Information processing device, program, and information processing method

Abstract

To provide a user with more desired information.SOLUTION: A contribution information processing unit collects event-related contribution information that is contribution information concerning an event. A scoring unit uses a predetermined classification model and calculates a score showing relevance to a classification item for a predetermined classification item by a type of the event from the event-related contribution information. A provision processing unit controls provision of the event-related contribution information based on the score. The embodiment of an invention can be any form of an information processing device, a program, and an information processing method.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present application relates to an information processing device, a program, and an information processing method, and more particularly to filtering, scoring, and providing posted information about events posted on a network, for example. [Background technology]

[0002] Attempts are being made to utilize information posted to social networking (SNS) sites in broadcasting and network distribution of event information showing live coverage of various events. Such events include, for example, auto racing, sporting events, entertainment events, and gaming events. Entertainment events include musical performances and plays. Gaming events are not limited to those held in real space, such as Go and Shogi, but can also include online games held in virtual space, such as e-sports.

[0003] Posts on social media typically contain short sentences describing events that occurred at the event and the person's own feelings. The quality of the content of the posts is generally mixed. For example, some contain useful insights, while others are simply expressions of emotion. As a result, the posts often contain unnecessary noise for the reader. On the other hand, they can also provide relevant information based on useful facts. In other words, the posts can sometimes supplement information that could not be fully provided in the live coverage or commentary of the event itself.

[0004] This is expected to enable the filtering and provision of substantial supplementary information related to the main event content. Also, by providing posted information collected simultaneously with the event based on posting time information, it is expected that the interest of the main event will be increased in terms of experience and immersion, and the content of the main event will be complemented. In this regard, for example, Patent Document 1 describes a posting information system that accumulates information about posted information to be processed at the user's instruction, receives posted information from an information providing server that transmits and receives posted information, and calculates a priority for each piece of received posted information based on the accumulated posted information. The posted information is rated based on extracted keywords, and information that matches the user's interests is provided. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Laid-Open No. 2014-035719 Summary of the Invention [Problem to be solved by the invention]

[0006] However, the rating described in Patent Document 1 is limited to classification based on keywords related to the user's interests, and therefore the background and context of the event are not necessarily reflected in the information provided. The present application has been made in view of the above points, and one of the objects of the present application is to provide an information processing device, a program, and an information processing method that can provide a user with more desired information. [Means for solving the problem]

[0007] (1) The present application has been made to solve the above-mentioned problems, and one aspect of an embodiment of the present application is an information processing device that includes a post information processing unit that collects event-related post information, which is post information related to events; a scoring unit that uses a predetermined classification model to calculate a score indicating the relevance of the event-related post information to each predetermined classification item for the type of event; and a provision processing unit that controls the provision of the event-related post information based on the score.

[0008] (2) Another aspect of the present application is a method in an information processing device, the information processing method in an information processing device having the steps of: collecting event-related posted information, which is posted information about an event; calculating, using a predetermined classification model, a score indicating the relevance of the event-related posted information to each predetermined classification item for the type of event; and controlling provision of the event-related posted information based on the score. [Effects of the Invention]

[0009] According to the present invention, it is possible to provide the user with further desired information. [Brief explanation of the drawings]

[0010] [Figure 1] 1 is a schematic block diagram illustrating an example of a network configuration of an information processing system according to an embodiment of the present invention. [Figure 2] 1 is a schematic block diagram illustrating an example of a functional configuration of an information processing device according to an embodiment of the present invention. [Figure 3] FIG. 4 is an explanatory diagram showing a first control example of speech characteristics according to the present embodiment. [Figure 4] FIG. 10 is an explanatory diagram showing a second control example of speech characteristics according to the present embodiment. [Figure 5] FIG. 10 is an explanatory diagram showing a third control example of speech characteristics according to the present embodiment. [Figure 6] FIG. 10 is an explanatory diagram showing a fourth control example of speech characteristics according to the present embodiment. [Figure 7] FIG. 10 is a diagram illustrating an example of a posting service screen. [Figure 8] FIG. 2 is an explanatory diagram for explaining learning and inference of a classification model according to the present embodiment. [Figure 9] FIG. 2 is a diagram illustrating classification items according to the embodiment. [Figure 10] FIG. 4 is a diagram showing a display example of a display screen according to the embodiment. [Figure 11] FIG. 10 is a diagram illustrating an example of a filtering information setting screen according to the embodiment. [Figure 12] 10 is a flowchart illustrating an example of a posted information providing process according to the present embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0011] (overview) Hereinafter, an embodiment of the present invention will be described with reference to the drawings. First, an overview of this embodiment will be described. FIG. 1 is a schematic block diagram showing an example of a network configuration of an information processing system S1 according to this embodiment. The information processing system S1 includes an information processing device 10, an event information server device 30, and a posting information server device 40. The information processing device 10, the event information server device 30, and the posting information server device 40 are connected to each other via a network NW so as to be able to send and receive various types of data wirelessly or via a wired connection. The network includes one or a combination of the Internet, a public network, a local area network (LAN), etc.

