JP2898733B2 - Statement generation method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Overview] Regarding the sentence generation method in the teaching material creation support system, the burden on the teaching material creator is reduced as much as possible, and the creation of a program for automatically generating a sentence is supported. The purpose of the present invention is to provide a sentence generation method that can improve the learning effect by giving appropriate explanations.Question sentence data in the form of natural language input by the learner and the question sentence data are A model with question data conversion template data, which is a framework for converting into question data in a format that can be easily processed, a model having an undecided value having an object including an attribute value that has not been filled, and the undecided value are obtained. And a demon of sentence generation data for obtaining sentence generation data,
An input device for inputting template data for generating a sentence, which is a framework for generating a sentence from the data for generating a sentence, and a monitoring unit for managing operations on the model; A question data conversion processing unit that generates question data by collating with template data, and inference is performed according to the question data, and inference is performed by using an official demon and a demon of sentence generation data to determine an undetermined value. An undecided value extraction processing unit that generates a model with sentence generation data and sentence generation data from a model having a sentence, and generates a sentence by matching the sentence generation data with the sentence generation template data. A processing unit comprising a sentence generation processing unit to be output, data,
A processing device consisting of a memory having a function as a work area for generating a model and a sentence is provided, and a description is automatically generated.

[Industrial Application Field] The present invention relates to a sentence generation method in a teaching material creation support system.

In the field of artificial intelligence (hereinafter referred to as AI) and the like, a program that implements a physical model or the like by an object-oriented technique is often used.

The technology is, for example, computer-assisted learning (hereinafter,
It is also applied to simulation type CAL teaching materials in the field of (CAL). From the viewpoint of creating teaching materials, there is a demand for a general-purpose framework for generating (explanatory) sentences from the model of the CAL teaching materials, and from the viewpoint of learning, there is a need for knowledge that provides appropriate explanations to learners. CAL teaching materials are desired.

[Problems to be Solved by Conventional Techniques and Inventions] As conventional sentence generation methods or methods similar thereto, there are the following major ones.

What is realized by the question answering system (related to the AI field) It is related to fields such as natural language processing and machine translation.
For example, a natural language is generated from data in a knowledge expression format or a format similar thereto.

What is implemented in some specific intelligent CAL systems.

Mainly, it has a concept that promotes the problem solving (inference) technique of the system as a teaching (explanation) technique for teachers.
For example, the inference process for any problem of the expert system is taught (explained) to the user (learner) by displaying the production rules and the like in the order in which they were activated.

However, there has been no attempt to grasp, as a general-purpose framework, a sentence generation from a physical model by an object-oriented technique of a simulation-type CAL teaching material (a process of a physical simulation is described in a natural language).

 This placed a heavy burden on the creators of the teaching materials.

The present invention has been made in view of such conventional problems, and minimizes the burden on a creator of a teaching material, and supports creation of a program for automatically generating a sentence.
It is an object of the present invention to provide a sentence generation method capable of improving a learning effect by giving an appropriate explanation to a learner by a program.

[Means for Solving the Problems] FIGS. 1 (A) and 1 (B) are views for explaining the principle of the present invention.

In FIGS. 1 (A) and 1 (B), reference numeral 10 denotes a question sentence data 11 in the form of a natural language input by a learner, and the question sentence data 11 is converted into an easily processable form of question data 31. Template data for question data conversion 12 which is a framework of the model, a model 13 having an undecided value having an object including an attribute value which is not yet filled, an official demon 14 for obtaining the undecided value, and a statement. It manages the operation of the demonstration 15 of the sentence generation data for obtaining the generation data 33, the sentence generation template data 16 which is a framework for generating the sentence 34 from the sentence generation data 33, and the model 13. The input device 20 for inputting the data to the monitoring unit 17 is a question data 31 by matching the question sentence data 11 with the template data 12 for question data conversion.
Data conversion processing unit 23 for generating the
In accordance with 31, an inference is made by using an official demon 14 and a sentence generating data demon 15, and a model 32 having undetermined values and a model 32 with sentence generating data and sentence generating data 33 are Generated undetermined value extraction processing unit 24
By matching the sentence generation data 33 with the sentence generation template data 16, a processing unit 22 including a sentence generation processing unit 25 that generates and outputs a sentence 34,
This is a processing device including a memory 30 having a function as a work area for generating a sentence.

[Operation] In the present invention, as shown in FIGS. 1 (A) and 1 (B), a question sentence data 11, a question data conversion template data 12, a model 13 having an undetermined value,
The official demon 14, the sentence generation data demon 15, the sentence generation template data 16, and the monitoring unit 17 are input,
Stored in the memory 30.

The processing unit 22 sequentially executes the question data conversion process, the undecided value extraction process, and the sentence generation process.

The question sentence data 11 is collated with the question data conversion template data 12 by the question data conversion processing unit 23, and question data 31 is generated.

