Information processing device and information processing method

Abstract

The information processing device according to the present disclosure comprises: a score calculation unit (130) that calculates a score indicating how first content data resembles a character of interest on the basis of feature quantities expressing features of the first content data and the degrees of association of the character of interest and another character different from the character of interest with the first content data; and an extraction unit (140) that extracts, from the first content data, data to be associated with the character of interest on the basis of the score calculated by the score calculation unit.

Description

[Technical Field]

[0001] The present disclosure relates to an information processing device and an information processing method. [Background technology]

[0002] There is known technology that enables interaction between a user and an artificial intelligence via a network such as the Internet, and further deepens the relationship between the user and the artificial intelligence by giving the artificial intelligence a character.

[0003] To generate dialogue responses that sound like a specific character, it is necessary to prepare the utterance text of the specific character as training data for response generation. Character utterance text can be obtained from novels or animation scripts, but the amount of data that can be obtained is very small, making it difficult to use as training data for dialogue response generation.

[0004] Therefore, in order to increase the amount of spoken text that has specific characteristics such as character-likeness, a method has been proposed in which text data is manually created using crowdsourcing or the like (for example, Non-Patent Document 1). [Prior art documents] [Non-patent literature]

[0005] [Non-Patent Document 1] Zhang, Saizheng, et al., "Personalizing Dialogue Agents: I have a dog, do you have pets too?", Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics, 2018, Volume 1: Long Papers Summary of the Invention [Problem to be solved by the invention]

[0006] However, creating data manually takes time, and it is difficult to collect a large amount of data in a short period of time.

[0007] An object of the present disclosure is to provide an information processing device and an information processing method that can easily collect spoken text having specific characteristics. [Means for solving the problem]

[0008] The information processing device of the present disclosure includes a score calculation unit that calculates a score indicating the likelihood that first content data is a target character based on a feature that indicates the characteristics of the first content data and the degree of association of the target character and other characters different from the target character with the first content data, and an extraction unit that extracts data from the first content data to be associated with the target character based on the score calculated by the score calculation unit. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a schematic diagram illustrating a configuration of an example of an information processing system applicable to an embodiment. [Figure 2] 1 is a functional block diagram of an example for explaining functions of an information processing system 1 according to an embodiment. [Figure 3] 1 is a flowchart illustrating an example of a process performed by an information processing system according to an embodiment. [Figure 4] FIG. 2 is a block diagram illustrating a hardware configuration of an example of a server applicable to the embodiment. [Figure 5] FIG. 2 is a block diagram illustrating a hardware configuration of an example of a terminal device applicable to the embodiment. [Figure 6] FIG. 2 is a schematic diagram illustrating an example of large-scale text data stored in a large-scale text data storage unit, which is applicable to an embodiment. [Figure 7] 3 is a schematic diagram showing an example of dialogue data stored in a dialogue data storage unit, which is applicable to the embodiment; FIG. [Figure 8] 10 is a flowchart illustrating an example of processing by a speaker character determination unit according to the embodiment. [Figure 9] 3 is a schematic diagram showing an example of the result of a speaker character determination process by a speaker character determination unit 110 according to the embodiment. FIG. [Figure 10] 10A and 10B are schematic diagrams illustrating an example of a result of determination of the presence or absence of versatility by a versatility determination unit according to the embodiment. [Figure 11] 10 is a flowchart illustrating an example of processing by a character score calculation unit according to the embodiment. [Figure 12] 10 is a schematic diagram showing an example of a result of obtaining a character score by a character score calculation unit according to the embodiment. FIG. [Figure 13] 10 is a schematic diagram showing an example of a data extraction screen as a user interface generated and presented by a data extraction unit, which is applicable to the embodiment; FIG. [Figure 14] FIG. 4 is a schematic diagram for explaining a data extraction process performed by a data extraction unit according to an embodiment. [Figure 15] FIG. 10 is a schematic diagram illustrating an example of an utterance text viewing screen according to the embodiment. [Figure 16] FIG. 10 is a schematic diagram showing another example of the spoken text viewing screen according to the embodiment. [Figure 17] 10 is a schematic diagram showing an example of text data for a target character output by a data extraction unit according to the embodiment; FIG. [Figure 18] 10 is a flowchart illustrating an example of processing by a character score calculation unit according to a second modified example of the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. In the following embodiments, the same components are denoted by the same reference numerals, and redundant description will be omitted.

[0011] Hereinafter, embodiments of the present disclosure will be described in the following order. 1. Overview of this Disclosure 2. Overall Configuration Example According to the Embodiment 3. Details of Processing According to the Embodiment 3-1. Processing by the speaker character determination unit 3-2. Processing by the generality judgment unit 3-3. Character score calculation 3-4. Processing by the data extraction unit 4. First Modification of the Embodiment 5. Second Modification of the Embodiment 6. Third Modification of the Embodiment

[0012] 1. Overview of this Disclosure In the present disclosure, by associating each piece of text data as content data collected in large quantities from the Internet, etc., with one or more characters that appear in a specific work (referred to as the target work), it is possible to automatically generate utterances from these characters.

[0013] More specifically, in this disclosure, for spoken text (first content data) included in a large amount of text data collected from the Internet or the like (referred to as large-scale text data), a score indicating the likelihood of the spoken text being a target character among characters in a target work is calculated. The score is calculated based on feature quantities indicating the characteristics of the spoken text and the degree of relevance of each character included in the target work to the spoken text. Based on the calculated score, spoken text to be associated with the target character is extracted from the large-scale text data.

[0014] With this configuration, the present disclosure makes it possible to easily collect spoken text with specific characteristics.

[0015] [2. Overall Configuration Example of the Embodiment] Next, an example of the overall configuration according to an embodiment of the present disclosure will be described. Fig. 1 is a schematic diagram showing the configuration of an example of an information processing system applicable to the embodiment.

[0016] 1, an information processing system 1 according to an embodiment includes a server 10 and a terminal device 20 that are connected to each other via a network 2, such as the Internet. The server 10 may be a single computer or multiple computers that are connected by cloud computing technology.

[0017] The terminal device 20 is, for example, a personal computer. Alternatively, the terminal device 20 may be a portable computer such as a tablet computer or a smartphone. The terminal device 20 is connected to the network 2 via wired or wireless communication.

[0018] 2 is a functional block diagram of an example illustrating functions of an information processing system 1 according to an embodiment. In FIG. 2, the information processing system 1 includes a large-scale text data storage unit 100, a dialogue data storage unit 101, a speaker character determination unit 110, a versatility determination unit 120, a character score calculation unit 130, and a data extraction unit 140. The information processing system 1, including the large-scale text data storage unit 100, the dialogue data storage unit 101, the speaker character determination unit 110, the versatility determination unit 120, the character score calculation unit 130, and the data extraction unit 140, constitutes an information processing device as a whole. Note that, of these, a portion of the data extraction unit 140 is included in the terminal device 20, and the remaining portion is included in the server 10.

