JPH08305708A - Method and device for generating explanatory document - Google Patents

Abstract

PURPOSE: To provide an explanatory document generating device and method capable of generating an explanatory document to be intuitively understood by a user. CONSTITUTION: This explanatory document generating device is provided with a knowledge base 4 for storing a name expressing a concept, the name of a concept upper or lower than the concept concerned and plural attributes expressing features included in the concepts, a similar concept searching means 1 which searches the knowledge base 4 and acquires a similar concept close to a concept to be explained, an attribute concept searching means 2 for searching a concept having a sense feature not included in the similar concept and a document generating means 3 for generating an explanatory document by combining the features of a concept searched by the means 1, 2.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for generating explanatory text, and more particularly to a method and apparatus for generating explanatory text for generating an explanatory text for explaining a concept. More specifically, for example, the present invention relates to a description sentence generation method and device for generating a description sentence that explains various terms and contents of a teaching material when using a natural language to support learning in an education system or the like.

[0002]

2. Description of the Related Art In the prior art, an explanation sentence generating apparatus for generating a sentence explaining a certain concept is explained based on related items such as an attribute of a concept, a superordinate concept and a subordinate concept in a knowledge base of the apparatus. Generate an explanation of the target concept.

The knowledge base describes only the lexical facts of a concept. Therefore, the conventional explanation sentence generation device generates the explanation sentence related to the explanation target concept by referring to this knowledge base. For example, if the knowledge base includes “carnivorous cats” as an explanatory sentence for the concept of “tiger”, the explanatory note generation device uses the concept of the object to be explained as the subject and “carnivorous cats” as the subject. The description is connected to generate a description “Tiger is a carnivorous feline”.

[0004]

However, since the explanation sentence generating apparatus in the above conventional technique uses the knowledge base in which only dictionary-like facts are described, it is possible to explain a certain concept like a dictionary. The problem is that we cannot explain what it is like intuitively. For example, when an explanation sentence of the concept of "tiger" is generated, the conventional explanation sentence generation device can generate a dictionary-like explanation sentence such as "Tiger is a carnivorous animal". A tiger is like a boss of a big scary cat. ”It is another thing that is not directly related, but there is a problem that it is not possible to generate a metaphorical expression comparing a similar object.

The present invention has been made in view of the above points, and an object thereof is to provide a description sentence generation method and a device capable of generating a description sentence that is intuitively understandable to a user of the device. And Further, a further object of the present invention is to
It is an object of the present invention to provide a description sentence generation method and device that allow even a person who is not familiar with handling a description sentence generation device to easily add or delete a description sentence.

[0006]

FIG. 1 is a block diagram showing the principle of the present invention. According to the present invention, in a description sentence generation device that generates a description sentence for explaining a concept in natural language, a name representing the concept, a name of a concept corresponding to a higher order or a lower order from the concept, and the concept has A knowledge base 4 that stores a plurality of attributes expressing features, a similar concept search unit 1 that searches the knowledge base 4 to obtain a similar concept close to the concept to be explained, and a similar concept that exists in the concept to be explained, Sentence generating means for generating an explanatory sentence by combining the features of the concepts searched by the attribute concept searching means 2 and the attribute concept searching means 2 for searching a concept having a sensory feature not included in 3 and 3.

Further, according to the present invention, in an explanation sentence generation device for generating an explanation sentence for explaining a concept in a natural language, information indicating a dictionary fact of a concept, a name indicating the concept, and the concept are used. In view of the name of the concept corresponding to the higher or lower order, the knowledge base 4 that stores a plurality of attributes expressing the features of the concept, the knowledge base 4 is searched, and the similar concept that acquires a similar concept close to the concept to be explained Search means 1;
Attribute concept search means 2 for searching for a concept that has a sensory feature that is present in the concept to be explained but is not included in similar concepts
And a fact information acquisition means for acquiring information representing a lexical fact other than the sensory features of the concept to be explained from the knowledge base 4, a fact information acquisition means, a similar concept searching means 1 and an attribute concept searching means 2. It has a sentence generation means 3 for generating an explanatory sentence by combining the features of the concepts.

The similar concept searching means 1 calculates the number of matching sensory features of the similar concepts for all the acquired similar concepts, and selects the attribute having the maximum sensory feature. If there are a plurality of attributes having the maximum number of sensory features in the sensory feature selecting means and the first sensory feature selecting means, the second sensory feature selecting one having a close affinity with the explanatory concept Including selection means.

