute a sentence, but not required by a DBMS 12 are described, a concept decision part 3 which judges what information each word in the input sentence means on the basis of the concept dictionary 4 and removes unnecessary words in the input sentences on the basis of the unnecessary word dictionary 5, a complementing process part 7 which complements omitted words in the input sentence according to a rule corresponding to the type of the input sentence, and a format conversion part and a main control part 11 which convert the input sentence into the representation form that the DBMS 12 can interprets on the basis of the process results of the concept decision part 3 and complementing process part 7.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an intelligent man-machine interface, and relates to a sentence format conversion device capable of issuing instructions and questions to various information processing systems in natural language.

2. Description of the Related Art In a conventional information processing system, instructions and inquiries to a computer are made using a language defined for each system. For example, in a database system which is one of information processing systems, a command such as select name from automobile where body weight> 2000 is given to the system by using SQL (structured search language) which is a relational database operation language. Then, the "automobile" table in the database is searched for an item having a value of the item "body weight" larger than 2000.

In the case of a man-machine interface, necessary items can be selected from a menu,
By inputting a sentence in accordance with a template given in advance, a computer was instructed or asked a question.

[0003]

However, in the conventional information processing system, since the input method is different for each system, the user needs to be familiar with the input format for each system. Therefore, when the number of types of systems used by users in the future information society increases, the burden on the users will increase. In addition, the man-machine interface designer must design the man-machine interface by assuming the knowledge that the user has for each system, and it is very difficult to decide the model of the user. Labor is required. These problems can be solved if a natural language sentence that can be used by any ordinary person can be used as an input language for any information processing system.

However, since the natural language sentence that a person uses everyday is premised on the knowledge of the grammar etc. that a person has,
Words are often omitted. Therefore, when the natural language sentence is used as the input language of the information processing system, the processing may be impossible due to lack of information due to omission of the word.

The present invention has been made under such a background, and a sentence format conversion device capable of converting an input sentence in natural language accompanied by omission of words into a format which can be processed by various information processing systems. It is intended to be provided.

[0006]

In order to solve the above-mentioned problems, the present invention is a sentence format conversion device for converting a directive sentence and a question sentence expressed in natural language into a format interpretable by an information processing system. There is a word division processing unit that divides the input sentence into words, a request type determination unit that determines the type of each word in the input sentence based on a keyword that indicates the type of the instruction or question, and A concept dictionary that describes information indicating whether it belongs to the concept of system, and unnecessary words that describe unnecessary words such as particles and auxiliary verbs that are necessary for constructing a sentence but do not include the information required by the information processing system. A dictionary and a judgment based on the concept dictionary as to what information each word in the input sentence means,
A concept determining unit that removes unnecessary words in the input sentence based on the unnecessary word dictionary; a complementing processing unit that complements words omitted in the input sentence according to a predetermined complementing rule according to the type of the input sentence; It is characterized by comprising a conversion means for converting the input sentence into an expression format interpretable by the information processing system on the basis of the information obtained by the concept determination section and the information obtained by the complementation processing section.

[0007]

According to the present invention, the word division processing unit divides the input sentence in natural language into words by the morphological analysis method and the like, and the request type determination unit determines the type of instruction or question in each word in the input sentence. The type is judged based on the keyword to be expressed, and the concept judgment unit determines what information each word in the input sentence means (for example, “car” is a target concept word representing a target, “heavy” is a quantity). Information that is a comparative conceptual word indicating the degree of) is determined based on the conceptual dictionary, and unnecessary words in the input sentence are removed based on the unnecessary word dictionary. Then, the complement processing unit complements the omitted word in the input sentence according to a predetermined complement rule according to the type of the input sentence, and the conversion unit combines the information obtained by the concept determination unit and the information obtained by the complement processing unit. Based on, the input sentence is converted into an expression format that can be interpreted by the information processing system.
This allows the user of the information processing system to give instructions and questions to the computer in natural language.

[0008]

Embodiments of the present invention will be described below with reference to the drawings. A: Overall Configuration FIG. 1 shows a sentence format conversion device 10 according to an embodiment of the present invention.
It is a block diagram which shows the structure of the database system which has 0 as an input interface.

In FIG. 1, reference numeral 1 is a word division processing section for dividing a natural language sentence input from the input device 9 into words by a morphological analysis method. Reference numeral 2 is a request type determination unit that determines the type of the instruction or question represented by the input sentence by matching a word in the input sentence with a keyword registered in advance.

Next, 3 is a concept determination unit. The concept determination unit 3 refers to the concept dictionary 4 to determine which concept each word word-divided by the word-division processing unit 1 belongs to, and also refers to the unnecessary word dictionary 5, particularly to a system (described later). A word that does not contain information necessary for the information processing system to perform processing is removed from the input sentence (that is,
The words such as particles and auxiliary verbs that are unnecessary for performing processing such as interpretation of the input sentence in the information processing system are removed from the input sentence).

