JPH096781A - Method and device for processing sentence - Google Patents

Abstract

PURPOSE: To properly rearrange sentences in the order of passage of time and to make clear the timewise connecting relation of respective sentences by dividing the sentences into a statement sentence and the other sentence. CONSTITUTION: This device is composed of an internal bus 1, CPU 2, ROM 3, RAM 4, input control part 5, keyboard 6, pointing device 7, display output control part 8, CRT 9, external storage device control part 10, hard disk 11 and floppy disk 12. First of all, a Japanese sentence is analyzed, an adverb, auxiliary verb, or verb showing the statement of a book is detected and the sentences are divided into the statement sentence including the statement of an author and the explanation sentences showing the other matters. Then, a link expressing the time sequential order of respective sentences describing the statement of the author, link expressing the time sequential order of respective explanation sentences and link expressing the time sequential order of the describing parts of respective matters are generated and further, a link expressing the time sequential order of the statement sentence, explanation sentences and respective matters is generated. Thus, the relation of the passage of time for each sentence included in the sentence can be grasped.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a sentence processing method and apparatus, and more particularly to a sentence processing method and apparatus for rearranging sentences in chronological order. Further, in particular, the present invention relates to a Japanese learning system for teaching Japanese and a Japanese sentence processing method and apparatus applicable to automatic generation of animation from natural language.

[0002]

2. Description of the Related Art Generally, as a method of analyzing a time relation such as a story, a sentence arranged in the order of appearance is rearranged based on information such as an adverb indicating time, a connective particle or a connective indicating a connection relation. Many. In these methods, one time flow is assumed, events (events) are placed in it, and the context and overlap are described.

[0003]

However, in the conventional method, the author's statement cannot be distinguished from the non-past tense as an expression method, and the author's statement is incorporated into the time series of matters. There was a possibility that For example, consider the example sentence shown in FIG. Underlined example sentence 1: "The mother's deer's neck slowly sinks into the water" is the same time as the past time around, but it is shown in non-past tense to make it stand out. There is.

Further, the underlined part of Example sentence 2: "There is no doubt that it is an act out of compassion as a hunter" is the statement of the author's statement. The thing about "acts of compassion as a hunter" is the same as the flow of time around me, but the judgment of the author "I am sure" happened at the same time as the flow of time around me. Has not been confirmed yet.

The underlined part of Example Sentence 3: "I didn't know then, but it was my grandfather's compassion," is also the statement of the author's statement. "I didn't know at that time" and "It was my grandfather's compassion", but it was the same as the flow of time around me, but the statement of the statement that made the decision "no" is later than the time of matter. It exists at the time of.

When the example sentence 1 is simply interpreted by the conventional method,
Only the underlined part is out of the time series. Taking the non-past tense as an expression method into consideration and adjusting the non-past tense to the surrounding time series, both the underlined part of the example sentence 2 and the underlined part of the example sentence 3 exist in the surrounding time series. As the part of matters such as "acts out of compassion as a hunter", "I did not understand at that time", "it was my grandfather's compassion" was positioned earlier than the statement part, the more past It may be placed incorrectly at the time of day.

The present invention has been made in view of the above-mentioned conventional example, and provides a sentence processing method and device for rearranging sentences correctly in the order of the flow of time and clarifying temporal connection of each sentence. With the goal.

[0008]

In order to achieve the above object, a text processing method and apparatus of the present invention have the following configurations. That is, a morpheme decomposing step of decomposing a sentence into morphemes and a dictionary that stores morphemes indicating a statement are searched, and if a morpheme indicating a statement matching the decomposed morpheme is detected, a sentence including the morpheme is extracted. A statement sentence determination step of determining a statement sentence is provided, and the sentence is divided into a statement sentence and other sentences.

