JP5387166B2 - Information processing apparatus and program - Google Patents

Description

  The present invention relates to an information processing apparatus and a program.

  There is a technique for converting a sentence described in a natural language such as Japanese into a sentence rephrased with another expression having a corresponding meaning (see, for example, Patent Documents 1, 2, and 3). In these techniques, a sentence is paraphrased using a conversion rule that converts a specific expression into another expression so as not to change the semantic content of the sentence. This conversion rule is created manually or automatically generated by analyzing a dependency structure of a sentence whose meaning is known to correspond beforehand.

JP 2003-022264 A JP 2005-149494 A JP 2008-234175 A

  An object of the present invention is to provide an information processing apparatus and a program for realizing a paraphrase into another sentence that corresponds to a semantic content but has a different dependency structure.

  The invention according to claim 1 is an information processing apparatus, which is a pair of sentence elements corresponding to each other from two sentences having meanings corresponding to each other, wherein one includes a verb and the other includes an appendix. A sentence element extracting means for extracting a set of elements; and a pair information output means for outputting pair information of a verb and an attached word that can be paraphrased based on the set of sentence elements to be extracted. .

  The invention according to claim 2 is the information processing apparatus according to claim 1, wherein one of the two sentences is a sentence written in a first language and the other is written in a second language different from the first language. The sentence element extraction means extracts a sentence element including the verb from a sentence described in the first language, extracts a sentence element including the adjunct from a sentence described in the second language, The pair information output means translates the adjunct word into the first language and outputs pair information of the verb and the translated adjunct word.

  The invention according to claim 3 is the information processing apparatus according to claim 2, wherein the pair information output means is a verb included in a sentence described in the first language, and the sentence element to be extracted Information on the type of an adjunct word that can modify a verb different from the verb included in is acquired, and the adjunct word is translated into the first language using the acquired information.

  The invention according to claim 4 is an information processing apparatus, which is paired information of a paraphrasable verb and an adjunct, and a pair in which a sentence element including the paraphrasable verb or an adjunct is associated with a role in the sentence. A pair information acquisition means for acquiring information and a sentence element specification for specifying a sentence element including one of the acquired pair information and having a role associated with the pair information among the sentence elements included in the processing target sentence Means, and sentence conversion means for converting the processing target sentence into another sentence having a corresponding meaning based on the acquired pair information and the specified sentence element.

  The invention according to claim 5 is a sentence that extracts a pair of sentence elements corresponding to each other from two sentences having meanings corresponding to each other, one of which includes a verb and the other includes an appendix. This is a program for causing a computer to function as element extraction means and pair information output means for outputting paraphrase verb / attachment pair information based on the extracted sentence element pairs.

  According to the first and fifth aspects of the present invention, it is possible to generate pair information for realizing a paraphrase into another sentence that corresponds to the semantic content but has a different dependency structure.

  According to the second aspect of the present invention, the pair information can be generated using two sentences having meanings corresponding to each other in different languages.

  According to the third aspect of the present invention, when pair information is generated using sentences in different languages, pair information that enables a paraphrase from a plurality of attached words having meanings to corresponding sentences. Can be generated.

  According to the fourth aspect of the present invention, the sentence to be processed can be converted into another sentence that corresponds to the semantic content but has a different dependency structure.

It is a figure which shows the structural example of the information processing apparatus which concerns on embodiment of this invention. It is a functional block diagram which shows an example of the function which the information processing apparatus which concerns on embodiment of this invention implement | achieves. It is a figure which shows an example of the conversion rule table containing pair information. It is a flowchart which shows an example of the flow of a pair information generation process. It is a flowchart which shows an example of the flow of a pair information generation process. It is a figure which shows an example of the alignment information provided to the sample sentence. It is a figure which shows an example of the syntax semantic analysis result with respect to a sample sentence. It is a figure which shows an example of the syntax semantic analysis result with respect to another sample sentence. It is a flowchart which shows an example of the flow of a preposition translation process. It is a flowchart which shows an example of the flow of a paraphrase process. It is a flowchart which shows an example of the flow of a paraphrase process. It is a figure which shows an example of the syntax semantic analysis result with respect to a process target sentence.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 is a diagram illustrating a configuration example of an information processing apparatus 1 according to an embodiment of the present invention. The information processing apparatus 1 is, for example, a personal computer or the like, and includes a control unit 11, a storage unit 12, an operation unit 13, and a display unit 14.