[0012] The information processing device 10 collects posted information related to events (sometimes referred to herein as "event-related posted information") that is provided using the network NW. The information processing device 10 receives posted information distributed from, for example, a posted information server device 40. The information processing device 10 calculates a score for the collected posted information using a classification model that assigns a score indicating the degree of relevance for each predetermined classification item to the type of event. The information processing device 10 determines whether to provide the event-related posted information based on the calculated score. The information processing device 10 presents the event-related posted information that has been determined to be provided. The posted information mainly includes text. The information processing device 10 may perform text-to-speech synthesis on the event-related posted information to be provided, and convert the text content representing the event-related posted information into spoken voice for presentation.

[0013] The event information server device 30 distributes content indicating the progress of an event using the network NW. In this application, "event" refers to an occasion, a festival, a musical or theatrical performance, a film screening, a promotional activity, and the like. Events are not limited to those held in real space, but may also be held in virtual space or across virtual and real space. Examples of such events include online exhibitions, online courses, and online games (including electronic competitions (so-called e-sports)). The progress of an event is represented by content mainly including video and audio.

[0014] The posted information server device 40 is a server device that provides a posting service using a network NW. The posting service may be provided as part of a social networking service (SNS). In providing the posting service, the posted information server device 40 receives posted information from a posting source device and transmits part or all of the received posted information in a predetermined data format to a destination device using the network NW. The posting source device can usually be part of the destination device. The information processing device 10 can also be part of the destination device. The posting service is provided in the form of, for example, a blog, an electronic bulletin board, or the like.

[0015] The information processing device 10 is primarily used by a user who receives posted information. The user of the information processing device 10 may be a physically able-bodied person or a visually impaired person. The information processing device 10 may be implemented as a general-purpose information device such as a personal computer (PC), a tablet terminal device, or a smartphone, or may be configured as a dedicated terminal device. The event information server device 30 and the posted information server device 40 may be configured as a web server using a general-purpose device such as a PC, a workstation, or a server device, or may be configured as a dedicated server device.

[0016] 1, there is one each of information processing device 10, event information server device 30, and posting information server device 40, but generally there will be a plurality of information processing devices 10. There may also be two or more each of event information server device 30 and posting information server device 40. Posting information server device 40 may provide one posting service at a time, or two or more posting services at a time.

[0017] Next, an example of the functional configuration of the information processing device 10 according to this embodiment will be described. Fig. 2 is a schematic block diagram showing an example of the functional configuration of the information processing device 10 according to this embodiment. The information processing device 10 includes a control unit 120 , a storage unit 140 , an input / output unit 150 , a playback unit 160 , a display unit 170 , an operation unit 180 , and a sound collection unit 190 . The control unit 120 is a computer system including, for example, a processor and a memory. The processor is a CPU (Central Processing Unit) or the like. The processor reads a predetermined program stored in advance in the storage unit 140 and executes the processing indicated by the instructions written in the read program. This execution realizes the functions of the information processing device 10, including the function of the control unit 120 described below. In this application, executing the processing indicated by the instructions written in the program may be referred to as "executing a program" or "running a program." The control unit 120 may be realized using a dedicated component.

[0018] The storage unit 140 stores various types of data used in the processes executed by the control unit 120. The storage unit 140 stores various types of data acquired by the control unit 120. The storage unit 140 includes storage media such as a RAM (Random Access Memory) and a ROM (Read Only Memory). The input / output unit 150 inputs or outputs various data to or from other devices wirelessly or via wire. The input / output unit 150 may be connected to other devices via a network NW. The input / output unit 150 may be configured to include, for example, either an input / output interface, a communication interface, or both.

[0019] The playback unit 160 plays back audio according to audio data input from the control unit 120. The playback unit 160 includes a speaker that emits sound according to the acquired audio data. The number of speakers is not limited to one, and may be two or more. Multiple speakers enable multi-channel playback. Display unit 170 displays various types of display information in accordance with display data input from control unit 120. Display unit 170 may be any of a liquid crystal display, an organic electroluminescence display, and the like.

[0020] The operation unit 180 receives a user operation and generates an operation signal in response to the received operation. The operation unit 180 outputs the generated operation signal to the control unit 120. The operation unit 180 may include a general-purpose input device such as a mouse, keyboard, or touch sensor, or may include dedicated components such as buttons, knobs, or dials. Note that the display functioning as the display unit 170 and the touch sensor functioning as the operation unit 180 may be configured as an integrated single touch panel.

[0021] The sound collection unit 190 collects sound arriving at the unit itself and generates a sound signal indicating the collected sound. The sound collection unit 190 has a microphone. The sound collection unit 190 outputs the generated sound signal to the control unit 120. Note that the control unit 120 may have a voice operation function. When providing the voice operation function, the control unit 120 performs known voice recognition processing on the sound signal input from the sound collection unit 190 and identifies a predetermined voice command. Then, the control unit 120 operates in accordance with the identified voice command.

[0022] Next, a description will be given of an example of the functional configuration of the control unit 120. The control unit 120 is configured to include a posted information processing unit 122, a scoring unit 124, and a provision processing unit 126. The posted information processing unit 122 acquires event-related posted information relating to a specific event provided via the network NW. The posted information processing unit 122 includes a posted information collecting unit 122a, an event date and time determining unit 122b, a posted information classifying unit 122c, and a similar event determining unit 122d.

