Control program, control apparatus, and control method

Abstract

To control arrangement of a plurality of words displayed, the words being included in a set of words.SOLUTION: A computer receives an instruction to designate a relationship between a position of a first word and a position of a second word displayed on a display device, out of a plurality of words included in a set of words. The computer causes a third word and a fourth word to be displayed on the display device so that a relationship between a position of the third word and a position of the fourth word out of the multiple words may be a predetermined relationship according to the designated relationship between the position of the first word and the position of the second word, on the basis of similarities between a first relation between words and a second relation between words. The third word is identified from among the multiple words other than the first word, on the basis of similarities between characteristics of the first word and characteristics of words other than the first word out of the multiple words. The first relationship between words is a relationship between the characteristics of the first word and the characteristics of the second word. The second relationship between words is a relationship between the characteristics of the third word and the characteristic of the fourth word.SELECTED DRAWING: Figure 9

Description

[Technical field]

[0001] The present invention relates to a control technique for controlling the display of words. [Background technology]

[0002] A word cloud is sometimes used to provide an overview of multiple words contained in text data. A word cloud is a collection of words arranged in a specified area of ​​a display screen, and the more frequently a word appears in the text data, the larger it is displayed.

[0003] Regarding word clouds, a technique for generating word clouds using Python (registered trademark) is known (see, for example, Non-Patent Document 1). A distributed representation of words is also known (see, for example, Non-Patent Document 2). A technique for identifying similarities in relationships between words using a language model is also known (see, for example, Non-Patent Document 3). [Prior art documents] [Non-patent literature]

[0004] [Non-Patent Document 1] “GitHub - amueller / word_cloud: A little word cloud generator in Python”, [online], GitHub, [Retrieved February 10, 2023], Internet<URL:https: / / github.com / amueller / word_cloud> [Non-Patent Document 2] T. Mikolov et al., “Distributed Representations of Words and Phrases and their Compositionality”, NIPS'13: Proceedings of the 26th International Conference on Neural Information Processing Systems - Volume 2, December 2013, pages 3111-3119. [Non-Patent Document 3] A. Ushio et al., “BERT is to NLP what AlexNet is to CV: Can Pre-Trained Language Models Identify Analogies?”, Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, August 2021, pages 3609-3624. Summary of the Invention [Problem to be solved by the invention]

[0005] When generating a word cloud using a program, it can be difficult to understand the arrangement of words contained in the word cloud.

[0006] This problem is not limited to word clouds, but occurs when displaying multiple words contained in various word sets.

[0007] In one aspect, the present invention aims to control the arrangement of displayed words when displaying a plurality of words included in a word set. [Means for solving the problem]

[0008] In one proposal, the control program causes the computer to execute the following processes.

[0009] The computer receives an instruction specifying a relationship between a position of a first word and a position of a second word displayed by the display device among a plurality of words included in the word set. The computer causes the display device to display the third word and the fourth word such that a relationship between a position of a third word and a position of a fourth word among the plurality of words is a predetermined relationship according to the relationship between the position of the first word and the position of the second word specified in the instruction, based on a similarity between the first relationship between the words and the second relationship between the words.

[0010] The third word is identified from among the multiple words other than the first word based on a similarity between a feature of the first word and a feature of the multiple words other than the first word. The first relationship between the words is a relationship between the feature of the first word and the feature of the second word. The second relationship between the words is a relationship between the feature of the third word and the feature of the fourth word. Effect of the Invention

[0011] According to one aspect, when displaying a plurality of words included in a word set, the arrangement of the displayed words can be controlled. [Brief description of the drawings]

[0012] [Figure 1] FIG. 1 illustrates a word cloud. [Diagram 2] FIG. 1 shows two word clouds for comparison. [Diagram 3] FIG. 2 is a functional configuration diagram of a control device according to the embodiment. [Figure 4] 13 is a flowchart of a control process. [Diagram 5] FIG. 2 is a functional configuration diagram showing a specific example of a control device. [Figure 6] FIG. 13 is a diagram showing a noteworthy word and surrounding words included in a change instruction. [Figure 7] FIG. 13 is a diagram showing a focus word and focus words similar to the focus word; [Figure 8] FIG. [Figure 9] FIG. 13 shows the modified word cloud. [Figure 10] FIG. 1 is a diagram showing word combinations. [Figure 11] FIG. 2 is a diagram illustrating word vectors. [Figure 12] FIG. 13 is a diagram showing a difference vector. [Figure 13] FIG. 13 is a diagram showing the arrangement of attention words. [Figure 14] FIG. 13 is a diagram showing the arrangement of surrounding words. [Figure 15] 13 is a flowchart of a word arrangement control process. [Figure 16] FIG. 13 is a diagram showing peripheral words common to a plurality of attention words. [Figure 17] FIG. 13 is a diagram showing words that are selected when no distinction is made between a focus word and peripheral words. [Figure 18] FIG. 13 is a diagram showing words selected when a focus word is distinguished from peripheral words. [Figure 19] FIG. 2 is a hardware configuration diagram of an information processing device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] Hereinafter, the embodiments will be described in detail with reference to the drawings.

[0014] When a user browses through multiple words in a word cloud, it is desirable to optimize the placement of words, for example, placing related words close to each other makes it easier to understand the relationships between the words.

[0015] FIG. 1 shows an example of a word cloud. FIG. 1(a) shows an example of a word cloud including "FFF", "CCC", "light", and "big". "FFF" and "CCC" are words that represent hypothetical product names of PCs (Personal Computers), and "light" and "big" are words that represent characteristics of PCs. In the example of FIG. 1, "light" is a characteristic that corresponds to "FFF", and "big" is a characteristic that corresponds to "CCC".