  The control unit 11 is, for example, a CPU or the like, and executes various types of information processing according to programs stored in the storage unit 12. A specific example of processing executed by the control unit 11 in the present embodiment will be described later.

  The storage unit 12 includes, for example, a memory element such as a RAM, a hard disk, and the like. The storage unit 12 holds programs executed by the control unit 11 and various data. The storage unit 12 also operates as a work memory for the control unit 11.

  The operation unit 13 is, for example, a keyboard or a mouse, and receives a user instruction operation, and outputs the content of the instruction operation to the control unit 11. The display unit 14 is, for example, a liquid crystal display, and displays an image in accordance with an instruction from the control unit 11.

  Hereinafter, functions realized by the information processing apparatus 1 according to the present embodiment will be described. As shown in FIG. 2, the information processing apparatus 1 functionally includes a sentence element pair extraction unit 21, a pair information output unit 22, a processing target sentence acquisition unit 23, a pair information acquisition unit 24, and a sentence element specification. A unit 25 and a sentence conversion processing unit 26 are included. These functions are realized, for example, when the control unit 11 executes a program stored in the storage unit 12. This program may be provided via communication means such as the Internet, or may be provided by being stored in various computer-readable information storage media such as an optical disk.

  The information processing apparatus 1 acquires a plurality of sample sentences from an existing corpus (sentence example database including a plurality of natural language sentences) and the like, and uses the analysis result to include paraphrase verb / attachment pair information. Generate transformation rules. Then, using this conversion rule, a paraphrase process for converting a natural language sentence to be processed (hereinafter referred to as a process target sentence) into another sentence having a corresponding meaning is performed. Specifically, the sentence element pair extraction unit 21 and the pair information output unit 22 perform a process of generating pair information, a processing target sentence acquisition unit 23, a pair information acquisition unit 24, a sentence element identification unit 25, and a sentence conversion process. The unit 26 executes a paraphrase process using the generated pair information. In the present embodiment, the processing target sentence is a Japanese sentence.

  The sentence element pair extraction unit 21 acquires two sample sentences (hereinafter referred to as sample sentence pairs) having meanings corresponding to each other. Then, by selecting one sentence element from each of the two sample sentences, a pair of sentence elements corresponding to each other (hereinafter referred to as a sentence element pair) is extracted. Here, the sentence element is an element constituting the sentence, and is composed of one or a plurality of morphemes and indicates a unit having a semantic unit. That is, the sentence element corresponds to a phrase or a clause.

  Here, the sentence element pair extraction unit 21 extracts a sentence element pair so that one of the sentence element pairs includes a verb and the other includes a sentence element including a noun and an attached word attached thereto. Note that the sentence element pair extraction unit 21 extracts such a set of corresponding sentence elements, and the role of each sentence element in the sentence (for example, the main character, the purpose case, etc. May be specified.

  As a specific example, the sentence element pair extraction unit 21 acquires a sample sentence pair representing the same meaning by a method such as reading an example sentence from an existing corpus. In the present embodiment, a bilingual corpus is used as a corpus including sample sentence pairs having corresponding meanings. In the bilingual corpus, two sample sentences having the same meaning and expressed in different first and second languages are stored in association with each other. You may use a corpus that collects pairs of sample sentences that have the same meaning but different expressions in one language. Compared to such a corpus, various types of bilingual corpora are known. Easy to use.

  The pair information output unit 22 outputs verb / attachment pair information that can be paraphrased based on the sentence element pair extracted from the sample sentence pair by the sentence element pair extraction unit 21. In the present embodiment, the pair information output unit 22 has a verb included in one of the extracted sentence element pairs, a case particle as an appendage included in the other, and a grammatical role in the sample sentence of these sentence elements. The information shown is associated with each other and output as pair information. A set of a plurality of pairs of information output in this way is stored in the storage unit 12 as a conversion rule used for the paraphrase process in the present embodiment. FIG. 3 shows an example of a conversion rule table composed of a plurality of pieces of pair information.

  The process target sentence acquisition unit 23 acquires a process target sentence that is a target of the paraphrase process. Specifically, the processing target sentence acquisition unit 23 may read and acquire the processing target sentence from various information storage media, or may acquire the processing target sentence by receiving a user input to the operation unit 13.