[0023] Posted information collection unit 122a collects posted information sent to network NW from posted information server device 40. Posted information collection unit 122a outputs the collected posted information to event date and time determination unit 122b. The event date and time determination unit 122b identifies an event period for which posted information is to be collected, and determines, based on the identified event period, posted information input from the posted information collection unit 122a during a collection period as an event for which posted information is to be provided from its own device (sometimes referred to herein as a "target event"). The event date and time determination unit 122b adopts posted information related to the target event as a candidate for posted information to be provided. The event date and time determination unit 122b outputs the adopted posted information to the posted information classification unit 122c. The event date and time determination unit 122b may store the unadopted posted information in the storage unit 140, distinguishing it from the target event, or may discard it.

[0024] Event information may be set in advance in the posted information processing unit 122. The event information may include, for example, one or more items such as the event date and time, event name, organizer, venue or address, and keywords for each event. The posted information processing unit 122 may, for example, refer to event information stored in advance in the storage unit 140 to identify the event information for a target event specified by an operation signal or voice command. The event date and time determination unit 122b may set the collection period for the target event to be equal to the event period, or may set a longer period that includes the event period. For example, the start of the collection period may precede the start of the event period by several minutes to several tens of minutes, and the end of the collection period may be delayed by several minutes. This is because the target event may be a topic of conversation among spectators and other users even before or after the start or end of the target event. Note that, in this application, executing various processes based on an operation signal or voice command, or triggering such processes, may be referred to as "in response to an operation."

[0025] The posted information classification unit 122c classifies posted information related to a target event from the posted information input from the event date and time determination unit 122b. The posted information classification unit 122c determines whether the event related to the input posted information corresponds to the target event. The posted information classification unit 122c can determine whether the event corresponds to the target event, for example, based on whether the message constituting the posted information or the metadata attached to the posted information includes one or more of the event name, venue, and keywords of the target event included in the event identification information. For example, a hashtag may be included as metadata. A hashtag includes a # followed by a character string indicating the agenda, keywords, or other heading phrases related to the posted information. The posted information classification unit 122c stores posted information related to the target event in the storage unit 140, distinguishing it from other posted information. Note that the posted information classification unit 122c may reject other posted information. The posted information processing unit 122 stores posted information for each target event, thereby saving posted information for each event held in the past in the storage unit 140.

[0026] The similar event determination unit 122d determines past events of the same type as the target event as similar events. Similar events may be different events that belong to a common series. For example, if the target event is an auto race, competitions held at different times or venues would be similar events. If the target event is an athletic competition, similar events would include events, matches between different athletes or teams belonging to the same league, and different matches belonging to the same year or tournament. If the target event is a concert performance, similar events would include different performances by the same musician or orchestra. If the target event is a play, movie, or drama, similar events would include works with different scene developments that share a common title or characters.

[0027] The similar event determination unit 122d can, for example, refer to event holding information and determine, as a similar event, a past event that has one or both of the event name and keywords in common with the target event. The similar event determination unit 122d may determine, instead of the event holding information, as a similar event, an event that has a word or phrase that appears more frequently than a predetermined frequency in messages that make up posted information or in metadata attached to posted information. The similar event determination unit 122d may analyze the correlation between the content (e.g., text, audio, video) of the target event and the keywords appearing in the posted information for each event, and determine that an event having a correlation that is close to the correlation related to the target event within a certain range is a similar event. The similar event determining unit 122d associates the event determined to be a similar event with similar event information indicating that the event is a similar event in the storage unit 140.

[0028] The scoring unit 124 uses a pre-trained classification model to calculate, as an explanatory variable, a score indicating the degree of association for each classification item of the event type related to the target event for the posted information acquired by the posted information processing unit 122. The scoring unit 124 includes a learned classification setting unit 124a, a model training unit 124b, and a score calculation unit 124c.

[0029] The learning category setting unit 124a identifies similar events by referring to similar event information stored in the storage unit 140, and identifies a corresponding category from among multiple category items from newly stored posted information associated with the similar events. The multiple category items include at least one fact category and at least one emotion category. The learning category setting unit 124a can use known text classification techniques to identify the category items. The learning category setting unit 124a associates a category label indicating the identified category with the posted information to be processed and stores it in the storage unit 140.

[0030] The model learning unit 124b generates training data including posted information to which classification labels are set for similar events, and learns a classification model using the generated training data. The model learning unit 124b stores a parameter set of the classification model obtained by learning in the storage unit 140. The score calculation unit 124c reads out the parameter set stored in the storage unit 140 and sets it in the classification model. The score calculation unit 124c uses the classification model to calculate a score for each classification item related to the target event for each piece of posted information. Learning of the classification model by the model learning unit 124b and calculation (inference) of the score by the score calculation unit 124c will be described later.

