JPH0157829B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a sentence generation system that generates sentences in a target language based on a conceptual structure using a word search device, a syntactic pattern search device, and a word relationship determination device in the target language. be.

Conventional English sentence generation systems that create English sentences from Japanese sentences generate English sentences from Japanese sentences by associating English words with Japanese words.
This conventional method has the drawbacks of being limited in the number of English sentence patterns it can handle and not being able to use proper English.

SUMMARY OF THE INVENTION It is an object of the present invention to eliminate the above-mentioned drawbacks and to provide a sentence generation system capable of generating complex target language sentences with optimal syntactic patterns. And for that reason,
The sentence generation system of the present invention generates a concept network from a concept network having a plurality of sets each consisting of two concept numbers representing two mutually related concepts and a concept number representing a relationship between the concepts. a target language word search device 2, which is a sentence generation system that outputs sentences in a target language that match the meaning of a sentence shown, and has a word dictionary that stores words in the target language corresponding to each of various concept numbers; A syntactic pattern search device 3 having a syntactic pattern dictionary that stores syntactic patterns for each of a plurality of types of target language words, attributes that the target language words can have, and a target when the target language words have certain attributes. A word relationship determination device 4 having a word association matrix showing relationships with other words in a language, a word search device 2 of the target language, a syntactic pattern search device 3, and a word relationship determination device 4 according to a pre-created procedure. The present invention is characterized in that it has a target language sentence generation device 1 which controls the device 4 and outputs sentences in the target language that are compatible with the input concept network.

Hereinafter, the present invention will be explained with reference to the drawings.

FIG. 1 shows an outline of an embodiment of the English sentence generation system of the present invention.

In FIG. 1, 1 is an English generation device, 2 is an English word search device, 3 is a syntax pattern search device, 4 is an inter-word relationship determination device, and 5 is a word order determination device.

The English generation device 1 controls the entire system. The English word search device 2 has a dictionary in which English words corresponding to numbers representing concepts are stored, as shown in FIG. The syntactic pattern search device 3 has a dictionary that stores syntactic patterns corresponding to English words, as shown in FIG. For example, the English word "create" has an action agent, an object,
Information indicating that a prepositional phrase is required is stored. Each field of the syntactic pattern search device has a number written in it, which represents the word order.
English words corresponding to prepositional phrases, etc. are searched. The inter-word relationship determination device 4 has an inter-word relationship matrix as shown in FIG. The inter-word relationship matrix shows the attributes that an English word can have, as well as the relationship with other English words when an English word has a certain attribute. The example in FIG. 4 shows that the word "She" has an action-dominant attribute, and the word "She", which has an action-direction attribute, is related to the word "create". Word order determining device 5
gives word order to words.

FIG. 5 shows a specific example of input/output of the English writing system. As shown in Figure 5, the Japanese sentence ``She makes butter from milk'' is converted into a concept association table as shown. The concept of “girlfriend” is numbered 10200, the attached word “ha” indicating the action is number 1, the concept of “milk” is numbered 16080,
The attached word “kara” indicating the source is number 10, the concept “butter” is number 16030, the attached word “wo” indicating the object is number 2, and the concept “making” is number 16030.
It is represented by the number 30500. Each entry in the concept association table contains data indicating two related concepts and the relationship between them. When such a conceptual relationship table is input to the English generation device 1, the English generation device 1 uses the English word search device 2, the syntactic pattern search device 3, and the word gap determination device 4 to
Outputs the English sentence "make butter from milh".

Next, an outline of the operation of the English generation device 1 will be shown below.

(1) Processing begins with the central concept in the conceptual relationship.

(2) First, use English word search device 2 for the selected concept.
, and search for English words that correspond to that concept. Multiple English words may be elicited.
If there are multiple English words, use the one with the highest priority, and if there is a mismatch, use the one with the next priority.

(3) When an English word is selected for a certain concept, a syntactic pattern for the English word is searched for using the syntactic pattern search device 3. If there is more than one, use the one with the highest priority, and if there is a non-conformity, use the one with the next priority.

(4) When a certain syntactic pattern is selected, the search priority for the surrounding concepts indicated by arrows is determined according to that syntactic pattern. The following processing is performed in order of priority. Select English words as in (2) for the prioritized concepts,
Select a syntax pattern as in (3). If there is a matching syntax pattern or English word, the word relation determining device 4 is used to check whether there is a relationship between the English word at the tip of the arrow and the English word at the origin of the arrow. If there is a relationship, the process is repeated for the next priority concept. If there is no relationship, the syntactic pattern and English word of the original concept of the arrow are changed and reprocessed.

(5) Finally, by arranging English words in the same order as the word order, an English sentence is created.

