Information verification program, information processing device and information processing system
The system accurately verifies the reliability of posted information by identifying events and their impact ranges, addressing the challenge of wide-impact events through geographical and temporal metadata analysis.
Patent Information
- Application Number
- JP2023085214
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-24
- Publication Date
- 2024-12-05
AI Technical Summary
Existing methods struggle to accurately verify the reliability of posted information, particularly during events with wide geographical or temporal impacts, such as disasters, due to mismatches in location and time between posted information and highly reliable sources.
An information processing system that utilizes posted information, including text, image, and audio, identifies the event and its impact range, calculating reliability based on geographical and temporal metadata compared to highly reliable information sources.
Enables high-accuracy reliability calculation of posted information by considering the event's geographical and temporal impact range, improving verification accuracy during widespread events.
Smart Images

Figure 2024168491000001_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to an information verification program, an information processing device, and an information processing system. [Background technology]
[0002] In recent years, posting services that display information posted by general users on a posting screen have become widespread. For example, some SNS (Social Networking Service, also known as Social Media) services allow users to post information that includes not only text information but also image information and audio information.
[0003] Such posted information can be a valuable source of information about the damage situation in various regions during emergencies such as disasters. However, there is a problem that the posted information may contain false information. For this reason, there is a demand for verifying the reliability and accuracy of posted information and detecting false information in real time.
[0004] As an example of such technology, the following posting evaluation method has been proposed, in which first information on the presence or absence of a predetermined phenomenon is extracted from text information included in the posted information, second information on the presence or absence of the phenomenon is extracted from highly reliable information, and an accuracy index of the posted information is calculated by comparing the first information with the second information.
[0005] Regarding the use of posted information, for example, a data analysis platform has been proposed that searches social networks for additional information related to trials. [Prior art documents] [Patent documents]
[0006] [Patent Document 1] Patent Publication No. 2022-122297 [Patent Document 2] US Patent Application Publication No. 2014 / 344172 Summary of the Invention [Problem to be solved by the invention]
[0007] By the way, in the method of verifying the reliability of posted information by comparing it with highly reliable information, for example, for an event that occurs in a specific location, such as a train delay or a crowded toilet, the reliability of the posted information indicating the event can be verified with a relatively high degree of accuracy. However, for an event that has a wide geographical or temporal impact, such as a disaster, even if the posted information is correct, the location and time of the event indicated by the posted information often do not match the location and time of the event obtained from the highly reliable information. For this reason, there is a problem that the reliability of posted information indicating such an event cannot be verified with a high degree of accuracy.
[0008] According to one aspect, the present invention has an object to provide an information verification program, an information processing device, and an information processing system that are capable of calculating the reliability of posted information with high accuracy. [Means for solving the problem]
[0009] In one proposal, an information verification program is provided that causes a computer to execute the following processes: acquire posted information including first information which is text information, second information which is at least one of image information or audio information, and metadata of the second information; identify an event that has occurred based on the first information; identify an impact range indicating the extent of at least one of the geographical or temporal impact of the event; and calculate a first reliability of the posted information based on the impact range and the metadata.
[0010] In addition, in one proposal, an information processing device is provided that executes a process similar to the process based on the above-mentioned information verification program. Furthermore, in one proposal, an information processing system having a first information processing device and a second information processing device is provided. The first information processing device accepts posting of posted information including first information, which is text information, and second information, which is at least one of image information and audio information, and displays the first information and the second information on a posting screen. The second information processing device acquires the posted information from the first information processing device, identifies an event that has occurred based on the first information, identifies an impact range indicating at least one of a geographical impact or a temporal impact of the event, and calculates a first reliability of the posted information based on the impact range and metadata of the second information. Effect of the Invention
[0011] On the one hand, the reliability of posted information can be calculated with high accuracy. [Brief description of the drawings]
[0012] [Figure 1] FIG. 2 is a diagram illustrating a configuration example and a processing example of a first information processing system. [Diagram 2] FIG. 11 illustrates an example of a configuration of an information processing system according to a second embodiment. [Diagram 3] FIG. 2 illustrates an example of a hardware configuration of a validation server. [Figure 4] FIG. 13 is a diagram illustrating a comparative example of calculation of reliability of posted information. [Diagram 5] FIG. 11 is a diagram illustrating an example of reliability calculation according to the second embodiment. [Figure 6] FIG. 2 illustrates an example of a configuration of processing functions included in a validation server; [Figure 7] FIG. 13 is a diagram illustrating an example of a data configuration of an occurrence event table. [Figure 8] FIG. 13 is a diagram illustrating an example of a data configuration of an event classification table. [Figure 9] 13 is a flowchart illustrating an example of an occurrence event information collection process. [Figure 10] 13 is a flowchart illustrating an example of a posted information acquisition process. [Figure 11] 13 is a flowchart illustrating an example of an influence range estimation process. [Figure 12] FIG. 13 is a diagram illustrating a specific example of an estimation process of an affected range. [Figure 13] 13 is a flowchart illustrating an example of a reliability calculation process. [Figure 14] FIG. 11 is a diagram showing another example of posted information. [Figure 15] FIG. 11 is a first diagram showing an example of output processing of a verification result. [Figure 16] FIG. 2 is a second diagram showing an example of output processing of a verification result. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0013] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. First Embodiment Fig. 1 is a diagram showing a configuration example and a processing example of a first information processing system. The information processing device 1 shown in Fig. 1 is a device that verifies the reliability of information posted to a posting service such as an SNS. The information processing device 1 has a processing unit 2. The processing unit 2 is, for example, a processor. In this case, the processing of the information processing device 1 shown below is realized by, for example, the processor executing a predetermined program.
[0014] The processing unit 2 acquires posted information including first information, which is text information, second information, which is at least one of image information and audio information, and metadata of the second information, as a verification target. In the example of FIG. 1, it is assumed that posted information 10 is acquired, which includes a posted statement 11, "My house was washed away by a flood," as the first information, and image information 12 as the second information. It is also assumed that metadata 13 included in the posted information 10 includes information on the shooting location of the image information 12. This information on the shooting location is expressed, for example, as latitude and longitude.