[0031] The provision processing unit 126 executes processing for providing the scored posted information based on the calculated score. The provision processing unit 126 includes a filtering unit 126a, a filtering setting unit 126b, a speech expression adjustment unit 126c, and a speech synthesis unit 126d. The filtering unit 126a and the filtering setting unit 126b are involved in filtering the posted information. The speech expression adjustment unit 126c and the speech synthesis unit 126d are involved in reading out the posted information. Here, reading out refers to providing speech that indicates the content of communication in text. If the function of reading out the posted information is not provided, the provision processing unit 126 may cause the display unit 170 to display a display screen including the posted information determined by the filtering unit 126a to be available for provision. In this case, the speech expression adjustment unit 126c and the speech synthesis unit 126d may be omitted.

[0032] The filtering unit 126a determines whether or not to provide posted information corresponding to the score based on the score newly stored in the storage unit 140 (filtering). The filtering unit 126a determines whether or not the score for each category item is equal to or greater than a predetermined reference value. The filtering unit 126a determines that posted information containing a category item with a score equal to or greater than the reference value is available for provision, and determines that posted information containing a category item with a score equal to or greater than the reference value is unavailable for provision. The filtering unit 126a may delete posted information determined to be unavailable for provision without saving it. The reference value may be common regardless of the category item, or may differ for each category item. The provision processing unit 126 stores provision availability information indicating availability in the storage unit 140 in association with the posted information to be determined. The provision processing unit 126 may include information on category items with scores equal to or greater than the reference value, i.e., information on category items related to the determination that provision is available (sometimes referred to as “determination categories” in this application), in the provision availability information.

[0033] The filtering unit 126a may determine whether or not to provide posted information based on a priority set for each determination classification. The filtering unit 126a may determine that some of the posted information determined to be available is unavailable for provision, so that the amount of provided information increases for posted information of a determination classification with a higher priority and decreases for posted information of a determination classification with a lower priority. The filtering unit 126a may apply the priority for each classification item set by the filtering setting unit 126b. As an indicator of the amount of provision, any indicator such as the number of individual pieces of posted information, the number of bits, the number of packets, the number of characters, etc. may be used. When the posted information is converted into spoken voice and provided (described later), any indicator such as the number of bits, the number of packets, the playback time, etc. of audio data of the converted spoken voice may be used as an indicator of the amount of provision.

[0034] The filtering unit 126a may determine whether to provide posted information by further considering the number of posts per unit time and the category. For example, the filtering unit 126a may limit the amount of provision by determining that some of the posted information determined to be available based on the score is unavailable so that the number of pieces of posted information to be provided per unit time is equal to or less than a predetermined maximum number. The filtering unit 126a may limit the amount of provision of posted information determined to be available for a category related to the fact category to a certain amount or less. In this case, even if the amount of provided posted information conveying facts increases sharply, limiting the amount to a certain amount or less allows the recipient to reliably understand the facts being conveyed. Furthermore, it is not necessary to limit the amount of provision of posted information determined to be available for a category related to the emotion category. In this case, no limit is imposed on the amount of provided posted information conveying emotions, and the sense of realism of the emotions regarding the target event is conveyed to the recipient along with the amount.

[0035] The filtering unit 126a may also control the number of pieces of posted information provided based on content related to the target event. For example, when the content to be distributed represents a current situation of the target event, the filtering unit 126a may prioritize posted information determined to be available for classification items related to fact over posted information determined to be available for classification items related to emotion. In other cases, the filtering unit 126a may cancel the priority of posted information determined to be available for classification items related to fact. When the content to be distributed represents a current situation of the target event while it is being carried out, the facts indicated by the posted information are transmitted to the recipient with priority, thereby enabling the recipient to reliably grasp the progress of the target event. For example, the filtering unit 126a may identify the attributes (e.g., type, performers, etc.) and progress status (e.g., before starting, during execution, after completion, etc.) of the target event by referring to metadata attached to the content of the target event distributed from the event information server device 30.

[0036] The filtering setting unit 126b sets the amount of provision related to the provision of posted information. The setting of the amount of provision may be applied not only to text indicating posted information but also to spoken voice (described later) indicating posted information. The filtering setting unit 126b waits for filtering setting information indicating the priority of each classification item from, for example, the terminal device 20 to which the posted information is to be provided. The filtering setting information is acquired in response to an operation. The filtering setting unit 126b may cause the display unit 170 to display a filtering setting screen.

[0037] When a user of the information processing device 10 desires analytical information based on facts, a higher value may be set as the priority of a classification item related to fact classification than other classification items. When a sense of realism (e.g., excitement) in a real or virtual venue is desired, a lower value may be set as the priority of a classification item related to emotion classification than other classification items.

[0038] The speech expression adjustment unit 126c identifies the score related to the emotion category among the classification items for posted information that has been determined to be newly available from among the posted information stored in the storage unit 140. The speech expression adjustment unit 126c uses a known prosody control method to determine speech expression control parameters for controlling the prosody of the speech related to the posted information so that the higher the score related to the emotion category, the higher the degree of emotional expression of the posted information. The speech expression adjustment unit 126c may control the degree of emotional expression by further considering the priority related to the emotion category set in the filtering setting information. In this case, the speech expression adjustment unit 126c determines the degree of emotional expression so that the higher the priority related to the set emotion category, the higher the degree of emotional expression.