In another invention, an adverb indicating the author's statement or an auxiliary verb or verb indicating the author's statement is detected from the sentence and divided into a statement including the author's statement and an explanatory statement indicating other matters. A classifying step, a step of creating two time series of a link expressing the time series of the statement and a link expressing the time series of the explanatory sentence, and a link for associating matters between the two time series. And a step of performing.

Another aspect of the invention is to search a morpheme decomposing means for decomposing a sentence into morphemes and a dictionary storing morphemes indicating a statement, and detect a morpheme indicating a statement matching the decomposed morpheme. For example, the statement including the morpheme is a statement statement determining unit that determines the statement as a statement, and the statement is divided into statements and other statements.

[0011]

[Operation] With the above configuration, the sentence is decomposed into morphemes,
If a morpheme indicating a statement that matches the decomposed morpheme is detected by searching a dictionary that stores morphemes indicating the statement, a sentence including the morpheme is determined to be a statement sentence, and the sentence is defined as the statement sentence and it. Other sentences.

Another invention is to detect an adverb indicating an author's statement or an auxiliary verb or a verb indicating an author's statement from a sentence and divide the statement into a statement including the author's statement and an explanatory statement indicating other matters. , Creating two time series of a link representing the time series of the statement and a link representing the time series of the explanatory text,
A link that correlates matters between the two time series is generated.

According to another aspect of the invention, the morpheme decomposition unit decomposes the sentence into morphemes, and the statement sentence determination unit searches the dictionary storing the morphemes indicating the statement, and makes a statement matching the decomposed morpheme. If the morpheme indicating is detected, the sentence including the morpheme is determined to be a statement sentence, and the sentence is divided into a statement sentence and other sentences.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS First, after summarizing one of the points of a text processing method and apparatus of an embodiment according to the present invention, a detailed description will be given. In the sentence processing method of the present embodiment, a Japanese sentence is analyzed to detect adverbs, auxiliary verbs, verbs, and the like indicating the author's statement, and a statement including the author's statement and an explanatory statement indicating other matters are formed. Divide. Then, a link expressing the time-series order of each sentence describing the author's statement, a link expressing the time-series order of each explanatory sentence, and a link expressing the time-series order of the description part of each matter are generated, Furthermore, by generating a link that expresses a time-series order between a statement and an explanation and each matter, it is possible to correctly grasp the time flow relationship of each sentence included in the sentence.

The present embodiment will be described below in detail with reference to the drawings. In order to facilitate the reader's understanding, the processing method in the present embodiment will be described with reference to the example sentence (FIG. 1) referred to in the description of the related art again. FIG. 2 is a system block diagram of the text processing apparatus of this embodiment. Reference numeral 1 denotes a system bus, and each constituent block to be described below is connected to the system bus 1.

2 is a CPU (Central Processing Uni
t), and executes the text processing program stored in advance in the ROM 3 (Read Only Memory) or the hard disk 11 or the floppy disk 12. 4 is RAM
In (Random Access Memory), input sentence data to be analyzed, statement sentence data accumulated during analysis, explanation sentence data, and the like are stored.

An input control unit 5 is connected with a keyboard 6 and an input device such as a pointing input device 7. A user operates the keyboard 6 to input operation instructions of the analysis program and text data.
A display output control unit 8 is connected to the CRT 9. An external storage device control unit 10 is connected with a hard disk 11, a floppy disk 12, and the like.

The hard disk 11 stores a text processing program and an analysis dictionary. The text to be analyzed may be stored in the hard disk 11 or the floppy disk 12. Next, FIG. 3 is a block diagram showing an outline of the document analysis processing structure in the sentence processing program according to the embodiment of the present invention.

Here, 201 is an analysis dictionary, which is stored in the hard disk 11 in FIG. 2 in advance. A text input unit 202 stores a text input using the keyboard 6 or the pointing input device 7 in the RAM 4, the hard disk 11, or the floppy disk 12.