The present invention also uses adjectives or adjective verbs as sensory features. Further, the attribute concept searching means 2 includes an attribute concept acquiring means for acquiring an attribute concept that maximizes the number of attribute concept sense feature matches of the concept in the knowledge base 4. The present invention, in an explanatory sentence generating method for generating a sentence explaining a certain concept, calculates the similarity of target concepts stored in a knowledge base based on the commonality of attributes, and does not prepare an explanatory sentence in advance. Then, a description using a simile is generated.

FIG. 2 is a diagram for explaining the principle of the present invention. According to the present invention, in an explanatory sentence generating method for generating a sentence that explains a certain concept, a knowledge base is searched (step 1), a similar concept close to the input explanatory target concept is acquired (step 2), and the explanatory target concept is acquired. , A concept having a sensory feature that is not included in a similar concept is searched (step 3), and the features of the searched concept are combined to generate an explanatory sentence (step 4).

When searching for a concept close to the concept to be explained (step 2), a knowledge base having a sensory feature represented by an adjective or an adjective verb is searched. The present invention is, in an explanatory sentence generating method for generating an explanation for explaining a concept in natural language, a name representing the concept, a name of a concept corresponding to a higher order or a lower order in view of the concept, and a feature of the concept. A concept to be explained is given to a knowledge base that stores a plurality of concepts having at least one attribute expressing, and the name of the concept corresponding to the upper direction or the lower direction viewed from the concept to be explained is referred by referring to the knowledge base. At least one is acquired, and the number of times the acquired attribute of the concept matches the attribute of the concept to be explained are counted,
By comparison, a concept close to the concept to be explained is acquired, and a description sentence related to the concept to be explained is generated using the attribute of at least one concept close to the acquired concept to be explained.

At least one of an adjective and an adjective verb is used as at least one attribute of the above concept.

[0013]

According to the present invention, an explanation sentence is generated and output similarly to a conventional explanation sentence generation device, and an explanation concept is extracted and inputted from the outputted explanation sentence, or an explanation inputted by an operator. Based on the concept, a concept similar to the explanatory concept is acquired, a concept having the difference between the sensory features of the explanatory concept and the similar concept as a sensory feature is searched from the knowledge base, and finally, the explanatory concept and the similar concept. By generating and outputting a sentence by combining knowledge of the attribute concept, it is possible to generate an explanatory sentence that is more intuitive and easy to imagine, in addition to generating a dictionary-like explanatory sentence. is there.

Further, since the adjective or the adjective verb is used as the sensory feature of the target concept, the explanation can be variously changed even if the user is not familiar with the device.

[0015]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 3 shows the configuration of the explanatory sentence generation apparatus according to an embodiment of the present invention. The explanatory note generation device 5 shown in the figure searches the knowledge base 4 and the knowledge base 4 that store upper and lower relations between target concepts, physical attributes of the target concepts, and sensory characteristics as slots, and explains Focusing on common attributes from the similar concept search unit 1 that acquires concept groups from the concept to the nth degree, and the concept groups obtained by the similar concept search unit 1, the concept with the maximum number of similar concept sense feature matches is found. An attribute concept search unit 2 to be selected, and a sentence generation unit 3 that extracts a sense feature that does not match and that generates a sentence by embedding a relation to an analogous concept or an adjective in a prepared template.

Here, the concept to be explained is a concept which is a target for generating an explanation sentence. The similar concept is a concept similar to the concept to be explained. For example, if the concept to be explained is “tiger”, the similar concept is “cat”. The sensory characteristics are expressed by adjectives or adjective verbs and have sensory meanings. For example, the sensory characteristics of “cat” are “small”, “pretty”, and the like.

The similar concept searching unit 1 searches the knowledge base 4 based on the input concept to be explained, and acquires a similar concept having a concept close to the concept to be explained. The similar concept searching unit 1 extracts and inputs the concept to be explained from the already generated explanation by the same method as the conventional method, or the explanation expressed in a more simplistic manner by the operator. When a sentence is desired, an explanation target concept to be explained is input. Also, the similar concept searching unit 1
May input a concept that the user does not know. This is to avoid using a concept that the user does not know when explaining the concept to be explained.