[0016] In the example of Figure 1(a), “light”, which is related to “FFF”, is placed near “FFF”, and “big”, which is related to “CCC”, is placed near “CCC”, making it easy to see the relationship between the words.

[0017] Figure 1(b) shows another example of a word cloud that includes the four words in Figure 1(a). In the example of Figure 1(b), "light," which is related to "FFF," is not placed near "FFF" but near "CCC," making it difficult to see the relationship between the words.

[0018] Furthermore, when comparing multiple word clouds, it is easier to compare them if the word arrangements in the word clouds are similar.

[0019] 2 shows examples of two word clouds to be compared. Word cloud 201 includes "FFF", "CCC", "light", and "big", and word cloud 202 includes "MMM", "SSS", "thin", and "light". "MMM" and "SSS" in word cloud 202 are words representing a hypothetical product name of a PC, and "thin" and "light" are words representing the characteristics of the PC.

[0020] In this example, there may be differences between the arrangement of words in word cloud 202 and the arrangement of words in word cloud 201, making comparison of word cloud 201 and word cloud 202 difficult.

[0021] In one method of generating a word cloud, the positions of words included in the word cloud are randomly determined each time the word cloud is generated. In this case, if the arrangement of words included in a word cloud once generated does not meet the user's intention, it is possible to regenerate the word cloud. However, even if the word cloud is generated multiple times, the arrangement of words may still be difficult to understand, or unnecessary words may still be included. Therefore, the word cloud that the user intends may not necessarily be obtained.

[0022] Furthermore, when a user modifies a program for generating the word cloud while checking the generated word cloud, the user's workload increases.

[0023] 3 shows an example of the functional configuration of a control device according to an embodiment. A control device 301 in FIG.

[0024] Fig. 4 is a flowchart showing an example of a control process performed by the control device 301 in Fig. 3. First, the receiving unit 311 receives an instruction to specify the relationship between the position of a first word and the position of a second word displayed by the display device among a plurality of words included in a word set (step 401).

[0025] Next, based on the similarity between the first relationship between the words and the second relationship between the words, the control unit 312 causes the display device to display the third word and the fourth word such that the relationship between the positions of the third word and the fourth word among the multiple words is a predetermined relationship that corresponds to the relationship between the positions of the first word and the second word specified in the instruction (step 402).

[0026] The third word is identified from among the multiple words other than the first word based on a similarity between a feature of the first word and a feature of the multiple words other than the first word. The first relationship between the words is a relationship between the feature of the first word and the feature of the second word. The second relationship between the words is a relationship between the feature of the third word and the feature of the fourth word.

[0027] According to the control device 301 in FIG. 3, when a plurality of words included in a word set are displayed, the arrangement of the displayed words can be controlled.

[0028] Fig. 5 shows a specific example of the control device 301 in Fig. 3. The control device 501 in Fig. 5 includes a display control unit 511, a reception unit 512, a focus word identification unit 513, a surrounding word identification unit 514, a display device 515, and a storage unit 516. The reception unit 512 corresponds to the reception unit 311 in Fig. 3, and the display control unit 511, the focus word identification unit 513, and the surrounding word identification unit 514 correspond to the control unit 312 in Fig. 3.

[0029] The storage unit 516 stores word cloud information 521. The word cloud information 521 includes word information for each of a plurality of words extracted from text data. As the text data, a part or all of the text of various documents is used. The text data may be text of a web page, a paper, a book, or the like.

[0030] Each piece of word information includes the character string of the word, the display form, and position information. The character string represents a string of one or more characters included in the word. The display form represents the size, color, font, etc. of the character string. The position information represents the position where the character string is placed within the display area of ​​the word cloud. The multiple words included in the word cloud information 521 correspond to a word set.

[0031] The display control unit 511 instructs the display device 515 to display a word cloud. Based on the instruction from the display control unit 511, the display device 515 displays the word cloud in a display area on the screen using the word cloud information 521. This allows the user to get an overview of the words included in the displayed word cloud.

[0032] It should be noted that not all words included in the word cloud information 521 are necessarily displayed within the display area. For example, if the area of ​​the display area is small compared to the number of words included in the word cloud information 521, and it is difficult to arrange all the words within the display area, some words may not be displayed.

[0033] When the user wishes to change the arrangement of words included in the word cloud, the user inputs a change instruction 522 to the control device 501. The receiving unit 512 receives the change instruction 522 input by the user and stores it in the storage unit 516.

[0034] The modification instruction 522 is, for example, an instruction to modify the relationship between the position of the attention word WA1 and the position of the peripheral word WB1 so that the distance between the position of the attention word WA1 and the position of the peripheral word WB1 among the words included in the word cloud is decreased. The attention word WA1 and the peripheral word WB1 correspond to the first word and the second word, respectively.

[0035] The attention word identification unit 513 refers to the change instruction 522 and calculates a similarity S1 between the feature of the attention word WA1 and the feature of other words included in the word cloud information 521. Then, the attention word identification unit 513 identifies an attention word WA2 similar to the attention word WA1 from among the other words based on the similarity S1.

[0036] The attention word identifying unit 513 identifies, for example, a word having a similarity S1 greater than a threshold value TH1 as an attention word WA2 from among other words included in the word cloud information 521. The attention word WA2 corresponds to the third word.

[0037] The surrounding word identification unit 514 calculates the similarity S2 between the relationship R1 between words and the relationship R2 between words. Then, the surrounding word identification unit 514 identifies the surrounding word WB2 from among the words included in the word cloud information 521 other than the attention word WA2 based on the similarity S2. The relationship R1 is the relationship between the characteristics of the attention word WA1 and the characteristics of the surrounding word WB1. The relationship R2 is the relationship between the characteristics of the attention word WA2 and the characteristics of other words included in the word cloud information 521.

