JP7718097B2 - Proofreading support device, proofreading support method, and program - Google Patents

Description

The present invention relates to a calibration support device, a calibration support method, and a program.

Sentences are proofread for spelling variations. For example, Patent Document 1 discloses a technique for extracting character strings that are candidates for spelling variations from the sentence to be proofread, and then comparing the extracted candidates with the spellings of other candidates to determine whether or not there is spelling variation.

Japanese Patent Application Publication No. 3-184162

The method described in Patent Document 1 compares a character string extracted from a sentence with all other candidate spellings. For example, if the number of character strings extracted from a sentence is K (K is any natural number), the number of comparisons is K x (K-1). Therefore, if the sentence to be proofread is short, around 100 words, only about 10,000 comparisons will be necessary. However, for a long sentence, around 1 million words, the number of comparisons will increase exponentially to around 100 million. This poses the problem that checking spelling variations in a long sentence can take several days, making it unrealistic.

The present invention was made in light of this situation, and aims to provide a proofreading support device, proofreading support method, and program that can extract character strings that may have spelling variations, even in long sentences, without increasing the time required for proofreading.

The proofreading support device of the present invention comprises an acquisition unit that acquires a target sentence to be proofread; a division unit that divides the target sentence into at least three or more segments to generate segmented sentences; a concatenation unit that concatenates different segments from the segmented sentences to generate concatenated sentences that are shorter than the target sentence; and an evaluation unit that extracts character strings that are candidates for spelling variations in the concatenated sentences.

The proofreading support method of the present invention is a computer-implemented proofreading support method in which an acquisition unit acquires a target sentence to be proofread, a division unit generates divided sentences by dividing the target sentence into at least three or more segments, a concatenation unit generates concatenated sentences by concatenating mutually different segments from the segmented sentences, the concatenated sentences being shorter than the target sentence, and an evaluation unit extracts character strings in the concatenated sentences that are candidates for spelling variations.

The program of the present invention is a program for causing a computer to operate as the above-described proofreading support device, and is a program for causing the computer to function as each unit of the proofreading support device.

This invention reduces the risk of overlooking spelling variations, and can prevent the time required for proofreading from increasing, even for long sentences.

3A to 3C are diagrams illustrating processing performed by the calibration support device 10 according to the embodiment. 1 is a block diagram showing an example of the configuration of a calibration support device 10 according to an embodiment. FIG. 10 is a diagram illustrating an example of list information 120 according to an embodiment. FIG. 10 is a diagram illustrating an example of list information 120 according to an embodiment. FIG. 10 is a diagram illustrating an example of list information 120 according to an embodiment. FIG. 10 is a diagram illustrating an example of list information 120 according to an embodiment. FIG. 10 is a diagram illustrating an example of list information 120 according to an embodiment. FIG. 10 is a diagram illustrating an example of list information 120 according to an embodiment. 1 is a flowchart showing the flow of processing performed by a calibration support device 10 according to an embodiment. FIG. 10 is a diagram illustrating a process for extracting spelling variation candidates. FIG. 10 is a diagram illustrating a process for extracting spelling variation candidates. FIG. 10 is a diagram illustrating a process for extracting spelling variation candidates. FIG. 10 is a diagram illustrating a process for extracting spelling variation candidates.

Embodiments of the present invention will be described below with reference to the drawings.

First, we will explain how to extract spelling variation candidates from a sentence. Figures 10A to 10D are diagrams that explain the process of extracting spelling variation candidates.

Figure 10A shows an example of a sentence. As shown in the example of Figure 10A, we will explain by way of example how to extract spelling variation candidates from the sentence "There is handling available. There is handling available. There is handling available. There is no handling available."

Figure 10B shows an example of a sentence divided into words (morphemes). As shown in the example of Figure 10B, if the sentence in Figure 10A is divided into words, it would look like this: "Handling/ga/ari/masu/./Handling/wa/ari/masu/./Handling/ga/ari/masu/./Handling/wa/nai/desu/ne/." Here, "/" is the symbol indicating a division.