[0039] The speech expression adjustment unit 126c outputs the determined speech expression control parameters and the posted information to the speech synthesis unit 126d. In controlling prosody, parameters such as volume, speech rate, pitch, and intonation are adjusted. For example, posted information with a high score in emotion classification tends to produce speech with a relatively strong tone and high pitch. Posted information with a low score in emotion classification and a high score in fact classification tends to produce speech with a relatively weak tone and low pitch. The stronger the speech rate, the more pronounced any of the volume, speech rate, and intonation is.

[0040] The speech synthesis unit 126d uses a known text-to-speech synthesis method to synthesize speech indicating the speech content indicated in the posted information input from the speech expression adjustment unit 126c, with prosody specified by the speech expression control parameters. The provision processing unit 126 causes the playback unit 160 to play back the speech synthesized by the speech synthesis unit 126d.

[0041] The provision processing unit 126 may set the volume of the uttered voice so that the volume increases as the priority of each judgment item specified in the filtering setting information increases. Generally, the amount of posts (number of posts) varies depending on the progress of the event. Therefore, multiple pieces of posted information acquired simultaneously may be provided. In this case, the providing processing unit 126 may adjust the playback characteristics of the speech sounds synthesized from the individual pieces of posted information so that they differ for each piece of posted information, and may cause the playback unit 160 to play back the output sound obtained by mixing the adjusted sounds. As the playback characteristics, for example, delay, volume, volume ratio between channels in multi-channel playback, etc. can be applied as control parameters.

[0042] For example, when delay is used as the playback characteristic, the provision processing unit 126 may delay the speech sounds associated with each piece of posted information by different delay amounts and set different start times. In the example of FIG. 3, the provision processing unit 126 may set different start times t 01 ~t 03 The provision of the speech voices is started from the speech voice 11. The difference in the start time between the speech voices is preset to, for example, 1 to 3 seconds or more. In the example of FIG. 4, after the provision of the speech voice 11 for one piece of posted information is completed, the provision processing unit 126 starts providing the speech voice 12 for another piece of posted information.

[0043] When volume is used as the playback characteristic, the provision processing unit 126 may adjust the volume of the speech associated with each piece of posted information using different gains to vary the volume. To support multi-channel playback, the provision processing unit 126 provides the output audio as audio signals on multiple channels. In this case, the provision processing unit 126 may assign the speech to a different channel for each piece of posted information, or may assign the speech with different volume ratios between the channels. This allows the speech to be perceived in different directions for each piece of posted information using the playback unit 160 that supports the multi-channel playback method.

[0044] FIG. 5 illustrates a case where, in two-channel stereo, the provision processing unit 126 assigns speech sounds 21 and 22 for two pieces of posted information to the left channel (L ch) and the right channel (R ch), respectively. The speech sounds 21 and 22 are presented from the left channel speaker and the right channel speaker, respectively. At this time, the user perceives the speech sounds 21 and 22 to the left and right, respectively. In the example of FIG. 6, the provision processing unit 126 assigns volume ratios between the left and right channels of 1:0, 1 / √2:1 / √2, and 0:1 to speech sounds 31, 32, and 33 for three pieces of posted information, respectively. At this time, the user perceives the speech sounds 31, 32, and 33 to the left, front, and right, respectively.

[0045] The method of adjusting the playback characteristics may differ depending on the classification of the posted information to be provided. For example, the provision processing unit 126 uses different channels for speech sounds related to the fact classification and speech sounds related to the emotion classification. In this case, the user can clearly distinguish between the two based on the direction in which the speech sounds are perceived. The provision processing unit 126 may perform the process of setting a different delay amount for each piece of posted information on speech sounds related to the fact classification, but may not perform the process on speech sounds related to the emotion classification. This is because, while there is a high need to clearly understand the content of speech sounds related to the fact classification, it is not necessarily necessary to understand the content of speech sounds related to the emotion classification.

[0046] Next, an example of posted information will be described. FIG. 7 is a diagram illustrating an example of a posting service screen. The posting service screen Im02 is displayed on the display unit 170 and is configured by arranging text representing individual pieces of posted information. The illustrated posting service screen Im02 displays examples of posted information posted by spectators who visited the venue during an auto race as the target event. The posting service screen Im02 is configured by vertically arranging the user name of the information processing device 10, a search box TI02, and multiple posted information fields TP02, TP04, etc. in that order. Each posted information field includes the poster's username, the posting time (year, month, date, hour, minute), posted information, a URL, and a hashtag. An icon is placed immediately before the user name. Posted information is usually represented by short text of about one to five lines. The URL indicates the location of content showing the progress of the event. Each hashtag includes the tournament name "Auto Race 2024" and the venue name "M Circuit" as headwords.