[0019] The speaker character determination unit 110, versatility determination unit 120, character score calculation unit 130, and data extraction unit 140 are configured by executing an information processing program according to the embodiment on, for example, a CPU (Central Processing Unit). However, without being limited to this, some or all of the speaker character determination unit 110, versatility determination unit 120, character score calculation unit 130, and data extraction unit 140 may be configured by hardware circuits that cooperate with each other.

[0020] The large-scale text data storage unit 100 stores large-scale text data including the above-mentioned large amount of utterance data. The dialogue data storage unit 101 stores dialogue data indicating lines spoken by each character appearing in the target work. Specific examples of the data stored in the large-scale text data storage unit 100 and the dialogue data storage unit 101 will be described later.

[0021] 3 is a flowchart illustrating an example of processing performed by the information processing system 1 according to the embodiment. Prior to the processing of the flowchart in FIG. 3, large-scale text data collected in large quantities via the Internet or the like is stored in the large-scale text data storage unit 100. Furthermore, text data of lines spoken by all characters appearing in the target work is stored in the dialogue data storage unit 101. Specific examples of the large-scale text data stored in the large-scale text data storage unit 100 and the dialogue data stored in the dialogue data storage unit 101 will be described later.

[0022] 3, in step S10, the speaker character determination unit 110 determines whether each of the utterance texts included in the large-scale text data stored in the large-scale text data storage unit 100 is likely to be an utterance of each character, based on the dialogue text (second content data) of each character stored in the dialogue data storage unit 101. In the next step S11, the versatility determination unit 120 determines whether each of the utterance texts has versatility. In the next step S12, the character score calculation unit 130 calculates a score indicating the likelihood of each of the utterance texts being the target character. In the next step S13, the data extraction unit 140 extracts target character text data 150 to be used as the utterance text of the target character, based on the processing results of the speaker character determination unit 110, the versatility determination unit 120, and the character score calculation unit 130.

[0023] Fig. 4 is a block diagram showing an example of a hardware configuration of a server 10 applicable to the embodiment. Note that, in this description, the server 10 is configured as a single computer. In Fig. 4, the server 10 includes a CPU (Central Processing Unit) 1000, a ROM (Read Only Memory) 1001, a RAM (Random Access Memory) 1002, a storage device 1003, and a communication interface (I / F) 1004, which are communicably connected to each other via a bus 1010.

[0024] The storage device 1003 is a non-volatile storage medium such as a hard disk drive or flash memory. The CPU 1000 controls the operation of the server 10 using the RAM 1002 as a work memory in accordance with programs stored in the ROM 1001 and the storage device 1003. The communication I / F 1004 performs communication via the network 2 under the control of the CPU 1000.

[0025] 5 is a block diagram showing an example of a hardware configuration of a terminal device 20 applicable to the embodiment. In the figure, the terminal device 20 includes a CPU 2000, a ROM 2001, a RAM 2002, a display control unit 2003, a storage device 2004, an input device 2005, a data I / F 2006, and a communication I / F 2007, which are communicably connected to each other via a bus 2010.

[0026] The storage device 2004 is a non-volatile storage medium such as a hard disk drive, flash memory, etc. The CPU 2000 controls the operation of the terminal device 20 in accordance with the programs stored in the ROM 2001 and the storage device 2004, using the RAM 2002 as a work memory.

[0027] The display control unit 2003 generates a display signal that can be displayed by the display 2020 based on the display control signal generated by the CPU 2000 in accordance with a program, and supplies the generated display signal to the display 2020. As a result, the display 2020 displays a screen in accordance with the display control signal.

[0028] The input device 2005 receives input operations from a user, generates a control signal in response to the input operation, and passes it to the CPU 2000. The CPU 2000 can control the operation of the terminal device 20 in accordance with the control signal. The input device 2005 may be configured to output a control signal in response to a contact position, and may be formed integrally with the display 2020 to form a touch panel.

[0029] The data I / F 2006 is an interface for transmitting and receiving data to and from external devices via a wired or wireless connection. The data I / F 2006 may be implemented using a universal serial bus (USB) or Bluetooth (registered trademark). The communication I / F 2007 communicates via the network 2 under the control of the CPU 2000.

[0030] As described above, the information processing system 1 according to the embodiment is configured on the server 10, except for a part of the data extraction unit 140. For example, the large-scale text data storage unit 100 and the dialogue data storage unit 101 are configured in a predetermined storage area in the storage device 2004 of the server 10. Furthermore, the speaker character determination unit 110, the versatility determination unit 120, the character score calculation unit 130, and the data extraction unit 140 are each configured as, for example, modules in a main storage area of ​​the RAM 1002 by the CPU 1000 executing the information processing program according to the embodiment in the server 10.

[0031] The information processing program can be acquired from outside via the network 2 by communication via the communication I / F 1004, and installed on the server 10. However, the information processing program may be provided by being stored in a removable storage medium such as a CD (Compact Disk), a DVD (Digital Versatile Disk), or a USB (Universal Serial Bus) memory.

[0032] As described above, the terminal device 20 is equipped with some of the functions of the data extraction unit 140. The terminal device 20 can realize some of the functions of the data extraction unit 140, for example, by a browser application installed in the terminal device 20. In this case, for example, the terminal device 20 reads a program for executing some of the functions of the data extraction unit 140 on the browser application from the server 10 via the network 2. However, the present invention is not limited to this, and a program for executing some of the functions of the data extraction unit 140 may be installed in the terminal device 20.

[0033] [3. Details of Processing According to the Embodiment] Next, the processing according to the embodiment will be described in more detail.

[0034] (3-1. Processing by the speaker character determination unit) The following describes the processing by the speaker character determination unit 110 according to the embodiment, which was explained in step S10 in the flowchart of Fig. 3. The speaker character determination unit 110 uses, as input, the large-scale text data stored in the large-scale text data storage unit 100 and the dialogue data of all characters appearing in the target work, which is stored in the dialogue data storage unit 101.

[0035] 6 is a schematic diagram showing an example of large-scale text data stored in the large-scale text data storage unit 100, which can be applied to an embodiment. Here, text data published on the Internet is collected as the large-scale text data. In particular, text posted by users on a social networking service (SNS), which is a service provided on the Internet, and replies (responses) to those posts are paired and collected as the large-scale text data.

[0036] In FIG. 6, the large-scale text data includes the items "Utterance No.", "Post," and "Reply to Post." The item "Post" indicates the posted text, and the item "Reply to Post" indicates a reply to the item "Post." The item "Utterance No." is a serial number for the pair of the item "Post" and the item "Reply to Post." In the example of FIG. 6, for example, in Utterance No. [1], the posted text "I'm really sorry" is paired with the utterance text replying to this post "No, no, I don't mind at all," and is assigned Utterance No. [1].