  The pair information acquisition unit 24 refers to the conversion rule table output from the pair information output unit 22 and stored in the storage unit 12 and acquires pair information used for paraphrase processing. The pair information acquired by the pair information acquisition unit 24 is used by a sentence element specifying unit 25 and a sentence conversion processing unit 26 described later.

  The sentence element specifying unit 25 performs a syntax semantic analysis of the processing target sentence acquired by the processing target sentence acquisition unit 23. Here, in the syntactic and semantic analysis, first, the part of speech and the utilization form of each morpheme constituting the processing target sentence are specified. Then, sentence elements composed of one or more morphemes are identified, and the dependency structure of each sentence element (modification relation between sentence elements) and the role of each sentence element (grammatical role, Semantic roles, etc.) are identified. This syntactic and semantic analysis may be realized using a known method. Furthermore, the sentence element identification unit 25 includes one of pair information (here, a verb) acquired by the pair information acquisition unit 24 among the sentence elements included in the processing target sentence using the result of the syntax semantic analysis. A sentence element having a role associated with the pair information is specified. This identified sentence element becomes the target of paraphrase in the paraphrase process of the process target sentence.

  The sentence conversion processing unit 26 converts the processing target sentence into another sentence having a corresponding meaning based on the pair information acquired by the pair information acquisition unit 24 and the sentence element specified by the sentence element specifying unit 25. Specifically, the sentence conversion processing unit 26 performs a conversion process of replacing one (verb) of the pair information included in the sentence element specified by the sentence element specifying unit 25 with the other (attached word) of the pair information. Thereby, the process target sentence acquired by the process target sentence acquiring unit 23 is paraphrased as another sentence having a corresponding meaning.

  Hereinafter, a specific example of the flow of processing executed by the control unit 11 of the information processing apparatus 1 according to the present embodiment will be described. First, a specific example of the pair information generation process executed by the sentence element pair extraction unit 21 and the pair information output unit 22 will be described with reference to the flowcharts of FIGS. 4 and 5.

In the following, as an example for explaining the flow of processing, pair information generation processing is executed for a sample sentence pair consisting of the following sample sentences 1 and 2, and as a result, one line in the conversion rule table of FIG. A case where the pair information shown in the eyes is generated will be described.
Sample 1: "He attended the party with his wife."
Sample sentence 2: "He attended the party with his wife."

  As is clear from the sample sentences 1 and 2, in the following example, the bilingual corpus referred to by the sentence element pair extraction unit 21 stores Japanese sentences and English sentences having the same meaning in association with each other. Sample sentence 1 is the same Japanese sentence as the sentence to be processed. Furthermore, in this example, it is assumed that alignment information is given in advance to the sample sentence pairs included in the bilingual corpus. The alignment information is information that associates sentence elements having corresponding meanings in two sentences. FIG. 6 is a diagram illustrating an example of the alignment information given to the sample sentences 1 and 2 described above. In this figure, P1 of sample sentence 1 and P1 of sample sentence 2, P2 and P3 of sample sentence 1 and P2 of sample sentence 2, P4 of sample sentence 1 and P3 of sample sentence 2, P5 and sample sentence 1 P4 of the text 2 corresponds to each. Such alignment information can be obtained by, for example, the technique disclosed in Hideki Tsujioka “Long word alignment considering clause boundaries” (Journal of the 11th Annual Conference of the Language Processing Society, March 2005). To be used.

  First, the sentence element pair extraction unit 21 reads the sample sentences 1 and 2 from the parallel corpus (S1). The bilingual corpus may be stored in the storage unit 12 or may be stored in an external information processing apparatus connected via communication means.

  Subsequently, the sentence element pair extraction unit 21 performs syntax and semantic analysis on each of the two sample sentences acquired in S1 (S2). This syntax and semantic analysis may be executed by the same algorithm as the process executed by the sentence element specifying unit 25 for the processing target sentence.