The similar concept searching unit 1 searches the knowledge base 4 with the input concept to be explained as a key. As shown in FIG. 4, the knowledge base 4 expresses the characteristic of one concept to be explained by an adjective or an adjective verb and is stored as an attribute. In the following description, the concept to be explained is represented by a frame-type knowledge representation in this way. Further, each concept has an arbitrary slot, and each slot has an arbitrary value. For example, "cat"
Has two slots, "seed" and "sensory feature".

Each concept slot can be inherited and referenced even if the concept slot does not actually exist in the concept, if the slot is defined in the superordinate concept. For example,
In the example of the knowledge base 4 shown in FIG. 4, the slot “seed” does not exist in the concept of “tiger”, but the slot “seed: Carnivora feline” is defined in the superordinate concept “cat”. Therefore, the “seed” of “tiger” is also a carnivorous cat. These superordinate and subordinate conceptual relationships can be constructed by using a general similar dictionary or classification system.

Some concepts in Knowledge Base 4 include:
There is a slot called "sensory feature", and as its value, an attribute specified by a general adjective or adjective verb is stored. The similar concept searching unit 1 extracts a similar concept in which the number of matching sensory features of the similar concept acquired by searching the knowledge base 4 based on the concept to be explained has the maximum value. If there are multiple similar concepts that take the maximum value,
In the thesaurus with the concept to be explained, those with close relatives should be selected. Further, the similar concept searching unit 1 extracts similar concepts having different sensory features from the similar concepts acquired from the knowledge base 4.

The attribute concept searching unit 2 searches for a concept having a sensory feature which is present in the concept to be explained but is not included in the similar concepts. The attribute concept search unit 2 examines a difference in sensory features between the explanation concept and a similar concept extracted for explaining the concept to be explained, and searches for an attribute concept used for explaining the sensory feature difference. Attribute concept search unit 2
Finds a sensory feature set that does not match between the explanatory concept and the similar concept, finds the number of sensory feature matches from the knowledge base 4 for concepts other than the explanatory concept, the similar concept, and the user unknown concept, and the value is the maximum. Search for a concept that is and use it as an attribute concept.

The attribute concept search unit 2 calculates the attribute concept sense feature difference, and if there is no feature difference, the sentence generation unit 3 to be described later.
In, a noun related to an attribute concept, a noun corresponding to a similar concept, and a noun corresponding to an explanation target concept are generated and held. If there is the feature difference, the sentence generation unit 3 creates a sentence using a slot other than the sensory feature of the concept to be explained.

The sentence generator 3 combines the features of the concepts searched by the similar concept search unit 1 and the attribute concept search unit 2 to generate an explanatory sentence. In the attribute concept search unit 2, when there is a difference in attribute concept feeling features, the explanation sentence is generated using slots other than the feeling features of the concept to be explained. In addition, when there is no difference in attribute concept sense characteristics, a noun corresponding to the attribute concept, a noun corresponding to the similar concept, and a noun corresponding to the concept to be explained are generated in advance, and a generation result is generated according to a predetermined template. Embed and assemble the description and output. Further, the knowledge base 4 is searched with the concept to be explained input to the similar concept searching unit 1, and the explanation sentence in which a dictionary-like fact expression is obtained is also output.

FIG. 5 is a flow chart showing the outline operation of the explanation sentence generation of one embodiment of the present invention. (Step 101) First, the similar concept searching unit 1 acquires a concept within n degrees of degree from the explanation target concept C _exp .

More specifically, the concept to be explained first (explanation)
Target concept: c_exp) 1 and the concept that the user does not know
More than one (user unknown concept set C_ukn[]) Similar concept
Input to the search unit 1. Similar concept search unit 1 is a knowledge base
4 is searched, and the explanation target concept c _expTo n less than relative
Get all. However, the explanation target concept c_expMyself and you
User Unknown Concept Set C_uknExcludes items included in [].
The remaining concepts in this way are similar concept sets
(C_sim[]).

Step 102) The similar concept searching unit 1
For all similar concept sets C _sim [], the number of similar concept sense feature matches Agr (Sf (C _exp ), Sf (C _sim [i
])) Is the maximum, the similar concept i (i _max ) is selected. However, Agr (Sf1, Sf2) is the sensory feature Sf
The number of coincidences of 1 and Sf2 is shown.