[0037] Large-scale text data is not limited to text posted on social networking sites collected from the Internet. For example, text posted on websites on the Internet may be extracted and collected as large-scale text data, or movie subtitle data may be collected. Furthermore, large-scale text data is not limited to text data on the Internet; text data stored in a local environment can also be collected. Furthermore, large-scale text data is not limited to pairs of posts and replies to those posts; replies alone may be collected. Large-scale text data is not limited to manually created text, such as social networking site posts or movie subtitles, but may also include text automatically generated by machines (such as artificial intelligence).

[0038] FIG. 7 is a schematic diagram showing an example of dialogue data stored in the dialogue data storage unit 101, which can be applied to the embodiment. As described above, the dialogue data storage unit 101 stores dialogue data for all characters appearing in the target work. The target work is content in which the target character appears, such as a novel or animation. In the following example, it is assumed that the characters appearing in the target work are six people: "Hero," "Princess," "Partner," "Villager," "Traveler," and "Demon King."

[0039] Note that the type of character is not particularly limited as long as it is a subject that speaks within the target work. For example, characters are not limited to people, but may include anthropomorphized animals, plants, inorganic objects, pseudo-personalities generated by programs, etc.

[0040] In FIG. 7, the item "Character" indicates a character, and the item "Line" indicates a line spoken by the corresponding character. The item "Line No." is a serial number for a pair of the items "Character" and "Line." In the example of FIG. 7, for example, the line "I'll save everyone! Leave it to me" is associated with the character "Hero," and the line "Thank you. I'll fight too!" is associated with the character "Princess." In the example of FIG. 7, for the sake of explanation, each character is associated with one line, but in reality, each character is associated with multiple lines and stored in the line data storage unit 101.

[0041] Returning to the explanation of the speaker character determination unit 110, the line data of all characters in the target work stored in the line data storage unit 101 is used as learning data, and a binary classifier (speaker character determiner) that determines the speaker character of the spoken text is created by some kind of machine learning method. This binary classifier corresponds to the speaker character determination unit 110.

[0042] Here, there are no particular limitations on the machine learning method and features used to create the binary classifier. For example, binary classification can be performed using logistic regression or a support vector machine, using the frequency of appearance or importance of words included in the spoken text (such as TF-IDF (Term Frequency-Inverse Document Frequency)) as features. Alternatively, a classifier can be created using a neural network.

[0043] As shown in Figure 7, when there are six characters, namely, "Hero," "Princess," "Partner," "Villager," "Traveler," and "Demon King," the speaker character determination unit 110 prepares a total of six binary classifiers that determine whether the speaker of the spoken text is "Hero," "Princess," ..., or "Demon King."

[0044] 8 is a flowchart illustrating an example of processing by the speaker character determination unit 110 according to the embodiment. FIG. 8 illustrates processing for one piece of utterance data among the utterance data stored in the large-scale text data storage unit 100. The utterance data used to determine the speaker character here is, as described above, data for a "reply to a post." This is to measure the character of the text on the replying side, assuming that it will be used as learning data for response generation.

[0045] Prior to the processing of the flowchart in FIG. 8, the character determination value for each character in the target work is initialized to a value of 0. In FIG. 8, in step S20, the speaker character determination unit 110 acquires one spoken text from the large-scale text data storage unit 100. In the next step S21, the speaker character determination unit 110 selects a character (referred to as a determination character) from among the characters appearing in the target work for which speaker character determination will be performed. In the example of FIG. 7, one (for example, "Hero") is selected from "Hero," "Princess," "Partner," "Village Girl," "Traveler," and "Demon King."

[0046] In the next step S22, the speaker character determination unit 110 uses a binary classifier corresponding to the determined character to estimate the speaker of the utterance text acquired in step S20. Specifically, assuming that the utterance text acquired in step S20 is "No, no, I don't mind at all" and the determined character is "Hero," the speaker character determination unit 110 uses a binary classifier for the determined character "Hero" to determine whether the utterance text is estimated to be from a "Hero," in other words, whether the content of the utterance text is likely to be from a "Hero."

[0047] In step S22, if the speaker character determination unit 110 estimates that the spoken text acquired in step S20 is from the determined character (step S22, "Yes"), the process proceeds to step S23. In step S23, the speaker character determination unit 110 sets the character determination value of the determined character to value = 1, and the process proceeds to step S24.

[0048] On the other hand, if the speaker character determination unit 110 does not estimate that the spoken text acquired in step S20 is from the determined character (step S22, "No"), it skips the processing of step S23 and proceeds to step S24.

[0049] In step S24, the speaker character determination unit 110 determines whether the processing of steps S21 to S23 has been completed for all characters included in the target work. If the speaker character determination unit 110 determines that there is a character among the characters included in the target work for which the processing of steps S21 to S23 has not yet been completed (step S24, "No"), the processing returns to step S21. The speaker character determination unit 110 selects one character from among the characters included in the next target work for which the processing of steps S21 to S23 has not yet been completed, and executes the processing from step S22 onwards.

[0050] On the other hand, if the speaker character determination unit 110 determines in step S24 that the processing of steps S21 to S23 has been completed for all characters included in the target work (step S24, "Yes"), it ends the series of processing steps according to the flowchart in Fig. 8. Then, the speaker character determination unit 110 obtains the next utterance text from the large-scale text data storage unit 100, and executes the processing again from step S20.

[0051] Fig. 9 is a schematic diagram showing an example of the result of a speaker character determination process by the speaker character determination unit 110 according to the embodiment. The example in Fig. 9 shows an example of the result of performing speaker character determination on each utterance text shown in Fig. 6. In Fig. 9, the text of the item "Reply to post" is acquired by the speaker character determination unit 110 as the utterance text.

[0052] 9, the item "Hero Determination" indicates the determination result (character determination value) in step S22 when the determined character is "Hero." The item "Hero Determination" has a character determination value of 1, which indicates that the corresponding spoken text is estimated to be that of the character "Hero," that is, the corresponding spoken text is likely to be that of the character "Hero." Similarly, the items "Princess Determination," "Partner Determination," "Traveler Determination," "Village Girl Determination," and "Demon King Determination" have character determination values ​​of 1, which indicates that the corresponding spoken text is estimated to be that of the "Princess," "Partner," "Traveler," "Village Girl," and "Demon King," respectively.

[0053] As an example, for the spoken text of utterance No. [1], "No, no, I don't mind at all," only the item "Hero Determination" has a character determination value of 1, and this utterance text is presumed to be from the character "Hero." In other words, the speaker character of the spoken text of utterance No. [1] is presumed to be the character "Hero." On the other hand, for the spoken text of utterance No. [6], "That's good! Let's do that," the items "Hero Determination" and "Village Girl Determination" each have a character determination value of 1, and this utterance text is presumed to be from the characters "Hero" and "Village Girl," and the speaker characters are the characters "Hero" and "Village Girl." In other words, the spoken text of utterance No. [6] is presumed to be like the character "Hero" and the character "Village Girl."