[0047] Next, an example of learning and inference of the classification model according to this embodiment will be described. Fig. 8 is an explanatory diagram for explaining learning and inference of the classification model according to this embodiment. The model learning unit 124b acquires training data prior to learning the classification model. The training data includes multiple data pairs, and each data pair includes an input value serving as an explanatory variable and an output value serving as a target variable. The model learning unit 124b generates training data by, for example, using posted information as an input value and accumulating data pairs including element values ​​for each category item as output values ​​corresponding to the posted information for each piece of posted information. Among the output values ​​of each data pair, the model learning unit 124b includes a maximum value (e.g., 1) as the element value for the category item indicated in the category label set by the learning category setting unit 124a, and a minimum value (e.g., 0) as the element value for the other category items.

[0048] In training the classification model, the model training unit 124b repeatedly updates the parameter set of the classification model so as to minimize the magnitude of the difference between the score for each classification item calculated for an input value using the classification model and the output value for a data pair including that input value. Here, "minimization" means searching for a parameter set that minimizes the difference as much as possible, and does not necessarily mean absolutely minimizing it.

[0049] The classification model may be, for example, a decision tree, a random forest, or a neural network. For updating the parameter set, methods such as the steepest gradient method or the stochastic gradient descent method can be used. The index of the magnitude of the difference may be, for example, a weighted mean square error or cross entropy. The model learning unit 124b can determine whether the parameter set has converged based on whether the change in the magnitude of the difference between each parameter set update is equal to or less than a certain value. When determining that the parameter set has converged, the model learning unit 124b stops updating the parameter set and stores the parameter set of the classification model at that time in the storage unit 140.

[0050] In estimating scores using a classification model, the score calculation unit 124c reads and sets a parameter set stored in the storage unit 140. Using the classification model based on the set parameter set, the score calculation unit 124c calculates a score for each category related to a target event for each piece of posted information. The score calculation unit 124c associates the calculated score with the posted information to be processed and stores it in the storage unit 140. As a result, even if the same wording is included in the posted information, different categories of events are scored for different categories. For example, if the target event is baseball, a baseball player's last name will receive a higher score than the same last name in another type of event. Even if the word "spin" is the same when the target event is figure skating, the word "spin" in auto racing will receive different scores in different contexts. Therefore, scoring is possible for each type of event according to its importance and frequency in the context.

[0051] Examples of classification items according to this embodiment will be described. The set of classification items differs depending on the event type, but is common to events of the same type. The set of classification items for each event includes at least one fact classification item and one emotion classification item. The classification items for each event are not limited to one level, but can be hierarchically organized into multiple levels. Figure 9 shows the classification items in the first level for an auto race event, including race information, car information, and other information. The second level, which relates to other information, includes an emotion expression system, a thinking / analysis system, and a fact / confirmation system. Of these, the thinking / analysis system and fact / confirmation system of race information, car information, and other information belong to the fact classification. The emotion expression system belongs to the emotion classification.

[0052] Next, an example of providing posted information will be described. Provided data including posted information may be provided simultaneously with the content of the target event, and both may be presented on the information processing device 10. FIG. 10 shows an example of a display screen that presents video showing a live commentary of an auto race as a target event and text of posted information side by side. This display screen is displayed on the display unit 170. The live commentary video and text of the posted information are arranged above and below the display screen. The playback unit 160 plays back the live commentary audio of the auto race superimposed with spoken voice related to the posted information. More specifically, the live commentary audio, which includes the sounds of racing car engines and horns, is mixed with spoken voices that read out in sequence, such as "Didn't you crash?" and "You went off the course at the S-curve," and then played back.

[0053] The filtering setting unit 126b may set the amount of posted information to be provided based on the event type of the target event or the status of the target event. The filtering setting unit 126b can acquire, as the event type, a similar event notified by the similar event determination unit 122d. The filtering setting unit 126b can identify the status of the target event by referring to metadata attached to the content of the target event.

[0054] For example, for auto racing, athletics, and the like, the filtering setting unit 126b presets a maximum value for the amount of information provided, and adjusts the amount of information provided so as not to exceed the maximum value. This is because posted information is expected to be provided within a range that does not hinder the recipient from understanding the progress of the live broadcast. For example, for a concert, a play, a movie, or the like, the filtering setting unit 126b may set the amount of information provided during the performance to zero, and set the maximum amount of information provided during intermission or preparation time to a standard amount greater than zero. This is because posted information provided during the performance would interfere with the appreciation of the target event. On the other hand, posted information can be expected to be provided during non-performance times.

[0055] The filtering setting unit 126b may set a maximum provision amount greater than the standard provision amount for games such as competitive games, Go, and Shogi, because in games, the importance of audio in the content of the target event is low, and active provision of posted information can be expected.

[0056] The above settings that do not involve a user operation may be treated as initial values ​​and may be variable in response to a user operation. The filtering setting unit 126b may associate the settings specified by the operation with the event type of the target event and store them in the storage unit 140 instead of the initial values. The filtering setting unit 126b may read and apply the settings stored in the storage unit 140 when providing audio data of posted information related to a different event belonging to the event type.

[0057] Next, an example of a filtering information setting screen will be described. FIG. 11 is a diagram illustrating an example of the filtering information setting screen according to this embodiment. The filtering information setting screen is a screen for setting a priority for each classification item of posted information in response to a user operation. In the example of FIG. 11, the title of the target event is "Auto Race 2024," the venue name is "M Circuit," and the character string "Read-Aloud Text Adjustment Mixer" is displayed as the title of the setting that can be made for each classification item. The priority is indicated by the relative position of the knob in the longitudinal direction of the slider bar arranged for each classification item. The further to the right the knob is positioned, the higher the priority.