[0038] The surrounding word specification unit 514 specifies, for example, words having a similarity S2 greater than a threshold value TH2 as surrounding words WB2 from among other words included in the word cloud information 521. The surrounding words WB2 correspond to the fourth word.

[0039] The display control unit 511 determines the positions of the attention word WA2 and the surrounding words WB2 so that the positions of the attention word WA2 and the surrounding words WB2 have a predetermined relationship, and changes the position information of the attention word WA2 and the surrounding words WB2 included in the word cloud information 521. The predetermined relationship is, for example, a relationship in which the surrounding words WB2 are arranged within a predetermined distance from the position of the attention word WA2. Then, the display control unit 511 instructs the display device 515 to display the word cloud.

[0040] Based on an instruction from the display control unit 511, the display device 515 uses the changed word cloud information 521 to display a word cloud within the display area on the screen.

[0041] When a change instruction 522 is input to move the surrounding word WB1 closer to the attention word WA1, the surrounding word WB2 can be positioned within a predetermined distance from the attention word WA2 to move the surrounding word WB2 closer to the attention word WA2 in accordance with the change instruction 522.

[0042] Fig. 6 shows an example of the attention word WA1 and the surrounding words WB1 included in the change instruction 522. "FFF", "fast", "delivery", and "big" are displayed in the display area 601 in Fig. 6. The user selects "FFF" as the attention word WA1 and "fast" as the surrounding word WB1. The user then inputs a change instruction 522 to move "fast" closer to "FFF". This changes the position of "fast" from P1 to P2.

[0043] 7 shows an example of an attention word WA2 similar to the attention word WA1. The attention word identification unit 513 calculates a similarity S1 between the feature of "FFF" included in the change instruction 522 in FIG. 6 and the features of other words included in the word cloud information 521. Then, the attention word identification unit 513 identifies "CCC", which is similar to "FFF", as the attention word WA2 based on the similarity S1. "CCC" is a word that was not displayed in the display area 601 in FIG. 6.

[0044] 8 shows an example of surrounding words WB2. The surrounding word identification unit 514 calculates a similarity S2 between the relationship R1 between words and the relationship R2 between words. The relationship R1 is a relationship between the feature of "FFF" and the feature of "fast". The relationship R2 is a relationship between the feature of "CCC" and the feature of another word included in the word cloud information 521. Then, the surrounding word identification unit 514 identifies "big" as a surrounding word WB2 from among the other words included in the word cloud information 521 based on the similarity S2.

[0045] The display control unit 511 performs an adjustment to change the position of "Large" so that "Large" approaches "CCC." As a result, the position of "Large" is changed from P3 to P4.

[0046] Fig. 9 shows an example of the word cloud after the change. The display control unit 511 hides "Delivery", which is unrelated to the change instruction 522 in Fig. 6. As a result, "Delivery" is deleted from the display area 601.

[0047] In the word cloud of FIG. 9, "fast" related to "FFF" is placed near "FFF", and "big" related to "CCC" is placed near "CCC". Furthermore, since the unnecessary word "delivery" has been deleted, the relationship between words is easy to understand. Note that, in the above example, the change instruction 522 is an instruction to change the words so as to decrease the distance between the words, but the change instruction may be an instruction to increase (move farther away) the distance between the words, and an adjustment may be made to increase the distance between words other than the target of the instruction in response to the instruction.

[0048] 5, the arrangement of words in the word cloud can be appropriately controlled in accordance with the change instruction 522 input by the user. Instead of the change instruction 522, the user may input to the control device 501 an instruction to fix the relationship between the position of the attention word WA1 and the position of the peripheral word WB1.

[0049] As a feature of a word, for example, a word vector that represents the word is used. Since a word vector is generated from the context in which the word is used, two words that are used in similar contexts will also have similar word vectors. The context represents other words or sentences in a sentence, words before and after, etc.

[0050] When a word vector is used as a feature of a word, the attention word identification unit 513 calculates, as a similarity S1, the inter-vector similarity between the word vector VA1 of the attention word WA1 and the word vector V of another word included in the word cloud information 521. As the inter-vector similarity, a cosine similarity, the inverse of the inter-vector distance, etc. can be used.

[0051] The attention word identification unit 513 may filter words using part-of-speech information of the attention word WA1, etc. For example, if the attention word WA1 is a noun, words other than nouns are excluded from the similarity calculation.

[0052] The surrounding word identification unit 514 can calculate the similarity S2 between the relationship R1 between words and the relationship R2 between words, for example, by using the method described in Non-Patent Document 2. In this case, the surrounding word identification unit 514 calculates a difference vector D1 representing the relationship R1 by the following formula using the word vector VA1 of the attention word WA1 and the word vector VB1 of the surrounding word WB1.

[0053] D1=VA1-VB1 (1)

[0054] Furthermore, the surrounding word specification unit 514 uses the word vector VA2 of the attention word WA2 and the word vector V of another word included in the word cloud information 521 to calculate a difference vector D2 representing a relationship R2 according to the following formula.

[0055] D2=VA2-V (2)

[0056] Then, the surrounding word specification unit 514 calculates the inter-vector similarity between the difference vector D1 and the difference vector D2 as the similarity S2.

[0057] By using word vectors as the features of words, the similarity S1 and the similarity S2 can be easily calculated.