  7 and 8 show an example of analysis results obtained as a result of executing syntax-semantic analysis on the sample sentences 1 and 2, respectively. In these drawings, each morpheme is shown surrounded by a frame. The symbols attached to each morpheme indicate the part of speech. N is a noun, PRON is a pronoun, DET is a determiner, V is a verb, AUX is an auxiliary verb, and PP is a particle (Japanese). ) Or preposition (in English). Here, AUX and PP are used together with independent words such as nouns and verbs, and are auxiliary words that do not constitute sentence elements by themselves. In addition, NP indicates a noun phrase including a noun as a main word, VP indicates a verb phrase including a verb as a main word, PP indicates a preposition phrase, and S has a structure as a sentence by itself. Indicates an element. Further, the subject attached to the NP indicates that the grammatical role of the noun phrase is the main case, oblique indicates the oblique case, and object indicates the objective case.

  Next, the sentence element pair extraction unit 21 determines whether the sample sentence 1 includes two or more verb phrases VP (S3). If two or more verb phrases VP are not included, it is presumed that a sentence element that can be paraphrased as an expression that does not include a verb is not included in the sample sentence, and thus the process ends.

  Here, since two verb phrases VP “accompanied” and “attended” are included in the sample sentence 1, the subsequent processing is executed. That is, the variable i is initialized with 1 (S4), and the processes described below are sequentially executed for the plurality of verb phrases VP. In the following, the i-th verb phrase VP in the sample sentence 1 is expressed as VPi using the variable i.

  First, the sentence element pair extraction unit 21 substitutes all the consecutive modification components related to the verb Vi that is the main word of the verb phrase VPi in the sample sentence 1 into the list NPLi (S5). For example, when i = 1, the verb V1 is “accompanied”, and there is only one consecutive modification component “wife” as shown in FIG. 7, so the element of NPL1 is {“wife”}. Thereafter, the sentence element pair extraction unit 21 initializes the variable j with 1 (S6), and sequentially executes the processes described below for all elements included in the element list NPLi. Hereinafter, the j-th element included in the list NPLi is expressed as NPij using the variable j.

  The sentence element pair extraction unit 21 acquires, as an element NPAij, the sentence element of the sample sentence 2 associated with the element NPij by the alignment information (S7). Since “with his wife” is associated with NP11 “wife” as shown in FIG. 6, this sentence element is acquired as NPA11.

  Subsequently, the sentence element pair extraction unit 21 determines whether or not a condition that the sentence element of the sample sentence 1 associated with the modification destination verb of the NPAij acquired in S7 matches the verb phrase VPi according to the alignment information. Determine (S8). For example, the modification destination of “with his wife” is the verb “attended” as shown in FIG. 8, and the sentence element of the alignment destination of this verb is “attend” as shown in FIG. Therefore, this verb phrase “attendance” does not match “accompanied” which is the verb phrase VP1, and the condition of S8 is not satisfied.

  If the condition of S8 is satisfied, the verb Vi is not determined to be paraphrased, and the process proceeds to S12, and the subsequent process for the next element in the element list NPLi is executed. On the other hand, when the condition of S8 is not satisfied, the verb phrase VPi of the sample sentence 1 corresponds to the element NPAij of the sample sentence 2, and the verb Vi included in the verb phrase VPi is an appendage included in the element NPAij. It is presumed that it can be paraphrased in the Japanese translation of the preposition. Therefore, the pair information output unit 22 first acquires the grammatical role GFij in the sample sentence 1 of the element NPij (S9). For example, in the case of NP11, since the grammatical role GF11 is “object” (objective) as shown in FIG. 7, this information is acquired.

  Further, the pair information output unit 22 uses the preposition PAii included in the element NPAij and the verb VAi that is the alignment destination of the modified verb of the element NPAij in order to specify the case particle that is an appendix corresponding to the verb Vi. The preposition translation process is executed (S10). The contents of this process will be described later. For example, in the case of element NPA11 ("with his wife"), the preposition PA11 is "with" and the verb VA1 is "attend". When the process is executed, the case particle “to” is obtained as a translation of the preposition “with”.