More specifically, the similar concept searching unit 1 calculates the similar concept sense feature matching number Agr (Sf (c _exp ), Sf (C _sim [i])) for all similar concepts C _sim []. Then, the similar concept i (i _max ) that maximizes it is determined. It should be noted that Sf (c) is a function for acquiring the sensory feature from the concept c, and Agr (sf (1, sf2) acquires the number in which the same adjective or adjective verb is used for the sensory features sf1 and sf2. For example, in the example of the knowledge base 4 in Fig. 4, in the case of the concept of "tiger" and the concept of "boss", Sf (tiger) = [large, strong, scary] Sf (boss) = [strong] , Fright], and Agr (Sf (tiger), Sf (parent)) = Agr ([large, strong, dreaded], [strong, dreaded]) = 2.

As described above, when there are a plurality of concepts in which the number of matching sensory features has the maximum value, the one having a close affinity with the concept to be explained is selected. Step 103) Similar concept sense feature difference Diff (Sf (C _exp ), Sf (C
_sim [i _max ])) is calculated and it is not an empty set (0), that is, when there is a concept in which the number of sensory feature matches is maximum,
If it is an empty set (0), the process proceeds to step 104.

Note that Diff (sf1, sf2) is a function that is included in the sensory feature sf1 and that acquires an adjective or a set of adjectives not included in sf2, and is called "tiger" in the knowledge base 4 in FIG. In the case of the concept of “boss”, Diff (Sf (tiger), Sf (boss)) = Diff ([large, strong, scary], [strong, scary]) = [large].

Step 104) The attribute concept searching unit 2
A concept having a sensory feature that is included in the concept to be explained but is not included in the similar concept is searched for. The attribute concept search unit 2
Of the concepts of the knowledge base 4, for all concepts C _rest [] other than the explanation target concept C _exp , the user unknown concept set C _ukn [], and the similar concept C _sim [], the attribute concept sense feature matching number Agr (Diff ( C _exp , C _sim [i _max ]), Sf
(C _rest [j])) is calculated, and the concept j (j _max ) that maximizes the number of matches is selected. This C _rest [j _max ] is called an attribute concept. Here, of the concepts of the knowledge base 4, all the concepts other than the explanation concept, the user unknown concept set, and the similar concept set are used when the concept to be explained and the similar concept other than the user unknown concept are determined. This is for examining the difference between the concept to be explained and its similar concept, and then for explaining the difference.

Step 105) The attribute concept searching unit 2
Attribute concept sense feature difference Diff (Diff (Sf (C _exp ), Sf (C
_sim [i _max ])), Sf (C _rest [j _max ])), and if it is not the empty set (0), then step 1
If it is the empty set (0), the process proceeds to step 06.
Move to 7.

Step 106) The sentence generator 3 uses the attribute concept sense feature difference Diff (Diff (c _exp , C _sim [i _max ]), S
f (C _rest [j _max ])) generates an adjective form of all adjectives or adjective verbs and temporarily holds them.

Step 107) The sentence generator 3 generates a noun corresponding to the attribute concept C _rest [j]. Step 108) The sentence generation unit 3 uses the similar concept C
Generate a noun corresponding to _sim [i].

Step 109) The sentence generator 3 generates a noun corresponding to the explanation target concept C _exp . Step 110) The sentence generator 3 uses the explanation target concept C _exp
Generate and output a sentence using a slot other than the sensory features of. That is, the explanation sentence of the dictionary-like fact information of the explanation target concept C _exp is acquired.

Step 111) If there is an attribute concept, the result of step 108, the result of step 109 and the case particle "no" are used by using the generation results held in the above steps 107, 108 and 109. When the result of step 107 is connected in the order, and there is no concept attribute, the generation result of step 108 is directly used in step 11
Move to 2.

Step 112) "* Ex in the template" For example, * Exp is like * Sim "
The generated result of step 108 is displayed in the "p" part as "* Si
The result of 109 is embedded in the m ″ part and this is held. The above contents will be described below by using a specific example.

[First Specific Example] In the following first specific example
Consider the case where you want to generate a description of the concept of "tiger".
It Here, the explanation target concept c_expIs a tiger,
User unknown concept set C _uknThere is no []. Therefore, you
The user knows everything except the concept of "tiger"
Suppose.