[0058] Word vector VA1 is an example of a first vector. Word vector VB1 is an example of a second vector. Word vector VA2 is an example of a third vector. Word vector V of another word is an example of a fourth vector. Of the word vectors V of another word, the word vector V of surrounding word WB2 is an example of a fifth vector. Difference vector D1 is an example of a sixth vector. Difference vector D2 representing the difference between word vector VA2 and word vector V of surrounding word WB2 is an example of a seventh vector. The subtraction between word vectors in equations (1) and (2) is an example of a predetermined calculation.

[0059] The surrounding word identification unit 514 can also calculate the similarity S2 using the method described in Non-Patent Document 3. In this case, the surrounding word identification unit 514 generates a given sentence X using a combination of the attention word WA1 and the surrounding words WB1, and a combination of the attention word WA2 and other words included in the word cloud information 521. Then, the surrounding word identification unit 514 calculates an evaluation value E(X) of the sentence X using a language model that is a trained machine learning model, thereby acquiring the evaluation value E(X), and uses the evaluation value E(X) as the similarity S2.

[0060] In machine learning for generating a language model, for example, a plurality of sentences prepared in advance and an evaluation value for each sentence are used as training data.

[0061] 10 shows an example of a word combination. C1 represents a combination of a focus word WA1 and a surrounding word WB1. "Japan" corresponds to the focus word WA1, and "Tokyo" corresponds to the surrounding word WB1.

[0062] C2 represents a combination of the attention word WA2 and other words. “China” corresponds to the attention word WA2, and “Beijing”, “Paris”, “apple”, and “go” correspond to other words included in the word cloud information 521.

[0063] The surrounding word specification unit 514 generates sentence X by embedding C1 and C2 in the following template.

[0064] ___ is to ___ as ___ is to ___

[0065] As a result, the following sentences X1 to X4 are generated.

[0066] X1: Japan is to Tokyo as China is to Beijing X2: Japan is to Tokyo as China is to Paris X3: Japan is to Tokyo as China is to apple X4: Japan is to Tokyo as China is to go

[0067] The surrounding word specification unit 514 calculates the evaluation value E(Xi) of the sentence Xi (i=1 to 4) using the language model. As a result, the following evaluation values ​​E(X1) to E(X4) are obtained.

[0068] E(X1)=0.9 E(X2)=0.8 E(X3)=0.2 E(X4)=0.1

[0069] By calculating the evaluation value E(Xi) of the sentence Xi using the language model, the similarity S2 can be calculated with high accuracy.

[0070] Next, a specific example of the word arrangement control process based on the change instruction 522 will be described. "FFF", "CCC", "PC", "fast", and "big" are used as words included in the word cloud information 521. The attention word WA1 included in the change instruction 522 is "FFF", and the surrounding word WB1 is "fast".

[0071] 11 shows examples of word vectors of words included in word cloud information 521. Word vector 1101 is the word vector of "FFF". Word vector 1102 is the word vector of "CCC". Word vector 1103 is the word vector of "PC". Word vector 1104 is the word vector of "fast". Word vector 1105 is the word vector of "big".

[0072] In this example, word vectors 1101 to 1105 are shown as two-dimensional vectors, but word vectors 1101 to 1105 may be N-dimensional vectors (N is an integer equal to or greater than 3).

[0073] The attention word identification unit 513 calculates the cosine of the angle between the word vector 1101 and the word vector of the other word as the similarity S1 between "FFF" and the other word. For example, the cosine of the angle α between the word vector 1101 and the word vector 1102, cos(α), is used as the similarity S1 between "FFF" and "CCC". The cosine of the angle β between the word vector 1101 and the word vector 1105, cos(β), is used as the similarity S1 between "FFF" and "big".

[0074] In this example, the following similarity S1 is calculated for "CCC", "PC", "fast", and "large".

[0075] CCC S1=0.98 PC S1=0.91 Fast S1=0.2 Large S1=0.1

[0076] The attention word specifying unit 513 selects, as the attention word WA2, words that satisfy the following conditions from among "CCC", "PC", "fast", and "loud" in descending order of similarity S1.

[0077] Condition Q1: Similarity S1 is greater than threshold TH1. Condition Q2: The number of attention words is equal to or less than the upper limit LA.

[0078] The threshold value TH1 is set to 0.9, and the upper limit value LA is set to 2. In this case, the similarity S1 of "CCC" is the largest and is greater than the threshold value TH1, so "CCC" is selected first as the attention word WA2. As a result, the attention words become "FFF" and "CCC", so the number of attention words becomes two.

[0079] The similarity S1 of "PC" is the second largest and is also larger than the threshold TH1. However, if "PC" is selected as the attention word WA2, the number of attention words will be three, which exceeds the upper limit LA, and therefore "PC" is not selected. The similarity S1 of "fast" and "big" is smaller than the threshold TH1, so "fast" and "big" are not selected.

[0080] 12 shows an example of difference vectors representing relationships between words. Difference vector 1201 is a difference vector representing relationship R1 between word vector 1101 of "FFF" and word vector 1104 of "fast". Difference vector 1202 is a difference vector representing relationship R2 between word vector 1102 of "CCC" and word vector 1105 of "big".

[0081] The surrounding word identification unit 514 uses the relationship between the word vector 1102 of "CCC" and the word vectors of words other than "FFF" and "CCC" as the relationship R2. Then, the surrounding word identification unit 514 calculates the cosine of the angle between the difference vector 1201 and the difference vector representing the relationship R2 as the similarity S2 between the relationship R1 and the relationship R2.

[0082] For example, when the relationship between the word vector 1102 of "CCC" and the word vector 1105 of "big" is used as the relationship R2, cos(γ), which is the cosine of the angle γ between the difference vector 1201 and the difference vector 1202, is used as the similarity S2.

[0083] In this example, the following similarity S2 is calculated for "PC", "fast", and "big".