[0015] When the processing unit 2 acquires such posted information 10, it identifies the event that has occurred (the occurring event) based on the posted text 11. For example, the processing unit 2 extracts words from the posted text 11 and identifies the occurring event from the posted text 11 by comparing the extracted words with predetermined words indicating each of a plurality of events. In the example of FIG. 1, for example, a flood is identified as the occurring event from the word "flood" included in the posted text 11. Note that the occurrence event may be identified by, for example, comparing information on the actually occurring event acquired from a highly reliable information source such as a public institution with information on the event extracted from the posted text 11.
[0016] Next, the processing unit 2 identifies the range of influence (area of influence) caused by the identified occurrence event. This range of influence indicates at least one of the range of geographical influence and time-wise influence caused by the occurrence event. In the example of Fig. 1, since the metadata 13 includes the shooting location, which is geographical information, the geographical range of influence caused by the occurrence event, i.e., the range in which a flood will occur, is identified.
[0017] Here, as an example, the processing unit 2 acquires affected area information 20 corresponding to the occurring event (flood) from a highly reliable information source such as a public institution, and identifies the affected area 30 based on the affected area information 20. As the affected area information 20, a hazard map 21 indicating an expected flood area and a precipitation measurement value 22 are acquired. In the hazard map 21, for example, the expected flood area when the accumulated precipitation in a certain time T is 100 mm and 200 mm are shown. Meanwhile, as the precipitation measurement value 22, for example, the measurement value of the accumulated precipitation in the most recent certain time T is acquired. In the example of FIG. 1, it is assumed that the precipitation measurement value 22 indicates that the accumulated precipitation is 100 mm. In this case, the processing unit 2 identifies the expected flood area when the accumulated precipitation is 100 mm, which is shown in the hazard map 21, as the affected area 30.
[0018] Next, the processing unit 2 calculates the reliability of the posted information 10 based on the identified affected area 30 and the metadata 13 of the image information 12. For example, the processing unit 2 calculates the reliability of the posted information 10 based on the positional relationship between the estimated flood area identified as the affected area 30 and the shooting location of the image information 12 indicated by the metadata 13. As a specific example, the closer the distance between the estimated flood area and the shooting location, the higher the calculated reliability value.
[0019] As a comparative example, a method of calculating reliability by determining whether the shooting location indicated by the metadata 13 matches highly reliable information indicating the location where flooding has occurred will be given. In this case, the highly reliable information indicates, for example, whether flooding has occurred for a number of predetermined locations. However, the shooting location indicated by the metadata 13 does not necessarily match any of these multiple locations. For this reason, for example, even if the posted information 10 is correct information (i.e., flooding has occurred at the shooting location), if the shooting location does not match the location where flooding has occurred among the above-mentioned multiple locations, the calculated value of the reliability of the posted information 10 will be low.
[0020] In contrast to this, in the above process by the processing unit 2, the reliability of the posted information 10 is calculated based on the range affected by the occurring event (the expected flooded area in FIG. 1) and the metadata of the second information (the image information 12 in FIG. 1) (the shooting location of the image information 12 in FIG. 1). As a result, the reliability is calculated by comparing the geographical or temporal range affected by the event with the metadata, rather than the discrete place or time at which the event occurred. Therefore, even if an event that may affect a wide geographical or temporal range, such as a disaster such as a flood, is identified from the posted text 11, the reliability of the posted information 10 can be calculated with high accuracy.
[0021] As the time range of influence, for example, the occurrence period of the occurrence event (the period from the start time of occurrence to the end time of occurrence) can be applied. In this case, the creation time of the second information (the shooting time in the case of image information) is used as metadata, and the processing unit 2 can accurately calculate the reliability based on the occurrence period and the creation time.
[0022] In addition, both the geographical impact range and the temporal impact range may be used as the impact range. In this case, the processing unit 2 can accurately calculate the reliability based on, for example, a comparison result between the geographical information included in the metadata and the geographical impact range, and a comparison result between the temporal information included in the metadata and the temporal impact range.
[0023] Second Embodiment Fig. 2 is a diagram showing a configuration example of an information processing system according to a second embodiment. The information processing system shown in Fig. 2 includes an SNS server 50, information source servers 60a, 60b, ..., and a verification server 100. The SNS server 50, the information source servers 60a, 60b, ..., and the verification server 100 are connected via a network 70. The network 70 is, for example, a wide area communication network such as the Internet.
[0024] The SNS server 50 is a server computer operated by an SNS operator. The SNS server 50 displays posted information posted by users on a posting screen. In the present embodiment, as an example, posted information including a posting statement (text) and an image is posted to the SNS server 50 and displayed on the posting screen. The image included in the posted information may be either a still image or a moving image.
[0025] The information source servers 60a, 60b,... are server computers that serve as information sources for the validation server 100 to collect information on various events. For example, the information source servers 60a, 60b,... are web servers that operate web sites that provide information on events. The information source servers 60a, 60b,... are operated by public institutions such as the state or local government, or private businesses that provide weather information, etc. This enables the validation server 100 to obtain highly reliable information on events from the information source servers 60a, 60b,....
[0026] The verification server 100 monitors posted information posted to the SNS server 50, verifies the reliability of the posted information, and outputs the verification result. For example, the verification server 100 extracts event information related to events such as weather, earthquakes, and disasters from the posted information, and collects related information related to the extracted event information from the information source servers 60a, 60b, .... The verification server 100 evaluates the reliability of the posted information by comparing the extracted event information with the collected related information, and calculates a reliability that is an index of reliability.
[0027] 3 is a diagram showing an example of a hardware configuration of the validation server 100. The validation server 100 is realized as a computer having the hardware configuration shown in FIG. The validation server 100 shown in FIG. 3 includes a processor 101, a random access memory (RAM) 102, a hard disk drive (HDD) 103, a graphics processing unit (GPU) 104, an input interface (I / F) 105, a reader 106, and a network interface (I / F) 107.
[0028] The processor 101 performs overall control of the entire verification server 100. The processor 101 is, for example, a central processing unit (CPU), a micro processing unit (MPU), a digital signal processor (DSP), an application specific integrated circuit (ASIC), or a programmable logic device (PLD). The processor 101 may also be a combination of two or more elements of a CPU, an MPU, a DSP, an ASIC, or a PLD.
[0029] The RAM 102 is used as a main storage device of the verification server 100. The RAM 102 temporarily stores at least a part of an OS (Operating System) program and application programs to be executed by the processor 101. The RAM 102 also stores various data necessary for processing by the processor 101.
[0030] The HDD 103 is used as an auxiliary storage device of the validation server 100. The OS program, application programs, and various data are stored in the HDD 103. Note that other types of non-volatile storage devices such as SSDs (Solid State Drives) can also be used as the auxiliary storage device.
[0031] A display device 104a is connected to the GPU 104. The GPU 104 displays an image on the display device 104a in accordance with an instruction from the processor 101. The display device 104a may be a liquid crystal display or an organic EL (ElectroLuminescence) display.
[0032] An input device 105a is connected to the input interface 105. The input interface 105 transmits a signal output from the input device 105a to the processor 101. The input device 105a may be a keyboard or a pointing device. The pointing device may be a mouse, a touch panel, a tablet, a touch pad, a trackball, or the like.
[0033] A portable recording medium 106a is detachably attached to the reading device 106. The reading device 106 reads data recorded on the portable recording medium 106a and transmits it to the processor 101. The portable recording medium 106a may be an optical disk, a semiconductor memory, or the like.
[0034] The network interface 107 transmits and receives data to and from other devices such as the SNS server 50 via the network 70 . The above hardware configuration can realize the processing functions of the validation server 100. The SNS server 50 and the information source servers 60a, 60b, . . . can also be realized as computers having the hardware configuration shown in FIG.
[0035] Incidentally, in the case of emergencies such as disasters, accurate information must be provided in real time to ensure people's safety. Currently, when a disaster occurs, news media reports the damage situation in each region as much as possible, but detailed information on the damage situation in each region is not provided in real time. On the other hand, on SNS, general users report events occurring in various regions in real time. For this reason, information posted on SNS can be a valuable source of information in emergencies such as disasters.
[0036] However, fake and misinformation is often spread on SNS. For this reason, if fake or misinformation is spread during an emergency such as a disaster, it may have a negative impact on people's behavior and psychology, causing confusion. In particular, posted information that includes images and audio is more likely to attract users' attention than posts that are text only, because viewers can understand the information visually and audibly. For this reason, when images or audio containing fake or misinformation are spread, their influence is strong. Therefore, it is important to enable users who view posted information to recognize whether posted information that includes not only text but also images and audio is correct information, or fake or misinformation.
[0037] 4 is a diagram showing a comparative example of calculation of reliability of posted information. As an example of a method for calculating reliability of posted information, a method for calculating reliability of posted information by determining whether metadata of an image included in the posted information (e.g., a shooting position) matches highly reliable information on event information extracted from the posted information.
[0038] In the example of Fig. 4, it is assumed that posted information 200 including a captured image 202 and metadata 203 related to the image 202 has been acquired. The posted text 201 is a sentence saying "Houses are sinking due to landslides," and the image 202 shows a scene as shown in the posted text 201. The metadata 203 includes, for example, the time when the image 202 was captured, "15:00," and the location where the image was captured, "Town C, District B."
[0039] Also, assume that highly reliable information 210 is acquired from a highly reliable information source to verify this posted information 200. The highly reliable information 210 illustrated in Fig. 4 is landslide warning information issued by the government. In the example of Fig. 4, the highly reliable information 210 indicates that a landslide warning information was issued for area A at 3 p.m.
[0040] Assume that the reliability of the posted information 200 is calculated by such a coincidence determination between the posted information 200 and the highly reliable information 210. In this case, the shooting time included in the metadata 203 matches the predicted landslide warning time (15:00) indicated by the highly reliable information 210, but the shooting location "District B, Town C" included in the metadata 203 is different from the landslide warning area "District A" indicated by the highly reliable information 210. For this reason, even if the content of the posted information 200 is correct, the reliability of the posted information 200 may be calculated to be a low value.
[0041] For example, events such as train delays and crowded restrooms occur in specific locations, so for such events, the reliability can be calculated with a relatively high degree of accuracy by determining whether highly reliable information indicating the location of the event matches the metadata 203.
[0042] On the other hand, the impact of a disaster tends to extend over a relatively wide area. The highly reliable information 210 provides information indicating whether or not a disaster has occurred at each of a number of predetermined locations. However, it is difficult to determine whether or not a disaster has occurred at a location that is a certain distance away from any of the multiple locations based only on the highly reliable information 210 as shown in FIG. 4. This reduces the accuracy of calculating the reliability.
[0043] In response to this, the validation server 100 of this embodiment collects information that can identify the range of influence (scope of influence) caused by the occurrence of an event such as a disaster from websites of public institutions, websites of service companies that provide information on events, etc. The validation server 100 identifies the scope of influence caused by the occurrence of the event from the collected information, and calculates the reliability of the posted information 200 based on the identified scope of influence and the metadata 203.
[0044] As the range of influence, at least one of a geographical range of influence and a temporal range of influence is used. When a geographical range of influence is used, the reliability is calculated based on the shooting location and the range of influence, which are geographical information in the metadata 203. When a temporal range of influence is used, the reliability is calculated based on the shooting time and the range of influence, which are temporal information in the metadata 203.
[0045] In the following description, a geographical range of influence is used as an example. The posted information to be verified includes a posted text and an image. If the posted information includes audio instead of an image, the audio pickup location and audio pickup time included in the audio metadata may be used instead of the shooting location and shooting time included in the image metadata.
[0046] Fig. 5 is a diagram showing an example of reliability calculation in the second embodiment. In Fig. 5, the posted information 200 shown in Fig. 4 is the target of reliability calculation. In the example of FIG. 5, a hazard map showing a risk area of a terrain collapse due to a landslide is used as the affected range information 221 for identifying the affected range of a landslide. The verification server 100, for example, identifies the collapse risk area shown in the hazard map as the affected range, and calculates the reliability based on the shortest distance between this collapse risk area and the shooting position of the image 202 included in the metadata 203. The shorter the shortest distance, the higher the calculated reliability value. In the example of FIG. 5, since the shortest distance between the collapse risk area and the shooting position is relatively short at 50 m, it is determined that the shooting position may be subject to landslide damage. Therefore, a relatively high value is calculated as the reliability of the posted information 200. In this way, by using the affected range of an event, it is possible to calculate the reliability with high accuracy.
[0047] The information on the extent of impact for identifying the extent of impact can be, for example, disaster occurrence simulation information, disaster occurrence risk information, information showing the extent of past disaster occurrence, etc. This information on the extent of impact is obtained from highly reliable information sources, such as the websites of public institutions and the websites of service companies that provide information on events.
[0048] As shown in FIG. 5, the validation server 100 can further calculate the reliability by combining various highly reliable information in addition to the affected area. For example, the highly reliable information 222 shown in FIG. 5 is landslide warning information issued by the government, and indicates that there is a possibility of a landslide occurring in area A. The validation server 100 can calculate the reliability based on, for example, the positional relationship (such as distance) between area A indicated in the highly reliable information 222 and the shooting location included in the metadata 203. In the example of FIG. 5, area A and area B, which is the shooting location, are adjacent to each other, and it is determined that the shooting location is close to a landslide warning area. For this reason, a relatively high value is calculated as the reliability of the posted information 200.
[0049] The alert area "Area A" indicated by the highly reliable information 222 can be considered as the range of influence of the disaster occurrence. That is, the validation server 100 can also acquire the highly reliable information 222 as the influence range information.
[0050] 5 is a measurement of the accumulated precipitation amount issued by the government, and indicates that the accumulated precipitation amount from the start of the rain to 3 p.m. was measured to be 250 mm. For example, since the accumulated precipitation amount up to the shooting time included in the metadata 203 is high (for example, exceeds a predetermined threshold), the verification server 100 determines that there is a risk of a landslide at the shooting position and calculates a relatively high value as the reliability of the posted information 200.
[0051] The validation server 100 can calculate the final reliability of the posted information 200 by integrating (for example, averaging) the three types of reliability calculated as described above.
[0052] 6 is a diagram showing an example of the configuration of the processing functions of the validation server 100. The validation server 100 includes a storage unit 110, an event information collection unit 121, a posted information collection unit 122, an influence range estimation unit 123, a reliability calculation unit 124, and a validation result output unit 125.
[0053] The storage unit 110 is a storage area secured in a storage device included in the validation server 100, such as the RAM 102 or the HDD 103. The storage unit 110 stores an occurrence event table 111, an event classification table 112, and a validation result table 113.
[0054] Information about an event that has occurred (occurring event) is registered in the occurring event table 111. The information registered in the occurring event table 111 is information collected from predetermined highly reliable information sources such as the information source servers 60a, 60b, . . .
[0055] Information indicating the relationship between a cause event and a result event is registered in the event classification table 112. A cause event is an event that can be the cause of a result event, and a result event is an event that can be caused by the occurrence of a cause event. In addition, information indicating the type of measurement value used to estimate the impact range of a result event and information indicating the reference destination of information referenced to estimate the impact range are also registered in the event classification table 112.
[0056] Information indicating the verification result of the posted information is registered in the verification result table 113. For example, an identification number for identifying the posted information and the reliability of the posted information are registered in the verification result table 113 in association with each other.
[0057] The processes of the event information collecting unit 121, the posted information collecting unit 122, the influence extent estimating unit 123, the reliability calculating unit 124, and the verification result output unit 125 are realized by, for example, the processor 101 executing a predetermined program.
[0058] The event information collection unit 121 collects information about the occurring event from predetermined highly reliable information sources such as the information source servers 60a, 60b, . . . and registers the information in the occurring event table 111 as occurring event information.
[0059] The posted information collection unit 122 acquires posted information posted to the SNS server 50 from the SNS server 50 . The impact range estimation unit 123 analyzes the posted text included in the posted information to extract information related to the event, and identifies an occurring event related to the event based on the occurring event table 111. The impact range estimation unit 123 collects impact range information and high reliability information corresponding to the identified occurring event, and estimates the impact range of the occurring event based on the collected information and metadata of the posted information.
[0060] The reliability calculation unit 124 calculates a first reliability based on the metadata of the image in the posted information and the estimated range of influence. The reliability calculation unit 124 also calculates a second reliability based on the distance between the shooting position of the image and the posting position of the posted information itself. Furthermore, the reliability calculation unit 124 calculates a third reliability based on the difference between the shooting time of the image and the posting time of the posted information itself. Then, the reliability calculation unit 124 calculates a total reliability by integrating the first to third reliability.
[0061] The verification result output unit 125 outputs the calculated overall reliability. For example, the verification result output unit 125 creates a reliability display screen on which the first to third reliability are written together with the overall reliability, and displays the reliability display screen on the user's terminal. The reliability display screen may be displayed on a posting screen by the SNS server 50 together with the posted information to be verified.
[0062] 7 is a diagram showing an example of the data configuration of the occurring event table 111. Information about the occurring event collected by the event information collection unit 121 is accumulated in the occurring event table 111. The occurring event ID, the occurring event name, the occurring event classification name, and the occurrence / end time are registered in the occurring event table 111 as information about the occurring event.
[0063] The occurrence event ID is identification information for identifying the occurrence event. The occurrence event name indicates the name of the occurrence event. For example, the name given to the occurrence event by a public institution is registered as the occurrence event name. The occurrence event classification name is registered as the classification name to which the occurrence event corresponds among the result event classification names that classify result events. The occurrence / end time indicates the occurrence time and end time of the occurrence event. In FIG. 7, 2:00 on July 25, 2022 is registered as the occurrence time, but the end time is not registered. This indicates that the occurrence event has not yet ended.
[0064] Fig. 8 is a diagram showing an example of the data configuration of an event classification table. Information indicating the correlation between cause event classification names that classify cause events and result event classification names that classify result events is registered in the event classification table 112. For example, in Fig. 8, result event classification names of "flood," "landslide," and "inundation" correspond to the cause event classification name of "typhoon." This indicates that a cause event of a typhoon can cause result events of flood, landslide, and inundation.
[0065] Other possible causal events include, for example, earthquakes, large-scale accidents, etc. Consequences of earthquakes include tsunamis, landslides, etc. Consequences of large-scale accidents include power outages, the arrival of harmful substances, etc.
[0066] In the event classification table 112, for each combination of a cause event classification name and a result event classification name, influencing factors and influence range standard information are registered in association with each other. The influencing factors and influence range standard information are information for estimating the influence range (influence range information). The influencing factors indicate the type of measurement value referenced to estimate the influence range of the result event. The influence range standard information is standard information referenced to estimate the influence range. For example, the influence range of a result event is identified from the measurement value indicated by the influencing factors and the influence range standard information. Note that in practice, link information indicating a predetermined reference destination (source of acquisition) of the influence range standard information is also registered in the item of influence range information.
[0067] Next, the process of each unit of the validation server 100 will be described. FIG. 9 is a flowchart showing an example of an occurring event information collection process. [Step S11] The event information collector 121 scans predetermined information sources to obtain, from those information sources, occurrence event information indicating a new occurrence event and update information of occurrence event information of an event that has already occurred. Note that, in this step S11, for example, information on events that have occurred throughout Japan may be obtained.
[0068] [Step S12] The event information collector 121 registers occurring event information indicating a new occurring event in the occurring event table 111. Furthermore, upon receiving update information of the occurring event information of an already-occurring event (e.g., the end time of the occurring event), the event information collector 121 causes the corresponding occurring event information in the occurring event table 111 to reflect the update information.
[0069] [Step S13] Determine whether a certain amount of time has passed since the execution of the process of step S11. If the certain amount of time has not passed, the process enters a wait state until the certain amount of time has passed. If the certain amount of time has passed, the process returns to step S111, and the process of acquiring the event information is executed.
[0070] FIG. 10 is a flowchart illustrating an example of the posted information acquisition process. [Step S21] The posted information collection unit 122 acquires, from the posting server 50, posted information that has been posted to the posting server 50.
[0071] [Step S22] The posted information collecting unit 122 verifies the authenticity of the image included in the acquired posted information. If the image is determined to be authentic as a result of the verification, the process proceeds to step S23, and if the image is determined to be inauthentic, the process ends. In the latter case, for example, the verification result output unit 125 outputs 0 points as the overall reliability. Also, information indicating that the image is fraudulent may be displayed together with the posted information.
[0072] In the verification process in step S22, the posted information collecting unit 122 verifies, for example, whether an image included in the posted information is a photograph or a video image taken in the real world, and not an artificially created or processed image such as an animation or an illustration. For example, a classifier that distinguishes between an artificially created or processed image and a photograph taken in the real world is created in advance by machine learning using the former image and the latter photograph. In step S22, the posted information collecting unit 122 inputs the image in the posted information to the created classifier, thereby verifying whether the image is a photograph taken in the real world.
[0073] Furthermore, when metadata is added to an image in the posted information, the posted information collecting unit 122 verifies whether the photographing location and photographing time of the image included in the metadata are legitimate information. This verification may be performed, for example, by using an image authorship verification function provided in commercially available image editing software. Alternatively, the following method may be adopted. A first hash value calculated based on a parameter embedded in a predetermined area of the image, a second hash value calculated based on a script for generating a hash value, and the script are recorded in the metadata. Then, in step S22, the posted information collecting unit 122 calculates a third hash value based on the script in the metadata, and calculates a fourth hash value based on the script in the metadata. The posted information collecting unit 122 compares the first and second hash values in the metadata with the calculated third and fourth hash values to determine whether the image has been tampered with.
[0074] [Step S23] The post information collection unit 122 determines whether the image in the post information has a shooting location and a shooting time added as metadata. If the corresponding metadata has been added, the process proceeds to step S25. If the corresponding metadata has not been added, the process proceeds to step S24.
[0075] [Step S24] The posted information collecting unit 122 creates information on the image shooting location and shooting time that are not added as metadata, for example, by analyzing the image in the posted information. As an example, the posted information collecting unit 122 compares the image in the posted information with an image previously taken from the sky, such as a satellite image or an aerial image, to estimate the shooting location of the image in the posted information and add it to the image as metadata. The posted information collecting unit 122 also estimates the shooting time of the image based on the shooting location included in the metadata or the way sunlight enters and the way shadows are cast at the shooting location estimated in the above procedure, and adds it to the image as metadata. The posted information collecting unit 122 may also perform a morphological analysis of the posted text to extract proper nouns, expressions indicating locations, and expressions indicating times, and consider the former as the shooting location and the latter as the shooting time.
[0076] [Step S25] The posted information collection unit 122 temporarily stores the acquired posted information in the storage unit 110. Fig. 11 is a flowchart showing an example of an influence range estimation process, and Fig. 12 is a diagram showing a specific example of the influence range estimation process.
[0077] [Step S31] The influence scope estimation unit 123 analyzes the posted text included in the posted information to identify the cause event category name, the result event category name, and the occurred event name associated with the posted information.
[0078] For example, the influence range estimation unit 123 extracts nouns from the posted text by performing a morphological analysis of the posted text. When the extracted nouns match the occurrence event names registered in the occurrence event table 111, the influence range estimation unit 123 identifies the occurrence event names as occurrence event names related to the posted information. In this case, the influence range estimation unit 123 extracts the occurrence event classification names associated with the identified occurrence event names from the occurrence event table 111, and identifies the extracted occurrence event classification names as cause event classification names related to the posted information. In addition, the influence range estimation unit 123 extracts the result event classification names associated with the identified cause event classification names from the event classification table 112, and determines whether any of the extracted result event classification names match the words extracted from the posted text. When any of the extracted result event classification names match the words extracted from the posted text, the influence range estimation unit 123 identifies the result event classification name as the result event classification name related to the posted information. Furthermore, if none of the extracted result event classification names match the words extracted from the posted text, the impact scope estimation unit 123 searches the posted text for a noun whose meaning is close to the extracted result business classification name, and if found, identifies the corresponding result business classification name as the result business classification name corresponding to the posted information. Note that the semantic closeness of nouns may be determined, for example, by a method such as word2vec, which is a distributed representation of words.
[0079] On the other hand, if the noun extracted from the posted text is not registered in the occurred event table 111 as the name of the occurring event, the following process is executed. The influence range estimation unit 123 judges whether or not there is a noun that matches a cause event classification name or a result event classification name registered in the event classification table 112 among the nouns extracted from the posted text. If there is a noun that matches a result event classification name among the extracted nouns, the influence range estimation unit 123 specifies the result event classification name as a result event classification name related to the posted information. In addition, the influence range estimation unit 123 specifies the cause event classification name associated with the result event classification name in the event classification table 112 as a cause event classification name related to the posted information. On the other hand, if there is a noun that matches a cause event classification name among the nouns extracted from the posted text, the influence range estimation unit 123 specifies the cause event classification name as a cause event classification name related to the posted information. In this case, the influence range estimation unit 123 specifies the result event classification name related to the posted information by the above-mentioned method.
[0080] In addition, if the noun extracted from the posted text does not match either the cause event classification name or the result event classification name, the cause event classification name and the result event classification name can be identified, for example, based on whether or not there is a noun among the extracted nouns that is close in meaning to the cause event classification name or the result event classification name.
[0081] When the cause event classification name related to the posted information is identified in this manner, the influence range estimation unit 123 extracts the occurring event information in which the identified cause event classification name is registered as the occurring event classification name from the occurring event table 111. When one corresponding occurring event information is extracted, the influence range estimation unit 123 identifies the occurring event name included in the occurring event information as the occurring event name related to the posted information. When multiple corresponding occurring event information is extracted, the influence range estimation unit 123 identifies from the occurring event information an occurring event whose noun related to the occurring event name or the occurring event classification name matches the noun extracted from the posted text, and identifies the occurring event name included in the identified occurring event information as the occurring event name related to the posted information.
[0082] In the example of FIG. 12, it is assumed that posted information 230 including posted text 231, image 232, metadata 233 of image 232, and metadata 234 of posted information 230 is acquired. In step S31, "flood" and "house" are extracted as nouns from posted text 231, and among these, "flood" matches the result event classification name "flood", so "flood" is identified as the result event classification name related to posted information 230. Furthermore, since the cause event classification name "typhoon" is associated with the result event classification name "flood" in the event classification table 112 (see FIG. 8), "typhoon" is identified as the cause event classification name related to posted information 230. Then, occurred event information in which "typhoon" is registered as the occurred event classification name is extracted from occurred event table 111, and the occurred event name "Typhoon No. 1" included in the extracted occurred event information is identified as the occurred event name related to posted information 230 (see FIG. 7).
[0083] For example, when multiple pieces of occurring event information in which "typhoon" is registered as the occurring event classification name are registered in the occurring event table 111, the influence range estimation unit 123 identifies one of the occurring event information based on the relationship between the occurrence time included in the occurring event information and the shooting time included in the metadata 233. For example, occurring event information whose occurrence time is included in the period from the shooting time to a certain time before is identified.
[0084] [Step S32] The impact extent estimation unit 123 collects impact extent standard information corresponding to the result event classification name identified in step S31. Here, the information source of the impact extent standard information is identified based on the link information of the impact extent standard information registered in the event classification table 112 in association with the combination of the cause event classification name and the result event classification name identified in step S31. As the impact extent standard information, for example, damage prediction information officially issued by a public institution is collected.
[0085] 12, "typhoon" and "flood" are specified as the cause event category name and the result event category name, respectively, corresponding to the posted information 230. Therefore, a flood hazard map is collected as the affected area standard information based on the event category table 112 (see FIG. 8).
[0086] [Step S33] The influence scope estimation unit 123 analyzes the posted message and determines whether the posted message is in the present tense or the past tense. [Step S34] The influence range estimation unit 123 determines the time and position for estimating the influence range based on the identified tense. When the tense is past tense, the influence range at the shooting position and shooting time of the image is estimated. This is because, when the tense is past tense, it is considered that information prior to the posting time of the posted information (mainly information at the time when the image was taken) is expressed in the posted message. On the other hand, when the tense is present tense, the influence range at the posting position and posting time of the posted information is specified. This is because, when the tense is present tense, it is considered that information at the time when the posted information is posted is expressed in the posted message. In this way, by determining the time and position for estimating the influence range according to the tense of the posted message, it becomes possible to accurately estimate the influence range in the next step S35.
[0087] In the example of Fig. 12, the tense of the posted statement 231 is the present tense. Therefore, the posting time "September 10, 2022, 15:00" included in the metadata 234 of the posted information 230 is determined as the time for estimating the impact range. In addition, although not shown, the posting position included in the metadata 234 is determined as the position for estimating the impact range.
[0088] The metadata 234 of the posted information 230 is not necessarily added to the posted information 230. When the metadata 234 is not added to the posted information 230, the influence range estimation unit 123 acquires the metadata 234 from the SNS server 50 separately from the posted information 230.
[0089] [Step S35] The influence range estimation unit 123 collects highly reliable information. Here, the measurement values of the influence factors at the positions and times identified in step S34 are collected as highly reliable information. The collected influence factors are those associated with the result event classification names identified in step S31 in the event classification table 112. In addition, the link destinations that are the information sources of the influence factors are highly reliable link destinations such as the websites of public institutions and companies that provide weather information services. Such link destinations may be determined in advance for the corresponding influence range reference information.
[0090] The influence extent estimation unit 123 estimates the influence extent at the time specified in step S34 based on the collected measurement values of the influencing factors and the influence extent reference information collected in step S32.
[0091] In the example of Fig. 12, as described above, a flood hazard map is collected as the affected area reference information. This flood hazard map indicates areas expected to be inundated when the 24-hour accumulated precipitation is 200 mm, 300 mm, and 400 mm, for example. In this case, in step S35, the measured value of the accumulated precipitation in the most recent 24 hours up to the posting time at the posting location (actually, the observation point closest to the posting location) is collected as an influencing factor. In Fig. 12, it is assumed that the measured value of the accumulated precipitation is 450 mm. In this case, the affected area estimation unit 123 identifies the area expected to be inundated when the accumulated precipitation is 400 mm from the flood hazard map, and estimates this area as the affected area.
[0092] For example, if the measured cumulative precipitation is less than the minimum cumulative precipitation (200 mm in FIG. 12) at which flooding is expected to occur in the flood hazard map, the affected area estimation unit 123 will estimate that there is “no affected area” (i.e., no flooding will occur at the relevant location due to Typhoon No. 1).
[0093] 13 is a flowchart showing an example of a reliability calculation process. In the following description, the first to third reliabilities are assumed to take values ranging from 0 points to 100 points. [Step S41] The reliability calculation unit 124 calculates a first reliability based on the difference between the affected area estimated in step S35 of Fig. 11 and the shooting position included in the metadata of the image. Basically, the reliability calculation unit 124 calculates the first reliability to be higher as the difference between the affected area and the shooting position is smaller.
[0094] As an example, when the shooting position is included in the range of influence, the reliability calculation unit 124 calculates the first reliability as 100 points. When the shooting position is outside the range of influence, the reliability calculation unit 124 calculates the first reliability to be lower as the distance between the shooting position and the boundary of the range of influence closest to the shooting position increases.
[0095] As another example, the reliability calculation unit 124 may calculate the first reliability based on the difference between the river that is the source of the flood and the shooting position. For example, a river that is adjacent to or included in the affected area and is the farthest from the shooting position is selected as this river. For example, when the shooting position is included in the affected area, the reliability calculation unit 124 calculates the first reliability to be higher within a range of 100 points to 70 points as the distance between the shooting position and the river becomes shorter.
[0096] As an example, if a straight line is drawn through the shooting position and perpendicular to the river, the distance from the position of the river on that line to the boundary of the affected area is divided into 30 equal parts. Then, every time the shooting position moves 1 / 30th of a second from the position of the river toward the boundary of the shooting position, 1 point is subtracted from 100 points to the first reliability. For example, if the shooting position is located on the boundary of the affected area and is the furthest away from the river on the above line, the first reliability is 70 points.
[0097] Furthermore, when the shooting position is outside the affected range, the reliability calculation unit 124 reduces the first reliability by 1 point for every 0.1 km that the distance between the shooting position and the river increases, with a maximum value of 70 points and a minimum value of 0 points. When it is estimated that there is no affected range due to low precipitation, the reliability calculation unit 124 sets the first reliability to a lower value when there is no affected range than when there is, when the distance between the shooting position and the river is the same.
[0098] In step S41, for example, if the tense of the posted message is the past tense in step S34 of Fig. 11, the shooting location may be used, and if the tense of the posted message is the present tense in step S34, the posting location of the posted information may be used instead of the shooting location. This can improve the calculation accuracy of the first reliability itself.
[0099] [Step S42] The reliability calculation unit 124 calculates the second reliability based on the distance between the shooting location included in the metadata of the image and the posting location of the posted information. Basically, the smaller the distance between the shooting location and the posting location, the higher the value of the second reliability.
[0100] For example, if the distance between the shooting location and the posting location is within 5 km, the reliability calculation unit 124 calculates the second reliability to be higher within a range of 100 points to 70 points as the distance decreases. Also, if the distance between the shooting location and the posting location is 5 km or more, the reliability calculation unit 124 deducts 1 point from the second reliability for every 0.1 km increase in the distance within a range of 70 points to 0 points. As an example, if the distance between the shooting location and the posting location is 2 km, the second reliability is calculated to be 88 points. Such a second reliability is an index showing the party involved in the posted information (or posted text).
[0101] [Step S43] The reliability calculation unit 124 calculates the third reliability based on the difference between the shooting time included in the metadata of the image and the posting time of the posted information. Basically, the reliability calculation unit 124 sets the third reliability to a lower value as the difference between the shooting time and the posting time increases.
[0102] As a specific example, the reliability calculation unit 124 determines the time when the causal event begins to affect the shooting position as the start time of the causal event. When the causal event is a typhoon as shown in FIG. 12, the start time of the causal event is, for example, the time when the shooting position enters the rainstorm area. The time when the shooting position enters the rainstorm area is acquired, for example, by highly reliable information. The reliability calculation unit 124 calculates the third reliability based on the difference between the shooting time and the projection time, taking into account such a start time of the causal event.
[0103] For example, when the tense of the posted message is past tense, if the difference between the posting time and the shooting time is within six hours, the reliability calculation unit 124 calculates the third reliability to be higher within a range of 50 points to 100 points, the smaller the time difference is, while if the time difference is more than six hours, the reliability calculation unit 124 reduces the third reliability by 10 points for each additional hour of the time difference within a range of 50 points to 0 points.
[0104] Furthermore, when the tense of the posted text is the present tense or the future tense, if the difference between the posting time and the shooting time is within three hours, the reliability calculation unit 124 calculates the third reliability to be higher within a range of 50 points to 100 points, the smaller the time difference, while if the time difference is more than three hours, the reliability calculation unit 124 reduces the third reliability by 10 points for each additional hour of the time difference within a range of 50 points to 0 points.
[0105] In addition, if the photographing time is before the start time of the causal event, there is a high possibility that the photographing time is inconsistent and a photograph unrelated to the causal event is used, so the third reliability is set to a low value. For example, if the photographing time is before the start time of the causal event, the reliability calculation unit 124 sets the third reliability to 0 points.
[0106] As an example, in the posted information 230 shown in Fig. 12, the photographed time is 14:30 on September 10, 2022, and the posted time is 15:00 on September 10, 2022. If the start time of the causal event is 9:00 on September 10, 2022, this start time is before the photographed time, and the tense of the posted statement 231 is the present tense, so the calculation method for when the difference between the posted time and the photographed time is within 3 hours is applied. Since the difference between the photographed time 14:30 and the posted time 15:00 is 30 minutes (= 0.5 hours), the third reliability is calculated as 92 points using the formula 100-(50 / 3) x 0.5.
[0107] [Step S44] The reliability calculation unit 124 calculates the overall reliability by integrating the calculated first to third reliability. For example, the overall reliability is calculated by averaging the first to third reliability. As in the above example, if the first reliability is 70 points, the second reliability is 88 points, and the third reliability is 92 points, the overall reliability is calculated as 83 points (rounded off to the nearest tenth).
[0108] Fig. 14 is a diagram showing another example of posted information. Posted information 240 shown in Fig. 14 includes posted text 241, image 242, metadata 243 of image 242, and metadata 244 of posted information 240. It is assumed that the posting position and posting time of metadata 244 are the same as those of posted information 230 shown in Fig. 12.
[0109] In the case of this posted information 240, the affected area is identified as the same affected area (area expected to be flooded) as that of the posted information 230 shown in Fig. 12. However, the first reliability is calculated as 0 points because the photographed location of image 242 is in Indonesia, which is not included in the affected area set in Japan and is more than 4000 km away from the affected area.
[0110] In addition, since the shooting location and the posting location are separated by a distance of more than 4000 km, the second reliability is also calculated as 0 points. Furthermore, if the start time of the causal event is 9:00 on September 10, 2022, as in the above example, the shooting time included in the metadata 243 is before the start time of the causal event, so the third reliability is also calculated as 0 points.
[0111] Therefore, the overall reliability is calculated as 0 points. Fig. 15 is a first diagram showing an example of an output process of a verification result. A reliability display screen 250a shown in Fig. 15 is a display screen based on the posted information 230 shown in Fig. 12. Fig. 16 is a second diagram showing an example of an output process of a verification result. A reliability display screen 250b shown in Fig. 16 is a display screen based on the posted information 240 shown in Fig. 14.
[0112] The verification result output unit 125 displays reliability display screens 250a and 250b as shown in Fig. 15 and Fig. 16 on the user's terminal device as a screen showing the verification result. It is preferable that these reliability display screens 250a and 250b are displayed as screens corresponding to the posted information when the posted information to be verified is displayed on the screen of the SNS site, for example. This allows a user who attempts to view the posted information to easily recognize the verification result of the reliability of the posted information.
[0113] The reliability display screens 250a and 250b include a first reliability display section 251, a second reliability display section 252, a third reliability display section 253, a total reliability display section 254, and a reliability label display section 255. These reliability display screens 250a and 250b display not only the total reliability but also the calculated values of the first to third reliability, so that the basis for the reliability calculation can also be presented to the user.
[0114] The reliability labels displayed on the display unit 255 divide the overall reliability into three ranges, for example, and display "high" for 70 points or more and "medium" for 50 points or more and less than 70 points, and "low" for less than 50 points. By displaying such reliability labels, the reliability can be presented to the user in an easy-to-understand manner.
[0115] Furthermore, the overall reliability is "low" in the reliability display screen 250b of Fig. 16. When the overall reliability is "low" in this way, a display section 256 showing the reason why the overall reliability is low may be displayed.
[0116] In the second embodiment described above, the first reliability is calculated taking into consideration the range of influence of an event related to the posted information, so that the reliability of the posted information can be calculated with high accuracy even when the impact of the event is widespread, such as a disaster. In addition, the final reliability is calculated taking into consideration not only the first reliability but also the second reliability indicating the party involved in the posted information and the third reliability indicating the lack of contradiction between the shooting time and the posting time, so that it is possible to improve the accuracy of the reliability calculation.
[0117] The processing functions of the devices (e.g., information processing device 1, SNS server 50, verification server 100, information source servers 60a, 60b, ...) shown in each of the above embodiments can be realized by a computer. In this case, a program describing the processing contents of the functions that each device should have is provided, and the above processing functions are realized on the computer by executing the program on a computer. The program describing the processing contents can be recorded on a computer-readable recording medium. Examples of computer-readable recording media include optical disks and semiconductor memories. Examples of magnetic storage devices include hard disk drives (HDDs) and magnetic tapes. Examples of optical disks include CDs (Compact Discs), DVDs (Digital Versatile Discs), and Blu-ray Discs (BD, registered trademark).
[0118] When distributing a program, for example, the program is recorded on a portable recording medium such as a DVD or CD and then sold. The program can also be stored in a storage device of a server computer and transferred from the server computer to other computers via a network.
[0119] A computer that executes a program stores, for example, a program recorded on a portable recording medium or a program transferred from a server computer in its own storage device. The computer then reads the program from its own storage device and executes processing according to the program. The computer can also read the program directly from a portable recording medium and execute processing according to the program. The computer can also execute processing according to the received program each time a program is transferred from a server computer connected via a network. [Explanation of symbols]
[0120] 1. Information processing device 2 Processing section 10 Posting Information 11 Posts 12 Image Information 13 Metadata 20. Information on the scope of impact 21 Hazard Map 22 Precipitation measurements 30 Scope of Impact
Claims
1. On the computer, Acquire posted information including first information which is text information, second information which is at least one of image information and audio information, and metadata of the second information; Identifying an event that has occurred based on the first information; identifying an impact area indicating the extent of at least one of the geographic and temporal impacts of the event; calculating a first reliability of the posted information based on the range of influence and the metadata; An information verification program that executes the process.
2. The metadata includes a creation location where the second information was created, The identification of the impact range includes identifying the impact range indicating a geographical impact range of the event; In the calculation of the first reliability, the first reliability is calculated based on a positional relationship between the influence range and the creation position.
2. The information verification program according to claim 1.
3. The metadata includes a creation location where the second information was created, In the identification of the impact range, an occurrence range of the event is identified as the impact range; In the calculation of the first reliability, the first reliability is calculated based on a positional relationship between the occurrence range and the creation position.
2. The information verification program according to claim 1.
4. The metadata includes a creation time when the second information was created, In the identification of the affected range, if the tense of the first information is past tense, the occurrence range at the creation time is identified, and if the tense of the first information is present tense, the occurrence range at the posting time when the posted information is posted is identified.
4. The information verification program according to claim 3.
5. Calculating a second reliability of the posted information based on a distance between the creation location and a posting location where the posted information is posted; calculating an overall reliability of the posted information based on the first reliability and the second reliability; 4. The information verification program according to claim 3, further causing the computer to execute a process.
6. The metadata includes a creation time when the second information was created, calculating a third reliability of the posted information based on a difference between the creation time and a posting time of the posted information; The overall reliability is calculated based on the first reliability, the second reliability, and the third reliability.
6. An information verification program according to claim 5.
7. Acquire posted information including first information which is text information, second information which is at least one of image information and audio information, and metadata of the second information; Identifying an event that has occurred based on the first information; identifying an impact area indicating the extent of at least one of the geographic and temporal impacts of the event; A processing unit that calculates a first reliability of the posted information based on the influence range and the metadata; An information processing device having the above configuration.
8. a first information processing device that accepts posting of posted information including first information that is text information and second information that is at least one of image information and audio information, and displays the first information and the second information on a posting screen; a second information processing device that acquires the posted information from the first information processing device, identifies an event that has occurred based on the first information, identifies an impact range indicating at least one of a geographical impact and a temporal impact of the event, and calculates a first reliability of the posted information based on the impact range and metadata of the second information; An information processing system having the above configuration.
Citation Information
Patent Citations
Data evaluation system and posts evaluation method
JP2022122297A
Data analytics platform using social network and web data to identify a pattern or anomaly among relevant events and entities
US20140344172A1
