Communication data log processing apparatus, method, and storage medium having a program stored therein

Abstract

This disclosure relates to a communication data log processing apparatus, method, and storage medium storing a program. The communication data log processing apparatus of one embodiment includes a communication data receiving unit and a partition determination unit. The communication data receiving unit receives communication data contained in a communication data log, the communication data including spoken statements and metadata, and the communication data log is a log of communication data. The partition determination unit determines the partition to which the communication data received by the communication data receiving unit belongs based on the spoken statements and metadata.

Description

Technical Field

[0001] The implementation methods relate to communication data log processing apparatus, methods, and storage media storing programs. Background Technology

[0002] The Internet of Things (IoT) is developing in various settings, including on-site operations like hotel cleaning and security, and meetings where people interact. For example, IoT enables the accumulation of conversation history—such as dialogues in on-site operations and meeting minutes—as communication data logs. These logs record the history of conversations between one or more speakers on one or more topics, regardless of input methods such as voice or text. For instance, communication data logs are sometimes accumulated when two speakers discuss the same topic, and sometimes when four speakers form pairs and each pair speaks on their respective topics.

[0003] Structured communication logs are easier to analyze. Furthermore, if information that can be extracted from these logs to improve business operations and increase efficiency, then they have been effectively utilized. Here, structuring the communication logs means, for example, dividing the communication data into sections, each section representing a series of summaries on the same topic. For instance, if "Business A is complete" is followed by "Understood," these two statements fall under the same topic of "Completion Report of Business A," and can therefore be grouped into the same section. Similarly, if "Where should I take spare part B?" is followed by "Please put spare part B in room C," these two statements fall under the same topic of "Moving Spare Part B," and can also be grouped into the same section.

[0004] In typical speeches, topics may change frequently or multiple topics may arise in parallel. Furthermore, even within consecutive speeches by the same speaker, topics may differ. Moreover, spoken and written language may be mixed together. These characteristics make it difficult to structure communication data logs. Summary of the Invention

[0005] This embodiment provides a structured communication data log processing apparatus, method, and storage medium storing programs that enable easy processing of communication data logs.

[0006] The communication data log processing apparatus of this embodiment includes a communication data receiving unit and a partition determination unit. The communication data receiving unit receives communication data contained in a communication data log, the communication data including spoken statements and metadata, and the communication data log is a log of the communication data. The partition determination unit determines the partition to which the communication data received by the communication data receiving unit belongs based on the spoken statements and metadata. Attached Figure Description

[0007] Figure 1 This diagram illustrates the structure of an example of a system including various embodiments of a communication data log processing apparatus.

[0008] Figure 2 A diagram illustrating an example of a communication data log.

[0009] Figure 3 This is a block diagram illustrating an example of the structure of the communication data log processing apparatus of the first embodiment.

[0010] Figure 4 A diagram showing an example of a partition list.

[0011] Figure 5 This diagram illustrates an example of the hardware structure of a communication data log processing device.

[0012] Figure 6 A flowchart illustrating the operation of the communication data log processing apparatus of the first embodiment.

[0013] Figure 7 The flowchart illustrates the partitioning process.

[0014] Figure 8 A block diagram illustrating an example of the structure of the communication data log processing apparatus of the second embodiment.

[0015] Figure 9 A flowchart illustrating the operation of the communication data log processing apparatus of the second embodiment.

[0016] Figure 10 A block diagram illustrating the structure of an example of a communication data log processing apparatus according to a variation of the second embodiment.

[0017] Figure 11 A flowchart illustrating the operation of the communication data log processing apparatus of a modified example of the second embodiment.

[0018] Figure Labels

[0019] 11-1m: Input device; 2: Communication data log storage device; 3: Communication data log processing device; 31: Communication data receiving unit; 32: Partition discrimination unit; 33: Partition storage unit; 34: Partition analysis unit; 35: Analysis result storage unit; 36: Output control unit; 37: Communication data log receiving unit; 101: Processor; 102: Memory; 103: Input device; 104: Output device; 105: Communication device; 106: Storage; 107: Bus; 1061: Communication data log processing program; 1062: Correlation evaluation model; 1063: Communication data; 1064: Partition list. Detailed Implementation

[0020] The embodiments are described below with reference to the accompanying drawings.

[0021] (First Embodiment)

[0022] Figure 1 This diagram illustrates the structure of an example of a system including communication data log processing apparatuses according to various embodiments. The system of this embodiment includes input devices 11-1m, a communication data log storage device 2, and a communication data log processing device 3.

[0023] Input devices 11-1m are m (where m is a natural number) input devices used for inputting spoken statements. For example, input devices 11-1m are input devices used for inputting spoken statements from corresponding users among users U1-Um. The input method of input devices 11-1m is not limited as long as the device is capable of inputting spoken statements. For example, input devices 11-1m can be devices that collect voice of spoken statements input by users via a microphone. Alternatively, for example, input devices 11-1m can be devices that collect text of spoken statements input by users via a keyboard.