[0058] The classification items are "Race Information," "Car Information," "Other Information," "Emotional Expression," "Thinking and Analysis," and "Fact and Verification." Of these, "Race Information," "Car Information," and "Other Information" are classification items belonging to the first hierarchical level. "Other Information" is a classification item in the first hierarchical level, and encompasses the classification items "Emotional Expression," "Thinking and Analysis," and "Fact and Verification," which belong to the second hierarchical level. A relative priority is set for each of the classification items belonging to the second hierarchical level, "Emotional Expression," "Thinking and Analysis," and "Fact and Verification," among these three items.

[0059] The provision processing unit 126 uses the instructed priority to control the amount of posted information provided, the volume of the speech, etc., as described above, thereby adjusting the amount of posted information provided in accordance with the user's preferences. The provision processing unit 126 may be instructed to start, stop, pause, resume, etc., playback of the spoken voice in response to an operation. Screen components (e.g., buttons) used for these operations may be arranged on the filtering information setting screen. Here, pausing means temporarily stopping the provision of posted information for a suspension period until the next instruction to resume is given. The provision processing unit 126 saves posted information or its audio data that is determined to be available during the suspension period in chronological order. After the suspension period ends, the provision processing unit 126 provides the saved posted information or audio data in chronological order.

[0060] Next, an example of the posted information providing process according to this embodiment will be described. FIG. 12 is a flowchart showing an example of the posted information providing process according to this embodiment. (Step S102) Posted information collection unit 122a collects posted information distributed from posted information server device 40 via network NW. (Step S104) The event date and time determination unit 122b refers to the event information indicating the duration of each event, and identifies the target event based on the posting time of the collected posting information.

[0061] (Step S106) The score calculation unit 124c calculates the score for each classification item for the event type related to the target event using a classification model that has been trained in advance. (Step S108) The filtering unit 126a selects posted information having a classification item whose calculated score is equal to or greater than a predetermined reference value. (Step S110) The speech expression adjuster 126c identifies the score related to the emotion classification of the selected posted information, adjusts the speech expression parameters so that the speech expression becomes more pronounced as the score related to the emotion classification becomes higher, and outputs the adjusted speech expression parameters to the speech synthesis unit 126d.

[0062] (Step S112) The speech synthesis unit 126d uses the adjusted speech expression parameters to The speech expression adjuster 126c performs text-to-speech synthesis processing on the posted information selected using the adjusted speech expression parameters to generate speech data. (Step S114) The provision processing unit 126 causes the speech voice indicated by the generated voice data to be reproduced by the reproduction unit 160. Thereafter, the processing of FIG. 12 ends.

[0063] As described above, the information processing device 10 according to this embodiment includes a posted information processing unit 122 that collects event-related posted information, which is posted information related to events; a scoring unit 124 that calculates a score indicating the relevance of the event-related posted information to each predetermined classification item of the type of event using a predetermined classification model; and a provision processing unit 126 that controls the provision of the event-related posted information based on the score. With this configuration, event-related posted information is evaluated according to the score for each predetermined classification item for the type of event, and the provision of the event-related posted information is controlled based on the evaluated score. By adjusting the provision state based on the classification appropriate for the type of event, more desired event-related posted information is provided to users.

[0064] The provision processing unit 126 may determine the category items whose scores are equal to or greater than a predetermined reference value, and control the provision of event-related posted information in accordance with the priority of each category item set in response to an operation. With this configuration, event-related posted information is categorized according to predetermined categories and provided based on the priority specified in response to an operation. Therefore, event-related posted information related to the category desired by the user is provided preferentially, thereby improving user satisfaction.

[0065] The provision processing unit 126 may convert text indicating the event-related posted information into spoken voice. This configuration allows event-related posted information expressed in text to be provided as spoken audio, which can satisfy users who desire information provided by audio, including, for example, visually impaired users.

[0066] The classification items may include at least one fact classification and one or more sentiment classifications. With this configuration, event-related posted information is classified into at least whether the transmitted information primarily contains facts or primarily contains emotions, thereby responding to user preferences that can differ significantly depending on whether they are based on facts or emotions, and providing clues for determining the delivery format appropriate to that preference.

[0067] The provision processing unit 126 may set the speech expression of the speech voice based on the score calculated for the emotion classification. With this configuration, the speech expression is adjusted based on the degree of emotional expression. Since the poster's emotions are conveyed to the user through the speech expression of the speech voice, the event-related posted information is conveyed with a higher sense of realism.

[0068] When providing a plurality of pieces of event-related posted information at the same time, the providing processor 126 may provide different playback characteristics to the uttered voice for each piece of event-related posted information. The reproduction characteristics are, for example, at least one of volume, delay amount, and volume ratio between multiple channels. With this configuration, the speech corresponding to each of the multiple pieces of event-related posted information provided simultaneously is played back with different playback characteristics, allowing the user to easily understand the content of the speech conveyed by each piece of event-related posted information based on the difference in playback characteristics.

[0069] The posted information processing unit 122 may determine an event based on at least one of the posting time of the posted information, a keyword included in the posted information, and an entry phrase associated with the posted information. With this configuration, events related to the collected posted information are determined by referencing the event information for each event. Posted information that is less relevant to the event and posted information related to events other than the target event being processed is eliminated, making it possible to provide consistent event-related posted information that is specific to the target event.

[0070] The scoring unit 124 may determine the classification items of posted information of similar events that have a common type with the event in question, and may train the classification model using training data including the determined classification items for each piece of posted information. This configuration allows the classification model to learn the association between posted information from similar events held in the past and the determined classification items. By using the trained classification model, collected posted information is classified according to the classification items determined for similar events. This results in more meaningful classification of posted information than when arbitrary classification items are used.

[0071] The embodiments of the present invention have been described in detail above with reference to the drawings, but the specific configuration is not limited to that described above, and various design changes can be made within the scope of the gist of the present invention.

[0072] For example, some or all of the components of the input / output unit 150, the playback unit 160, the display unit 170, the operation unit 180, and the sound collection unit 190 may be omitted from the information processing device 10. The omitted components may be connected to the control unit 120 via a wired or wireless connection so as to be able to send and receive various types of data. In the above description, it is assumed that the information processing device 10 is a user device used by an individual user. However, the information processing device 10 may be configured as a server device accessible from other terminal devices via a network NW. In this case, the input source of the above-mentioned operation signals and voice commands is another terminal device. The output destination of the text, voice data, and other data of the posted information to be provided is also another terminal device. In this case, the providing processing unit 126 may perform filtering according to the operation independently for each accessing terminal device, and transmit the text, etc. of the posted information obtained by filtering.

[0073] Furthermore, the event information server device 30 may simultaneously broadcast the content of the target event to unspecified destination devices, or may distribute the content to specific destination devices. The event information server device 30 may transmit the content to the destination device, which is the request source device, in response to a distribution request from the request source device. The information processing device 10 may be a destination device for the event content. The event content may be transmitted over a broadcast transmission path instead of or together with the network NW, which is a communication transmission path. If an event is open to specified or unspecified audience members and audience members can post information about the event, the content of the event does not necessarily need to be distributed over the network NW. In this case, event information server device 30 may be omitted.

[0074] Furthermore, part or all of the information processing device 10 in the above-described embodiment may be realized as an integrated circuit such as an LSI (Large Scale Integration). Each functional block of the information processing device 10 may be individually implemented as a processor, or part or all of the functional blocks may be integrated into a processor. Furthermore, the integrated circuit implementation method is not limited to LSI, and may be implemented using a dedicated circuit or a general-purpose processor. Furthermore, if an integrated circuit implementation technology that can replace LSI emerges due to advances in semiconductor technology, an integrated circuit based on that technology may be used. [Explanation of symbols]

[0075] 10...information processing device, 30...event information server device, 40...posted information server device, 120...control unit, 122...posted information processing unit, 124...scoring unit, 126...provision processing unit, 140...storage unit, 150...input / output unit, 160...playback unit, 170...display unit, 180...operation unit, 190...sound collection unit

Claims

1. a post information processing unit that collects event-related post information, which is post information related to an event; a scoring unit that calculates a score indicating a relevance of the event-related posted information to each of predetermined classification items for the type of the event using a predetermined classification model; a providing processing unit that controls providing of the event-related posted information based on the score; An information processing device comprising:

2. The providing processing unit Determine the classification items for which the score is equal to or greater than a predetermined standard value; According to the priority of each classification item set according to the operation, Control provision of the event-related posted information The information processing device according to claim 1 .

3. The providing processing unit Converting the text indicating the event-related posted information into spoken speech The information processing device according to claim 2 .

4. The classification items include at least one fact classification and one emotion classification. The information processing device according to claim 3 .

5. The providing processing unit setting a speech expression for the speech based on the score calculated for the emotion classification; The information processing device according to claim 4 .

6. The providing processing unit When a plurality of pieces of event-related posted information are provided at the same time, different playback characteristics are given to the uttered voice for each piece of event-related posted information. The information processing device according to claim 3 .

7. The regeneration characteristics are At least one of volume, delay, and volume ratio between multiple channels The information processing device according to claim 6 .

8. The posted information processing unit The event is determined based on at least one of a posting time of the posted information, a keyword included in the posted information, and an entry phrase associated with the posted information. The information processing device according to claim 1 .

9. The scoring unit determining a classification item of posted information of a similar event that has a common type with the event; The classification model is trained using training data including the classification items determined for each piece of posted information. The information processing device according to claim 1 .

10. On the computer, A program for causing the information processing device according to claim 1 to function.

11. An information processing method in an information processing device, The information processing device, collecting event-related posted information, which is posted information about the event; calculating a score indicating a relevance of the event-related posted information to each predetermined classification item of the event type using a predetermined classification model; controlling provision of the event-related posted information based on the score; An information processing method that performs the above.

Images