Reference numeral 203 denotes a sentence analysis unit, which is a central processing unit of the sentence processing of this embodiment. A statement sentence storage unit 204 stores the sentence analysis processing result of the sentence analysis unit 203 and the sentence determined to be a statement sentence. Here, the statement storage unit 204
Are assigned to the RAM 4. Reference numeral 205 denotes a time relation analysis unit, which is a statement stored in the statement storage unit 204, an explanatory text stored in the explanatory text storage unit 206 described below, and a detailed description that will be described later. The temporal dependence of matters is determined by referring to the analysis dictionary 201.
An example of the analysis dictionary 201 is shown in FIG.

Reference numeral 206 denotes an explanation sentence storage unit, which is a sentence analysis unit 2.
As a result of the sentence analysis processing in 03, the sentence determined to be a statement is stored. Here, the explanatory note storage unit 204 is assigned to the RAM 4. 207 is an analysis result output unit,
The time relation of each sentence analyzed by the time relation analysis unit 205 and each related sentence are displayed.

Next, FIG. 4 is a flow chart showing the procedure of the Japanese sentence analysis process of this embodiment which embodies the concept of the sentence process of FIG. Further, FIG. 5 shows an example of the dictionary data of the auxiliary verb indicating the author's statement, the adverb indicating the author's statement, and the verb indicating the author's judgment stored in the analysis dictionary 11. 4 will be described below.

In step S301, the input of each sentence is accepted. If there is an input, the process proceeds to step S302. If not, the process ends. In step S302, the input sentence is analyzed. Specifically, first, the input sentence is morphologically analyzed and decomposed into morphemes. Next, the dependency analysis is performed to collect the phrases, and the dependency destination is determined. And
The semantic analysis process is performed to analyze the deep structure of the sentence. As these analysis results, a morpheme that is a constituent element of a sentence, a syntax tree created by a phrase that combines morphemes, and a semantic network that expresses the deep structure of the sentence are generated.

In step S303, it is determined whether or not the analysis result of the sentence (sentence network of the sentence) is a statement sentence. This judgment is made by searching the dictionary data shown in FIG. 5 for the “auxiliary verb dictionary indicating the author's statement” and the “adverb dictionary indicating the author's statement” and whether there is a match in the sentence. Determine whether If there is a match, the sentence is determined to be a statement and step S3 is performed.
Go to 04.

If there is a match with the "verb dictionary indicating the author's judgment" in the dictionary data shown in FIG. 5, it is presumed that the sentence end is the non-past tense and the judgment subject is the author. Even if it is done, it is determined to be a statement and the process proceeds to step S304. Here, the judgment as to whether the judgment subject is the author is made in FIG.
In the sentence, the subject corresponding to one of the verbs in the "verb dictionary indicating the author's judgment" is omitted in the sentence: 1) A word indicating the first person such as "I", "I", "I", etc. 3) If the word indicates an expression subject such as “author” or “author”, the judgment subject is judged to be the author.

For example, in the underlined part of Example 2 in FIG. 1: “There is no doubt that it is an act of compassion as a fisherman”, there is an auxiliary verb that indicates the author's statement “There is no mistake”. Is judged as a statement.

In addition, the underlined part of Example 3 in FIG. 1: "I didn't know at that time, it must have been my grandfather's compassion." In that, there is an auxiliary verb indicating the author's statement, "no wonder". , Similarly, this sentence is determined as a statement.

In step S304, the statement determined to be a statement is added to the statement already determined to be a statement. Here, with reference to FIG. 6, a data structure generated by the addition of the statement sentence, a data structure generated by the addition of the later-described description sentence, and a data structure showing the relationship between them will be described. The left side of FIG. 6 shows how each statement is added in a list structure from the top to the bottom of FIG. Also,
The right side of FIG. 6 shows how each description is added in a list structure from the top to the bottom of FIG. Further, in the center of FIG. 6, as will be described later, the matters extracted from the statement are linked in chronological order, and a part of an ellipse of a dot-and-dash line that surrounds the statement is enclosed in a frame enclosing the statement containing these matters. Means that those things are included in the statement. To realize this relationship on a computer, for example, a preceding pointer may be attached to the corresponding statement from those matters. Also, solid arrows from the explanatory note to the matter extracted from the central statement indicate the correspondence of the temporal relationship between the explanatory note and the matter pointed to from the explanatory note. This data structure is RAM
(4) (FIG. 2). Statement storage (204)
Then, the list structure of the statement is generated, and the list structure of the statement is generated in the description storage unit (205). Also,
The matter extracted from the statement and the link structure for the statement and the description are also generated in the RAM (4).

In step S305, the matter in the statement is retrieved. Specifically, searching for "auxiliary verb dictionary showing author's statement", "adverb dictionary showing author's statement", and "verb dictionary showing author's statement", and searching Judge things as things. For example, FIG.
In the underlined part of Example 2 of "the action taken out of compassion as a hunter", and in the underlined part of Example 3 of Fig. 1 "I did not understand at that time", "It was the compassion of my grandfather." Is extracted as a matter.

In step S306, the extracted matters are associated with the explanatory texts stored in the explanatory text storage unit (205) in time series. That is, in the extracted matter, whether there is an adverb existing in the "adverb dictionary showing simultaneity" in FIG. 5, or if it is a past tense sentence with no morpheme that limits time, the matter is Corresponding to the time series of the explanatory text as if it occurred at the same time as the immediately preceding explanatory text.
In the case of the past tense with adverbs indicating the times before and after, the time series anteroposterior relationship of the explanatory text is examined and associated with an appropriate place. In the case of non-past, it is not associated with the explanation. Then, the process returns to step S301, and the same processing for the next sentence is repeated.

As described above, FIG. 6 shows how this correspondence is made. The above-mentioned adverbs indicating simultaneity and adverbs indicating times before and after indicate a time relationship with the matter described in the preceding sentence. "Adverb dictionary showing simultaneity"
5 shows examples of "adverb dictionary showing relative past" and "adverb dictionary showing relative future". For example, in the case of “She started crying. It would have been very painful at that time.” There is an adverb “then”, so there are two things, “she started crying” and “it was painful”. It can be estimated that the times are almost the same.

In addition, in the case of "She started crying. It must have been so painful.""Suffering" does not have an adverb indicating time and is in the past tense. It can be inferred that the times of the two things, "I had a hard time" were almost the same.

Furthermore, in the case of "She began to cry. It must have been so painful until then." There is an adverb "until then" that indicates the temporal past. I started crying. "It can be presumed that this was before time.

Furthermore, in the case of "She started crying. It would have been easier from then on." There is an adverb "after that" which indicates the temporal future, so "easy" means "she I started crying. "It can be presumed to be later in time.

Furthermore, in the case of "She started crying. You must think that she should cry", "I think that she should cry" is a non-past sentence because the time is limited. "She started crying."
, And "I want to cry" are not associated with each other. In other words, "I want to cry" cannot be limited to "she started crying."

In the example of FIG. 6, "I didn't know at that time" has an adverb "at that time" that indicates simultaneity, and "it was my grandfather's compassion" is a past tense sentence with no time limit. . These two things happened at the same time as the last sentence, "My grandfather pointed his hurt gun to his mother, and triggered it,""As I approached, the fawn went into the forest. It has disappeared. "

Next, step S30 which is a processing step when it is determined that the statement is not a statement in step S303.
7 and subsequent explanations. In step S307, a sentence that is determined not to be a statement sentence, that is, an explanatory sentence is added to the sentence that has already been determined to be an explanatory sentence (stored in the explanatory sentence accumulating unit 205). In step S308, the added explanatory note is rearranged in chronological order if necessary, and the process returns to step S301.

Whether or not the rearrangement is necessary is determined as follows. That is, the "adverb dictionary showing the relative past" in FIG.
Or when there is an adverb indicating before and after the time as registered in "Adverb dictionary showing relative future", or when before and after being registered in "connective particle dictionary indicating retrospective" If there is a compound sentence etc. connected by a connective particle, rearrange the matter considering the context.

[0039] For example, in the case of "Field is muddy. It rained the day before", the order of appearance is that "Field is muddy" before "It rained the day before". Has been placed,
Here, since there is an adverb that indicates the relative past, "the day before," it is determined that rearrangement is necessary.

Then, as a time series, the rearrangement is performed so that "it rained on the previous day" comes before "the field is muddy." Also, in the case of "I called you before leaving home.", "Leaving home" was before "Called" in the order of appearance, but there was a retrospective connection particle "before". So I rearrange it so that "I called you" comes before "I left home."

In the example of the sentence of FIG. 6, it is not necessary to rearrange the explanatory sentence and the time series. As described above, according to the sentence analysis method of the present embodiment, the statement describing the author's statement and the explanatory note are arranged in chronological order, and the matters in the statement are arranged in chronological order to make the statement. By generating the temporal and content dependency relationships between sentences, explanations, and matters, more plausible time-series arrangements of sentences can be obtained. (Other Examples) 1. In the above embodiment, the time series of statements or explanatory sentences is updated for each statement, but the analysis results of multiple statements are accumulated for each statement or explanatory statement in the order of appearance, and when the input of the statement is completed. Alternatively, the time series creation processing may be collectively performed. 2. In addition, instead of accumulating the analysis result of a sentence into a statement sentence and an explanatory sentence separately, accumulate the analysis results in the order of appearance, and create only the time series separately for the statement sentence and the explanation sentence. You may 3. The information of the word for judging the statement may be held as the data of the word in the analysis dictionary instead of being held by the processing system at the time of judging the statement. 4. If the target sentence is a sentence in which the first person is the main character, the judgment criteria that it is a statement is strict, and it is a statement only when an adverb or auxiliary verb indicating the author's statement appears in the non-past tense. You may do it. 5. Alternatively, the correspondences may be cut off, and the time series of the statement and the explanatory text may be extracted as separate time series and treated as separate sentences. For example, it can be used for the task of dividing an animation scenario into two scenarios of character movement and narration.

When this embodiment is applied to the narration system, the position where the extracted statement is present is used as the narration part to give the display timing of a predetermined image of the animation, or the image is switched at the position where the statement is present. Can be slowed down to enhance the impression of narration. Further, when this embodiment is applied to an interactive Japanese language learning system that teaches Japanese to people, a basic classification of a predetermined sentence, that is, each sentence is a statement sentence or an explanatory sentence, and further, a matter in the statement sentence. It is possible to display to the learner whether or not there is, and by presenting the temporal flow of sentences, it is possible for the learner to easily learn Japanese.

The present invention may be applied to a system composed of a plurality of devices or an apparatus composed of a single device. Needless to say, the present invention can be applied to a case where the present invention is achieved by supplying a program to a system or an apparatus. As described above, according to the present embodiment, two time series of a link expressing a time series describing an author's statement and a link expressing a time series describing a matter are created, and two time series of By creating an appropriate correspondence, a more plausible time series arrangement can be obtained.

[0044]

As described above, according to the present invention, it is possible to correctly rearrange sentences in the order of the flow of time and clarify the temporal connection relation of each sentence.

[Brief description of drawings]

FIG. 1 is a diagram showing an example of a sentence processed by a sentence processing apparatus of this embodiment.

FIG. 2 is a block diagram showing a configuration of a text processing device of this embodiment.

FIG. 3 is a diagram showing a rough processing flow of a text processing method according to the present embodiment and a relationship between the processings.

FIG. 4 is a flowchart showing a procedure of Japanese sentence processing of the present embodiment.

FIG. 5 is a diagram showing contents of various dictionaries searched by the Japanese sentence processing of the present embodiment.

FIG. 6 is a diagram showing a data structure which is generated by the Japanese sentence processing of the present embodiment and which shows a temporal relation and a content relation of a statement, an explanation, and a matter of the statement.

[Simple code description]

 1 Internal Bus 2 CPU 3 ROM 4 RAM 5 Input Control Unit 6 Keyboard 7 Pointing Device 8 Display Output Control Unit 9 CRT 10 External Storage Device Control Unit 11 Hard Disk 12 Floppy Disk

Claims (14)

[Claims] 1. A morpheme decomposing step of decomposing a sentence into morphemes, and a dictionary storing morphemes indicating a statement are searched, and if a morpheme indicating a statement matching the decomposed morpheme is detected, the morpheme is extracted. A statement processing method, comprising: a statement statement determining step of determining a statement including the statement statement, wherein the statement statement is divided into a statement statement and other statements. 2. The sentence processing method according to claim 1, wherein the dictionary storing morphemes indicating the statement includes a dictionary storing auxiliary verbs indicating the statement of the author. 3. The sentence processing method according to claim 1, wherein the dictionary storing a morpheme indicating the statement includes a dictionary storing an adverb indicating the statement of the author. 4. The statement deciding step further searches a dictionary storing a verb indicating the author's decision, a verb matching the decomposed morpheme is detected, and a sentence including the matching morpheme is detected. The sentence processing method according to claim 1, wherein the sentence is determined to be a statement sentence if the end of the sentence is a non-past tense. 5. The explanatory sentence determination step of determining a sentence other than the sentence determined to be a statement sentence in the statement sentence determination step as an explanatory sentence, further comprising: an explanatory sentence determination step. Sentence processing method described in. 6. The statement determination step of determining a matter other than a morpheme determined to be a morpheme in a statement determined to be a statement in the statement determination step, further comprising a matter determination step. The sentence processing method described in any one of the above. 7. The method further comprises: storing sentences including morphemes indicating the statement determined in the statement determination step in the order of appearance, and storing explanatory sentences determined in the description statement determination step in the order of appearance. The sentence processing method according to any one of claims 1 to 4. 8. The matter determined in the matter determining step,
The sentence processing method according to claim 7, further comprising a step of storing in an appearance order. 9. If an adverb indicating simultaneity exists in the stored matter, it is determined that it has occurred at the same time as the explanation sentence immediately before the sentence including the matter, and the correspondence with the explanation sentence. The sentence processing method according to claim 8, further comprising: adding. 10. If the sentence containing the stored matter is a past tense sentence that has no time-limiting morpheme, it is determined that it occurred at the same time as the explanation sentence immediately before the sentence containing the matter. ,
The sentence processing method according to claim 8, further comprising associating with the description. 11. The sentence processing according to claim 7, wherein when the stored description includes an adverb indicating the time before and after, the text is further rearranged with the description before and after the description. Method. 12. When the stored explanation sentence is a compound sentence connected by auxiliary verbs indicating the time before and after, the explanation sentence before and after the explanation sentence is further rearranged. The sentence processing method described in 7. 13. A classification step of detecting an adverb indicating an author's statement or an auxiliary verb or verb indicating an author's statement from a sentence and dividing the statement into a statement including the author's statement and an explanatory statement indicating other matters. A step of creating two time series of a link expressing a time series of a statement, a link expressing a time series of an explanatory sentence, and a step of generating a link associating matters between the two time series. A text processing method characterized by that. 14. A morpheme decomposing means for decomposing a sentence into morphemes, and a dictionary storing a morpheme showing a statement are searched, and if a morpheme showing a statement matching the decomposed morpheme is detected, the morpheme is extracted. A sentence processing apparatus, comprising: a statement sentence determination unit that determines a sentence that includes a statement sentence, and that divides the sentence into a statement sentence and other sentences.