[0054] In this way, the character determination value can be considered to be a value indicating the degree of relevance of each character included in the target work to the spoken text.

[0055] Information indicating the result of the speaker character determination process shown in FIG. 9 is stored as an output of the speaker character determination unit 110 in the storage device 1003 or RAM 1002 of the server 10, for example.

[0056] In the above description, the binary classifier is created using the dialogue data of "all characters" included in the target work as learning data, but this is not limited to this example. For example, the binary classifier may be created using dialogue data of characters with a predetermined number of lines among all characters included in the target work. For example, all characters included in the target work may be limited to characters with a number of lines of 100 or more, and the binary classifier may be created based on the dialogue data of the limited characters. In this case, in step S21 of FIG. 8, it is possible to exclude from selection characters characters whose lines were not used to create the binary classifier.

[0057] In the above example, a binary classifier is created based on the dialogue data of all characters in one target work, but this is not limited to this example. For example, multiple works may be considered as target works, and the characters from each of the multiple works may be treated in an integrated manner.

[0058] (3-2. Processing by the Versatility Judgment Unit) Next, we will explain the processing by the versatility determination unit 120, which was explained in step S11 in the flowchart of Fig. 3. The versatility determination unit 120 uses the determination result by the speaker character determination unit 110 shown in Fig. 9 as input. That is, the versatility determination unit 120 determines whether or not each utterance text has versatility, based on the result of the speaker character determination by the speaker character determination unit 110.

[0059] Here, the versatility of a speech text refers to whether the speech text depends on a specific character. That is, speech text with high versatility has low dependency on a specific character, while speech text with low versatility has high dependency on a specific character. More specifically, for example, speech text with high versatility is expected to have little sense of incongruity whether it is spoken by any of the above-mentioned characters "Hero," "Princess," "Partner," "Traveler," "Villager Girl," and "Demon King." On the other hand, speech text with low versatility, for example, particularly high dependency on the character "Hero," is expected to have a large sense of incongruity when used in the speech of a character other than the character "Hero."

[0060] The versatility determination unit 120 calculates the proportion of characters estimated to be the speaker of the utterance text from the result of the speaker character determination for the utterance text to be determined. If the calculated proportion exceeds a threshold, the versatility determination unit 120 determines that the utterance text has versatility.

[0061] Fig. 10 is a schematic diagram showing an example of a determination result of the presence or absence of versatility by the versatility determination unit 120 according to the embodiment. Fig. 10 adds an item "versatility" indicating the determination result of the presence or absence of versatility to Fig. 9 described above. For spoken text determined to have low versatility, the item "versatility" is set to a value of [0]. On the other hand, for spoken text determined to have high versatility, the item "versatility" is set to a value of [1].

[0062] In the example of FIG. 10, the threshold for determining whether or not the utterance has versatility is set to 60%. For example, the utterance text "Thank you" of utterance No. [3] is determined to be spoken by four characters out of six (characters "Princess," "Partner," "Traveler," and "Village Girl"). The percentage of characters determined to be spoken by the utterance text of this utterance No. [3] is approximately 67%, which is larger than the threshold. Therefore, the utterance text "Thank you" of this utterance No. [3] is determined to have versatility.

[0063] On the other hand, for example, the spoken text of utterance No. 2, "Is that so? I wonder if I should wear a scarf," is determined to be spoken by two characters out of a total of six characters (the characters "Princess" and "Village Girl"). The proportion of characters determined to be spoken by the spoken text of this utterance No. [2] is approximately 33[%], which is smaller than the threshold. Therefore, the spoken text of this utterance No. [2], "Is that so? I wonder if I should wear a scarf," is determined to have no versatility.

[0064] The result of the determination by the versatility determination unit 120 is stored in the storage device 1003 or RAM 1002 of the server 10, for example.

[0065] In the above description, the threshold value for determining whether or not a given item has versatility is set to 60%. However, this is merely an example, and the present invention is not limited to this example.

[0066] In the above example, any character included in the target work may be excluded from the characters used in the versatility determination. For example, in the above example, among all the characters included in the target work, "Hero," "Princess," "Partner," "Traveler," "Village Girl," and "Demon King," the speech characteristics of the character "Demon King" are extremely different from those of the other characters. In such a case, this character "Demon King" may be excluded from the versatility determination.

[0067] (3-3. Processing by the Character Score Calculation Unit) Next, we will explain the processing by the character score calculation unit 130, which was explained in step S12 in the flowchart of Fig. 3. The character score calculation unit 130 uses as input the determination result by the speaker character determination unit 110 shown in Fig. 9. That is, the character score calculation unit 130 calculates a score indicating the character for each utterance text based on the result of the speaker character determination by the speaker character determination unit 110.

[0068] The character of a spoken text refers to the resemblance of the spoken text to the target character. For example, for a spoken text in which the target character is not the speaker, if the spoken text would not feel strange even if the target character were to speak the spoken text, the resemblance of the spoken text to the target character is considered to be high. The character score of a spoken text is a value indicating the resemblance of the spoken text to the target character.

[0069] 11 is a flowchart illustrating an example of processing by the character score calculation unit 130 according to the embodiment. Prior to the processing according to the flowchart in FIG. 11, the character score calculation unit 130 designates one character from among all characters included in the target work as a target character. For example, in the above example, one of the "Hero," "Princess," "Partner," "Village Girl," "Traveler," and "Demon King" (e.g., "Hero") is selected as the target character.

[0070] In step S30, the character score calculation unit 130 acquires one utterance text from the large-scale text data storage unit 100. The acquired utterance text is called a target utterance text.

[0071] In the next step S31, the character score calculation unit 130 sets an initial score for the target character. For example, when the total number of characters in the target work is N, if the target character has been determined to be the speaker of the target utterance text, the character score calculation unit 130 sets N points as the initial score for the target character. On the other hand, if the target character has not been determined to be the speaker of the target utterance text, the character score calculation unit 130 sets 0 points as the initial score for the target character.

[0072] In the next step S32, the character score calculation unit 130 selects one character from among the remaining characters after excluding the target character from all characters included in the target work.

[0073] In the next step S33, the character score calculation unit 130 determines, based on the determination result by the speaker character determination unit 110, whether or not the character selected in step S32 is estimated to be the speaker of the target utterance text.

[0074] If the character score calculation unit 130 determines in step S33 that the selected character is estimated to be the speaker of the target utterance text (step S33, "Yes"), it proceeds to step S34. In step S34, the character score calculation unit 130 subtracts 1 point from the N points allocated to the target character, sets the value as the new allocated point for the target character, and proceeds to step S35.

[0075] On the other hand, if the character score calculation unit 130 determines in step S33 that the selected character is not estimated to be the speaker of the target utterance text (step S33, "No"), it cancels step S34 and moves the processing to step S35.