  In response to this, the pair information output unit 22 associates the verb Vi to be processed, the case particle obtained by the preposition translation processing of S10, and the grammatical role GFij obtained in S9 with each other. It outputs as pair information (S11). In the examples given so far, as shown in the first line of the conversion rule table of FIG. 3, the verb “accompanied” and the case particle “to” are output as counter information associated with the grammatical role of the objective case. The

  Thereafter, the sentence element pair extraction unit 21 adds 1 to the variable j (S12), and determines whether the value of j exceeds the number of elements in the list NPLi (S13). If not, since the list NPLi contains elements that are not yet processed, the process returns to S7 to continue the process for the next element NPij. On the other hand, if it exceeds, all the elements included in the list NPLi have been processed. Next, 1 is added to the variable i (S14), and the verb phrase in which the value of i is included in the sample sentence 1 It is determined whether the number of VPs has been exceeded (S15). If not, since there is a verb phrase that has not yet been processed, the process returns to S5 to continue the process for the next verb phrase VPi. On the contrary, if all the verb phrases are processed, the pair information of the verb and case particle satisfying the predetermined condition is output, so the pair information using the sample sentences 1 and 2 is output. Complete the generation process.

  Note that when the verb phrase VP2 (“attend”) is to be processed, the component of the verb V2 (“attend”) included in the verb phrase VP2 is only “to party”, so the element of NPL2 is “ Only one at a party. The alignment element of the element NP21 is “the party”, and the modification destination is “attended” as shown in FIG. The alignment destination of “attended” is “Attend” and matches the verb phrase VP2. Therefore, unlike the case of element NP11, in the case of element NP21, it is determined that the condition is satisfied in the determination of S8. As a result, the verb “attend” is not informative.

  Next, a specific example of the flow of preposition translation processing in S10 of the flow described above will be described with reference to the flowchart of FIG.

  First, the pair information output unit 22 acquires an English preposition PP and a Japanese verb V as input of processing (S21). For example, in the example of the element NP11 (“wife”) described above, the verb “attend” corresponding to the prefix “with” included in the element NPA11 (“with his wife”) and the modification destination “attended” of the element NPA11 is included. And are acquired.

  Next, the pair information output unit 22 inquires of the bilingual dictionary about the preposition PP acquired in S21, and acquires a translation word candidate list PPL including translation word candidates of the preposition PP (S22). Here, as a specific example, it is assumed that a list composed of two case particles “de” and “to” is obtained as a translation candidate list for the preposition “with”.

  Subsequently, the pair information output unit 22 inquires the valence dictionary for the verb V acquired in S21, and acquires a case frame that the verb V can take (S23). Here, the valence dictionary is a dictionary that defines what case frame (the type of case element that modifies the verb) is used with the verb. Case frames in Japanese are defined by the type of case particles attached to nouns. Here, as specific examples, it is assumed that three case frames “N”, “N”, and “N” are obtained as the case frames of the verb “attend”.

  Next, the pair information output unit 22 initializes the variable n with 1 (S24), and repeatedly executes the following processing in order for each translation word candidate included in the translation word candidate list PPL acquired in S22.

  Specifically, the pair information output unit 22 first determines whether or not the nth element Tn of the list PPL is included in the case frame acquired in S23 (S25). If not included, the element Tn is not set as a translation candidate and the process proceeds to S27. On the other hand, if it is included, the element Tn is output as a translation of the preposition PP (S26). In the example of “with” described above, since the second element T2 “to” is included in the case frame of the verb “attend”, the case particle “to” is output as a translation candidate for “with”. .

  Thereafter, the pair information output unit 22 adds 1 to the variable n (S27), and determines whether or not the value of the variable n exceeds the number of elements in the list PPL (S28). If not, the process returns to S25 to process the next element Tn. If it exceeds, the process has been completed for all elements included in the list PPL, and the process proceeds to the next process.

  Next, the pair information output unit 22 determines whether or not one or more translated words have been output in S26 (S29). If one or more translated words are being output, the process ends. On the other hand, if there is no element Tn that satisfies one condition and no translation is output, all elements included in the list PPL are output as translations of the preposition PP (S30), and the process is terminated. This is a procedure for outputting as a translated word a case particle that may be used in other words even when a translated word included in the case frame cannot be found.

  The processing for generating pair information when using a bilingual corpus has been described so far, but when using sample sentence pairs with different expressions in the same language, the flow is almost the same as the flow shown in FIGS. Pair information is generated by the processing. Specifically, in this case, the preposition translation process shown in FIG. 9 is not necessary, and the process of S10 is simply replaced with a process of acquiring a case particle PAij included in the element NPAij. In the process of S11, the verb Vi and the case particle PAij are output as pair information. Other processing may be the same processing as the flow shown in FIGS. 4 and 5.