[0084] Large S2=0.83 Fast S2=0.81 PC S2=0.2

[0085] The surrounding word specification unit 514 selects, as surrounding words WB2, words that satisfy the following conditions from among "PC", "fast", and "loud" in descending order of similarity S2.

[0086] Condition Q3: Similarity S2 is greater than threshold TH2. Condition Q4: The number of peripheral words WB2 for each attention word WA2 is equal to or less than the upper limit LB.

[0087] The threshold value TH2 is set to 0.7, and the upper limit value LB is set to 1. In this case, since the similarity S2 of "big" is the largest and is greater than the threshold value TH2, "big" is selected first as the surrounding word WB2.

[0088] The similarity S2 of "fast" is the second largest and is also larger than the threshold TH2. However, if "fast" is selected as the surrounding word WB2, the number of surrounding words WB2 for "CCC" becomes two, which exceeds the upper limit LB, and therefore "fast" is not selected. The similarity S2 of "PC" is smaller than the threshold TH2, so "PC" is not selected.

[0089] The surrounding word specification unit 514 can also select surrounding words for "FFF" in the same way as the surrounding words for "CCC." In this case as well, LB or less surrounding words are selected.

[0090] 13 shows an example of the arrangement of attention words. The display control unit 511 changes the position information of "FFF" and "CCC" so that the attention words "FFF" and "CCC" are sparsely arranged in the display area 1301. As a result, "FFF" and "CCC" are arranged at positions that are an appropriate distance apart.

[0091] 14 shows an example of the arrangement of peripheral words. The display control unit 511 changes the position information of "fast", which is a peripheral word for "FFF", so that "fast" is arranged near "FFF". Furthermore, the display control unit 511 changes the position information of "big", which is a peripheral word for "CCC", so that "big" is arranged near "CCC".

[0092] In this way, the user only needs to specify the relationship between the position of "FFF" and the position of "fast", and "CCC", which is similar to "FFF", is automatically extracted, and "FFF" and "CCC" are placed in appropriate positions. Furthermore, "big", which has a similar relationship with "CCC" as the relationship between the characteristics of "FFF" and "fast", is automatically extracted, and "fast" and "big" are placed in appropriate positions.

[0093] The change instruction 522 is not limited to an instruction to move the peripheral word WB1 closer to the attention word WA1, and may be an instruction representing another change. For example, the change instruction 522 may be an instruction to move the peripheral word WB1 to the left or right of the attention word WA1, or an instruction to move the peripheral word WB1 away from the attention word WA1.

[0094] Fig. 15 is a flowchart showing an example of a word arrangement control process performed by the control device 501 of Fig. 5. First, the display device 515 uses the word cloud information 521 to display a word cloud within a display area on the screen (step 1501).

[0095] Next, the user inputs a change instruction 522 to the control device 501, and the receiving unit 512 receives the input change instruction 522 (step 1502).

[0096] Next, the attention word identification unit 513 calculates a similarity S1 between the characteristics of the attention word WA1 included in the change instruction 522 and the characteristics of other words included in the word cloud information 521 (step 1503). Then, the attention word identification unit 513 identifies an attention word WA2 similar to the attention word WA1 based on the similarity S1 (step 1504).

[0097] Next, the surrounding word identification unit 514 calculates the similarity S2 between the relationship R1 between words and the relationship R2 between words (step 1505). Then, the surrounding word identification unit 514 identifies surrounding words WB2 from among the words included in the word cloud information 521 other than the attention word WA2 based on the similarity S2 (step 1506).

[0098] Next, the display control unit 511 determines the positions of the attention word WA2 and the surrounding words WB2 so that the positions of the attention word WA2 and the surrounding words WB2 have a predetermined relationship (step 1507). Then, the display control unit 511 changes the position information of the attention word WA2 and the surrounding words WB2 included in the word cloud information 521. The display control unit 511 also changes the position information of the attention word WA1 and the surrounding words WB1 as necessary.

[0099] Next, the display control unit 511 instructs the display device 515 to display the word cloud, and the display device 515 uses the changed word cloud information 521 to display the word cloud within the display area (step 1508).

[0100] Fig. 16 shows an example of peripheral words common to multiple attention words. In the specific example of the word placement control process described above, assume that "CCC" is selected as the attention word WA2 similar to "FFF", and "big" and "fast" are selected as peripheral words WB2 for "CCC". "Fast" is a peripheral word common to "FFF" and "CCC".

[0101] In this case, the display control unit 511 arranges two "fast"s in the display area 1301. One "fast" is arranged near "FFF", and the other "fast" is arranged near "CCC".

[0102] When determining the size of a string, the two "fast"s are treated separately: for example, a "fast" placed near "FFF" is treated as the word "fast-FMV".

[0103] Figure 17 shows an example of words selected when attention words and peripheral words are not distinguished. If attention words and peripheral words are not distinguished and all words are treated equally, the influence of unnecessary relationships may become noise.

[0104] In the example of Fig. 17, it is assumed that a word cloud is generated from text data of product reviews in order to obtain an overview of evaluations of a PC brand. The user selects "FFF", "fast", and "thin", and inputs a modification instruction 522 to move "fast" and "thin" closer to "FFF". In this case, "thick" is highly likely to be selected as a word W that satisfies the following condition:

[0105] Condition Z1 The relationship between the characteristics of W and the characteristics of “slow” is similar to the relationship between the characteristics of “thin” and the characteristics of “fast”. The relationship between the characteristics of condition Z2 W and the characteristics of “CCC” is similar to the relationship between the characteristics of “thin” and the characteristics of “FFF”.

[0106] However, since the user's change instruction 522 does not intend to make "thin" closer to "fast", the relationship between the feature of "thin" and the feature of "fast" is an unnecessary relationship. Similarly, the relationship between the feature of W and the feature of "slow" is also an unnecessary relationship.