The process for extracting spelling variation candidates will be explained using Figures 10C and 10D. Figure 10C shows the process for extracting spelling variation candidates. Figure 10D shows an example of a list used for the evaluation of Figure 10C.

The table in Figure 10C shows items such as words, the number of words registered in the list, evaluation results, and whether or not the word is registered in the list. The words are listed in order as they appear in the sentence. The number of words registered in the list shows the number of words registered in the list. The list is a list used when evaluating words included in a sentence, as shown in the example in Figure 10D. Evaluation here means determining whether the words included in the sentence match or are similar to the words listed in the list.

The evaluation results show the results of evaluating the words contained in the sentence. The evaluation results here show the results of determining whether the words contained in the sentence match (or are similar) to the words registered in the list.

List registration indicates whether a word contained in a sentence has been registered in a list. Here, if it is determined based on the evaluation results that the word contained in the sentence does not match (or is not similar to) a word registered in the list, the word contained in the sentence is registered in the list.

The table in Figure 10D shows items such as list number, word, element 1, element 2, etc. The list number is identification information such as a number that uniquely identifies a word registered in the list. The word indicates the word identified by the list number. The element indicates a word determined to be similar to the word identified by the list number.

First, evaluation begins with no words registered in the list. That is, it is determined whether the word "handle" shown at the top of the table in Figure 10C matches or is similar to a word registered in the list. At this point, no words are registered in the list. Therefore, the evaluation result for the word "handle" is "no match or similar." Words that are evaluated as "no match or similar" are registered in the list. In the example in Figure 10C, the word "handle" is registered in No. 1 of the list. And, in Figure 10D, an example is shown in which the word "handle" is registered in No. 1 of the list.

Next, it is determined whether the word "ga" matches or is similar to words registered in the list. Because only the word "handling" is registered in the list, the evaluation result for the word "ga" is "no match or similar." Words that are evaluated as "no match or similar" are registered in the list. In the example of Figure 10C, the word "ga" is registered in No. 2 of the list. And in Figure 10D, an example is shown in which the word "ga" is registered in No. 2 of the list.

Next, it is determined whether the word "ari" matches or is similar to words registered in the list. Only the words "handle" and "ga" are registered in the list. Therefore, the evaluation result for the word "ari" is "no match or similar." Words that are evaluated as "no match or similar" are registered in the list. In the example of Figure 10C, the word "ari" is registered in No. 3 of the list. And in Figure 10D, an example is shown in which the word "ari" is registered in No. 3 of the list.

Next, it is determined whether the word "masu" matches or is similar to words registered in the list. Only the words "handle," "ga," and "ari" are registered in the list. Therefore, the evaluation result for the word "masu" is "no match or similar." Words that are evaluated as "no match or similar" are registered in the list. In the example of Figure 10C, the word "masu" is registered in No. 4 of the list. And in Figure 10D, an example is shown in which the word "masu" is registered in No. 4 of the list.

Next, it is determined whether the word "." matches or is similar to words registered in the list. Only the words "handle," "ga," "ari," and "masu" are registered in the list. Therefore, the evaluation result for the word "." is "no match or similar." Words that are evaluated as "no match or similar" are registered in the list. In the example of Figure 10C, the word "." is registered in No. 5 of the list. And in Figure 10D, an example is shown in which the word "." is registered in No. 5 of the list.

Next, it is determined whether the word "handle" matches or is similar to words registered in the list. The list currently contains only the words "handle," "ga," "ari," "masu," and "." Therefore, the evaluation result for the word "handle" is "match No. 1." Words that are evaluated as "matching" are not newly registered in the list.

Next, it is determined whether the word "wa" matches or is similar to words registered in the list. Only the words "handle," "ga," "ari," "masu," and "." are registered in the list. Therefore, the evaluation result for the word "wa" is "no match or similar." Words that are evaluated as "no match or similar" are registered in the list. In the example of Figure 10C, the word "wa" is registered in No. 6 of the list. And in Figure 10D, an example is shown in which the word "wa" is registered in No. 6 of the list.

Next, it is determined whether the word "ari" matches or is similar to words registered in the list. The list currently contains the words "handle," "ga," "ari," "masu," ".", and "wa." Therefore, the word "ari" is evaluated as being "similar" to "ari" in list No. 3. Words that are evaluated as being "similar" are not newly registered in the list. Words that are evaluated as being "similar" are added to the elements of words that are determined to be similar. Here, an example is shown in Figure 10D where the word "ari" has been added to element 1 of list No. 3.

Whether words are similar or not may be determined arbitrarily. For example, consider a case in which the proofreading support device 10 extracts candidates for spelling variations, and the user decides whether to proofread them to ensure a consistent spelling. In this case, it is preferable for the words in a sentence that are the subject of proofreading as spelling variations to be determined to be "similar."

For example, in the sentence in Figure 10A, it is preferable that "handai", "handai", and "handai" be determined to be similar words. This is because if the sentence contains a mixture of "handai", "handai", and "handai", it is likely that a proofreader would determine these as spelling variations and would proofread the sentence to unify them into one of the spellings. Also, it is preferable that "ari" and "ari" be determined to be similar words. This is because if the sentence contains a mixture of "ari" and "ari", it is likely that a proofreader would determine these as spelling variations and would proofread the sentence to unify them into one of the spellings.

Note that the example in Figure 10C illustrates a case where "ari" and "nai" are determined to be dissimilar words. However, this is not limiting. "ari" and "nai" may also be determined to be similar words. For example, if a sentence contains a mixture of "ari" and "nai," and the proofreader may determine that the spellings of "ari" and "nai" are variations in spelling and proofread the sentence to unify them into one spelling, the proofreading support device 10 is programmed to determine that "ari" and "nai" are similar words.

As shown in the example of Figure 10C, when extracting spelling variations, words in a sentence are evaluated in order. Then, based on the evaluation results, words are registered in a list. When evaluating a word in a sentence, it must be compared with all the words registered in the list. For this reason, the time required for evaluation increases depending on the number of words registered in the list. For example, as shown in the "Number of words registered in list" item in Figure 10C, in the sentence shown in the example of Figure 10A, the final number of words registered in the list is "9."

If the sentence is short and the number of extracted words is small, the evaluation does not take much time. However, if the sentence is long and the number of extracted words is large, the evaluation will take an enormous amount of time, making it unrealistic.

To address this issue, this embodiment divides the sentence. Dividing the sentence shortens it, making it possible to reduce the time required for evaluation. However, simply evaluating the divided sentences does not ensure consistency in spelling variations throughout the entire sentence. For example, when a sentence is divided into two, it is possible that the word "handling" is written without any spelling variations in the first half of the sentence, and the word "handling" is written without any spelling variations in the second half of the sentence. In this case, there is no spelling variation in the two divided sentences. However, there is spelling variation between "handling" and "handling" throughout the entire sentence. There is a risk that spelling variations will be overlooked if only the divided sentences are evaluated.

To address this issue, in this embodiment, the divided sentences are concatenated and the concatenated sentence is evaluated. This ensures that the spelling variations in the concatenated sentence are consistent. Furthermore, by performing a brute force search on the combinations of concatenated sentences, the spelling variations in the entire sentence are consistent.

Figure 1 is a diagram illustrating the processing performed by the proofreading support device 10 according to an embodiment. The upper part of Figure 1 shows target sentences ABCD. Target sentences ABCD are sentences to be proofread. In the example shown in this figure, the sentence containing the words "There is handling. There is handling. There is handling. There is no handling." is taken as target sentence ABCD.

First, the proofreading support device 10 generates segmented sentences by dividing the target sentence. In the example shown in this figure, the target sentence ABCD is divided into four segmented sentences A to D.

Next, the proofreading support device 10 generates concatenated sentences by concatenating the segmented sentences. In the example shown in this figure, concatenated sentences AB, AC, AD, BC, BD, and CD are generated by concatenating two different segmented sentences from segmented sentences A to D.

The proofreading support device 10 then evaluates each of the concatenated sentences AB, AC, AD, BC, BD, and CD, and generates a list for each (list information 120, described below). If there is a word that is registered twice in each list, the proofreading support device 10 merges that word and its element group. The merged list matches the list that would be generated if the entire sentence were evaluated.

Here, the proofreading support device 10 may process the evaluation of each of the concatenated sentences AB, AC, AD, BC, BD, and CD in parallel. For example, the proofreading support device 10 processes the evaluation of each concatenated sentence in parallel. In this case, a number of evaluations corresponding to the number of divisions are processed in parallel. By increasing the number of divisions, it is possible to shorten the time required for evaluation. On the other hand, it is conceivable that the total number of calculations will increase compared to when the entire sentence is evaluated. However, due to recent improvements in computer performance, the impact of an increase in the total number of calculations on processing time is not that great. In other words, even if the total amount of calculations increases, the effect of shortening processing time by processing in parallel is significant. Therefore, even for long sentences, the time required for proofreading does not increase.

The proofreading support device 10 may arbitrarily decide how many sentences to divide the target sentence into. Considering the possibility of concatenation, it would not make sense to divide the target sentence into two sentences and then concatenate the two sentences. For this reason, the proofreading support device 10 should divide the target sentence into at least three or more sentences.

Figure 2 is a block diagram showing an example configuration of a proofreading support device 10 according to an embodiment. The proofreading support device 10 is a computer device that extracts character strings that are candidates for spelling variations in a sentence to be proofread (target sentence). The proofreading support device 10 can be implemented, for example, as a server device, a cloud service, or a PC (Personal Computer).

The proofreading support device 10 includes, for example, a communication unit 11, a memory unit 12, and a control unit 13. The communication unit 11 communicates with external devices via a communication network or the like. The communication unit 11 receives text information indicating the target sentence from, for example, an external server device. The communication unit 11 also transmits character strings that are candidates for spelling variations extracted from the target sentence to the external server device or the like as proofreading results.

The memory unit 12 is composed of a storage medium such as a hard disk drive (HDD), flash memory, electrically erasable programmable read-only memory (EEPROM), random access read/write memory (RAM), or read-only memory (ROM), or a combination of these. The memory unit 12 stores programs for executing various processes of the proofreading support device 10, as well as temporary data used when performing various processes. The memory unit 12 stores, for example, list information 120. The list information 120 is information about candidate spelling variations extracted from the target sentence.

The control unit 13 is realized by causing a CPU (Central Processing Unit) provided as hardware in the proofreading support device 10 to execute a program. The control unit 13 comprehensively controls the proofreading support device 10. The control unit 13 includes, for example, an acquisition unit 130, a division unit 131, a connection unit 132, an evaluation unit 133, and a device control unit 134.

The acquisition unit 130 acquires the target sentence. For example, the acquisition unit 130 acquires information indicating the target sentence from an external server device or the like via the communication unit 11. The acquisition unit 130 outputs the acquired information indicating the target sentence to the division unit 131.

The segmentation unit 131 generates segmented sentences based on the target sentence. For example, the segmentation unit 131 generates segmented sentences so that each segmented sentence contains approximately the same number of words. In this case, the segmentation unit 131 divides the target sentence into words and calculates the number of words contained in the target sentence. The segmentation unit 131 searches for character strings before and after the statement corresponding to the number of words obtained by dividing the calculated number of words by the number of segments, and searches for boundaries that divide the target sentence.

The segmentation unit 131 uses, for example, periods, symbols, commas, etc. as boundaries for segmenting the target sentence. Examples of symbols include exclamation marks "!", question marks "?", musical notes "♪", and colons ":". Periods, symbols, commas, etc. are used as boundaries because they are unlikely to be the subject of orthographic variations. The segmentation unit 131 at least avoids setting boundaries in the middle of words. For example, if a boundary were set between "tori" and "i" in the word "tori", what was originally the word "tori" in the previous segmented sentence would be evaluated as the word "tori", which could result in an incorrect evaluation.

The segmentation unit 131 determines boundaries for segmenting the target sentence and generates multiple segmented sentences by segmenting the target sentence at the determined boundaries. The segmentation unit 131 outputs information indicating the segmented sentences that have been generated to the concatenation unit 132.

The concatenation unit 132 generates concatenated sentences. For example, the concatenation unit 132 generates all possible combinations of two different divided sentences as concatenated sentences. The concatenation unit 132 outputs information indicating the generated concatenated sentences to the evaluation unit 133.

The evaluation unit 133 evaluates each concatenated sentence. For example, the evaluation unit 133 divides the concatenated sentence into words and determines, in order, whether each divided word matches or is similar to a word registered in the list (list information 120). Furthermore, if a word neither matches nor is similar to a word registered in the list, the evaluation unit 133 registers the word in the list. Furthermore, if a word is similar to a word registered in the list, the evaluation unit 133 adds the word as an element of the words registered in the list (similar words).

The device control unit 134 exercises overall control over the proofreading support device 10. For example, the device control unit 134 outputs text information indicating the target sentence received by the communication unit 11 to the acquisition unit 130. The device control unit 134 merges the list (list information 120) generated as a result of the evaluation by the evaluation unit 133. The device control unit 134 transmits the merged list to an external server device as a list indicating possible spelling variations.

Figures 3 to 8 are diagrams showing examples of list information 120 according to an embodiment. Figures 3 to 8 show items such as list number, word, element 1, element 2, etc. These items are similar to the items shown in the table in Figure 10D, so their description will be omitted.

In Figure 3, an example of a list generated for concatenated sentence AB is shown as list information 120A. List information 120A shows an example in which the word "ari" written at the 6th and 7th characters and the word "ari" written at the 16th and 17th characters in concatenated sentence AB have been extracted as candidates for spelling variations.

In Figure 4, an example of a list generated in response to concatenated sentence AC is shown as list information 120B. List information 120B shows an example in which the word "hanari" (handling) written in characters 1-4 and the word "hanari" (handling) written in characters 21-23 of concatenated sentence AC have been extracted as candidates for spelling variations.

In Figure 5, an example of a list generated in response to concatenated sentence AD is shown as list information 120C. List information 120C shows an example in which the word "handling" written at characters 1-4 and the word "handling" written at characters 30-31 in concatenated sentence AD have been extracted as candidates for spelling variations.

In Figure 6, an example of a list generated for concatenated sentence BC is shown as list information 120D. List information 120D shows an example in which the word "手割 (handling)" written at characters 11-14 and the word "手返 (handling)" written at characters 21-23 in concatenated sentence BC have been extracted as spelling variation candidates. List information 120D also shows an example in which the word "あり (ari)" written at characters 16-17 and the word "あり (ari)" written at characters 25-26 in concatenated sentence BC have been extracted as spelling variation candidates.

In Figure 7, an example of a list generated in response to concatenated sentence BD is shown as list information 120E. List information 120E shows an example in which the word "handling" written at characters 11-14 and the word "handling" written at characters 30-31 in concatenated sentence BD have been extracted as candidates for spelling variations.

In Figure 8, an example of a list generated in response to concatenated sentence CD is shown as list information 120F. List information 120F shows an example in which the word "手帳" (handling) written at characters 21-23 and the word "手帳" (handling) written at characters 30-31 in concatenated sentence CD have been extracted as candidates for spelling variations.

Here, we will explain how the device control unit 134 merges lists (list information 120). First, the device control unit 134 identifies the positions where words that are candidates for spelling variations indicated in each list are written, and merges those where the identified positions overlap.

For example, when list information 120A to 120F as shown in Figures 3 to 8 are generated, the device control unit 134 identifies the positions where words are written in each list for the spelling variations "ari" and "ari" in list information 120A and the spelling variations "ari" and "ari" in list information 120D.

In list information 120A, the word "yes" is written at the 6th and 7th characters. In list information 120A, the word "yes" is written at the 16th and 17th characters. In list information 120D, the word "yes" is written at the 16th and 17th characters. In list information 120D, the word "yes" is written at the 25th and 26th characters.

The positions where the word "ari" is written in each list overlap at the 17th and 18th characters. In this case, the device control unit 134 considers the spelling variations "ari" and "ari" in list information 120A and the spelling variations "ari" and "ari" in list information 120D to be in the same group and merges them. As a result, the "ari" at the 7th and 8th characters, the "ari" at the 16th and 17th characters, and the "ari" at the 25th and 26th characters become candidates for spelling variations in the target sentence.

In addition, the device control unit 134 identifies the positions where words are written in each list for the spelling variations between "手持ち" and "手持ち" in list information 120B, the spelling variations between "手食べる" and "手介" in list information 120C, the spelling variations between "手介" and "手介" in list information 120D, the spelling variations between "手介" and "手介" in list information 120E, and the spelling variations between "手介" and "手介" in list information 120F.

In list information 120B, the word "handling" is written at characters 1-4. In list information 120B, the word "handling" is written at characters 21-23. In list information 120C, the word "handling" is written at characters 1-4. In list information 120C, the word "handling" is written at characters 30-31.

In list information 120D, the word "handling" is written at characters 11-14. In list information 120D, the word "handling" is written at characters 21-23. In list information 120E, the word "handling" is written at characters 11-14. In list information 120E, the word "handling" is written at characters 30-31.

The word "handling" is written at characters 21-23 in list information 120F. The word "handling" is written at characters 30-31 in list information 120F.

The device control unit 134 merges the spelling variations of "手帳" and "手帳" in list information 120B and the spelling variations of "手帳" and "手帳" in list information 120C, treating them as belonging to the same group. As a result, "手帳" at characters 1-4, "手帳" at characters 21-23, and "手帳" at characters 30-31 become candidates for spelling variations in the target sentence.

The device control unit 134 also merges overlapping spellings of the word "handling" in the 21st-23rd characters in each list. The device control unit 134 considers the spelling variations of "handling" and "handling" in list information 120B, the spelling variations of "handling" and "handling" in list information 120D, and the spelling variations of "handling" and "handling" in list information 120F to be in the same group and merges them. As a result, "handling" at characters 1-4, "handling" at characters 11-14, "handling" at characters 21-23, and "handling" at characters 30-31 become candidates for spelling variations in the target sentence.

The device control unit 134 also merges overlapping spelling variations of the word "handling" that occur at positions 11-14 of the characters in each list. The device control unit 134 considers the spelling variations of "handling" and "handling" in list information 120D and the spelling variations of "handling" and "handling" in list information 120E to be in the same group and merges them. As a result, "handling" at characters 11-14, "handling" at characters 21-23, and "handling" at characters 30-31 become candidates for spelling variations in the target sentence.

From the above, a list (first list) is generated in which "Handling" at characters 1-4, "Handling" at characters 21-23, and "Handling" at characters 30-31 are candidates for spelling variations in the target sentence. Furthermore, a list (second list) is generated in which "Handling" at characters 1-4, "Handling" at characters 11-14, "Handling" at characters 21-23, and "Handling" at characters 30-31 are candidates for spelling variations in the target sentence. A list (third list) is generated in which "Handling" at characters 11-14, "Handling" at characters 21-23, and "Handling" at characters 30-31 are candidates for spelling variations in the target sentence. In this case, the elements in the first list and the third list overlap. In this case, the device control unit 134 merges the first list, the second list, and the third list into a single list. As a result, "Handling" at characters 1-4, "Handling" at characters 11-14, "Handling" at characters 21-23, and "Handling" at characters 30-31 are candidates for spelling variations in the target sentence.

Figure 9 is a flowchart showing the flow of processing performed by the proofreading support device 10 according to an embodiment. The proofreading support device 10 acquires a proofreading target (target sentence) (step S10) and divides the acquired target sentence (step S11) to generate divided sentences. The proofreading support device 10 then concatenates the divided sentences (step S12) to generate a concatenated sentence. The proofreading support device 10 acquires the concatenated sentence (step S13) and evaluates the acquired concatenated sentence to extract candidates for spelling variations in the concatenated sentence (step S14). The proofreading support device 10 determines whether all concatenated sentences have been evaluated (step S15). If there are concatenated sentences that have not yet been evaluated, the proofreading support device 10 returns to step S13. If all concatenated sentences have been evaluated, the proofreading support device 10 merges duplicate words in the list (list information 120) generated as a result of the evaluation (step S16). The proofreading support device 10 transmits the merged list as the result of extracting candidates for spelling variations to, for example, an external server device.

As described above, the proofreading support device 10 of the embodiment includes an acquisition unit 130, a division unit 131, a concatenation unit 132, and an evaluation unit 133. The acquisition unit 130 acquires a target sentence. The division unit 131 generates divided sentences by dividing the target sentence into at least three or more segments. The concatenation unit 132 generates concatenated sentences that are shorter than the target sentence, by concatenating different segments from the segments. The evaluation unit 133 extracts character strings that are candidates for spelling variations in the concatenated sentence. This allows the proofreading support device 10 of the embodiment to extract character strings that are candidates for spelling variations in concatenated sentences that are shorter than the target sentence. Therefore, even if the target sentence is long, a concatenated sentence that is shorter than the target sentence can be proofread. Therefore, it is possible to extract character strings that are candidates for spelling variations even for long sentences without increasing the time required for proofreading.

In addition, in the proofreading support device 10 of the embodiment, the evaluation unit 133 compares the target word included in the concatenated sentence with registered words registered in list information 120, which lists words for which the presence or absence of spelling variations is to be determined. If the target word does not match or is not similar to the registered word, the evaluation unit 133 registers the target word in list information 120. If the target word is similar to the registered word, the evaluation unit 133 adds the target word to an element of the registered word similar to the target word in list information 120. Of the registered words registered in list information 120, the evaluation unit 133 regards the registered word and words added to the element of the registered word as candidates for spelling variations. This makes it possible for the proofreading support device 10 of the embodiment to extract candidates for spelling variations by a simple method of comparing the target word included in the concatenated sentence with list information 120, registering the target word if it does not match or is similar, and adding the target word to an element if it is similar.

The proofreading support device 10 of the embodiment further includes a device control unit 134. Based on the list information 120 generated corresponding to each linked sentence, the device control unit 134 merges registered words that are registered in multiple list information 120, among the registered words registered in each list information 120. This allows the proofreading support device 10 of the embodiment to extract spelling variations in the target sentence without duplication, based on the list information 120 of each linked sentence.

Now, let's consider a comparative example. One possible measure to shorten the time required for evaluation is to prepare a list of words that are prone to spelling variations, and then check only the words registered on the list for spelling variations, thereby speeding up proofreading. However, this method does not allow for checking spelling variations in words that are not registered on the list in advance. As a result, there is a possibility that spelling variations may be overlooked.

In contrast, in this embodiment, list information 120 corresponding to the concatenated sentence is generated. List information 120 is created depending on whether the words included in the concatenated sentence match or are similar to other words. This generates the corresponding concatenated sentence. This reduces the possibility of overlooking a spelling variation, thereby reducing the risk of overlooking a spelling variation.

Furthermore, in the proofreading support device 10 of the embodiment, the concatenation unit 132 generates concatenated sentences corresponding to all combinations of two different divided sentences. As a result, the proofreading support device 10 of the embodiment can extract the spelling variations between all divided sentences, even in cases where a consistent spelling in one divided sentence corresponds to a spelling variation in the other divided sentence, thereby reducing the risk of overlooking a spelling variation.

Furthermore, in the proofreading support device 10 of the embodiment, the evaluation unit 133 executes the process of extracting character strings that are candidates for spelling variations in parallel for each concatenated sentence. This enables the proofreading support device 10 of the embodiment to shorten the time required for evaluation.

In addition, in the proofreading support device 10 of the embodiment, the division unit 131 divides the target sentence using any of the periods, symbols, or commas in the target sentence as boundaries. As a result, in the proofreading support device 10 of the embodiment, boundaries are not set in the middle of words, making it possible to reduce the risk of overlooking spelling variations.

All or part of the calibration support device 10 in the above-described embodiment may be implemented by a computer. In this case, a program for implementing this function may be recorded on a computer-readable recording medium, and the program may be loaded into a computer system and executed. Note that the term "computer system" here includes hardware such as an OS and peripheral devices. Furthermore, "computer-readable recording medium" refers to portable media such as flexible disks, optical magnetic disks, ROMs, and CD-ROMs, as well as storage devices such as hard disks built into a computer system. Furthermore, "computer-readable recording medium" may also include devices that dynamically store programs for a short period of time, such as communication lines used when transmitting programs over a network such as the Internet or over a communication line such as a telephone line, or devices that store programs for a fixed period of time, such as volatile memory within a computer system that serves as a server or client. The program may also be designed to implement some of the above-described functions, or may be capable of implementing the above-described functions in combination with a program already stored in the computer system, or may be implemented using a programmable logic device such as an FPGA.

The above describes in detail an embodiment of the present invention with reference to the drawings, but the specific configuration is not limited to this embodiment and includes designs that do not deviate from the gist of the present invention.

10: Calibration support device 130: Acquisition unit 131: Division unit 132: Connection unit 133: Evaluation unit

Claims (9)

an acquisition unit that acquires a target sentence to be proofread; a division unit that divides the target sentence into at least three or more divided sentences;
a concatenation unit that generates a concatenated sentence that is formed by concatenating different divided sentences from among the divided sentences and that is shorter than the target sentence;
an evaluation unit that extracts character strings that are candidates for spelling variations in the concatenated sentence;
A calibration support device comprising: The evaluation unit
comparing the target word included in the concatenated sentence with registered words registered in list information indicating a list of words for which spelling variations are to be determined;
If the target word does not match or is not similar to the registered word, register the target word in the list information;
If the target word is similar to the registered word, add the target word to an element of the registered word similar to the target word in the list information.
The calibration support device according to claim 1 . the evaluation unit determines, among the registered words registered in the list information, the registered words and words added to the elements of the registered words as candidates for spelling variations;
The calibration support device according to claim 2 . and a device control unit that merges, based on the list information generated corresponding to each of the linked sentences, registered words that are registered in multiple pieces of list information, among the registered words registered in each piece of list information.
The calibration support device according to claim 3 . the concatenation unit generates the concatenated sentences corresponding to all combinations of the two mutually different divided sentences.
The calibration support device according to any one of claims 1 to 4. the evaluation unit executes in parallel a process of extracting character strings that are candidates for spelling variations for each of the concatenated sentences;
The calibration support device according to any one of claims 1 to 5. the dividing unit divides the target sentence using any one of a period, a symbol, or a comma in the target sentence as a boundary;
The calibration support device according to any one of claims 1 to 6. A proofreading support method performed by a computer, comprising: an acquisition unit acquiring a target sentence to be proofread;
a division unit that divides the target sentence into at least three or more divided sentences;
a concatenation unit that concatenates different segments of the segmented sentences to generate a concatenated sentence that is shorter than the target sentence;
an evaluation unit extracting character strings that are candidates for spelling variations from the concatenated sentence;
Calibration support method. A program for causing a computer to operate as the calibration support device described in any one of claims 1 to 7, and for causing the computer to function as each unit included in the calibration support device.