[0076] In step S35, the character score calculation unit 130 determines whether or not the processing of steps S32 to S34 has been completed for all characters remaining in the target work, excluding the target character. If the character score calculation unit 130 determines that the processing has not been completed for all characters (step S35, "No"), the processing returns to step S32, and the processing of steps S32 to S34 is executed for the next character.

[0077] On the other hand, if it is determined in step S35 that the processing has ended (step S35, "Yes"), the character score calculation unit 130 proceeds to step S36. In step S36, the character score calculation unit 130 obtains a character score for the target character of the target utterance text based on the allocation of N points.

[0078] After the processing of step S36, the series of processing according to the flowchart of Fig. 11 is terminated. The processing according to the flowchart of Fig. 11 is repeatedly executed for all utterance texts for which speaker determination has been performed by the speaker character determination unit 110 for the target character. Furthermore, the processing according to the flowchart of Fig. 11 is repeatedly executed in sequence for all characters included in the target work as target characters.

[0079] Fig. 12 is a schematic diagram showing an example of the result of character scores obtained by the character score calculation unit 130 according to the embodiment. Fig. 12 adds an item "Score" indicating the character score to Fig. 10 described above.

[0080] As explained in steps S33 and S34 of the flowchart in Fig. 11, when it is determined that a character selected from among the target character is estimated to be the speaker of the target utterance text, one point is subtracted from the initial score of N points of the target character to obtain a new score. If the number of characters other than the target character who are estimated to be the speakers of the target utterance text is M, the score [NM] obtained by subtracting M points from the initial score of N points is obtained as the character score of the target utterance text relative to the target character and others.

[0081] An example of a character score calculated by the character score calculation unit 130 will be described in more detail with reference to Figure 12. As an example, in the case of the spoken text "Me?" of utterance No. [4], the target character (assumed to be the character "Hero") is determined to be the speaker, and since there are six characters in total, the initial score N is 6 points. Next, since the spoken text is determined to be the speaker of one character other than the target character, "Partner," one point is deducted from the initial score, and the character score of the target character, "Hero," becomes 5 points.

[0082] As another example, in the case of the spoken text of utterance No. [2], "Really? I wonder if I should wear a scarf," the target character (character "Hero") is determined not to be the speaker, and the initial score N is set to 0. Next, the two characters other than the target character, "Princess" and "Village Girl," are determined to be the speakers of the spoken text, and 2 points are deducted from the initial score, resulting in the character score of the target character, character "Hero," being -2 points.

[0083] (3-4. Processing by the data extraction unit) Next, the processing by the data extraction unit 140, which was explained in step S13 in the flowchart of Fig. 3, will be explained. The data extraction unit 140 uses, as input, the determination result by the speaker character determination unit 110 (see Fig. 9), the determination result by the versatility determination unit 120 (see Fig. 10), and the character score calculated by the character score calculation unit 130 (see Fig. 12). The data extraction unit 140 extracts spoken text for the target character based on these determination results and the character score.

[0084] The data extraction unit 140 extracts the spoken text for the target character using a screen that presents these determination results and the character score as a user interface. However, the data extraction unit 140 may automatically extract the spoken text for the target character based on conditions that are specified in advance for these determination results and the character score.

[0085] Here, a case will be described in which utterance text for a target character is extracted in accordance with a user's instruction using a user interface screen.

[0086] 13 is a schematic diagram showing an example of a data extraction screen as a user interface that can be applied to an embodiment and is generated and presented by the data extraction unit 140. In Fig. 13, a data extraction screen 50 includes an extraction condition setting area 51 for setting data extraction conditions, and an extraction result display area 52 in which the data extraction results are displayed.

[0087] The extraction condition setting area 51 is provided with input sections 510, 511, 512, and 513 for the user to input data extraction conditions, and buttons 514 and 515 for executing processes according to user operations.

[0088] The character name of the target character is input to the input unit 510. The input unit 510 can be configured to select the target character from a list of all characters in the target work using, for example, a drop-down list.

[0089] The likeliness of the target character, that is, the condition of the character score, is input to input unit 511. In the example of Fig. 13, the data extraction target is a spoken text whose character score value is equal to or greater than the value input to input unit 511. Input unit 511 can also be configured to select a desired value from a list of character scores that can be specified using, for example, a drop-down list or the like.

[0090] An input unit 512 receives an input indicating whether or not the utterance text determined to be versatile is to be included in the extracted data.

[0091] Input unit 513 excludes from the data extraction targets any spoken text that includes the input character as a speaker character. Input unit 513 can be configured to select the target character from a list of all characters of the target work using, for example, a drop-down list. In response to a user operation on button 514, button 514 displays an utterance text viewing screen (described later) for viewing the utterance text of the character input to input unit 513.

[0092] In response to a user operation on button 515, spoken text included in the large scale text data stored in the large scale text data storage unit 100 is extracted in accordance with the conditions input to the input units 510 to 513. By operating button 515, spoken text selected based on the conditions input to the input units 510 to 513 is extracted as extracted data 520 from the large scale text data stored in the large scale text data storage unit 100. The extracted data 520 is displayed in extraction result display area 52.

[0093] Fig. 14 is a schematic diagram for explaining the data extraction process by the data extraction unit 140 according to the embodiment. In Fig. 14, the item "input utterance" corresponds to, for example, the item "post" in Fig. 9. The item "response utterance" is an utterance in response to the item "input utterance," and corresponds to, for example, the item "reply to post" in Fig. 9, and indicates utterance text.

[0094] The item "Characteristics of response utterance" includes three items: "Versatility," "Likeness of target character," and "Speaker character." The item "Versatility" corresponds to, for example, the item "Versatility" in FIG. 10, and indicates whether or not the corresponding utterance text is versatile. The item "Likeness of target character" corresponds to, for example, the item "Score" in FIG. 12. The item "Speaker character" is a compilation of, for example, the items "Hero determination," "Princess determination," "Partner determination," "Traveler determination," "Village girl determination," and "Demon king determination" in FIG. 9, and lists the characters determined as speaker characters for the utterance text.

[0095] Note that FIG. 14 shows each piece of data summarized for the purpose of explanation, and does not mean that the data extraction unit 140 holds each piece of data in this format.

[0096] In the example of Fig. 13 described above, the character "Hero" is set as the target character in input unit 510, and the value = 1 is set as the target character's likeness in input unit 511. Also, input unit 512 is set to include versatile utterances, and input unit 513 is set to the character "Village Girl" as an excluded character.

[0097] The data extraction unit 140 selects an utterance text to extract from each utterance text shown in the item "Response Utterance" in FIG. 14 based on each input in FIG.

[0098] The data extraction unit 140 first extracts data that satisfies the following conditions, according to the settings of the input units 510, 511, and 513: the target character likelihood is 1 or more, the speaker characters include the character "Hero," and the speaker characters do not include the character "Villager." In the example of FIG. 14, utterances No. [1] and [4] satisfy this condition. The data extraction unit 140 further extracts utterance texts that are deemed to have versatility according to the settings of the input unit 512. In the example of FIG. 14, utterance No. [3] satisfies this condition.