[0107] 18 shows an example of a word selected when the attention word and the peripheral words are distinguished. In this example, the user selects "FFF" as the attention word WA1, selects "fast" and "thin" as the peripheral words WB1, and inputs a modification instruction 522 to move "fast" and "thin" closer to "FFF". In this case, only the condition Z2 is used, and the possibility that "thick" will be selected as the word W increases.

[0108] In this way, by distinguishing between the word of interest and the surrounding words, it is possible to eliminate the influence of unnecessary relationships and select appropriate words in line with the user's intention.

[0109] In addition, when only one peripheral word is included in the change instruction 522, it is not necessarily necessary to distinguish between the attention word and the peripheral words. However, in reality, the word cloud after the change may change depending on which word is designated as the attention word, so there is a high possibility that a more appropriate word cloud will be generated by distinguishing between the two.

[0110] The configurations of the control device 301 in Fig. 3 and the control device 501 in Fig. 5 are merely examples, and some of the components may be omitted or changed depending on the application or conditions of the control device. For example, the display device 515 in Fig. 5 may be provided in a user terminal device outside the control device 501.

[0111] The flowcharts of FIG. 4 and FIG. 15 are merely examples, and some of the processes may be omitted or changed depending on the configuration or conditions of the control device.

[0112] 1, 2, 6 to 9, 13, 14, and 16 to 18 are merely examples, and the word clouds change depending on the words included in the word cloud information 521. The word combinations shown in FIG. 10 are merely examples, and the word combinations change depending on the words included in the word cloud information 521. The word vectors shown in FIG. 11 and the difference vectors shown in FIG. 12 are merely examples, and the word vectors and difference vectors change depending on the words included in the word cloud information 521.

[0113] The formulas (1) and (2) are merely examples, and the surrounding word specification unit 514 may use other formulas to calculate the vectors representing the relations R1 and R2.

[0114] Fig. 19 shows an example of the hardware configuration of an information processing device (computer) used as the control device 301 in Fig. 3 and the control device 501 in Fig. 5. The information processing device in Fig. 19 includes a CPU (Central Processing Unit) 1901, a memory 1902, an input device 1903, an output device 1904, an auxiliary storage device 1905, a medium drive device 1906, and a network connection device 1907. These components are hardware and are connected to each other by a bus 1908.

[0115] The memory 1902 is, for example, a semiconductor memory such as a Read Only Memory (ROM) or a Random Access Memory (RAM), and stores programs and data used in processing. The memory 1902 may operate as the storage unit 516 in FIG.

[0116] The CPU 1901 (processor) executes a program using the memory 1902, for example, to operate as the reception unit 311 and the control unit 312 in Fig. 3. The CPU 1901 executes a program using the memory 1902, to operate also as the display control unit 511, the reception unit 512, the attention word identification unit 513, and the surrounding word identification unit 514 in Fig. 5.

[0117] The input device 1903 is, for example, a keyboard, a pointing device, etc., and is used to input instructions or information from a user. The output device 1904 is, for example, a display device, a printer, etc., and is used to output inquiries or instructions to a user and processing results. The output device 1904 may operate as the display device 515 in Fig. 5, and the processing results may be a word cloud.

[0118] The auxiliary storage device 1905 is, for example, a magnetic disk device, an optical disk device, a magneto-optical disk device, a tape device, or the like. The auxiliary storage device 1905 may be a hard disk drive or an SSD (Solid State Drive). The information processing device can store programs and data in the auxiliary storage device 1905 and use them by loading them into the memory 1902. The auxiliary storage device 1905 may operate as the storage unit 516 in FIG. 5.

[0119] The medium drive device 1906 drives the portable recording medium 1909 and accesses the recorded contents. The portable recording medium 1909 is a memory device, a flexible disk, an optical disk, a magneto-optical disk, etc. The portable recording medium 1909 may be a CD-ROM (Compact Disk Read Only Memory), a DVD (Digital Versatile Disk), a USB (Universal Serial Bus) memory, etc. A user can store programs and data in the portable recording medium 1909 and load them into the memory 1902 for use.

[0120] Thus, the computer-readable recording medium that stores the program and data used in the processing is a physical (non-transitory) recording medium such as the memory 1902, the auxiliary storage device 1905, or the portable recording medium 1909.

[0121] The network connection device 1907 is a communication interface circuit that is connected to a communication network such as a WAN (Wide Area Network) or a LAN (Local Area Network) and performs data conversion associated with communication. The information processing device can receive programs and data from an external device via the network connection device 1907 and load them into the memory 1902 for use.

[0122] 5 is provided in a user terminal device outside the control device 501, the CPU 1901 receives a change instruction 522 from the user terminal device via the network connection device 1907. Then, the CPU 1901 transmits control information instructing the user terminal device to display the word cloud via the network connection device 1907.

[0123] 19, some of the components may be omitted or changed depending on the purpose or conditions of the information processing device. For example, when the portable recording medium 1909 or the communication network is not used, the medium drive device 1906 or the network connection device 1907 can be omitted.

[0124] Although the disclosed embodiments and their advantages have been described in detail, it will be appreciated that those skilled in the art may make various modifications, additions and omissions therein without departing from the scope of the invention as clearly set forth in the claims.