The undetermined value extraction processing unit 24 makes an inference according to the question data 31 and by using the official demon 14 and the sentence generation data demon 15 to determine from the model 13 having undetermined values. The model 32 and the sentence generation data 33 in which the values are embedded are generated.

The sentence generation data 33 is compared with the sentence generation template data 16 by the sentence generation processing unit 25 (explanation).
Statement 34 is generated and output.

 Therefore, the following effects can be obtained.

A program that automatically generates sentences by minimizing the burden on the creator (teacher, etc.) (simulation CA
Support the creation of L teaching materials).

Improve learning effects by providing appropriate explanations to learners (students, etc.) using the above programs (simulation type CAL teaching materials, etc.).

[Example] Hereinafter, an example of the present invention will be described with reference to the drawings.

 2 to 17 are views showing an embodiment of the present invention.

In FIG. 2, reference numeral 10 denotes an input device, and the input device 10 includes question sentence data 11, template data 12 for question data conversion, a model 13 having an undecided value, an official demon 14, and a demonic data 15 for sentence generation data. , Template data for statement generation
16 and the monitoring unit 17 are input.

The question sentence data 11 is information in the form of a natural language input by a learner.

The question data conversion template data 12 is a framework for converting the question sentence data 11 into an easily processable format called question data 31.

The model 13 having an undecided value has an object including an attribute value that is not yet filled (the object of the question sentence data 11).

The official demon 14 is for determining its undecided value.

The sentence generation data demon 15 is for obtaining (explanation) sentence generation data 33.

The sentence generation template data 16 is a framework for generating an (explanatory) sentence 34 from the sentence generation data 33.

The monitoring unit 17 manages operations on the model 13 such as replacement, deletion, addition, comparison, and insertion of parts (objects) constituting the model 13, and display of the model 13.
An erroneous operation on the model 13 is performed through the monitoring unit 17.

Reference numeral 20 denotes a processing device. The processing device 20 includes an input unit 21, a processing unit 22, an output unit 26, and a memory 30.
Is a question data conversion processing unit 23, an undecided value extraction processing unit 24,
And a sentence generation processing unit 25.

The input unit 21 receives, from the input device 10, question sentence data 11, question data conversion template data 12, a model 13 having an undetermined value, an official demon 14, and a sentence generation data demon 1.
5, statement generation template data 16, and monitoring unit 17
Is received and stored in the memory 30.

The processing unit 22 sequentially executes a question data conversion process, an undecided value extraction process, and a sentence generation process.

The question data conversion processing unit 23 generates the question data 31 by matching the question sentence data 11 with the question data conversion template data 12.

The undecided value extraction processing unit 24, according to the question data 31,
In addition, the official demon 14 and the sentence generation data demon 15
Is used to generate a model 32 in which the undetermined value is embedded and the sentence generation data 33 from the model 13 having the undetermined value.

The sentence generation processing unit 25 creates and outputs a (description) sentence 34 by comparing the sentence generation data 33 with the sentence generation template data 16.

The output unit 26 responds to the instruction from the monitoring unit 17 and
34 is output to the output area 41 of the output device 40.

The memory 30 is accessed from the question data conversion processing unit 23, the undecided value extraction processing unit 24, and the sentence generation processing unit 25,
It has a function as a work area.

 The output device 40 has an output area 41.

FIG. 3 shows an outline of the procedure module (function) name of the monitoring unit 17.

 Next, the operation will be described.

As shown in FIGS. 4 and 5, there is a problem that "when the frictional force of the object A is 50, the moving distance is 10, and the time is 2, explain how to find the power of the person A." The processing will be described using an example.

As shown in FIGS. 6 and 7, by comparing the question sentence data 11 "Please explain how to obtain the power of person A." with the question data conversion template data 12, the object name is obtained. (Person A) and attribute name (power) are extracted, and question data 31 is generated.

Next, as shown in FIG. 8, by accessing the attribute value (which is the target of the question sentence) in the model 13 which has not been determined by the question data 31, the official demon 14 is also activated. , Undetermined values are extracted. In other words, the work 500 of the object A is executed by the
And the power 250 of the person A is extracted.

Here, the demons set for the object X are shown in FIG. In Figure 9, the formulas for work, frictional force, and travel distance are
Display data 1 and 2 and sentence generation data demo 15 are set. The demon set for person X is
Shown in Figure 10. In FIG. 10, similarly, work, power,
Each formula of time, display data 1 and 2 and demonstration data 15 for sentence generation data are set.

The setting process of the sentence generation data 33 by starting the sentence generation data demo 15 is shown in FIGS. 11 and 12, and the generation of the sentence generation data 33 is not shown in FIG. It is performed in parallel with the extraction (inference) of the decision value. Here, fact attribute names (1,1), (1,2), (2,1), (2,2), fact attribute values (1,1), (1,2), (2,1) , (2,2), evidence data (1), (2), and conclusion attribute name (1),
(2) Conclusion attribute values (1) and (2) are generated.