[0099] The data extraction unit 140 extracts data for utterance Nos. [1], [3], and [4] in accordance with the settings of these input units 510 to 513. Extracted data 520, in which the data for utterance Nos. [1], [3], and [4] have been extracted in this way, is displayed in the extraction result display area 52 of the data extraction screen 50 shown in FIG.

[0100] In addition, in utterance No. [3], the speaker character includes the excluded character "village girl." In this case, when extracting because the item "versatility" is "yes," it is explained that the speaker character may include the excluded character. This is because utterance text with versatility is not limited to the excluded character "village girl," and may also be spoken by other characters.

[0101] 15 is a schematic diagram showing an example of an utterance text viewing screen according to the embodiment. In FIG. 15, an utterance text viewing screen 60a is displayed by operating the button 514 on the data extraction screen 50 in FIG.

[0102] The spoken text viewing screen 60a is provided with display areas 600 and 601, an input section 602, and a button 603. The display area 600 displays each character included in the target work. In the example of FIG. 15, each character is represented by a circle C1 to C6, and among the circles C1 to C6, the circle C1 of the target character, "Hero," is displayed the largest.

[0103] Also, in FIG. 15, when circles C1 to C6 are pointed to by cursor display 610 in response to a user's operation of cursor display 610, data extraction unit 140 displays a list of spoken texts estimated to be those of the pointed-to circle in display area 601.

[0104] At this time, when an overlapping portion of circles is pointed to, the data extraction unit 140 displays a list of spoken texts inferred to be spoken by the characters corresponding to each circle that shares the overlapping portion in the display area 601.

[0105] In the example of FIG. 15, cursor display 610 points to the overlapping portion between circle C1 of character "Hero" and circle C4 of character "Village Girl." Data extraction unit 140 displays spoken text estimated to be a common speaker of character "Hero" and character "Village Girl" in display area 601. On the other hand, when cursor display 610 points to the overlapping portion between circle C1 of character "Hero" and circle C3 of character "Traveler," where no speaker text estimated to be a common speaker exists, nothing is displayed in display area 601.

[0106] Furthermore, the user can select a desired circle from among the circles C1 to C6 displayed in the display area 600 by operating the cursor display 610, and move the selected circle within the display area 600. This allows the user to check, for example, the spoken text that is estimated to be spoken by the character "Hero" and another arbitrary character in common.

[0107] On the utterance text viewing screen 60a, the input unit 602 inputs a character that should be excluded as a speaker of the utterance text in the overlapping portion of the circles. In the example of FIG. 15, the utterance text in the overlapping portion between the circle C1 of the character "Hero" and the circle C4 of the character "Village Girl" is displayed in the display area 601. The user checks the utterance text displayed in the display area 601 and thinks that even if the utterance text is estimated to be spoken by the character "Hero," the utterance text that is common to the character "Village Girl" should be excluded, and inputs the character "Village Girl" into the input unit 602 as a character to be excluded. The data extraction unit 140 reflects the input content into the input unit 602 in the data extraction result in response to an operation on the button 603.

[0108] 16 is a schematic diagram showing another example of the utterance text viewing screen according to the embodiment. In FIG. 16, the utterance text viewing screen 60b is an example in which an input unit 604 and a button 605 are added to the utterance text viewing screen 60a shown in FIG. 15. The input unit 604 inputs a character that has been added to the data extraction target by the data extraction unit 140. In the example of FIG. 16, when the user looks at the utterance text in which the character "Traveler" is assumed to be the speaker, the user thinks that most of it would be fine to use as an utterance for the character "Hero," and so inputs the character "Traveler" into the input unit 604.

[0109] 17 is a schematic diagram showing an example of text data 150 for a target character output by the data extraction unit 140 according to the embodiment. This text data 150 for a target character is output by the data extraction unit 140, for example, in response to an operation on button 521 on the data extraction screen 50 of FIG. 13. In the example of FIG. 17, the text data 150 for a target character includes the items "Utterance No.", "Input Utterance," "Response Utterance," and "Target Character." The text data 150 for a target character includes each of the data that is to be extracted in FIG. 14.

[0110] The text data 150 for the target character may be stored in the terminal device 20, or may be stored in a storage device connected to the terminal device 20. The text data 150 for the target character may also be transferred to the server 10 via the network 2 and stored in the server 10.

[0111] 4. First Modification of the Embodiment Next, a first modified example of the embodiment will be described. In the above-described embodiment, the character score calculation unit 130 calculates the character score using the determination result by the speaker character determination unit 110. In contrast, in the first modified example of the embodiment, the character score calculation unit 130 calculates the character score without using the determination result by the speaker character determination unit 110. More specifically, in the first modified example of the embodiment, the probability that the target character is the speaker of the spoken text is calculated, and the calculated probability is used as the score.

[0112] The character score calculation unit 130 obtains the probability that the speaker of a certain utterance is a target character by creating a multi-class classifier that estimates which character the speaker of each utterance text is. The multi-class classifier can be created using logistic regression, for example, with the appearance frequency and importance (TF-IDF, etc.) of words included in the utterance text as features.

[0113] For example, a multi-value classifier is used to estimate which character in the target work is the speaker of the target utterance text "I'm fine." As an example, assume that the probability that each character is the speaker of the target utterance text is: character "Hero" = 0.5, character "Princess" = 0.0, character "Partner" = 0.3, character "Traveler" = 0.2, character "Villager Girl" = 0.0, and character "Demon King" = 0.0. In this case, the character score of the character "Hero" for the target utterance text is set to 0.5.

[0114] The probability that each character is the speaker of the target spoken text can be considered as a value indicating the degree of relevance of each character included in the target work to the spoken text.

[0115] 5. Second Modification of the Embodiment Next, a second modified example of the embodiment will be described. In the second modified example of the embodiment, as in the first modified example of the embodiment described above, the character score calculation unit 130 calculates the character score without using the determination result by the speaker character determination unit 110. More specifically, in the second modified example of the embodiment, the character score of the spoken text is calculated using the importance of each word for each character.

[0116] 18 is a flowchart illustrating an example of processing by the character score calculation unit 130 according to a second modified example of the embodiment. Prior to the processing according to the flowchart in FIG. 18, the character score calculation unit 130 designates one character from among all characters included in the target work as the target character. For example, in the above example, one of the "Hero," "Princess," "Partner," "Village Girl," "Traveler," and "Demon King" (e.g., "Hero") is selected as the target character.

[0117] In step S40, the character score calculation unit 130 acquires the dialogue texts of all characters in the target work, and extracts words as elements that make up the dialogue texts from each dialogue text.