[0125] The following notes are further disclosed regarding the embodiment described with reference to FIGS. (Appendix 1) accepting an instruction to specify a relationship between a position of a first word among a plurality of words included in a word set and a position of a second word among the plurality of words and displayed by the display device; displaying the third word and the fourth word on the display device such that a relationship between a position of a third word among the plurality of words and a position of a fourth word among the plurality of words has a predetermined relationship according to a relationship between the position of the first word and the position of the second word specified in the instruction, based on a similarity between the first relation between the words and the second relation between the words; The process is executed by a computer, the third word is identified from among the plurality of words other than the first word based on a similarity between a feature of the first word and a feature of a word other than the first word among the plurality of words; the first relationship between the words is a relationship between a feature of the first word and a feature of the second word; The control program is characterized in that the second relationship between the words is a relationship between a feature of the third word and a feature of the fourth word. (Appendix 2) the feature of the first word is a first vector representing the first word; the feature of the second word is a second vector representing the second word; the feature of the third word is a third vector representing the third word; the feature of the word other than the first word is a fourth vector representing the word other than the first word; the feature of the fourth word is a fifth vector representing the fourth word; The control program includes: performing a predetermined calculation using the first vector and the second vector to obtain a sixth vector that represents a first relationship between the words; performing the predetermined operation using the third vector and the fifth vector to obtain a seventh vector that represents a second relationship between the words; 2. The control program according to claim 1, further causing the computer to execute a process. (Appendix 3) The control program includes: obtaining an evaluation value of a predetermined sentence generated using a combination of the first word and the second word and a combination of the third word and the fourth word, using a language model; The process is further performed by the computer. 2. The control program according to claim 1, wherein the evaluation value of the specified sentence is used as a similarity between the first relationship between the words and the second relationship between the words. (Appendix 4) The control program includes: identifying the fourth word from among the words other than the third word based on a plurality of third relations including the second relation, each of the third relations being between a feature of the third word and a feature of a word other than the third word among the plurality of words; 2. The control program according to claim 1, which causes the computer to execute a process. (Appendix 5) the instruction is an instruction to change a relationship between a position of the first word and a position of the second word such that a distance between the position of the first word and the position of the second word decreases; The control program according to any one of appendices 1 to 4, characterized in that the process of displaying the third word and the fourth word on the display device includes a process of displaying the third word and the fourth word on the display device so that the fourth word is positioned within a predetermined distance from the position of the third word when the similarity between the first relationship between the words and the second relationship between the words is greater than a threshold value. (Appendix 6) a receiving unit that receives an instruction specifying a relationship between a position of a first word that is displayed by a display device among a plurality of words included in a word set, and a position of a second word that is displayed by the display device among the plurality of words; a control unit that displays the third word and the fourth word on the display device based on a similarity between a first relation between words and a second relation between words such that a relationship between a position of a third word among the plurality of words and a position of a fourth word among the plurality of words has a predetermined relationship according to a relationship between a position of the first word and a position of the second word specified in the instruction; Equipped with the third word is identified from among the plurality of words other than the first word based on a similarity between a feature of the first word and a feature of a word other than the first word among the plurality of words; the first relationship between the words is a relationship between a feature of the first word and a feature of the second word; The control device is characterized in that the second relationship between the words is a relationship between a feature of the third word and a feature of the fourth word. (Appendix 7) the feature of the first word is a first vector representing the first word; the feature of the second word is a second vector representing the second word; the feature of the third word is a third vector representing the third word; the feature of the word other than the first word is a fourth vector representing the word other than the first word; the feature of the fourth word is a fifth vector representing the fourth word; The control device described in Appendix 6, characterized in that the control unit performs a predetermined calculation using the first vector and the second vector to obtain a sixth vector representing a first relationship between the words, and performs the predetermined calculation using the third vector and the fifth vector to obtain a seventh vector representing a second relationship between the words. (Appendix 8) The control device described in Appendix 6, characterized in that the control unit uses a language model to obtain an evaluation value of a predetermined sentence generated using a combination of the first word and the second word and a combination of the third word and the fourth word, and uses the evaluation value of the predetermined sentence as a similarity between the first relationship between the words and the second relationship between the words. (Appendix 9) The control device according to claim 6, characterized in that the control unit includes the second relation and identifies the fourth word from among the words other than the third word based on a plurality of third relations, each of which is between a feature of the third word and a feature of a word other than the third word among the plurality of words. (Appendix 10) the instruction is an instruction to change a relationship between a position of the first word and a position of the second word such that a distance between the position of the first word and the position of the second word decreases; The control device described in any one of Appendices 6 to 9, characterized in that, when the similarity between the first relationship between the words and the second relationship between the words is greater than a threshold, the control unit displays the third word and the fourth word on the display device so that the fourth word is positioned within a predetermined distance from the position of the third word. (Appendix 11) accepting an instruction to specify a relationship between a position of a first word among a plurality of words included in a word set and a position of a second word among the plurality of words and displayed by the display device; displaying the third word and the fourth word on the display device such that a relationship between a position of a third word among the plurality of words and a position of a fourth word among the plurality of words has a predetermined relationship according to a relationship between the position of the first word and the position of the second word specified in the instruction, based on a similarity between the first relation between the words and the second relation between the words; The computer executes the process, the third word is identified from among the plurality of words other than the first word based on a similarity between a feature of the first word and a feature of a word other than the first word among the plurality of words; the first relationship between the words is a relationship between a feature of the first word and a feature of the second word; The control method, wherein the second relationship between the words is a relationship between a feature of the third word and a feature of the fourth word. (Appendix 12) the feature of the first word is a first vector representing the first word; the feature of the second word is a second vector representing the second word; the feature of the third word is a third vector representing the third word; the feature of the word other than the first word is a fourth vector representing the word other than the first word; the feature of the fourth word is a fifth vector representing the fourth word; performing a predetermined calculation using the first vector and the second vector to obtain a sixth vector that represents a first relationship between the words; performing the predetermined operation using the third vector and the fifth vector to obtain a seventh vector that represents a second relationship between the words; 12. The control method according to claim 11, further comprising the steps of: (Appendix 13) obtaining an evaluation value of a predetermined sentence generated using a combination of the first word and the second word and a combination of the third word and the fourth word, using a language model; The computer further executes the process, The control method according to claim 11, wherein the evaluation value of the specified sentence is used as a similarity between the first relationship between the words and the second relationship between the words. (Appendix 14) identifying the fourth word from among the words other than the third word based on a plurality of third relations including the second relation, each of the third relations being between a feature of the third word and a feature of a word other than the third word among the plurality of words; 12. The control method according to claim 11, wherein the processing is executed by the computer. (Appendix 15) the instruction is an instruction to change a relationship between a position of the first word and a position of the second word such that a distance between the position of the first word and the position of the second word decreases; The control method described in any one of Appendices 11 to 14, characterized in that the process of displaying the third word and the fourth word on the display device includes a process of displaying the third word and the fourth word on the display device so that the fourth word is positioned within a predetermined distance from the position of the third word when the similarity between the first relationship between the words and the second relationship between the words is greater than a threshold value. [Explanation of symbols]