Here, the concept dictionary 4 is a table in which information indicating which concept a word belongs to is described, and is stored in a memory or the like (not shown). For example, as shown in FIG. 2, for each concept type (in the figure, “target concept”, “quantity attribute”, ...) Representing the attribute of the concept of a word, the concept name (“car”, “mountain” in the figure) ”,“ Weight ”, ……)
Corresponding to some conceptual words ("automobile", "car", "car", ..., "mountain", "mountain", ..., "weight", "weight", ...) in the figure Is described. This concept dictionary 4 is within the range when the information processing system (database system in this embodiment) to be connected to the sentence format conversion device 100 is defined, because the range of the input sentence for which the format conversion is performed is limited. It is created by hand in advance.

The unnecessary word dictionary 5 is a table in which a list of words such as particles and auxiliary verbs that are included in the input sentence but do not include information necessary for the information processing system to perform processing is described. Like the concept dictionary 4, it is stored in a memory or the like.

Next, 6 is a format converter. The format conversion unit 6 sets each word or the concept to which each word belongs in a table of a predetermined format for each concept type based on the determination result of the concept determination unit 3. For example, as shown in FIG. 3, a CS having a storage area (hereinafter, referred to as a slot) in which each word or a concept to which each word belongs is stored for each concept type.
Set in the T (concept set) table. Also, when the input sentence is composed of a plurality of word string units by connecting words that mean “or”, “and”, etc., a plurality of cst tables are created for each word string unit, and the connection relation of each cst table is created. Is the above connection word and the identifier ID of each CST table
Create a rel (relation) table represented by.
Here, each item of the cst table is the following (1) to (10).
Represents the content shown in. (1) ID: identifier of cst table (2) Target concept: concept representing a judgment target (3) Quantity attribute: attribute related to quantity scale (4) Comparison concept: expressing degree of quantity value, tendency of value change (5) Modification concept: Concept that modifies the comparison concept (6) Time concept: Concept that represents time (7) Time modification concept: Concept that modifies the tense concept (8) Comparison target: What is the target of comparison ( 9) Comparison value: Value that serves as a reference for comparison (10) Kind of comparison: Kind of comparison

Next, 7 is a complement processing section. The complementation processing unit 7 complements the data created by the format conversion unit 6 with information omitted in the input sentence as necessary. This complementing process is performed based on the rule stored and held in the complementing processing unit 7 in advance.

FIG. 4 is a diagram showing an example of rules applied in the complementary processing. In the figure, first, rule 1 defines the termination condition of the complementary processing. In this example, when the request type is what type, it is stipulated that the complementing process is ended when data is stored in the slots of the target concept, the quantity attribute, and the comparison concept in each cst table.

Rule 2 defines the processing conditions for the complementary processing. In this example, data is not stored in the target concept slot of the cst table whose identifier ID is "X", and c whose identifier ID is "X + 1"
When data is stored in the target concept slot of the st table, the identifier ID of the data of the target concept slot of the cst table whose identifier ID is “X + 1” is “X”.
Is set in the target concept slot of the cst table.

That is, the rule 1 is what type, whic
Since the information required for the system to perform processing differs depending on the question sentence type such as h type or yes / no type, it is manually created in advance for each question sentence type. Rule 2 is applied according to the omitted concept type, and this is also created manually in advance. Since this rule 2 does not depend on the above-mentioned question sentence type, it is commonly applied to each question sentence type.

FIG. 5 shows information required for each question sentence type and example sentences of each question sentence type. That is, wha, such as which type or yes / no type
In addition to the information required for the t type (required slot in the figure), there is also a question sentence type for which information on the comparison target or comparison value is required.

Now, returning to FIG. 1, the structure of the embodiment will be described. In the figure, reference numeral 8 denotes a system interface unit, which transmits the request type given by the request type determination unit 2 and the data in the format created by the format conversion unit 6 and the complement processing unit 7 to the information processing system. The sentence format conversion device 100 is configured by the above units 1 to 8.

The input device 9 also takes in a natural language sentence input by the user to operate the information processing system and supplies it to the word division processing section 1. Output device 10
Outputs the information required by the user, the processing result, and the like.

Further, 11 is a main control unit composed of a CPU, a memory, etc., and controls the entire system including the sentence format conversion apparatus 100 described above. 12 is a DBMS
It is a (database management system), and operates a DB (database) 13 in which various data are stored according to an instruction from the main control unit 11. The above is the overall configuration of the present embodiment.