[0118] In the next step S41, the character score calculation unit 130 calculates the importance of word t to the target character for one of the words extracted in step S40, in accordance with the following formula (1): In formula (1), Im(t) represents the importance of word t, Fr(t) represents the frequency with which word t appears in the dialogue text of the target character, and R(t) represents the rarity of characters uttering word t. Im(t) = Fr(t) × R(t) …(1)

[0119] Here, the rarity R(t) of a character who utters word t can be calculated, for example, using the following formula (2): In formula (2), N represents the total number of characters included in the target work, and M(t) represents the number of characters who utter word t. R(t)=log(N / M(t)) …(2)

[0120] The importance Im(t) of word t is a value based on the rarity R(t) of the character who speaks word t, and can be considered to be a value indicating the degree of relevance to the spoken text of each character included in the target work.

[0121] In the next step S42, the character score calculation unit 130 determines whether or not the processing has been completed for all words extracted in step S40. If the character score calculation unit 130 determines that there are words among the words extracted in step S40 for which the processing of step S41 has not yet been executed (step S42, "No"), the processing returns to step S41, and the processing of step S41 is executed for the next word.

[0122] On the other hand, if the character score calculation unit 130 determines in step S42 that the processing has been completed for all words extracted in step S40 (step S42, "Yes"), it moves the processing to step S43.

[0123] In this way, by calculating the importance to the target character for each word extracted from the dialogue text of all characters in the target work, words that rarely appear in the dialogue text of the target character will be considered less important to the target character. Also, by introducing the rarity R(t) of characters that utter word t, the importance of words that are not rare and are uttered by any character will be lowered. If a word is frequently uttered only by the target character, its importance to the target character will be higher.

[0124] The frequency of appearance of word t in the dialogue text of the target character can be determined by the proportion of the number of appearances of word t in all words included in the dialogue text of the target character. Alternatively, the frequency of appearance of word t in the dialogue text of the target character can be determined by the proportion of dialogue texts containing word t in all dialogue texts of the target character.

[0125] In step S43, the character score calculation unit 130 calculates a character score indicating the likeliness of each utterance text to the target character based on the importance Im(t) of each word calculated in the processes of steps S41 and S42. More specifically, the character score calculation unit 130 calculates, for example, the average value of the importance of each word included in each utterance text to the target character as the character score of the entire utterance text.

[0126] For example, suppose the spoken text to be the target of score calculation is "I'm fine," which includes the words "I," "I'm fine," "I'm fine," and "I'm fine." If the importance of each word to the target character is as follows: "I" = 0.007, "I'm fine" = 0, "I'm fine" = 0, and "I'm fine" = 0.001, the character score calculation unit 130 calculates 0.002, which is the average of the importance of these four words, as the character score for the target character of this spoken text.

[0127] In the above description, the word "word" may be replaced with a "word string" consisting of multiple words. For example, a word string consisting of two words may be used. In the case of an utterance of "I'm fine," the word string consisting of two words would be "I'm fine," "I'm fine," and "I'm fine."

[0128] As described above, the information processing system 1 according to the embodiment and its first and second modifications can easily and automatically collect spoken texts that resemble a specific character (characteristics). Furthermore, it can automatically evaluate the character of spoken texts included in posts on SNS, movie subtitle data, etc., and extract texts that have a specific character.

[0129] Even with existing technologies, character utterance text can be obtained from novels or animation scripts, but the amount of data obtained is very small, making it difficult to use as learning data for generating dialogue responses.In contrast, the information processing system 1 according to the embodiment and its first and second modifications can automatically collect a large amount of text that resembles a specific character, making it possible to increase the amount of data.

[0130] Furthermore, in the information processing system 1 according to the embodiment and its first and second modifications, in the automatic evaluation of the character of the spoken text, it is possible to identify utterances that could have been spoken by any character (versatile utterances). Specifically, a binary classifier (character determiner) is provided that determines whether the speaker of the spoken text is each character, and if the proportion of characters determined to be speakers exceeds a threshold, the spoken text is determined to be versatile.

[0131] Existing technologies only consider the character characteristics of specific characters. That is, existing technologies only use character classifiers for specific characters. Therefore, due to a lack of learning data for the character classifier, a general utterance such as "thank you" (an utterance that could actually be an utterance of the character) may be determined as a negative example. In contrast, by applying the information processing system 1 according to the embodiment and its first and second modifications, it is possible to capture general utterances and add them to the utterance text of each character.

[0132] Furthermore, according to the information processing system 1 of the embodiment and its first and second variants, a user interface is included that selects and rejects spoken text based on the speaker character determination results, the automatic character evaluation value, and whether or not it is versatile, making it easy to create a corpus of spoken text.

[0133] 6. Third Modification of the Embodiment Next, a third modified example of the embodiment will be described. In the above-described embodiment and its first and second modified examples, large-scale text data is collected as content data from the Internet or the like, and character scores of utterance texts included in the collected large-scale text data are calculated. Content data applicable to the embodiment is not limited to text data. The third modified example of the embodiment is an example in which video data or music data (audio data) is applied as content data.

[0134] The information processing system 1 according to the above-described embodiment and its first and second modifications associates characters from a target work with spoken text included in large-scale text data. In contrast, the information processing system according to the third modification of the embodiment collects video data and music data that are made public on the Internet, etc.

[0135] An information processing system according to a third modification of the embodiment fragments collected video data or music data into predetermined units and labels each of the fragments. The information processing system determines the authorship likelihood of each labeled fragment for one or more authors of the predetermined video data or music data, in a manner similar to the processing according to the embodiment described above.

[0136] The specified video data or music data referred to here is data whose creator is known. On the other hand, fragmented video data or music data collected from the Internet, etc., may not necessarily have a clear creator.

[0137] That is, each fragment in the third modified example of the embodiment corresponds to the spoken text in the above-mentioned embodiment and its first and second modified examples. Also, in the third modified example of the embodiment, the author whose author-likeness is determined corresponds to the target character in the above-mentioned embodiment and its first and second modified examples.

[0138] Here, the fragments in video data can be clips or scenes that make up the video data, while the fragments in music data can be parts of a song, such as the introduction, first melody, second melody, chorus, interlude, and postlude, as well as phrases, sections separated by key changes, beat changes, etc.

[0139] In this way, in the third modified example of the embodiment, it is possible to associate a specific artist with a fragment of video data or music data collected from the Internet, etc. Note that when using video data or music data associated with a specific artist, it is necessary to give sufficient consideration to copyrights, etc.

[0140] The effects described in this specification are merely examples and are not limiting, and other effects may also be present.