When the knowledge base 4 has the contents shown in FIG. 4, the explanatory sentence generating device first searches the knowledge base 4 by the similar concept searching unit 1 as shown in step 101, and explains the concept c _exp. To get a concept within n relatives. Now, assuming that n = 2, C _sim [] = [cat, Persian cat, calico cat, lion] is obtained as a similar concept set.

Next, as shown in step 102, the similar concept searching unit 1 sets the similar concept sense feature matching numbers Agr (Sf (c _exp ), Sf (C) for all similar concepts C _{si m} [].
Select i that maximizes _sim [i])). Now, the number of matching sensory features of each similar concept is C _sim [0] = Agr (Sf (tiger), Sf (cat)) = Agr ([big, strong, scary], [small, cute]) = 0 C _sim [1] = Agr (Sf (tiger), Sf (Persian cat)) = Agr ([big, strong, scary], [small, cute]) = 0 C _sim [2] = Agr (Sf (tiger), Sf (calico cat)) = Agr ([big, strong, scary], [small, cute]) = 0 C _sim [3] = Agr (Sf (tiger), Sf (lion)) = Agr ([large, Strong, awful], [large, strong, awful]) = 3, so i that maximizes this is i _max = 3.

At this time, in step 103, the similar concept sense feature differences Diff (Sf (c _exp ), Sf (C
_{Since sim} [i _max ])) → Φ, the process proceeds to step 108. Step 10
In 8, the noun corresponding to the similar concept C _sim [i _max ] is generated. Since C _sim [i _max ] is “tiger” now,
The noun "tiger" is generated. Similarly, in step 109, a noun corresponding to the explanation target concept c _exp is generated. Now, since the explained concept c _exp is “lion”, the noun “lion” is generated.

Next, in step 110, the sentence generator 3
Generates and outputs a sentence using a slot other than the sensory feature of the concept to be explained c _exp . Now, the slots other than the sensory features of "Tiger" are "Species: Carnivora,""Size: about 3 m,""Color: yellow, black stripes," so for example, "Tiger is a carnivorous cat. It is an animal, the size is about 3 m, and the color is yellow with black stripes. "Is generated and output.

Next, in step 111, since there is no attribute concept, the generation result “lion” of step 109 is passed to step 112 as it is. Finally, in step 112, the generation result “tiger” of step 108 is included in the * Exp portion of the template “For example, * Exp is like * sim.” And step 111 is included in the * sim portion.
As a result, the "lion" is embedded and output. Since the generation result of step 110 is "tiger" and the result of step 111 is "lion", the sentence "a tiger is like a lion" is output.

As a result, the explanatory sentence generation device generates the sentence generated in step 110, "Tiger is a carnivorous feline, the size is about 3 m, and the color is yellow with black stripes." Combined with the sentence generated by, "To compare, a tiger is like a lion.", "A tiger is a carnivorous cat, about 3 m in size, and yellow with black stripes. The tiger is like a lion. "

[Second Concrete Example] Consider a case where a user does not know the concept of “lion” and an explanation sentence of the concept of “tiger” is generated. In this case, the explanation target concept c
_exp = tiger and C _ukn [] = [lion].

Similar to the first specific example above, when the knowledge base 4 has the contents shown in FIG. 4, when n = 2, first, in step 101, as a similar concept set, C _sim [] = [Persian cat, calico cat, cat] are obtained.

[0046] In step 102, all number of similar concept sensory characteristics consistent with the description target concept c _exp from similar concepts to select a concept which maximizes, now similar concept sensory features number match, C _sim [0] = Agr (Sf (tiger), Sf (cat)) = 0 C _sim [1] = Agr (Sf (tiger), Sf (Persian cat)) = 0 C _sim [2] = Agr (Sf (tiger), Sf (tiger) Calico)) = 0 and all are equal, so choose the one with the closest relative degree. That is, i _max = 0.

At this time, the similar concept sense feature difference Diff
(Sf (c_exp), Sf (C_sim[I _max])) Is
[Big, strong, scary] and not empty so
Step 104 and subsequent steps are executed (step 103). Steps
In 104, the explanation target concept in the knowledge base 4 and the user
All Concepts C except Knowledge Concept Set and Similar Concept Set
_restFor [], the number of attribute concept sense feature matches Agr (Di
ff (Sf (c_exp), Sf (C_sim[I_max])),
Sf (C _rest[J])) is maximized.
Now, if the knowledge base 4 has the contents of FIG. 4,
C_restSuppose [] is only [human relationships, parent, child]
Then, Agr (Diff (Sf (tiger), Sf (cat)), Sf (human relationship)) = Agr ([big, strong, scary], []) = 0 Agr (Diff (Sf (tiger), Sf (Cat)), Sf (parent) = Agr ([big, strong, dreaded], [strong, dreaded]) = 2 Agr (Diff (Sf (tiger), Sf (cat)), Sf (child) = Agr ([Big, strong, scary], [weak]) = 0, so C_rest[J] = sense of attribute concept when parent
The number of feature matches is maximum. That is, j_maxC where = 1
_rest[J_max] Is called an attribute concept.

Next, at step 105, the attribute concept sense feature difference Diff (Diff (Sf (c _exp ), Sf (C
_sim [i _max ])), Sf (C _rest [j _max ])) is checked for an empty set, but Diff (Diff (Sf (tiger), Sf (cat)), Sf (parent)) = Diff ([ Big, strong, scary], [strong, scary]) = [big], so this is not an empty set, so step 10
6 and below are executed.

In step 106, all adjectives or adnominal verb adnominal forms of the attribute concept sense feature differences are generated. Now, since Diff (Diff (Sf (tiger), Sf (cat)), Sf (parent)) = [large], this is made into a connected form, and “large” is generated.

Subsequently, in step 107, the attribute concept C
A noun corresponding to _rest [j _max ] is generated in step 108, a noun corresponding to the similar concept C _sim [i _max ] is generated in step 109, and a noun corresponding to the explanation target concept c _exp is generated in step 109. Since C _rest [j _max ] = boss, C _sim [j _max ] = cat, and c _exp = tiger, the nouns corresponding to them are “boss”.
It is assumed that "cat" and "tiger" are generated.

Next, at step 110, a sentence is generated using slots other than the sensory features of the concept c _{exp to be} explained. Assuming that it is the same as in Example 1, the generated sentence is "Tiger is a carnivorous feline animal, with a size of about 3 m and a yellow stripe with a black color."

Next, in step 111, since there is an attribute concept, the generation results of steps 107, 108, and 109 are used to calculate "generation result of step 107, step 109".
Is generated, the case particle “no”, and the generation result of step 108 ”. That is, the phrase "a big cat boss" is generated.

Finally, in step 112, the template is used to embed the generation result “tiger” in step 109 and the generation result “large cat boss” in step 111.
Output this. In other words, the sentence “a tiger is like the master of a big cat” is output. As a result, the explanatory sentence generation device 5
Produces and outputs the sentence "Tiger is a carnivorous feline, about 3 m in size, and yellow with black stripes. For example, a tiger is like the boss of a large cat." To do.

In this way, finally, a lexical explanation of the concept registered in the knowledge base 4 "Tiger is a carnivorous cat, the size is about 3 m, and the color is yellow with black stripes. After generating "," a tiger is like the boss of a big cat, which is compared with another concept that is not directly related to the concept to be explained. ] Can be generated, and by combining these two sentences, it is possible to generate a sentence that is intuitively easy to understand.

The present invention is not limited to the above-mentioned embodiments, and various modifications and applications are possible within the scope of the claims.

[0056]

As described above, according to the explanation sentence generation device and the explanation sentence generation method of the present invention, when a concept is explained in the education / learning promotion device or the like, it is stored in the knowledge base. Not only is it possible to generate lexical explanations, but we also search for specific attributes such as color and size of the target concept, and similar concepts when focusing on upper and lower concepts, and Since a sentence similar to the concept can be generated, it is possible to generate a descriptive sentence that is more intuitive and easy for the user of this device to imagine.

Further, since the explanation sentence generating apparatus and the explanation sentence generating method of the present invention use the adjective or the adjective verb as the sensory feature of the target concept, it is easy for a person who is not familiar with the apparatus to easily use the adjective or the adjective verb. Can be added or deleted.

[Brief description of drawings]

FIG. 1 is a principle configuration diagram of the present invention.

FIG. 2 is a diagram illustrating the principle of the present invention.

FIG. 3 is a configuration diagram of an explanatory sentence generation device according to an embodiment of the present invention.

FIG. 4 is a flowchart showing a description sentence generation processing operation according to an embodiment of the present invention.

FIG. 5 is a diagram showing an example of a knowledge base according to an embodiment of the present invention.

[Explanation of symbols]

 1 Similar Concept Search Unit, Similar Concept Search Unit 2 Attribute Concept Search Unit, Attribute Concept Search Unit 3 Sentence Generation Unit, Sentence Generation Unit 4 Knowledge Base 5 Explanatory Sentence Generation Device

Claims (10)

[Claims] 1. An explanatory sentence generation method for generating a sentence for explaining a certain concept, wherein similarity between target concepts stored in a knowledge base is calculated based on commonality of attributes, and an explanatory sentence is prepared in advance. Without
An explanation sentence generation method for generating an explanation sentence using a metaphor. 2. An explanatory sentence generation method for generating a sentence for explaining a certain concept, wherein a knowledge base is searched to obtain a similar concept close to the concept to be explained, and a concept existing in the concept to be explained exists in the similar concept. 2. The explanation sentence generating method according to claim 1, wherein a concept having a sensory feature that is not included is searched, and the explanation sentence is generated by combining the features of the searched concepts. 3. The explanatory sentence generating method according to claim 2, wherein a knowledge base having a sensory feature expressed by an adjective or an adjective is searched when searching for a concept close to the concept to be explained. 4. The explanation sentence generating method according to claim 3, wherein at least one of an adjective and an adjective is used as at least one attribute of the concept. 5. A description sentence generation method for generating an explanation for explaining a concept in natural language, the name representing a concept, the name of a concept corresponding to a higher order or a lower order from the concept, and the concept having A concept corresponding to an explanation target concept is given to a knowledge base that stores a plurality of concepts having at least one attribute expressing a characteristic, and the concept corresponds to an upper direction or a lower direction viewed from the concept of the explanation target with reference to the knowledge base. At least one of the names of the above-mentioned concept and the number of times the acquired attribute of the concept matches the attribute of the explained concept are counted and compared to obtain a concept close to the explained concept. An explanation sentence generation method, wherein an explanation sentence relating to the explanation subject concept is generated using an attribute of at least one concept close to the explanation subject concept. 6. An explanatory sentence generating apparatus for generating an explanatory sentence for explaining a concept in natural language, wherein a name representing the concept, a name of a concept corresponding to a higher order or a lower order from the concept, and the concept A knowledge base that stores a plurality of attributes expressing the sensory characteristics that it has, a similar concept search unit that searches the knowledge base and acquires a similar concept that is close to the concept to be explained, but exists in the concept to be explained, A sentence for generating an explanatory sentence by combining attribute concept searching means for searching a concept having a sensory feature not included in the similar concept, the similar concept searching means, and the feature of the concept searched by the attribute concept searching means An explanation sentence generation device comprising: a generation unit. 7. An explanation sentence generating apparatus for generating an explanation sentence for explaining a concept in natural language, information indicating a dictionary fact of a concept, a name indicating the concept,
A name of a concept corresponding to a higher order or a lower order in terms of the concept, a knowledge base storing a plurality of attributes expressing the features of the concept, and a search of the knowledge base to obtain a similar concept close to the concept to be explained. Similar concept search means, attribute concept search means for searching for concepts having sensory features that are present in the concept to be explained but are not included in the similar concept, and dictionary-like other than the sensory features of the concept to be explained A sentence for generating an explanatory sentence by combining factual information acquisition means for acquiring information indicating facts from the knowledge base, and the features of the concepts searched by the factual information acquisition means, the similar concept searching means, and the attribute concept searching means. An explanation sentence generation device comprising: a generation unit. 8. The first similar concept search means calculates the number of matching sensory features of similar concepts for all the acquired similar concepts, and selects the attribute having the maximum sensory feature. In the sensory feature selection means and the first sensory feature selection means, when there are a plurality of attributes having the maximum sensory feature, a second sensory feature that selects one having a close affinity with the explanatory concept The explanatory sentence generation apparatus according to claim 6, further comprising a selection unit. 9. The explanation sentence generation device according to claim 8, wherein an adjective or an adjective verb is used as the sensory feature. 10. The explanatory sentence generation according to claim 6, wherein the attribute concept searching means includes an attribute concept acquiring means for acquiring an attribute concept that maximizes the number of matching of the attribute concept feeling features of the concept in the knowledge base. apparatus.