  Next, a specific example of the flow of paraphrase processing executed by the processing target sentence acquisition unit 23, the pair information acquisition unit 24, the sentence element identification unit 25, and the sentence conversion processing unit 26 will be described with reference to the flowcharts of FIGS. I will explain.

  First, the process target sentence acquisition unit 23 acquires a process target sentence (S41). Next, the sentence element specifying unit 25 executes syntax semantic analysis of the processing target sentence acquired in S41 (S42). Hereinafter, as an example, a description will be given of a paraphrase process in which a Japanese sentence “a tsunami has occurred and a surfer is drowned” is a sentence to be processed. FIG. 12 is a diagram illustrating an example of a result of performing syntax and semantic analysis on the processing target sentence.

  Next, the sentence element specifying unit 25 acquires a list VL including all the verbs included in the processing target sentence (S43), and initializes the variable i with 1 and the variable p with 0 (S44). Thereafter, the processing described below is repeatedly executed for each verb included in the verb list VL. In the following, the i-th verb included in the verb list VL is denoted as Vi. In the example of the processing target sentence described above, V1 is “occurs” and V2 is “drown”.

  Next, the sentence element specifying unit 25 determines whether there is one noun phrase NPi that modifies the verb Vi (S45). When two or more noun phrases NPi modify the verb Vi, it is considered that the verb Vi cannot be rephrased as a case particle using the conversion rule according to the present embodiment, so that the process proceeds to S52 and the next verb Processing is performed.

  On the other hand, when there is one noun phrase NPi that modifies the verb Vi, the sentence element specifying unit 25 inquires whether there is pair information including the verb Vi (S46). In response to this inquiry, the pair information acquisition unit 24 refers to the conversion rule table illustrated in FIG. 3 and determines whether there is pair information corresponding to the verb Vi (S47). If it does not exist, it is determined that the verb Vi cannot be rephrased, and the process proceeds to S52 to process the next verb. If it exists, the pair information acquisition unit 24 reads and acquires information on the case particle Ci and the grammatical role GFi associated with the verb Vi from the conversion rule table (S48). In the case of the verb V1 (“occurs”), information indicating “de” as the case particle Ci and “SUBJECT” (major) as the grammatical role GFi is obtained from the conversion rule table of FIG.

  Next, the sentence element specifying unit 25 determines whether or not the grammatical role GFi acquired in S48 corresponds to the grammatical role of the noun phrase NPi that modifies the verb Vi in the processing target sentence (S49). If the grammatical roles are different as a result of the determination, it is determined that the verb Vi cannot be rephrased, so the process proceeds to S52 and the process for the next verb is performed. On the other hand, when the grammatical roles are the same, the verb phrase VPi including the verb Vi is specified as a paraphrase target sentence element. In the case of the verb V1, the grammatical role of the noun phrase NP1 (“tsunami ga”) that modifies it is specified as “subject” (major) by syntactic and semantic analysis as shown in FIG. Therefore, it is determined that paraphrasing is possible.

  In response to this, the sentence conversion processing unit 26 replaces the verb phrase VPi and the case particle existing at the end of the noun phrase NPi modifying the verb phrase VPi with the case particle Ci acquired in S48 (S50). In the case of the verb V1, the case particle “ga” and the verb phrase “occurs” included in the noun phrase NP1 are replaced with the case particle “de”. Further, the sentence conversion processing unit 26 adds 1 to the variable p (S51).

  Thereafter, the sentence element specifying unit 25 determines whether or not it is necessary to continue the process with another verb as a target of the paraphrase process. Specifically, the sentence element specifying unit 25 determines whether the value of the variable p is equal to or greater than the number obtained by subtracting 1 from the number of elements in the verb list VL (S52). Here, the variable p functions as a counter variable indicating the number of verbs that have undergone paraphrasing. When the determination condition of S52 is satisfied, paraphrasing is performed for other verbs excluding one verb included in the processing target sentence. It shows that it was broken. In this case, since it is determined that further paraphrase processing is unnecessary, the process proceeds to S55 and a paraphrase text is output. On the other hand, when the determination condition of S52 is not satisfied, 1 is added to the variable i (S53), and it is determined whether or not the value of the variable i exceeds the number of elements in the verb list VL (S54). If the value of the variable i has not exceeded the number of elements in the verb list VL, the process returns to S45 to determine whether or not the next verb Vi is to be paraphrased. If exceeded, the process proceeds to the next step S55.

  Finally, the sentence conversion processing unit 26 outputs the paraphrase sentence generated by performing the character string replacement in S50 (S55). In the example of the processing target sentence described above, as a result of replacing “occurs” with “de”, a sentence “surfer drowned in the tsunami” is output as the paraphrase sentence, and the paraphrase process ends.

  In the above description, the paraphrasing process is performed using the pair information of the verb and the case particle, but the embodiment of the present invention is not limited to this. In the above description, a sample sentence pair is obtained using a corpus that includes Japanese and English translations as a bilingual corpus. However, other than this, a bilingual corpus of various languages is used as a standard. You may get a body pair.

  In the above description, one information processing apparatus 1 executes both the pair information generation process and the paraphrase process. However, these processes may be executed by independent information processing apparatuses. . For example, an information processing device generates a conversion rule table including pair information using a bilingual corpus, etc., and another information processing device refers to the generated conversion rule table to rephrase the processing target sentence Processing may be executed.

  DESCRIPTION OF SYMBOLS 1 Information processing apparatus, 11 Control part, 12 Storage part, 13 Operation part, 14 Display part, 21 Sentence element pair extraction part, 22 Pair information output part, 23 Process target sentence acquisition part, 24 Pair information acquisition part, 25 Sentence element Specific part, 26 sentence conversion processing part.

Claims (5)

Means for acquiring two sentences of a first sentence and a second sentence having meanings corresponding to each other, and sentence element related information associating sentence elements corresponding to meanings between the two sentences;
A first sentence element including a verb in the first sentence from the two acquired sentences, and an appendix in the second sentence associated with a sentence element that modifies the verb by the sentence element related information. Sentence element extraction means for extracting a set of second sentence elements;
It is determined whether or not the verb to be modified by the second sentence element is associated with the verb included in the first sentence element based on the sentence element related information. a verb contains the grammatical role of sentence elements for modifying the verb, a case particle corresponding to accessory word contained in the second sentence elements, pairs of output as pairs information paraphrase possible verbs and case particle Information output means;
An information processing apparatus comprising: The first sentence is a first language, and the second sentence is a sentence written in a second language different from the first language ,
Before Symbol pair information output means, according to claim 1, wherein the outputting the accessory words a case particle obtained by translating the first language, as a counter information together with the verb contained in the first sentence element Information processing equipment. The pair information output means is a verb included in a sentence described in the first language, and is used as a case element that modifies the verb for a verb different from the verb included in the first sentence element. A case particle corresponding to the acquired case particle candidate is obtained from among the case particles that are candidates for acquiring a case particle candidate and translating the ancillary word included in the second sentence element into the first language. The information processing apparatus according to claim 2, wherein the information processing apparatus outputs the pair information . A pair information acquisition means for acquiring pair information of paraphrasable verbs and case particles , wherein the pair information is associated with a grammatical role in a sentence of a sentence element that modifies the paraphrasable verb;
Of the sentence elements included in the processing target sentence includes verbs which constitute the pair information the acquired grammatical role in the context of sentence elements for modifying the verb, the grammatical role associated with said pair information A sentence element identification means for identifying a corresponding sentence element;
By substituting a verb included in the specified sentence element and a case particle included in a sentence element that modifies the verb with a case particle associated with a paraphrasable verb in the paired information, Sentence conversion means for converting the processing target sentence into another sentence having a corresponding meaning;
An information processing apparatus comprising: Means for acquiring two sentences of a first sentence and a second sentence having meanings corresponding to each other, and sentence element related information for associating sentence elements corresponding to meanings between the two sentences;
A first sentence element including a verb in the first sentence from the two acquired sentences, and an appendix in the second sentence associated with a sentence element that modifies the verb by the sentence element related information. Sentence element extraction means for extracting a set of second sentence elements; and
It is determined whether or not the verb to be modified by the second sentence element is associated with the verb included in the first sentence element based on the sentence element related information. a verb contains the grammatical role of sentence elements for modifying the verb, a case particle corresponding to accessory word contained in the second sentence elements, pairs of output as pairs information paraphrase possible verbs and case particle Information output means,
As a program to make the computer function as.

Images