Next, the sentence generation data 33 is collated with the sentence generation template data 16 shown in FIG.

Thus, as shown in FIG. 14, from clauses 1, 2, and 3,
The first sentence is generated, "Because the frictional force is 50 and the moving distance is 10, the work is 500 from" power = frictional force x moving distance "." In addition, as shown in FIG.
According to 3, "Because the frictional force is 50 and the moving distance is 10, the work is 500 from" Relationship between work, frictional force and moving distance "."
Is generated.

In addition, as shown in FIG. 16, from the phrases 1, 2, and 3, the following sentence reads, "Because work is 500 and time is 2, the work rate is 250 from" work rate = work / hour "." Is generated. In addition, as shown in FIG. 17, according to the clauses 1, 2, and 3, the work rate is 250 according to "Relationship between work rate and work and time, because work is 500 and time is 2". A statement is generated.

These explanations 34 are displayed on the screen as shown in FIGS. 4 and 5, respectively.

[Effects of the Invention] As described above, according to the present invention, it is possible to support the creation of a program for automatically generating a sentence by minimizing the burden on the creator of the teaching material, and By providing an appropriate explanation to the learner, the learning effect can be improved.

[Brief description of the drawings]

FIGS. 1A and 1B are explanatory diagrams of the principle of the present invention, FIG. 2 is a block diagram showing an embodiment of the present invention, FIG. 3 is an explanatory diagram of procedure module names of a monitoring unit, FIG. FIG. 5 shows a screen of a CAL teaching material, FIG. 5 shows a screen of another CAL teaching material, FIG. 6 shows a template mechanism for question data conversion, and FIG. 7 shows question data conversion by the template mechanism. FIG. 8, FIG. 8 is a diagram showing a process of extracting (inferring) undetermined values by demons, FIG. 9 is a diagram showing demons set on the object X, FIG. 10 is a diagram showing demons set on the person X FIG. 11 is an explanatory diagram showing a setting process of sentence generating data, FIG. 12 is an explanatory diagram showing a setting process of other sentence generating data, FIG. 13 is a diagram showing a template mechanism of sentence generation, FIG. The figure shows the first sentence generation by the template mechanism. Shows a sentence generation, FIG. 16 shows the next sentence generation by the template mechanism, FIG. 17 is a diagram showing the next sentence generation by another template mechanism. In the figure, 10 ... input device, 11 ... question sentence data, 12 ... template data for question data conversion, 13 ... model with undecided values, 14 ... official demon, 15 ... for sentence generation Data demo, 16 ... template data for sentence generation, 17 ... monitoring unit, 20 ... processing unit, 21 ... input unit, 22 ... processing unit, 23 ... question data conversion processing unit, 24 ... Undecided value extraction processing unit, 25 …… Sentence generation processing unit, 26 …… Output unit, 30 …… Memory, 31 …… Question data, 32 …… Model with embedded undecided values, 33 …… Sentence generation Data, 34 ... (description) sentence, 40 ... output device, 41 ... output area.

────────────────────────────────────────────────── ─── Continuation of the front page (56) References Noguchi, et al., “Qualitative and quantitative information in guidance strategies”, Information Processing Society of Japan Research Report (89-CE-4-12), Vol. 18, 1989, p. 85-92 (Hei 1-2-5) (58) Fields surveyed (Int. Cl. ⁶ , DB name) G06F 17/27-17/30 JICST Science and Technology Reference File

Claims (1)

(57) [Claims] 1. A question text data (11) in the form of a natural language input by a learner and a framework for converting the question text data (11) into an easy-to-process format called question data (31). A template data for question data conversion (12), a model with an undecided value (13) having an object that has not yet been filled in with attribute values, and an official demon for finding the undecided value (14) And a demonstration of sentence generation data (1
5), a sentence generation template data (16) as a framework for generating a sentence (34) from the sentence generation data (33), and a monitoring unit for managing operations on the model (13). An input device (10) for inputting (17) and a question sentence data (11) are converted into a template for question data conversion.
By collating with the data (12), the question data (3
According to the question data conversion processing unit (23) that generates 1) and the question data (31), inference is performed by using an official demon (14) and a sentence generation data demon (15), From models with undecided values (13) to models with undecided values embedded (32) and sentence generation data (33)
Sentence generation processing that generates and outputs a sentence (34) by matching the sentence generation data (33) with the sentence generation template data (16) A processing unit (22) consisting of a processing unit (22) consisting of a unit (25) and a processing unit (20) consisting of a memory (30) having a function as a work area for generating data, a model, and a sentence. A sentence generation method characterized by generating (34).