Next, regarding the specific example shown in FIG. 5, how the English generation device 1 operates is shown in FIG.
This will be explained with reference to (b) and (c). As shown in Figure 6, the process begins with the central concept ``Create ³⁰⁵⁰⁰ ''.

Γ Search for an English word corresponding to "30500" using the English word search device 2. (1) create, (2) make... are searched.

Γ Considering the priority, select (1) create, and use the syntax pattern search device 3 to search for a syntax pattern for create.

Since there are a plurality of Γ syntax patterns 1 and 2, syntax pattern 1 having a higher priority is used.

Γ Based on syntactic pattern 1, add the word order to the concepts indicated by the arrows from ``Tsuru ³⁰⁵⁰⁰ ''.

Since Γ syntax pattern 1 has "actor" written as the word order, the word order is set to "Kojo ¹⁰²⁰⁰ " at the end of the arrow indicating "actor".

The English word search device 2 is used to search for a corresponding English word for Γ " ^Kare10200 ". Here (1)
She is sought out.

Find the syntactic pattern of Γ She. Since the syntactic pattern for She is only written as She, the syntactic pattern cannot be applied to surrounding concepts. However, since there are no concepts with arrows around "girlfriend ¹⁰²⁰⁰ ," this syntax pattern is suitable.

Γ The inter-word relationship determining device 4 is used to check whether the English word She has a relationship of action ¹ with respect to the English word create. Since there is a relationship, we return to "Tsuru ³⁰⁵⁰⁰ " and apply syntax pattern 1 to search for the next word order.

Γ Since create is entered as the word order,
Let create itself be the word order.

Γ Next, the word order is "butter ¹⁶⁰³⁰ " because it is an object. "Butter ^160
30 "
For ``Kojo ¹⁰²⁰⁰ '', select English words and syntax patterns.

Γ As shown in FIG. 6B, butter uses the word relationship determination device 4 to check whether there is a relationship called object ² with respect to create. Since it is unrelated
Syntax pattern 1 is not suitable for create, so syntax pattern 2 is used for create next.

Γ Syntax pattern 2 is also nonconforming.

Γ English word with priority 1 for "Tsukuru ³⁰⁵⁰⁰ "
Since create was unsuitable, it was reprocessed using make with priority 2.

Γ If we perform the same processing as before, “She has a main relationship with make” and “butter has a main relationship with make”.
There is an object relationship to make. ” and the process proceeds.

Γ Using the syntactic pattern of "make", search for an English word using the word order "milk ¹⁶⁰⁸⁰ " and search for the syntactic pattern. As before, milk is
The word relation determining device 4 is used to check whether a relation "source ¹⁰ " exists for make. There is a relationship.

Γ Now that the word order has been assigned to the concepts surrounding "Tsuru ³⁰⁵⁰⁰ " and all the concepts have been processed, an English sentence is generated based on the word order. Then “She
make butter from milk”.

As is clear from the above description, according to the present invention, sentences in the target language having appropriate words and syntactic patterns can be generated from conceptual structures. Note that the above explanation uses English sentence generation as an example, but
It goes without saying that the present invention can also be applied to sentences in other target languages, such as French sentence generation and German sentence generation.

[Brief explanation of drawings]

FIG. 1 is a diagram showing an overview of an embodiment of the English sentence generation system of the present invention, FIG. 2 is a diagram explaining an English word dictionary, FIG. 3 is a diagram explaining a syntax pattern storage dictionary, and FIG. FIG. 5 is a diagram illustrating a word relationship matrix, FIG. 5 is a diagram showing a specific example of input/output in the English sentence generation system, and FIG. 6 is a diagram showing the processing flow of the English sentence generation system of the present invention. 1...English generation device, 2...English word search device,
3... Syntactic pattern search device, 4... Inter-word relationship determination device.

Claims (1)

[Claims] 1. From a concept network having a plurality of sets consisting of two concept numbers representing two mutually related concepts and a concept number representing a relationship between the concepts, a sentence generation system that outputs a sentence in a target language that matches the meaning of the sentence, the target language word search device 2 having a word dictionary storing words in the target language corresponding to each of various concept numbers; a syntactic pattern retrieval device 3 having a syntactic pattern dictionary storing syntactic patterns for each type of target language word, attributes that the target language word can have, and a target language when the target language word has a certain attribute. A word relationship determination device 4 having a word association matrix showing relationships with other words, and a target language word search device 2, a syntactic pattern search device 3, and a word relationship determination device according to a pre-created procedure. 1. A sentence generation system comprising a target language sentence generation device 1 which controls a target language sentence generation device 4 and outputs a sentence in a target language adapted to an input concept network.