B: Operation of Sentence Format Converter 100 Next, the operation of the sentence format converter 100 will be described with reference to the flowchart shown in FIG. First, the word division processing unit 1
Divides the input sentence into words by the morphological analysis method (step S101). Next, the request type determination unit 2 matches the word string received from the word division processing unit 1 with the keyword dictionary held therein to determine the request type of the input sentence, and complements the determination result. It is supplied to the unit 7 and the system interface unit 8 (step S102).

Then, the concept determination unit 3 determines the concept dictionary 4 for each word of the word string received from the word division processing unit 1.
And associate the concept type and concept name using
By referring to the unnecessary word dictionary 5, unnecessary words such as particles and auxiliary verbs are removed from the input sentence (step S103).

Next, the format conversion section 6 includes the concept determination section 3 described above.
Based on the processing result of
l table and cst table are created (step S
104). Then, the complementary processing unit 7 determines whether the complementary processing is necessary for the cst table created by the format conversion unit 6 (step S105).

Here, when the complementary processing is not necessary, the complementary processing unit 7 regards the cst table as completed and supplies the cst table and the rel table to the format conversion unit 6. Then, the format conversion unit 6 supplies the completed cst table and rel table to the system interface unit 8, and the interface unit 8 combines these tables and the request type received from the request type determination unit 2 together. , To the main controller 11 (step S1
09).

On the other hand, in step S105, when the complementing process is necessary, the complementing processing unit 7 internally determines based on the request type received from the request type determining unit 2 and the state of the cst table received from the format converting unit 6. It is determined whether or not there is an applicable rule 2 among the held rules 2 (see FIG. 4) (step S106).

Here, if there is an applicable rule 2, the complementary processing section 7 applies the rule 2 to execute the complementary processing (step S107). Then, it is again determined whether or not the complementary processing is necessary (step S105). That is, if the input sentences are connected by a plurality of connecting words and there are three or more cst tables, the rule 1
Rule 2 is repeatedly applied until the condition of is satisfied, and the complementing process is repeated a plurality of times.

On the other hand, in step S106, if there is no applicable rule 2 such as the input sentence is grammatically inconsistent and the complementary processing cannot be performed, c
Since the st table cannot be completed, the system interface unit 8 is notified that the input sentence is incomprehensible. As a result, the interface unit 8 becomes the main control unit 1
The fact that the input sentence is incomprehensible is transmitted to 1 (step S108).

C: Overall Operation by Specific Example of Input Sentence Next, the operation of the database system (see FIG. 1) having the sentence format conversion device 100 as an input interface will be described.
The case where the input sentence "What is a car that is heavy and is very cheap" is given as an example.

First, when an input sentence is given from the input device 9, the word division processing unit 1 divides words by a morphological analysis method. In this example, "weight""ga""heavy""te"
"Price""ga""very""cheap""car""ha""what"
It becomes "is""ka". In this case, a word to be utilized like "heavy" is transformed into an end form like "heavy".

Next, the request type judgment unit 2 matches the keyword dictionary held internally with the words in this word string, and searches for "What" and "Ka" that satisfy this condition. Determine what type of statement you are requesting. Then, the complement processing unit 7 sets the request type to what type.
To the system interface unit 8.

Next, the concept determination unit 3 refers to the concept dictionary 4 shown in FIG. 2 and associates each word with a concept type and a concept name. In this example, "weight (quantity attribute, weight)"
"Heavy (comparative concept, heavy)""Te (connection word, and)"
"Price (quantity attribute, price)""Very (modification concept, ver
y) ”“ car (target concept, automobile) ”, and other words such as particles and auxiliary verbs are removed as unnecessary words according to the unnecessary word dictionary 5.

Then, the format conversion unit 6 fills in the concept name as the value of each slot in the internal table shown in FIG. 3 based on the processing result by the concept determination unit 3. At this time,
Separate the word string at the connecting word part, and set each word string unit to 1
One cst table is used, and the connection relation of these cst tables is represented by the rel table.

In this example, the cst table with the identifier ID = 1 is created by "weight (quantity attribute, weight)" and "heavy (comparative concept, heavy)" before "te (connection word, and)". , "Te (connecting word, and)" followed by "price (quantity attribute, price)", "very (modification concept, very)" and "car (target concept, automobile)", the cst of identifier ID = 2 The table is created.

Next, the complementary processing section 7 first determines whether or not the complementary processing is necessary according to the rule 1 shown in FIG. In this case, the target concept slot of the cst table with the identifier ID = 1 is blank and does not satisfy the completion condition of the complementing process, so that the rule 2 (FIG. Search). And
According to this rule 2, the complementary processing as shown in FIG. 7 is performed. In this example, the target concept slot of the cst table with the identifier ID = 1 is supplemented with the value “automobile” of the target concept slot of the cst table with the identifier ID = 2.
As a result, the termination condition of rule 1 is satisfied, and the complementing process is completed.