[0024] Input device 11-1m is equipped with a communication device. Therefore, when a speech statement is input from the corresponding user, input device 11-1m sends the speech statement data to communication data log storage device 2. The speech statement data includes the voice or text data of the input speech statement, the date and time when the voice or text input began, the date and time when the input ended, and the device ID. The device ID is the ID assigned to each input device in input device 11-1m.

[0025] The input device 11-1m can be configured to not only send the corresponding user's speech data to the communication data log storage device 2, but also send the corresponding user's speech data to other input devices, or reproduce speech data received from other input devices.

[0026] As a voice input device, the 11-1m can be, for example, a field input device fitted to the head of the user U1-Um. Alternatively, as a text input device, the 11-1m can be, for example, a personal computer or tablet terminal equipped with a text chat application. The input device 11-1m is not limited to these; furthermore... Figure 1 In this context, there is a one-to-one correspondence between input devices and users. However, in other words, the correspondence between input devices and users may not be one-to-one. In cases where the input devices are not one-to-one with users, it is preferable to use methods such as speech recognition to identify the speaking user.

[0027] The communication data log storage device 2 stores communication data logs generated based on speech data sent from input devices 11-1m. The communication data logs are obtained by recording the speech history of each user U1-Um. Figure 2 A diagram illustrating an example of a communication data log. (e.g.) Figure 2 As shown, the communication data log includes n (n is a natural number) communication data points 21-2n.

[0028] The communication data 21-2n include data ID, spoken statements, and metadata, respectively.

[0029] The data ID is the ID assigned to each communication data in communication data 21-2n.

[0030] The spoken statement is text data that shows the content of the spoken statement. When the spoken statement data includes voice data, the communication data log storage device 2 extracts the spoken statement through voice recognition of the voice data. Alternatively, when the spoken statement data includes text data, the communication data log storage device 2 extracts the spoken statement from the text data.

[0031] Meta-information refers to the information accompanying the communication data, including various information associated with the speech. Meta-information may include, for example, date, speech start time, speech end time, speaker ID, speaker role, and input method. The date is the date the corresponding speech statement was uttered. The speech start time is the time when the corresponding speech statement begins. The speech end time is the time when the corresponding speech statement ends. The speaker ID is an ID assigned to each user to identify the user who uttered the corresponding speech statement. The speaker role is information indicating the user's position or status relative to other users. The input method is the input method of the input device used to collect the corresponding speech statement. The date, speech start time, and speech end time can be determined based on, for example, the date and time of speech statement data collection accompanying the speech statement data. The speaker ID, speaker role, and input method can be determined by checking the device ID accompanying the speech statement data, the speaker ID pre-registered and associated with the device ID, the speaker role, and the input method. Here, the meta-information only needs to include... Figure 2 At least one of the six types shown is sufficient. Alternatively, the metadata may also include... Figure 2 Information other than the six types shown.

[0032] The communication data log processing device 3 performs processing to structure the communication data logs stored in the communication data log storage device 2 for analysis. One implementation method of structuring the communication data logs is, for example, dividing the communication data logs into partitions that summarize a series of topics related to the same subject.

[0033] Figure 3 This is a block diagram illustrating an example of the structure of the communication data log processing apparatus 3 according to the first embodiment. The communication data log processing apparatus 3 includes a communication data receiving unit 31, a partition discrimination unit 32, and a partition storage unit 33. The communication data log processing apparatus 3 can be integrally configured with the communication data log storage apparatus 2.

[0034] The communication data receiving unit 31 receives one piece of communication data from the communication data log storage device 2. For example, whenever new communication data is registered to the communication data log storage device 2, the communication data receiving unit 31 receives the new communication data.

[0035] The partition determination unit 32 determines the partition to which the communication data received by the communication data receiving unit 31 belongs. When a partition is determined, the partition determination unit 32 stores the communication data in the corresponding partition. Conversely, when a partition cannot be determined, the partition determination unit 32 creates a new partition and stores the communication data in that new partition. The partition determination unit 32 makes the determination based on factors such as the speech interval determined from the metadata of the communication data and the relevance of the content of the speech sentences determined from the speech sentences in the communication data. Details regarding the partition determination unit 32 will be explained later.

[0036] The partition storage unit 33 aggregates communication data into partitions for storage. Additionally, the partition storage unit 33 stores a list of partitions. This list is management data representing the contents of each partition. Figure 4 A diagram illustrating an example of a partition list. (See diagram below.) Figure 4 As shown, the partition list includes N (N is a natural number) partitions 331-33N. Partitions 331-33N each include a partition ID and a communication data list. Furthermore, the partition storage unit 33 can be implemented using a separate storage medium external to the communication data log processing device 3. In this case, the communication data log processing device 3 accesses the external storage medium as needed to retrieve information from the partition storage unit 33 and forwards the information to the partition storage unit 33.

[0037] The partition ID is the ID assigned to each partition from 331 to 33N.

[0038] The communication data list is a list of data IDs for communication data belonging to the corresponding partition.

[0039] Figure 5 This figure illustrates an example of the hardware structure of the communication data log processing device 3. The communication data log processing device 3 is a computer, having, for example, a processor 101, a memory 102, an input device 103, an output device 104, a communication device 105, and a storage unit 106 as hardware. The processor 101, memory 102, input device 103, output device 104, communication device 105, and storage unit 106 are connected to a bus 107.

[0040] Processor 101 is a processor that controls the overall operation of the communication data log processing device 3. Processor 101 operates as the communication data receiving unit 31 and the partitioning unit 32, for example, by executing programs stored in memory 106. Processor 101 is, for example, a CPU. Processor 101 can be an MPU, GPU, ASIC, FPGA, etc. Processor 101 can be a single CPU or multiple CPUs.

[0041] The memory 102 includes ROM and RAM. ROM is non-volatile memory. ROM stores the startup program of the communication data log processing device 3, etc. RAM is volatile memory. RAM is used as working memory, for example, when performing processing in the processor 101.

[0042] The input device 103 is a touch panel, keyboard, mouse, or other similar input device. When the input device 103 is operated, a signal corresponding to the operation is input to the processor 101 via the bus 107. The processor 101 performs various processes based on the signal.

[0043] The output device 104 is a display device such as a liquid crystal display, an organic EL display, and a printer that outputs various types of information.

[0044] The communication device 105 is a communication device used by the communication data log processing device 3 to communicate with external instruments, such as the communication data log storage device 2. The communication device 105 can be a communication device for wired communication or a communication device for wireless communication.

[0045] Storage 106 is a storage device such as a hard disk drive or a solid-state drive. Storage 106 stores various programs executed by processor 101, such as communication data log processing program 1061.

[0046] Additionally, storage 106 stores a relevance evaluation model 1062. The relevance evaluation model 1062 is a machine learning model used to evaluate the relevance between spoken statements. The relevance evaluation model 1062 can, for example, be a deep learning model obtained by learning the weights of each layer in a way that outputs a score representing the relevance between two spoken statements by comparing feature vectors extracted from the two input spoken statements. Feature vectors are extracted, for example, through morphological analysis. Information related to the relevance between words assumed to be spoken statements is provided to the relevance evaluation model 106 as teacher data.

[0047] The relevance evaluation model 1062 can be a deep learning model that takes a string formed by concatenating spoken statements with all or part of the meta-information as input, and outputs a score representing the relevance. Alternatively, the relevance evaluation model 1062 can also be a deep learning model that takes three or more spoken statements as input and outputs a score representing the relevance. By concatenating meta-information with spoken statements as strings, it is possible to process meta-information of different forms. Furthermore, by taking three or more spoken statements as input, a wider range of information can be processed. Moreover, the relevance evaluation model 1062 does not necessarily have to be a deep learning model. For example, the relevance evaluation model 1062 can be a model configured to evaluate the relevance between spoken statements based on pre-determined rules.

[0048] In addition, storage 106 also functions as a partitioned storage unit 33, summarizing and storing communication data 1063 for each partition. Furthermore, storage 106 stores a partition list 1064.

[0049] Here, the correlation evaluation model 1062, communication data 1063, and partition list 1064 can also be stored in a different device than the communication data log processing device 3. In this case, the communication data log processing device 3 obtains the required information by accessing the other device through the communication device 105.

[0050] Bus 107 is a data forwarding path for exchanging data between processor 101, memory 102, input device 103, output device 104, communication device 105 and storage 106.

[0051] Next, the operation of the communication data log processing device of the first embodiment will be described. Figure 6 A flowchart illustrating the operation of the communication data log processing apparatus of the first embodiment. For example, the process is performed whenever the communication data receiving unit 31 receives communication data. Figure 6 The processing.

[0052] In step S101, the partition determination unit 32 performs partition determination processing to determine which partition ID of the partition storage unit 33 the communication data received by the communication data receiving unit 31 should belong to. Details regarding the partition determination processing will be explained later. For now, we will assume that the partition to which the communication data should belong has already been determined.

[0053] In step S102, the partition discrimination unit 32 stores the communication data received by the communication data receiving unit 31 at, for example, the end of the partition of the partition ID determined by the partition discrimination process in the partition storage unit 33. Additionally, the partition discrimination unit 32 stores the data ID of the newly stored communication data in the communication data list corresponding to the partition ID in the partition list. Afterwards, Figure 6 The processing ends here. Here, for communication data that is determined to be unclassified in the partitioning process described later, the partitioning unit 32 does not store it in a partition. In this case, Figure 6 The processing ends there. Furthermore, for communication data that is determined to be classified into a new partition in the partition determination process described later, the partition determination unit 32 stores the communication data in that partition after creating a new partition ID. Afterwards, Figure 6 The processing is now complete.

[0054] Figure 7The flowchart of the partitioning and discrimination process is shown below. In step S201, the partitioning and discrimination unit 32 determines whether the spoken statement in the communication data received by the communication data receiving unit 31 is an unwanted spoken statement. Unwanted spoken statements include, for example, meaningless spoken statements that are unintentionally entered by the user at a set time, and spoken statements containing words that are explicitly indicated as unwanted spoken statements, such as "test". The partitioning and discrimination unit 32 stores a list of words that are explicitly indicated as unwanted spoken statements in advance. The partitioning and discrimination unit 32 can determine whether a spoken statement is meaningless or contains words that are explicitly indicated as unwanted spoken statements by performing morphological analysis or other methods to decompose the spoken statement into word units. Alternatively, the partitioning and discrimination unit 32 can also use a deep learning model to determine whether a spoken statement is unwanted. In step S201, if the spoken statement is determined to be an unwanted spoken statement, the process proceeds to step S202. In step S201, if the spoken statement is determined not to be an unwanted spoken statement, the process proceeds to step S203.

[0055] In step S202, the partitioning unit 32 determines that the communication data received by the communication data receiving unit 31 will not be classified into a partition. Afterwards, Figure 7 The processing is complete. In other words, in this implementation, unnecessary speech statements are not categorized into partitions.

[0056] In step S203, the partition determination unit 32 determines whether the speech interval from the end time of the last speech data added to the partition storage unit 33 to the start time of the speech data received by the communication data receiving unit 31 is longer than or equal to a threshold. This threshold can be determined as an appropriate value, such as 5 minutes. In step S203, if the speech interval of the communication data is determined to be long, the process proceeds to step S204. Additionally, if no partition has been created in the partition storage 33, i.e., no communication data is stored, the process also proceeds to step S204. In step S203, if the speech interval of the communication data is determined to be short, the process proceeds to step S205.

[0057] In step S204, the partitioning unit 32 determines that the communication data received by the communication data receiving unit 31 should be classified into a newly created partition. Afterwards, Figure 7 The processing is complete. That is, in the implementation, a speech statement that is spoken after a sufficiently long speech interval relative to the immediately preceding speech statement is considered to have low relevance to other speech statements and is classified into a new partition.

[0058] In step S205, the partitioning determination unit 32 determines whether the communication data received by the communication data receiving unit 31 is communication data containing a specific word that should be classified into a new partition. The specific word is a word used to achieve topic switching, such as "so" or "by the way." The specific word can be a word predetermined between users. In step S205, if it is determined that the communication data contains a specific word, the process proceeds to step S204. In this case, the communication data received by the communication data receiving unit 31 is classified into a new partition. In step S205, if it is determined that the communication data does not contain a specific word, the process proceeds to step S206.

[0059] In step S206, the partition discrimination unit 32 selects one or more communication data from the partition storage unit 33. For example, it selects the communication data that was last added to the partition storage unit 33. When the correlation evaluation model 1062 is configured to output a score based on three or more speech statements, the partition discrimination unit 32 can select two or more communication data from the partition storage unit 33.

[0060] In step S207, the partition discrimination unit 32 evaluates the correlation between the spoken sentences of the communication data received by the communication data receiving unit 31 and the spoken sentences of the communication data selected from the partition storage unit 33. The correlation can be evaluated based on the score obtained by inputting the spoken sentences of two or more communication data into the correlation evaluation model. Alternatively, as described above, a string composed of spoken sentences and metadata can be input into the correlation evaluation model.

[0061] In step S208, the partition discrimination unit 32 determines whether the relevance between the spoken phrases of the communication data received by the communication data receiving unit 31 and the spoken phrases of the communication data selected from the partition storage unit 33 is high. For example, a high relevance is determined when the score is above a threshold. In step S208, if the relevance is determined to be low, the process proceeds to step S209. In step S208, if the relevance is determined to be high, the process proceeds to step S210.

[0062] In step S209, the partitioning determination unit 32 determines whether to terminate the correlation evaluation. For example, when all communication data of the correlation evaluation object is selected from the partition storage unit 33, it is determined that the correlation evaluation should be terminated. The communication data of the correlation evaluation object is, for example, the communication data last added to the partition storage unit 33. Alternatively, the communication data of the correlation evaluation object can be, for example, communication data stored in the partition storage unit 33 that has a predetermined time elapsed since the start time of the communication data received from the communication data receiving unit 31, for example, two hours, before the end time of the speech. Furthermore, the communication data of the correlation evaluation object can also be all the communication data stored in the partition storage unit 33. In step S209, if it is determined that the correlation evaluation should not be terminated, the process returns to step S206. In this case, the partitioning determination unit 32 selects other communication data from the partition storage unit 33. In step S209, if it is determined that the correlation evaluation should be terminated, the process proceeds to step S204. In this case, the communication data received by the communication data receiving unit 31 is classified into a newly created partition.

[0063] In step S210, the partitioning unit classifies the communication data received by the communication data receiving unit 31 into the partitions belonging to the communication data that is evaluated as having high relevance. Afterwards, Figure 7 The processing is now complete.

[0064] The following is a detailed explanation. Figure 7 The processing. For example, suppose the communication data receiving unit 31 receives data from... Figure 2 The communication data log receives communication data 21. At this time, the partition determination unit 32 refers to the partition list stored in the partition storage unit 33. For example, when the partition list is empty, communication data 21 belongs to a newly created partition. In this case, the partition determination unit 32 creates a new partition with partition ID 1 in the partition storage unit 33, and stores communication data 21 at the beginning of the created partition. Furthermore, the partition determination unit 32 appends the data ID 1 of communication data 21 to the communication data list of partition 331 with partition ID 1 in the partition list.

[0065] Next, let's assume that the communication data receiving unit 31 receives data from... Figure 2 The communication data log receives communication data 22. Since the speech interval from the end of the speech in communication data 21 to the start of the speech in communication data 22 is shorter than a threshold, the partitioning discrimination unit 32 evaluates the correlation between the speech statements in communication data 22 and communication data 21. Here, regarding... Figure 2The communication data 21 contains the statement "Mr. A, the homework for August 11..." from [User 4], and the communication data 22 contains the statement "Understood" from [User 2]. Since [User 2] (i.e., [Mr. A]) is responding to a communication from [User 4] to [User 2], the statement in communication data 22 is determined to have a high degree of correlation with the statement in communication data 21. Furthermore, when the statement contains a string of metadata, [User 2] is [User 4]'s superior based on the speaker's role, and the input method is the same; therefore, the statement in communication data 21 is also determined to have a high degree of correlation with the statement in communication data 22. Therefore, the partitioning unit 32 appends communication data 22 to the end of partition ID 1. Furthermore, the partitioning unit 32 appends data ID 22 of communication data 22 to the communication data list of partition ID 1, partition 331, in the partition list.

[0066] Furthermore, assuming that the communication data receiving unit 31 receives data from... Figure 2 The communication data log receives communication data 23. Since the speech interval from the end of the speech in communication data 22 to the start of the speech in communication data 23 is shorter than a threshold, the partitioning discrimination unit 32 evaluates the correlation between the speech statements in communication data 23 and communication data 22. Here, because... Figure 2 The "Understood." statement from [User 2] in communication data 22 and the "Mr. A, I am B. Regarding 1220..." statement from [User 5] in communication data 23 are statements about different topics. Therefore, it is determined that the statement in communication data 23 has a low correlation with the statement in communication data 22. Thus, communication data 23 belongs to a newly created partition. In this case, the partition determination unit 32 creates a new partition with partition ID 2 in the partition storage unit 33, and stores communication data 23 at the beginning of the created partition. Furthermore, the partition determination unit 32 appends the data ID 3 of communication data 23 to the communication data list of partition 332 with partition ID 2 in the partition list.

[0067] After this, the partition discrimination unit 32 processes the communication data received by the communication data receiving unit 31 in the same manner and stores it in the partition storage unit 33.

[0068] As explained above, according to the first embodiment, not only the content of the spoken statements accumulated in the communication data log is used, but also the metadata attached to the communication data is used, thereby easily implementing the structuring of the communication data log. For example, for communication data with long speaking intervals, it is first classified into a newly created partition before evaluating the relevance of the spoken statements, thereby reducing the amount of communication data that needs to be evaluated for relevance. Accordingly, the processing of the partitioning unit 32 can be accelerated. In addition, by considering metadata in addition to spoken statements, communication data can be aggregated into appropriate partitions even when the topic changes at any time or when multiple topics are discussed in parallel.

[0069] Furthermore, communication data including unnecessary speech phrases is not categorized into partitions, and communication data including speech phrases containing specific words that should be categorized into newly created partitions are categorized into the newly created partitions before the relevance of speech phrases is evaluated. This can also reduce the amount of communication data that needs to be evaluated for the relevance of speech phrases. As a result, the processing of the partitioning unit 32 can also be accelerated.

[0070] In the first embodiment, the partition determination unit 32 may also cause the display device, which is the output device 104, to display the partition list, or print the partition list using a printer. Furthermore, when the partition list is displayed on the display device, if any data ID of the communication data list is selected, the partition determination unit 32 may further cause the content of the communication data for that data ID to be displayed.

[0071] (Second Implementation)

[0072] The second embodiment will now be described. Here, the same structure and operation as the first embodiment in the second embodiment will be omitted or simplified in the description. Figure 8 This is a block diagram illustrating an example of the structure of the communication data log processing apparatus 3 according to the second embodiment. In addition to the communication data receiving unit 31, the partition discrimination unit 32, and the partition storage unit 33, the communication data log processing apparatus 3 also includes a partition analysis unit 34, an analysis result storage unit 35, and an output control unit 36. The communication data receiving unit 31, the partition discrimination unit 32, and the partition storage unit 33 are the same as those described in the first embodiment, and therefore their description is omitted. The processor 101, for example, executes a program stored in the memory 106, and functions not only as the communication data receiving unit 31 and the partition discrimination unit 32, but also as the partition analysis unit 34 and the output control unit 36. Furthermore, the memory 106 functions not only as the partition storage unit 33, but also as the analysis result storage unit 35.

[0073] The partition analysis unit 34 analyzes communication data for each partition by referring to the partition list stored in the partition storage unit 33. For example, the partition analysis unit 34 analyzes the communication data of each partition and assigns attributes to the partitions, such as attributes representing the dialogue content within the partition. Furthermore, the analysis used for attribute assignment can be performed using a deep learning model obtained by learning the relationship between communication data and attributes. Here, the attributes assigned by the partition analysis unit 34 are predetermined attributes such as "instruction," "completion report," and "fault report." In addition, the partition analysis unit 34 can also analyze the busyness, fault frequency, etc., based on the time period by summarizing the spoken statements in the communication data for each time period. Furthermore, the partition analysis unit 34 can extract tasks by analyzing the communication data of each partition, visualize the tasks by associating the extracted tasks with speaker IDs, and perform other analyses such as linking related partitions and summarizing metadata for each related partition.

[0074] The analysis results storage unit 35 stores the analysis results of the partitions input from the partition analysis unit 34.

[0075] The output control unit 36 ​​displays the analysis results stored in the analysis result storage unit 35 on the screen of the display device 104, which is also an output device, or prints the analysis results on paper using a printer. The analysis results can be displayed on a browser. In addition, the output control unit 36 ​​can display or print information that greatly contributes to the determination of the analysis results along with the analysis results.

[0076] Next, the operation of the communication data log processing device of the second embodiment will be described. Figure 9 A flowchart illustrating the operation of the communication data log processing apparatus according to the second embodiment. For example, whenever the communication data receiving unit 31 receives communication data, the process is performed... Figure 9 The handling of this. Here. Figure 9 Processing steps S301-S302 and Figure 6 The processes in steps S101-S102 are performed in the same way. Therefore, the explanation is omitted.

[0077] In step S303, after the communication data is stored in the partitions, the partition analysis unit 34 analyzes the communication data by referring to the partition list stored in the partition storage unit 33. For example, it performs attribute assignment analysis for the partitions by concatenating all the metadata and spoken statements contained in the communication data within each partition and inputting them into a deep learning model. Alternatively, it can perform attribute assignment analysis for the partitions by inputting the communication data one by one into the deep learning model and using a majority vote of the attributes assigned to each piece of communication data.

[0078] Alternatively, the attributes assigned to partitions can be categorized, and the frequency of occurrence of partitions with attributes included in each category can be aggregated for each time period. Based on this, analysis results for each category in each time period can be obtained. For example, if the assigned attributes are "Instruction," "Completion Report," and "Fault Report," and the regular business category includes "Instruction" and "Completion Report," while the non-regular business category includes "Fault Report," the analysis results obtained from the regular business category indicate the task's workload, while the analysis results obtained from the non-regular business category indicate whether a fault has occurred. Categories can be set manually in advance or using statistical methods.

[0079] In step S304, the partition analysis unit 34 stores the analysis results in the analysis result storage unit 35.

[0080] In step S305, the output control unit 36 ​​displays the analysis results stored in the analysis result storage unit 35 on the display device, which is the output device 104. When displaying the analysis results, the output control unit 36 ​​may also display information that significantly contributes to the determination of the analysis results. Information that significantly contributes to the determination of the analysis results includes, for example, words or meta-information that significantly contribute to the determination of the partition's attributes. Furthermore, in addition to displaying information that significantly contributes to the determination, the output control unit 36 ​​may also display specific words or meta-information. Moreover, the output control unit 36 ​​may not only display information simply, but also emphasize information. By using emphasized display, the analysis results are easier for users to intuitively interpret and are also easier to understand visually.

[0081] As explained above, according to the second embodiment, in addition to structuring the communication data logs, the structured communication data logs are also analyzed. This allows for more direct provision of information useful for business improvement and efficiency enhancement to users. Furthermore, by sequentially classifying the communication data logs into sections as a structured format each time communication data is received, real-time analysis of the communication data logs is also possible. Real-time analysis of the communication data logs also enables the detection of temporary load spikes and load distribution.

[0082] (A variation of the second embodiment)

[0083] Next, a variation of the second embodiment will be described. Here, the same structures and operations as in the variation of the second embodiment will be omitted or simplified in the description. Figure 10This is a block diagram illustrating an example of the structure of the communication data log processing apparatus 3, a modified example of the second embodiment. In addition to the communication data receiving unit 31, partition discrimination unit 32, partition storage unit 33, partition analysis unit 34, analysis result storage unit 35, and output control unit 36, the communication data log processing apparatus 3 also includes a communication data log receiving unit 37.

[0084] The communication data log receiving unit 37 receives communication data logs containing two or more communication data from the communication data log storage device 2. Then, the communication data log receiving unit 37 inputs the communication data contained in the received communication data logs one by one into the communication data receiving unit 31.

[0085] Next, the operation of the communication data log processing apparatus of the modified example of the second embodiment will be described. Figure 11 This is a flowchart illustrating the operation of a communication data log processing apparatus according to a variation of the second embodiment. For example, the process begins whenever the communication data log receiving unit 37 receives a communication data log. Figure 11 The processing.

[0086] In step S401, the communication data log receiving unit 37 selects one piece of communication data from the communication data log. The communication data log receiving unit 37 selects the communication data according to, for example, the order of data IDs. Then, the communication data log receiving unit 37 inputs the selected communication data into the communication data receiving unit 31.

[0087] The processing steps S402-403 after the communication data receiving unit 31 receives the communication data are as follows: Figure 6 The processes in steps S101-S102 are performed in the same way. Therefore, the explanation is omitted.

[0088] In step S404, the partition determination unit 32 determines whether the processing of storing all communication data contained in the communication data log received by the communication data log receiving unit 37 into the partition has been completed. In step S404, if it is determined that the processing of all communication data is not complete, the process returns to step S401. In this case, the communication data log receiving unit 37 selects other communication data. In step S404, if it is determined that the processing of all communication data is complete, the process proceeds to step S405.

[0089] Figure 11 Steps S405-407 processing and Figure 9 The processes in steps S303-S305 are performed in the same way. Therefore, the explanation is omitted.

[0090] As explained above, in a variation of the second embodiment, the communication data log processing device 3 can accept communication data logs as input and can aggregate and structure multiple communication data sets. Accordingly, it can also be aggregated for analysis. For example, by aggregating and analyzing data over a day, the workload and task distribution of each user can be analyzed. Such analysis results can be effectively utilized for staffing and task assignment for the following day.

[0091] The instructions shown in the processing sequence described in the above embodiments can be executed based on a program, which is software. A general-purpose computer system pre-stores this program and reads it in, thereby achieving the same effect as the data log processing apparatus described above. The instructions described in the above embodiments are recorded as programs that can be executed by a computer on a disk (floppy disk, hard disk, etc.), optical disk (CD-ROM, CD-R, CD-RW, DVD-ROM, DVD±R, DVD±RW, Blu-ray Disc, etc.), semiconductor memory, or similar recording media. The storage format can be any form, as long as the recording medium is readable by a computer or embedded system. If the computer reads the program from the recording medium and uses the CPU to execute the instructions described in the program, the same operation as the data log processing apparatus of the above embodiments can be achieved. Of course, when the computer obtains or reads the program, it can obtain or read it via a network.

[0092] Alternatively, a portion of the processing for implementing this embodiment can be executed by instructions from a program installed on a computer or embedded in the system, such as an OS (operating system), database management software, or network (middleware) running on a computer.

[0093] Furthermore, the recording medium in this embodiment is not limited to media independent of the computer or embedded system, but also includes recording media that download and store or temporarily store programs transmitted via LAN, Internet, etc.

[0094] Furthermore, the recording medium is not limited to one; the processing in this embodiment can be performed from multiple media as well. The structure of the media can be arbitrary.

[0095] Furthermore, the computer or embedded system of this embodiment is used to execute the various processes of this embodiment based on the program stored in the recording medium, and can be any structure such as a device consisting of a personal computer, a microcomputer, or a system consisting of multiple devices connected by a network.

[0096] Furthermore, the computer in this embodiment is not limited to personal computers, but also includes arithmetic processing devices and microcomputers contained in information processing instruments, collectively referred to as instruments and devices capable of implementing the functions of this embodiment using programs.

[0097] While several embodiments of the invention have been described, these embodiments are provided by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope or spirit of the invention, and are included within the scope of the invention as set forth in the claims and its equivalents.

Claims

1. A communication data log processing device, comprising: A communication data receiving unit receives communication data contained in a communication data log, wherein the communication data includes spoken statements and metadata, and the communication data log is a log of the communication data; and The partitioning unit determines the partition to which the communication data received by the communication data receiving unit belongs, based on the spoken statements and the metadata. wherein When the correlation between a first speech statement contained in the first communication data received by the communication data receiving unit and a second speech statement contained in the second communication data summarized in the first partition is high, the partitioning discrimination unit determines that the partition to which the first communication data should belong is the first partition. When the correlation between the first speech statement and the second speech statement is low, the partition to which the first communication data should belong is determined as a newly created second partition.

2. A communication data log processing device, comprising: A communication data receiving unit receives communication data contained in a communication data log, wherein the communication data includes spoken statements and metadata, and the communication data log is a log of the communication data; and The partitioning unit determines the partition to which the communication data received by the communication data receiving unit belongs, based on the spoken statements and the metadata. When the correlation between the first string formed by concatenating the first speech statement and the first meta-information in the first communication data received by the communication data receiving unit and the second string formed by concatenating the second speech statement and the second meta-information in the second communication data summarized in the first partition is high, the partitioning unit will determine the partition to which the communication data received by the communication data receiving unit should belong as the first partition. When the correlation between the first string and the second string is low, the partitioning unit will determine the partition to which the communication data received by the communication data receiving unit should belong as the newly created second partition.

3. The communication data log processing apparatus according to claim 1 or 2, wherein, It also includes a partition storage unit, which summarizes and stores the communication data received by the communication data receiving unit for each partition identified by the partition discrimination unit.

4. The communication data log processing apparatus according to claim 1 or 2, wherein, The metadata includes at least one of the following: The date on which the statement was made; The speech begins at the moment the speech begins; The speech ends at the moment the speech concludes; Identify the speaker ID of the speaker who uttered the stated statement; The speaker role refers to the role of the speaker who utters the spoken statements; and The input method for the spoken statements.

5. The communication data log processing apparatus according to claim 1 or 2, wherein, It also includes a partition analysis unit that analyzes the communication data aggregated for each partition.

6. The communication data log processing apparatus according to claim 5, wherein, It also includes an output control unit that outputs the results of the analysis to an output device.

7. The communication data log processing apparatus according to claim 1 or 2, wherein, The partitioning unit sequentially determines the partition to which the communication data received by the communication data receiving unit should belong.

8. The communication data log processing apparatus according to claim 1 or 2, wherein, It also has a communication data log receiving unit, which receives the communication data log.

9. The communication data log processing apparatus according to claim 8, wherein, The communication data log receiving unit selects the communication data one by one from the communication data log and inputs it into the communication data receiving unit.

10. The communication data log processing apparatus according to claim 1 or 2, wherein, If the spoken words contained in the communication data received by the communication data receiving unit contain specific words, the partitioning unit prioritizes the determination based on the spoken words and the metadata, and determines the partition to which the communication data received by the communication data receiving unit should belong as a newly created partition.

11. The communication data log processing apparatus according to claim 1 or 2, wherein, When the communication data received by the communication data receiving unit includes unwanted speech statements, the partitioning unit does not determine the partition to which the communication data received by the communication data receiving unit should belong.

12. The communication data log processing apparatus according to claim 1 or 2, wherein, The metadata includes the speech start time of the speech statement and the speech end time of the speech statement. When the speech interval is above a threshold, the partitioning unit prioritizes the discrimination based on the speech statement and the metadata, and identifies the partition to which the communication data received by the communication data receiving unit should belong as a newly created partition. The speech interval is the time interval from the start time of the speech of the communication data received by the communication data receiving unit to the end time of the speech of the last communication data summarized into the partition.

13. A method for processing communication data logs, comprising: Accept communication data contained in a communication data log, the communication data including spoken statements and metadata, and the communication data log being a log of the communication data; and The received communication data is determined to belong to the partition based on the spoken statements and the metadata. wherein When the correlation between the first speech statement contained in the first communication data and the second speech statement contained in the second communication data summarized in the first partition is high, the partition to which the first communication data should belong is identified as the first partition. When the correlation between the first speech statement and the second speech statement is low, the partition to which the first communication data should belong is identified as the newly created second partition.

14. A method for processing communication data logs, comprising: Accept communication data contained in a communication data log, the communication data including spoken statements and metadata, and the communication data log being a log of the communication data; and The received communication data is determined to belong to the partition based on the spoken statements and the metadata. wherein When the correlation between the first string formed by concatenating the first speech statement and the first meta-information in the first received communication data and the second string formed by concatenating the second speech statement and the second meta-information in the second communication data summarized in the first partition is high, the partition to which the received communication data should belong is determined to be the first partition. When the correlation between the first string and the second string is low, the partition to which the received communication data should belong is determined to be the newly created second partition.

15. A computer-readable storage medium storing a communication data log processing program for causing a computer to perform the following operations: Accept communication data contained in the communication data log, the communication data including spoken statements and metadata, and the communication data log being a log of the communication data; The received communication data is used to determine the partition to which it belongs based on the spoken statements and the metadata. as well as When the correlation between the first speech statement contained in the first communication data and the second speech statement contained in the second communication data summarized in the first partition is high, the partition to which the first communication data should belong is identified as the first partition. When the correlation between the first speech statement and the second speech statement is low, the partition to which the first communication data should belong is identified as the newly created second partition.

16. A computer-readable storage medium storing a communication data log processing program for causing a computer to perform the following operations: Accept communication data contained in the communication data log, the communication data including spoken statements and metadata, and the communication data log being a log of the communication data; The received communication data is used to determine the partition to which it belongs based on the spoken statements and the metadata. as well as When the correlation between the first string formed by concatenating the first speech statement and the first meta-information in the first received communication data and the second string formed by concatenating the second speech statement and the second meta-information in the second communication data summarized in the first partition is high, the partition to which the received communication data should belong is determined to be the first partition. When the correlation between the first string and the second string is low, the partition to which the received communication data should belong is determined to be the newly created second partition.

Images