[0141] The present technology can also be configured as follows. (1) a score calculation unit that calculates a score indicating the likeliness of a first content data item to be a target character based on a feature amount indicating a feature of the first content data item and a degree of association of the target character and other characters different from the target character with the first content data item; an extraction unit that extracts data to be associated with the target character from the first content data based on the score calculated by the score calculation unit; An information processing device comprising: (2) a versatility determination unit that determines whether the first content data has versatility based on the degree of relevance; Furthermore, The score calculation unit calculating the score by excluding, from the first content data, the first content data determined by the versatility determination unit to have versatility; The information processing device according to (1) above. (3) The extraction unit extracting the data based on the score and the presence or absence of the versatility determined by the versatility determination unit; The information processing device according to (2) above. (4) an association determination unit that performs association determination to determine to which of the target character and the other character the first content data is associated; Further provided with The information processing device according to any one of (1) to (3). (5) The association determination unit The association determination is performed by classifying the target character and the other characters into binary values ​​with respect to the first content data. The information processing device according to (4) above. (6) The score calculation unit giving an initial score to each of the target character and the other characters, and calculating the score by subtracting the result of the binary classification by the association determination from the initial score; The information processing device according to (5) above. (7) The score calculation unit calculating the scores for the target character and the other characters based on the results of multi-value classification of the first content data; The information processing device according to any one of (1) to (3). (8) The score calculation unit determining the importance of each element constituting second content data corresponding to the target character and each of the other characters with respect to the target character, and calculating the score based on the importance; The information processing device according to any one of (1) to (3). (9) The score calculation unit determining the importance level based on the number of times the second content data is associated with the target character and the rarity of the second content data for the target character and the other characters; The information processing device according to (8). (10) The association determination unit performing the association determination when the number of second content data associated with the target character is equal to or greater than a predetermined number, among the second content data associated with the target character and the other characters; The information processing device according to any one of (4) to (8). (11) The extraction unit generating a specification screen having a condition specification section for specifying at least the target character and the lower limit value of the score as conditions for extracting the data; The information processing device according to any one of (1) to (10). (12) The extraction unit, with respect to the specified screen, an excluded character designation unit that designates a character to be excluded from the target of the data extraction from among the target character and the other characters; a display designation unit that designates the display of a viewing screen for viewing second content data associated with each of the target character and the other characters; Further providing: The information processing device according to (11) above. (13) The extraction unit providing an addition designation unit for designating a character to be added as a new target character to the target character among the other characters on the viewing screen; The information processing device according to (12) above. (14) the first content data is text data; The information processing device according to any one of (1) to (13). (15) The target character and the other characters are characters that appear in the target work, The score calculation unit determining the degree of relevance using dialogue data indicating dialogues of the target character and the other characters in the target work; The information processing device according to (14) above. (16) the first content data is text data posted on a social networking service (SNS); The information processing device according to (14) or (15). (17) The score calculation unit calculating the score using a response to the post as the first content data; The information processing device according to (16) above. (18) the first content data is video data; The information processing device according to any one of (1) to (13). (19) the first content data is music data; The information processing device according to any one of (1) to (13). (20) The first content data is data published on the Internet. The information processing device according to any one of (1) to (19). (twenty one) Executed by a processor, a score calculation step of calculating a score indicating the likelihood of a first content data being a target character based on a feature amount indicating a feature of the first content data and a degree of association of the target character and other characters different from the target character with the first content data; an extraction step of extracting data to be associated with the target character from the first content data based on the score calculated by the score calculation step; An information processing method comprising: [Explanation of symbols]

[0142] 1. Information Processing Systems 2 Network 10 Servers 20 Terminal equipment 50 Data extraction screen 51 Extraction condition setting area 52 Extraction result display area 60a, 60b Speech text viewing screen 100 Large-scale text data storage 101 Dialogue data storage section 110 Speaker character determination unit 120 Versatility judgment section 130 Character Score Calculation Unit 140 Data Extraction Unit 150 Text data for target character 510,511,512,513,602,604 Input section 514,515,521,603,605 buttons 520 Extracted Data 600,601 display area

Claims

1. a score calculation unit that calculates a score indicating the likelihood of a first content data item being a target character based on a feature amount indicating a feature of the first content data item and a degree of association of the target character and other characters different from the target character with the first content data item; an extraction unit that extracts data to be associated with the target character from the first content data based on the score calculated by the score calculation unit; An information processing device comprising:

2. a versatility determination unit that determines whether the first content data has versatility based on the degree of relevance; Furthermore, The score calculation unit calculating the score by excluding, from the first content data, the first content data determined by the versatility determination unit to have versatility; The information processing device according to claim 1 .

3. The extraction unit extracting the data based on the score and the presence or absence of the versatility determined by the versatility determination unit; The information processing device according to claim 2 .

4. an association determination unit that performs association determination to determine to which of the target character and the other character the first content data is associated; Further provided with The information processing device according to claim 1 .

5. The association determination unit The association determination is performed by classifying the target character and the other characters into binary values ​​with respect to the first content data. The information processing device according to claim 4 .

6. The score calculation unit giving an initial score to each of the target character and the other characters, and calculating the score by subtracting the result of the binary classification by the association determination from the initial score; The information processing device according to claim 5 .

7. The score calculation unit calculating the scores for the target character and the other characters based on the results of multi-value classification of the first content data; The information processing device according to claim 1 .

8. The score calculation unit determining the importance of each element constituting each of second content data corresponding to the target character and each of the other characters with respect to the target character, and calculating the score based on the importance; The information processing device according to claim 1 .

9. The score calculation unit determining the importance level based on the number of times the second content data is associated with the target character and the rarity of the second content data for the target character and the other characters; The information processing device according to claim 8 .

10. The association determination unit performing the association determination when the number of second content data associated with the target character is equal to or greater than a predetermined number, among second content data associated with the target character and the other characters; The information processing device according to claim 4 .

11. The extraction unit generating a specification screen having a condition specification section for specifying at least the target character and the lower limit value of the score as conditions for extracting the data; The information processing device according to claim 1 .

12. The extraction unit, with respect to the specified screen, an excluded character designation unit that designates a character to be excluded from the target of the data extraction from among the target character and the other characters; a display designation unit that designates the display of a viewing screen for viewing second content data associated with each of the target character and the other characters; Further providing The information processing device according to claim 11.

13. The extraction unit providing an addition designation unit for designating a character to be added as a new target character to the target character among the other characters on the viewing screen; The information processing device according to claim 12.

14. the first content data is text data; The information processing device according to claim 1 .

15. The target character and the other characters are characters that appear in the target work, The score calculation unit determining the degree of relevance using dialogue data indicating dialogues of the target character and the other characters in the target work; The information processing device according to claim 14.

16. the first content data is text data posted on a social networking service (SNS); The information processing device according to claim 14.

17. The score calculation unit calculating the score using a response to the post as the first content data; The information processing device according to claim 16.

18. the first content data is video data; The information processing device according to claim 1 .

19. the first content data is music data; The information processing device according to claim 1 .

20. the first content data is data published on the Internet; The information processing device according to claim 1 .

21. Executed by a processor, a score calculation step of calculating a score indicating the likelihood of a first content data being a target character based on a feature amount indicating a feature of the first content data and a degree of association of the target character and other characters different from the target character with the first content data; an extraction step of extracting data to be associated with the target character from the first content data based on the score calculated by the score calculation step; An information processing method comprising:

Images