When the complementary processing is completed, the system interface section 8 converts the information of each completed table into the format conversion section 6
From the request type determining unit 2 and sends the data to the main control unit 11 as data in the format shown in FIG. Then, the main control unit 11 quantifies necessary search conditions for each cst table, selects a database table and items, creates a database control statement based on the request type, and issues a search command to the DBMS 12.

That is, in this example, the main control unit 11 quantifies the comparative concept "heavy" in the cst table with the identifier ID = 1, and obtains weight> 2000. And the request type is wh
Since it is an at type, create the following database control statement to search the database for cars that meet the conditions. select name from car where weight> 2000

Similarly, the following database control statement is created based on the information in the cst table with the identifier ID = 2. select name from automobile where price <140

Further, the following database control statement is created from the above two database control statements according to the information in the rel table. select name from automobile where weight> 2000 and price <1
40

Next, the DBMS 12 searches the DB 13 according to the database control statement given from the main control unit 11 and returns the search result to the main control unit 11. Main controller 11
Passes the obtained search result to the output device 10. As a result, the output device 10 outputs the search result and informs the user.

As described above, according to the present embodiment, the request type of the directive sentence or question sentence input to the information processing system in natural language is first determined based on the keyword included in the sentence. , Cst table and rel table (in this case, unnecessary words such as particles and auxiliary verbs are removed). Then, it is determined according to the rule 1 whether necessary information is omitted according to the request type, and if the necessary information is omitted, the rule 2 is applied according to the omitted concept type. In this way, the cst table is complemented. The cst table, rel table, and request type information thus completed are converted into a format (for example, SQL format) suitable for each information processing system. This allows the user of the information processing system to give instructions and questions to the system using natural language sentences.

The present invention is not limited to the embodiment described above. For example, the information processing system connected to the sentence format conversion device 100 is not limited to the database system, but the information necessary for the system to operate can be specified, and the fixed format as illustrated in FIG. 3 can be determined. So long as it is applicable to any system.

Further, in the above-described embodiment, since it is premised that the cst table is processed one by one as an interface with the information processing system, the above-mentioned format conversion and complementary processing are performed. However, the format of the table and the rules applied to the complementary processing are appropriately changed according to the target information processing system.

[0044]

As described above, according to the present invention, it is possible to give instructions and questions to various information processing systems in natural language sentences, so that the system user can use each system. Eliminates the need to learn and reduces the burden.

[Brief description of drawings]

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

FIG. 2 is a diagram showing a description example of a concept dictionary 4 according to the same embodiment.

FIG. 3 is a diagram showing an example of a data format adopted in a format conversion unit 6 according to the same embodiment.

FIG. 4 is a diagram showing an example of rules applied in a complementing process according to the embodiment.

FIG. 5 is a diagram showing information required for each question sentence type corresponding to a request type and an example sentence of each question sentence type.

FIG. 6 is a flowchart showing an operation of the sentence format conversion device 100 according to the same embodiment.

FIG. 7 is a diagram showing an example of complementary processing according to the embodiment.

FIG. 8 is a diagram showing a format example when a processing result of the sentence format conversion apparatus 100 according to the embodiment is passed from the system interface section 8 to the main control section 11.

[Explanation of symbols]

 1 word division processing unit 2 request type determination unit 3 concept determination unit 4 concept dictionary 5 unnecessary word dictionary 6 format conversion unit 7 complementation processing unit 8 system interface unit 9 input device 10 output device 11 main control unit 12 DBMS 13 DB

 ─────────────────────────────────────────────────── ─── Continuation of front page (72) Inventor Shigeo Shimada 1-1-6 Uchisaiwaicho, Chiyoda-ku, Tokyo Nihon Telegraph and Telephone Corporation

Claims (1)

[Claims] 1. A sentence format conversion device for converting a directive sentence and a question sentence expressed in natural language into a format that can be interpreted by an information processing system, and a word division processing unit for dividing an input sentence into words. Of the words in the input sentence, a request type determination unit that determines the type of the instruction or question based on the keyword that represents the type, a concept dictionary that describes information indicating which system each word belongs to, and a concept dictionary An unnecessary word dictionary that describes unnecessary words such as auxiliary particles and auxiliary verbs that do not include information required by the information processing system, but what information does each word in the input sentence mean? Based on the concept dictionary and removes unnecessary words in the input sentence based on the unnecessary word dictionary, and a word omitted in the input sentence according to the type of the input sentence. The input sentence is converted into an expression format that can be interpreted by the information processing system, based on a complementary processing unit that complements the same predetermined complementary rule, and information obtained by the concept determination unit and information obtained by the complementary processing unit. A sentence format conversion device, comprising: