JP2005025257A - Document creation support processing program utilizing paraphrase, recording medium for this program, and document creation support processing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a processing program that supports document creation by means of sentence paraphrasing. <P>SOLUTION: An input/output processing means 11 acquires a sentence 2 to be operated from the original sentence data. A sentence analysis processing means 12 analyzes the internal structure and meaning of the sentence 2 using an indeclinable word dictionary 21 and a declinable word dictionary 22, and determines the propositional type or the inter-propositional type of the sentence 2 by referring to a proposition dictionary 23 and a propositional juncture dictionary 24. A paraphrase processing means 13 paraphrases the sentence 2 by referring to a paraphrase dictionary 25 and creates a group of paraphrases. A context matching processing means 14 performs the processing of matching the context of model sentence data in a template dictionary 26 with the context of case sentence data resulting from the paraphrasing of a predetermined range of document data including the sentence 2 with the group of paraphrases, and outputs a similar paraphrase as a paraphrase candidate. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a natural language processing technique using a computer. More specifically, the present invention relates to a processing program that supports sentence creation using a paraphrase for converting a sentence described in a natural language into a sentence in another expression, a recording medium storing the processing program, and a processing method.
[0002]
[Prior art]
Paraphrasing is a process of converting a sentence written in one natural language into a sentence with another expression, and converts the sentence length, style, expression, etc. while retaining the core meaning of the sentence. Say things.
[0003]
Paraphrases can be roughly classified into syntactic paraphrases, semantic paraphrases, and pragmatic paraphrases (see Non-Patent Document 1).
[0004]
Syntactic paraphrasing refers to a process of changing the expression of a sentence by replacing a word or phrase appearing in the sentence with a synonym or a synonym, a process of transforming a sentence based on a sentence structure mapping, and the like. For example, the original expression “generate system” is replaced with <create> and transformed into a new expression “create system”.
[0005]
Semantic paraphrasing refers to a process of transforming a sentence by referring to the context or situation in which the sentence is spoken, such as a process of changing the expression of a sentence by replacing a reference expression with its instruction content. For example, when the original expression “event of last year” was uttered in 1999, it is analyzed that <last year> means <1998] and transformed into a new expression “event of 1998”. is there.
[0006]
The pragmatic paraphrasing is a process of performing more complicated paraphrasing other than the above two types of paraphrasing and transforming into an expression having the same effect in a certain situation. For example, in the original expression "Do you know if the source of gcc exists?"<Someone-don't know? > From the semantic analysis of <where is the source of gcc>, etc., this expression is analyzed as a context for inquiring about the location of the object, and a new expression "tell the ftp site where the source of gcc is located Please transform to "Please".
[0007]
Here, as an example of using the sentence editing function of the sentence creation processing software executed by a computer or word processor, the following two sentences are exchanged in order.
[0008]
“Taro did not walk in the park because it was raining.”
The user first moves the preamble “Taro did not walk in the park” to the tail of “because it was raining” by a cut-and-paste operation. The two sentences are as follows:
[0009]
“Because it was raining. Taro did not walk in the park.”
However, the expression of the two sentences after rearrangement is unnatural as a natural language (here, Japanese), and the expression of the sentence needs to be changed. Therefore, the user deletes unnecessary characters and phrases and inserts necessary characters and phrases by operating a character key (hereinafter referred to as a key type). After the key type, the two sentences are as follows:
[0010]
“<> It was raining. <So> Taro didn't walk in the park.” (<> Is the change by key type)
Furthermore, a paraphrase that combines two rearranged sentences into one sentence is assumed. Similarly, the user deletes / inserts a phrase by key type. After the key type, the two sentences are as follows:
[0011]
“Because it was raining, Taro did not walk in the park.”
As described above, in the paraphrasing of sentences by rearranging and combining sentences, key types such as deletion and insertion of words are often required so that the expression of sentences is natural.
[0012]
[Non-Patent Document 1]
Kentaro Inui, “Technique to rephrase language expression”,
Proc. Of the 8th Annual Conference of the Association for Natural Language Processing, Proc.
March 2002, pp. 1-21
[0013]
[Problems to be solved by the invention]
Existing text data is often used when creating text using text creation software. In this case, when a sentence in existing sentence data is used by being transformed by combining, dividing, etc., conventionally, the user has performed key types such as deletion or insertion of words and phrases so that the expression of the sentence is natural. The sentence must be transformed so that the sentence to be used matches the surrounding context.
[0014]
However, it is cumbersome to perform a key type to change the expression of a sentence. In particular, in a computer or tablet computer in which a mouse operation, which has started to spread rapidly in recent years, is the center of input operation, character input by key type is rather complicated.
[0015]
Therefore, there is a strong demand for the implementation of a mechanism that supports creating a sentence without using a key type as much as possible in a computer whose pointing function is the center of input operation.
[0016]
Also, if you simply use syntactic or semantic paraphrasing as in the past, the sentence you want to paraphrase may be transformed into an unnatural expression that does not match the context in which it was spoken, and actual sentence creation support Could not be used.
[0017]
An object of the present invention is to enable text creation support by a simple operation without performing a key type in text creation support using sentence paraphrasing.
[0018]
Another object of the present invention is to provide highly useful sentence creation support by performing natural paraphrasing in consideration of sentence context in sentence creation support using sentence paraphrasing. It is.
[0019]
[Means for Solving the Problems]
In order to solve the above problems, the present invention is a program for causing a computer to execute a sentence creation support process using a paraphrase process that converts sentence data described in a natural language into sentence data in another expression. The following processing is performed.
[0020]
In the present invention, a proposition dictionary describing a proposition type that defines an action or state described by sentence data, a propositional association dictionary describing a relationship between a clue expression in the sentence and a type between propositions indicating a correspondence relationship between sentences, A paraphrase dictionary describing a paraphrase pattern based on dependency relations or clue expressions, or a template dictionary that stores sentence data that is recognized as having a good sentence structure and to which the proposition type and interproposition type are assigned. Is obtained by accessing the data storage means for storing. Also, sentence operation information indicating the paragraph operation target part such as a clause, sentence, or sentence to be subjected to the paraphrase operation and the type of the paraphrase operation is acquired from the text data displayed to the user.
[0021]
And analyzing the internal structure and meaning of the operation target part, identifying the proposition type of the operation target part, or the proposition type and inter-proposition type from the analysis result with reference to the proposition dictionary or the proposition concatenation dictionary, In accordance with the sentence operation information, the paraphrasing pattern is generated by referring to the paraphrase pattern and paraphrasing the operation target portion.
[0022]
Thereafter, the model sentence data including the propositional type of the operation target part is extracted from the template dictionary, and the text data in a predetermined range including the context of the model sentence data and the operation target part is extracted from the paraphrase group. The context of the example sentence data paraphrased by paraphrase is compared to evaluate the similarity between the sentence structure of the model sentence data and the sentence structure of the case sentence data, and the paraphrase obtained with a predetermined evaluation is output as a paraphrase candidate.
[0023]
In the present invention, when the above processing is performed, an arbitrary section or sentence of the sentence data displayed to the user or a predetermined selection item is selected by using the input pointing function.
[0024]
Note that when a desired candidate is selected from the paraphrase candidates output to the display device or the like by the user, the present invention sends the selected paraphrase candidate to the corresponding part of the original sentence data with respect to the predetermined sentence creation processing means. As input data for.
[0025]
According to the present invention, a user can select an operation target part (character string) to be used in other words from existing sentence data by using a pointing function provided in input means such as a mouse, a tablet pen, and an input panel. Furthermore, a desired candidate can be selected from the paraphrase candidates.
[0026]
Since the selected paraphrase candidate is used as an input sentence (input character string) in a predetermined sentence creation process, the user can acquire an input sentence without performing a key type, and can easily create a sentence. Become.
[0027]
In addition, if there is some text data that can be read by the computer, it is possible to create new text data by using the text data, and it is possible to improve the efficiency of the text creation work.
[0028]
Furthermore, the present invention performs a paraphrase process on the operation target part (character string), and then considers the context of the text to be paraphrased from the paraphrase group generated by the paraphrase process. Narrow down evaluations as candidates for paraphrasing. Thus, the user can select a desired paraphrase candidate from those narrowed down to appropriate paraphrase candidates, and can realize practical sentence creation support.
[0029]
The processing program according to the present invention can be stored in an appropriate recording medium such as a portable medium memory, a semiconductor memory, or a hard disk, which can be read by a computer, and is provided by being recorded on these recording media. These are provided by transmission / reception using various communication networks via a communication interface.
[0030]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described below.
[0031]
FIG. 1 is a diagram illustrating a configuration example of the present invention in one embodiment.
[0032]
The sentence creation support device 1 includes an input / output processing unit 11, a sentence analysis processing unit 12, a paraphrase processing unit 13, a context matching processing unit 14, an input / output unit 17, and a storage unit 20. 21, data such as a predicate dictionary 22, a proposition dictionary 23, a proposition concatenation dictionary 24, a paraphrase dictionary 25, and a template dictionary 26 are stored.
[0033]
The body dictionary 21 is data that describes semantic information and syntactic information such as predicates that are likely to co-occur as parts of speech and terms for nouns and pronouns that are generally classified as body words. The prescriptive dictionary 22 is data that describes semantic information and syntactic information such as part-of-speech, inflection forms, particles that are easy to co-occur, and predicates that are easy to co-occur as terms for verbs and adjectives. . The body language dictionary 21 and the word dictionary 22 are configured to be able to refer to each other. The body language dictionary 21 and the word dictionary 22 are realized by existing Japanese dictionary data for computers.
[0034]
The proposition dictionary 23 is data describing the relationship between sentence end expressions, sentence structures, and sentence proposition types. A proposition type is classification label information that defines an action or a state described by a sentence by a combination of tense, modality, aspect, etc. appearing in the sentence. As examples of proposition types, themes, phenomena, facts, judgments, etc. are defined.
[0035]
In the case of taking the simplest structure, the proposition dictionary 23 is composed of data in which a predetermined proposition type is labeled on the sentence end expression. In general, when verbs, adjectives, etc. are used at the end of a sentence or compound sentences, various information such as tense, modality, aspect, etc. is displayed in these sentences. The proposition type is defined in
[0036]
"Expression" [proposition type],
“Must do” [duty, objective],
"Maybe" [forecast] ...
[0037]
Alternatively, the proposition dictionary 23 includes data in which a predetermined proposition type is labeled on a sentence described by a combination of a plurality of terms and predicates.
[0038]
Generally, “Taro walked in the park.” Is described in the form <term, term,…, predicate> such as <Taro walks in the park>. A predetermined proposition type is assigned based on the tense, modality, and aspect displayed on the screen.
[0039]
<Taro took a walk in the park> [The past of factual action]
<Taro wanted to take a walk in the park> [Aspirational Past]
[0040]
Furthermore, in the case of a compound sentence such as “Taro wanted to take a walk in the park but it was raining outside,” the following paragraph (the antecedent) and the latter section (the latter): Each proposition type is labeled.
[0041]
<Taro wanted to take a walk in the park> [2 duplicate sentence antecedents-desire past];
<Outside, it was raining> [Two duplicate sentence consequents-natural phenomenon past description].
[0042]
The propositional connection dictionary 24 is data in which a relationship between clue expressions such as dependency and connection and a predetermined propositional type is described. The interproposition type is classification label information that defines a connection relationship between a plurality of propositions by clue expression or the like. As examples of the inter-proposition type, causality, parallelism, time passage, refinement, and illustration are defined.
[0043]
For example, in the case of the sentence “Taro wanted to walk in the park but it was raining outside”, the internal structure is as follows.
[0044]
<Taro wanted to walk in the park><but><outside it was raining>
Since the conjunctive word <is> between the previous section (predecessor) and the subsequent section (subsequent) is a clue expression, data to which the following interproposition type is assigned is stored in the propositional connection dictionary 24.
[0045]
<Taro wanted to take a walk in the park> [2 duplicate sentence antecedents-desire past];
<G> [Proposition type = paradox];
<Outside, it was raining> [Two duplicate sentence consequents-natural phenomenon past description]
The paraphrase dictionary 25 is data describing expression patterns that can be paraphrased on the basis of a dependency relationship or the like.
[0046]
The template dictionary 26 is text data (model text) in a predetermined unit that is recognized as having a good text structure, and is data to which information such as a proposition type and an inter-proposition type is given.
[0047]
Here, the criteria for determining whether or not the sentence structure is good differs depending on the kind and use of the sentence. Therefore, depending on the usage and type of text that the user wants to create, for example, it corresponds to each text type classified according to usage or type in advance, such as newspaper article template dictionary, academic paper template dictionary, legal text template dictionary, etc. A template dictionary 26 may be prepared, and the template dictionary 26 may be switched by the context matching processing unit 14.
[0048]
The input / output processing means 11 is means for performing data input / output processing with the input / output device 17. The input function of the input / output device 17 can be realized by a combination of a mouse, a tablet pen, and an input panel.
[0049]
The input / output processing unit 11 displays text data (original text) that is a selection source of character string data (operation target part) that is the target of the paraphrase operation on the input / output device 17, and the operation target part that is selected by the user And sentence operation information 3 for specifying the type of paraphrase operation.
[0050]
The part selected as the operation target part is a clause, a sentence, or a sentence. In this example, it is assumed that the part is selected in units of sentences, and the input / output processing unit 11 sets the operation target sentence 2 as the operation target part. get.
[0051]
In addition, as an example of the sentence operation information 3, the following operation is defined.
[0052]
Split: The process of splitting a sentence and generating multiple sentences,
Combining: Processing to combine multiple sentences and generate one sentence,
Delete: The process of deleting a specified part of a sentence,
State conversion: Processing to convert sentence state (active or passive),
Thematicization: A process of generating a sentence whose subject is a specified part of a sentence,...
[0053]
In addition, the input / output processing unit 11 displays one or more paraphrase candidates 4 that are evaluated that the paraphrase matches the context on the input / output device 17 and acquires the paraphrase candidate 4 selected by the user.
[0054]
In a predetermined sentence creation processing unit (not shown), the original sentence data (original sentence) from which the operation target sentence 2 is selected is rephrased by the paraphrase candidate 4 selected by the user, and is used as new sentence data.
[0055]
Note that the input / output device 17 in which the user selects the operation target sentence 2, the sentence operation information 3, and the paraphrase candidate 4 can select at least items displayed on the display screen and specify character string data. What is necessary is just to have a function.
[0056]
The sentence analysis processing means 12 refers to the body dictionary 21 and the predicate dictionary 22, analyzes the internal structure and meaning of the operation target sentence 2 using a predetermined analysis method, and operates the operation target sentence based on the proposition dictionary 23. This is means for identifying two proposition types and identifying the interproposition type of the operation target sentence 2 based on the propositional connection dictionary 24.
[0057]
The paraphrase processing unit 13 is a processing unit that performs a paraphrase process for the operation target sentence 2 corresponding to the sentence operation information 3 with reference to the paraphrase dictionary 25 and generates a paraphrase group. In the paraphrase process, syntactic or semantic paraphrasing is performed using syntactic information and semantic information, and many paraphrases are generated.
[0058]
The context matching processing unit 14 refers to the template dictionary 26 and compares the context of sentence data (example sentence) to be paraphrased for each paraphrase of the paraphrase group with the context of the corresponding model sentence in the template dictionary 26. This is a processing means for extracting the paraphrase used in the case sentence when the sentence is similar to a good context to some extent.
[0059]
FIG. 2 shows the flow of processing of the text creation support apparatus 1. First, the flow of processing will be described taking the case of sentence division as an example.
[0060]
Step S 10: The input / output processing means 11 displays an operation target selection screen for prompting selection of the operation target sentence 2 on the input / output device 17.
[0061]
FIG. 3 shows an example of the operation target selection screen 50. On the original text display area 51 for selecting the operation target sentence on the operation target selection screen 50, the operation target sentence 2 indicates that the sentence “Yes, yesterday's meeting was not profitable for both opinions and the discussion passed.” Choose as.
[0062]
Step S11: The sentence analysis processing means 12 refers to the body dictionary 21 and the predicate dictionary 22 to perform syntax analysis / semantic analysis of each element constituting the operation target sentence 2 to obtain sentence analysis information, semantic analysis information, and the like. get. The sentence analysis process and the semantic analysis process performed by the sentence analysis processing unit 12 are performed using an existing general processing method.
[0063]
Step S12: The sentence analysis processing means 12 extracts matching sentence end expressions and sentences from the proposition dictionary 23 based on the sentence analysis information and semantic analysis information of the operation target sentence 2, and is given to the sentence end expressions or sentences. Is identified as the proposition type of the operation target sentence 2.
[0064]
Step S13: The sentence analysis processing means 12 extracts a matching relation from the propositional connection dictionary 24 based on information indicating the dependency relation in the sentence analysis information of the operation target sentence 2, and determines the interproposition type. Is identified as the inter-proposition type of the operation target sentence 2.
[0065]
In the processing of step S11 to step S13, the sentence analysis processing means 12 acquires information such as sentence analysis information, semantic analysis information, proposition type, and interproposition type as shown in FIG. To do.
[0066]
“Preamble a:
<Yesterday's meeting> [Theme]
<Both opinions and profits do not meet each other>
Subsequent b:
<Discussions pass each other> [Past-objective observation]
Interproposition type: causal "
Step S14: Further, the input / output processing means 11 acquires the sentence operation information 3 from the sentence operation designation field 52 of the operation target selection screen 50. Here, since “sentence division” is selected in the sentence operation designation field 52, “division” is acquired as the sentence operation information 3.
[0067]
Step S15: The paraphrase processing unit 13 generates a paraphrase of the operation target sentence 2 corresponding to the sentence operation information 3 with reference to the paraphrase pattern of the paraphrase dictionary 25. Let the generated paraphrase be a paraphrase group.
[0068]
For example, it is assumed that the following paraphrase pattern is stored in the paraphrase dictionary 25 as shown in FIG.
[0069]
"... depending on ..."
"... So ..."
"... So, so ..."
"... No. As a result, I didn't ..."
[0070]
The paraphrase processing unit 13 uses these paraphrase patterns to generate various paraphrases of the operation target sentence 2 as described below to form a paraphrase group.
[0071]
“… The profits did n’t come together, so the discussion…”,
"... the profits did not match, so the discussion ...",
"... it didn't match the profit, so the argument ..."
[0072]
Step S16: The context matching processing means 14 narrows down, from the paraphrase group, text data (example sentence) to be paraphrased having an appropriate context from the paraphrase group by context matching with reference to the template dictionary 26.
[0073]
The context matching processing means 14 refers to the template dictionary 26 and extracts an exemplary sentence including the proposition of the operation target sentence 2. Further, each paraphrase is taken out from the paraphrase group, and a predetermined range (for example, a paragraph) of the original sentence data from which the operation target sentence 2 is selected is cut out as example sentence data (example sentence). Paraphrasing using the paraphrase obtained by extracting the case sentence, the syntactic information and the semantic information are acquired, and further, the proposition type and the interproposition type are acquired.
[0074]
Then, matching processing is performed on the sequence of proposition types in the case sentences and the sequence of proposition types in the extracted model sentences, and how similar the arrangement of the proposition types and the arrangement of language expressions in the entire predetermined range of both are. Evaluate what to do. Extract the paraphrases of the case sentences that have obtained a predetermined evaluation.
[0075]
The similarity evaluation is performed for all paraphrases in the paraphrase group, and the extracted paraphrases are used as paraphrase candidates. It should be noted that a predetermined number of paraphrases with good evaluation may be selected from the extracted paraphrases and used as paraphrase candidates.
[0076]
As shown in FIG. 4C, since the interproposition type of the operation target sentence 2 is “causal”, an exemplary sentence that includes the interproposition type = causal is extracted from the template dictionary 26. For example, it is assumed that the template dictionary 26 has an exemplary sentence including the following sentence.
[0077]
Causal: “The negotiations were over, as both sides did not agree.”
The degree of similarity between the arrangement of the propositional types of the model sentences and the arrangement of the propositional types of the example sentences is evaluated, and the paraphrase evaluated as having a predetermined similarity is set as a paraphrase candidate c (c1, c2,...).
[0078]
Thereafter, the input / output processing means 11 displays the paraphrase candidate display screen 60 on the input / output device 17.
[0079]
FIG. 5 shows an example of the paraphrase candidate display screen 60. The following paraphrase candidates c1 and c2 are displayed in the paraphrase candidate display area 61 of the paraphrase candidate display screen 60.
[0080]
c1: “Yes, yesterday's meeting was not profitable for both sides.
c2: “Yes, yesterday's meeting was not profitable for both parties. As a result, the discussions passed.”
When the user selects a desired paraphrase candidate from the paraphrase candidate display area 61 by using the pointing function of the input / output device 17, the sentence data in which the corresponding part of the original sentence data is paraphrased by the selected paraphrase candidate c is newly added. Displayed as text data in the text creation field 62.
[0081]
Next, the flow of processing when two sentences are combined and rephrased as one sentence will be described with reference to FIGS.
[0082]
As shown in FIG. 6, the user selects the following two operation target sentences 2 on the original sentence display area 51 of the operation target selection screen 50.
[0083]
Sentence 2a: “A soldier fired at a checkpoint on a wagon car that did not follow the stop order on the 31st.”
Sentence 2b: “Therefore, seven of the 13 women and children in the car died.”
The sentence analysis processing means 12 performs processing such as sentence / sentence analysis and semantic analysis for the selected operation target sentence 2 (sentence 2a and sentence 2b) with reference to the body dictionary 21 and the predicate dictionary 22. Furthermore, the proposition type and the interproposition type are identified based on sentence analysis information and semantic analysis information.
[0084]
For example, analysis information, a proposition type, and an interproposition type as shown in FIG.
[0085]
Sentence 2a: <At the checkpoint, soldiers shoot at a wagon car that does not follow the stop order on the 31st. > [Past-Facts]
Sentence 2b: <Therefore, 13 women, children and 7 of them who are in the car die. > [Past-Facts]
[Proposition type = factual context]
Further, from the “sentence combination” selected in the sentence operation designation field 52 of the operation target selection screen 50 and the user's mouse operation or tablet pen operation, as shown in FIG. And “direction of coupling = sentence 2a → sentence 2b”.
[0086]
Then, the paraphrase processing unit 13 refers to the paraphrase pattern in the paraphrase dictionary 25 and generates a paraphrase corresponding to the sentence combination. Further, the context matching processing means 14, similarly to the processing in the “sentence division” example, arranges the proposition types of the model sentences in the template dictionary 26 and the proposition types of the case sentences using the paraphrasing of the paraphrasing group. Similarity evaluation is performed. When it is evaluated that the arrangement of both proposition types over a predetermined range is similar to a certain level, the paraphrase is set as a paraphrase candidate as shown in FIG.
[0087]
c1: “On March 31, at the checkpoint, a soldier fired at a wagon car that did not obey the stop order, and seven of 13 women and children in the wagon car died.”, c2: “31 On the day, at the checkpoint, seven of the 13 women and children in a car who did not follow the stop order were killed by soldiers firing. "
c3: “On March 31, at the checkpoint, a soldier fired and seven of 13 women and children in a car who did not obey the stop order died.”
[0088]
As shown in FIG. 8, the input / output processing unit 11 displays the extracted paraphrase candidates in the paraphrase candidate display area 61 of the paraphrase candidate display screen 60.
[0089]
Next, with reference to FIG. 9, the flow of processing in the case of passivation among the paraphrasing operations for changing the sentence format will be described.
[0090]
The selection of the operation target sentence 2 is the same as in the case of sentence division and sentence combination. Assume that the user selects an operation target sentence 2 as shown in FIG. 9A on the original text display area 51 of the operation target selection screen 50.
[0091]
“The crowd gathered in the square pulled down the big statue.”
The sentence analysis processing means 12 acquires sentence analysis information and semantic analysis information for the operation target sentence 2 and identifies a proposition type and the like, as in the case of sentence division processing.
[0092]
<The crowd gathered in the square, and the big statue was pulled down. > [Past-Facts]
Further, from the selection of “passivation” in the sentence operation designation field 52 on the same screen as the operation target selection screen 50 shown in FIG. Thus, sentence operation information = passivation is acquired.
[0093]
Thereafter, the paraphrase processing unit 13 refers to the paraphrase pattern in the paraphrase dictionary 25 and generates a paraphrase for passivation. Then, the context matching processing means 14 evaluates the similarity between the model sentence proposition type sequence and the example sentence proposition type sequence in the template dictionary 26 as in the case of the sentence division process, and FIG. As shown in FIG. 6, a paraphrase obtained with a predetermined evaluation is set as a paraphrase candidate.
[0094]
c1: “The big statue was pulled down by the crowd gathered in the square.” c2: “The big statue was pulled down by the crowd gathered in the square.”
[0095]
The input / output processing unit 11 displays the extracted paraphrase candidates in the paraphrase candidate display area 61 of the paraphrase candidate display screen 60.
[0096]
Next, with reference to FIG. 10, a processing flow in the case of thematicization in the paraphrase operation for changing the sentence format will be described.
[0097]
As in the case of sentence division and sentence combination, it is assumed that the user selects an operation target sentence 2 as shown in FIG. 10A on the original sentence display area 51 of the operation target selection screen 50.
[0098]
“At the checkpoint, seven out of 13 women and children in a car who did not follow the stop order died.”
The sentence analysis processing unit 12 acquires sentence analysis information, semantic analysis information, and the like of the operation target sentence 2 and identifies a proposition type and the like.
[0099]
<At the checkpoint, 13 women, children and 7 of them who died in the car, who did not follow the stop order, died. > [Past-Facts]
Furthermore, from the selection of “theme” in the sentence operation designation field 52 on the same screen as the operation target selection screen 50 shown in FIG. 6 and the user's mouse operation, tablet pen operation, and the like, FIG. Thus, sentence operation information = thematicization and “the subject of thematicization = <7 people died>” are acquired.
[0100]
The paraphrase processing unit 13 refers to the paraphrase pattern in the paraphrase dictionary 25 and generates a thematic paraphrase. Then, the context matching processing means 14 performs the similarity evaluation between the sequence of the propositional type of the model sentence and the sequence of the propositional type of the case sentence in the template dictionary 26 as in the case of the sentence division, and FIG. As shown in FIG. 4, a paraphrase obtained with a predetermined evaluation is set as a paraphrase candidate.
[0101]
c1: “7 deaths occurred at a checkpoint among 13 women and children in a car who did not follow the stop order.”
c2: “Seven deaths occurred at the checkpoint among the 13 women and children in the car who did not follow the stop order.”
[0102]
Next, the flow of processing for deleting a part of a sentence will be described with reference to FIG.
[0103]
As in the case of sentence division and sentence combination, it is assumed that the user selects an operation target sentence 2 as shown in FIG. 11A on the original sentence display area 51 of the operation target selection screen 50.
[0104]
“He entered school, but he quit immediately because the lecture was boring.” The sentence analysis processing means 12 obtains sentence analysis information, semantic analysis information, etc. for the operation target sentence 2, and determines the proposition type, between propositions Identify the type.
[0105]
<He entered school>[Past-Facts];
<But> = Paradox;
<Lecture is boring>[Past-Judgment];
<I immediately quit> [Past-Fact]
Furthermore, sentence operation information = deletion is acquired by selecting “delete” in the sentence operation designation field 52 of the operation target selection screen 50.
[0106]
The paraphrase processing unit 13 refers to the paraphrase pattern in the paraphrase dictionary 25 and generates a paraphrase for deletion. Then, the context matching processing means 14 performs similar evaluation between the sequence of the propositional type of the model sentence and the sequence of the propositional type of the case sentence in the template dictionary 26 as in the case of the sentence division, and FIG. As shown in FIG. 4, a paraphrase evaluated as similar is used as a paraphrase candidate.
[0107]
c1: “He entered school, but he quit soon.”
c2: “The school lecture was boring and he quit immediately.”
[0108]
As described above, in the present invention, the user simply selects sentence data to be paraphrased from the existing sentence data displayed on the operation target selection screen 50 by a mouse operation or a tablet pen operation, and the paraphrase operation is performed. An appropriate expression paraphrase candidate group that matches the context before and after the desired sentence can be presented. Thus, the user can change the expression of the text data without performing the key type, and can efficiently perform the text creation work.
[0109]
As mentioned above, although this invention was demonstrated by the embodiment, various deformation | transformation are possible for this invention in the range of the main point.
[0110]
In the embodiment of the present invention, when the sentence creation support apparatus 1 is realized as a processing program that is read and executed by a computer, the input / output processing means 11 is an operating system of the computer that executes the sentence creation support apparatus 1. It is possible to use the processing function provided by.
[0111]
【The invention's effect】
As described above, according to the present invention, the selection of the sentence to be rephrased and the type of operation to be rephrased and the selection of the desired candidate from the paraphrasing candidates are performed by designating the item to be selected on the display screen. The user does not need to perform a key type for paraphrase, and can create a sentence by using the paraphrased expression only by a simple item selection operation. Therefore, the present invention can reduce the operation burden in the text creation support process.
[0112]
Furthermore, according to the present invention, in the paraphrase process, in consideration of the context flow of the paragraph including the sentence to be paraphrased and presenting only the paraphrase candidates evaluated as having a good sentence structure in the context, the user can Since only natural paraphrase candidates are presented, it is possible to obtain a suitable paraphrase candidate and create a sentence. Therefore, the present invention can provide a more practical sentence creation support process.
[0113]
Further, the present invention can provide an efficient text creation support process using existing text data.
[Brief description of the drawings]
FIG. 1 is a diagram illustrating a configuration example of the present invention.
FIG. 2 is a diagram showing a flow of processing of a text creation support apparatus.
FIG. 3 is a diagram illustrating an example of an operation target selection screen.
FIG. 4 is a diagram for explaining processing in the case of sentence division;
FIG. 5 is a diagram illustrating an example of a paraphrase candidate display screen.
FIG. 6 is a diagram illustrating an example of an operation target selection screen.
FIG. 7 is a diagram for explaining processing in the case of sentence combination;
FIG. 8 is a diagram illustrating an example of a paraphrase candidate display screen.
FIG. 9 is a diagram for explaining processing in the case of passivation.
FIG. 10 is a diagram for explaining processing in the case of subjecting.
FIG. 11 is a diagram for explaining processing in the case of deletion;
[Explanation of symbols]
1 Writing support device
11 Input / output processing means
12 sentence analysis processing means
13 Paraphrase processing means
14 Context matching processing means
17 I / O devices
20 storage unit
21 vocabulary dictionary
22 Precautionary dictionary
23 Proposition Dictionary
24 Propositional Articulated Dictionary
25 Paraphrasing dictionary
26 Template dictionary

Claims (5)

A program for causing a computer to execute a sentence creation support process using a paraphrase process for converting sentence data described in a natural language into sentence data in another expression,
A propositional dictionary that describes the propositional type that defines the actions or states that the sentence data describes, a propositional connection dictionary that describes the relationship between the clue expression in the sentence and the type between the propositions that indicate the correspondence between sentences, a dependency relationship or Data for storing a paraphrase dictionary describing a paraphrase pattern based on clue expression, or a template dictionary for storing exemplary sentence data to which the propositional type and the interpropositional type are given, which is sentence data recognized as having a good sentence structure Processing to access the storage means and acquire data;
A process of acquiring sentence operation information indicating an operation target part to be subjected to a paraphrase operation and a type of paraphrase operation from sentence data displayed to the user;
Analyzing the internal structure and meaning of the operation target part, and referring to the proposition dictionary or the proposition concatenation dictionary to identify the proposition type of the operation target part, or the proposition type and inter-proposition type from the analysis result;
In accordance with the sentence operation information, referring to the paraphrase pattern, paraphrasing the operation target part and generating a paraphrase group;
By extracting paradigm text data including the propositional type of the operation target part from the template dictionary, the text data in a predetermined range including the context of the model text data and the operation target part, and by paraphrase extracted from the paraphrase group A process of comparing the context of the paraphrase example sentence data and evaluating the similarity between the sentence structure of the model sentence data and the sentence structure of the case sentence data, and outputting a paraphrase obtained as a predetermined evaluation as a paraphrase candidate; ,
A sentence creation support processing program using paraphrasing that is executed by a computer. In the process of acquiring the operation target part and the sentence operation information,
A sentence creation support processing program using paraphrasing, which is performed by selecting an arbitrary section or sentence of sentence data displayed to a user or a predetermined selection item by using an input pointing function. A recording medium recording a program for causing a computer to execute a sentence creation support process using a paraphrase process that converts sentence data described in a natural language into sentence data in another expression,
A propositional dictionary that describes the propositional type that defines the actions or states that the sentence data describes, a propositional connection dictionary that describes the relationship between the clue expression in the sentence and the type between the propositions that indicate the correspondence between sentences, a dependency relationship or Data for storing a paraphrase dictionary describing a paraphrase pattern based on clue expression, or a template dictionary for storing exemplary sentence data to which the propositional type and the interpropositional type are given, which is sentence data recognized as having a good sentence structure Processing to access the storage means and acquire data;
A process of acquiring sentence operation information indicating an operation target part to be subjected to a paraphrase operation and a type of paraphrase operation from sentence data displayed to the user;
Analyzing the internal structure and meaning of the operation target part, and referring to the proposition dictionary or the proposition concatenation dictionary to identify the proposition type of the operation target part, or the proposition type and inter-proposition type from the analysis result;
In accordance with the sentence operation information, referring to the paraphrase pattern, paraphrasing the operation target part and generating a paraphrase group;
By extracting paradigm text data including the propositional type of the operation target part from the template dictionary, the text data in a predetermined range including the context of the model text data and the operation target part, and by paraphrase extracted from the paraphrase group A process of comparing the context of the paraphrase example sentence data and evaluating the similarity between the sentence structure of the model sentence data and the sentence structure of the case sentence data, and outputting a paraphrase obtained as a predetermined evaluation as a paraphrase candidate; ,
A recording medium on which a sentence creation support processing program using paraphrasing to be executed by a computer is recorded. In the process of acquiring the operation target part and the sentence operation information,
Records a text creation support processing program using paraphrasing, which is performed by selecting an arbitrary section or sentence of text data displayed to the user or a predetermined selection item by using an input pointing function. Recording medium. A sentence creation support processing method using a paraphrase process for converting sentence data described in a natural language into sentence data in another expression,
A propositional dictionary that describes the propositional type that defines the actions or states that the sentence data describes, a propositional connection dictionary that describes the relationship between the clue expression in the sentence and the type between the propositions that indicate the correspondence between sentences, a dependency relationship or Data for storing a paraphrase dictionary describing a paraphrase pattern based on clue expression, or a template dictionary for storing exemplary sentence data to which the propositional type and the interpropositional type are given, which is sentence data recognized as having a good sentence structure Access the storage means to get the data,
Obtain sentence operation information indicating the operation target part to be subjected to the paraphrase operation and the type of paraphrase operation from the text data displayed for the user,
Analyzing the internal structure and meaning of the operation target part, referring to the proposition dictionary or the proposition concatenation dictionary to identify the proposition type of the operation target part, or the proposition type and interproposition type from the analysis result,
In accordance with the sentence operation information, refer to the paraphrase pattern, paraphrase the operation target portion to generate a paraphrase group,
By extracting paradigm text data including the propositional type of the operation target part from the template dictionary, the text data in a predetermined range including the context of the model text data and the operation target part, and by paraphrase extracted from the paraphrase group Compare the context of the paraphrased case sentence data and evaluate the similarity between the sentence structure of the exemplary sentence data and the sentence structure of the case sentence data, and output the paraphrase obtained as a predetermined evaluation as a paraphrase candidate.
Text creation support processing method using paraphrasing.

Images