[0126] 201, 202 Word Cloud 301, 501 Control device 311, 512 Reception 312 Control Unit 511 Display control unit 513 Attention Word Identification Unit 514 Surrounding Word Identification Unit 515 Display device 516 Storage section 521 Word Cloud Information 522 Change Order 601, 1301 display area 1101~1105 Word Vector 1201, 1202 Difference Vector 1901 CPU 1902 Memory 1903 Input Device 1904 Output Device 1905 Auxiliary storage 1906 Media Drive Unit 1907 Network connection device 1908 Bus 1909 Portable recording media

Claims

1. accepting an instruction to specify a relationship between a position of a first word among a plurality of words included in a word set and a position of a second word among the plurality of words and displayed by the display device; displaying the third word and the fourth word on the display device such that a relationship between a position of a third word among the plurality of words and a position of a fourth word among the plurality of words has a predetermined relationship according to a relationship between the position of the first word and the position of the second word specified in the instruction, based on a similarity between the first relation between the words and the second relation between the words; The process is executed by a computer, the third word is identified from among the plurality of words other than the first word based on a similarity between a feature of the first word and a feature of a word other than the first word among the plurality of words; the first relationship between the words is a relationship between a feature of the first word and a feature of the second word; The control program is characterized in that the second relationship between the words is a relationship between a feature of the third word and a feature of the fourth word.

2. the feature of the first word is a first vector representing the first word; the feature of the second word is a second vector representing the second word; the feature of the third word is a third vector representing the third word; the feature of the word other than the first word is a fourth vector representing the word other than the first word; the feature of the fourth word is a fifth vector representing the fourth word; The control program includes: performing a predetermined operation using the first vector and the second vector to obtain a sixth vector that represents a first relationship between the words; performing the predetermined operation using the third vector and the fifth vector to obtain a seventh vector representing a second relationship between the words; 2. The control program according to claim 1, further causing the computer to execute a process.

3. The control program includes: obtaining an evaluation value of a predetermined sentence generated using a combination of the first word and the second word and a combination of the third word and the fourth word, using a language model; The process is further performed by the computer.

2. The control program according to claim 1, wherein the evaluation value of the predetermined sentence is used as a similarity between the first relationship between the words and the second relationship between the words.

4. The control program includes: identifying the fourth word from among the words other than the third word based on a plurality of third relations including the second relation, each of the third relations being between a feature of the third word and a feature of a word other than the third word among the plurality of words; 2. The control program according to claim 1, wherein the control program causes the computer to execute a process.

5. the instruction is an instruction to change a relationship between a position of the first word and a position of the second word such that a distance between the position of the first word and the position of the second word decreases; A control program according to any one of claims 1 to 4, characterized in that the process of displaying the third word and the fourth word on the display device includes a process of displaying the third word and the fourth word on the display device so that the fourth word is positioned within a predetermined distance from the position of the third word when the similarity between the first relationship between the words and the second relationship between the words is greater than a threshold value.

6. a receiving unit that receives an instruction specifying a relationship between a position of a first word displayed by a display device among a plurality of words included in a word set and a position of a second word displayed by the display device among the plurality of words; a control unit that displays the third word and the fourth word on the display device based on a similarity between a first relation between words and a second relation between words such that a relationship between a position of a third word among the plurality of words and a position of a fourth word among the plurality of words has a predetermined relationship according to a relationship between a position of the first word and a position of the second word specified in the instruction; Equipped with the third word is identified from among the plurality of words other than the first word based on a similarity between a feature of the first word and a feature of a word other than the first word among the plurality of words; the first relationship between the words is a relationship between a feature of the first word and a feature of the second word; The control device according to claim 1, wherein the second relationship between the words is a relationship between a feature of the third word and a feature of the fourth word.

7. accepting an instruction to specify a relationship between a position of a first word among a plurality of words included in a word set and a position of a second word among the plurality of words and displayed by the display device; displaying the third word and the fourth word on the display device such that a relationship between a position of a third word among the plurality of words and a position of a fourth word among the plurality of words has a predetermined relationship according to a relationship between the position of the first word and the position of the second word specified in the instruction, based on a similarity between the first relation between the words and the second relation between the words; The computer executes the process, the third word is identified from among the plurality of words other than the first word based on a similarity between a feature of the first word and a feature of a word other than the first word among the plurality of words; the first relationship between the words is a relationship between a feature of the first word and a feature of the second word; The control method according to claim 1, wherein the second relationship between the words is a relationship between a feature of the third word and a feature of the fourth word.

Images