JP7288701B2 - voice recognition device - Google Patents

Description

The present invention relates to a speech recognition device, a management system, a management program, and a speech recognition method. More specifically, the present invention relates to a device, system, program, and method capable of preventing and grasping troubles arising from conversation by converting voice data into text, generating utterance data, and comparing it with predetermined prohibited word data.

At work sites, especially at sites where people come into direct contact with people, troubles may arise from conversations. There are often troubles in which inappropriate remarks constitute harassment, and troubles in which the presence or absence of remarks regarding the contents of contracts and the details of the contents are disputed later. Among the troubles caused by conversation, there are also serious cases such as threatening acts by words and actions, and suggestions or abetting crimes.

Trouble caused by conversation is an important problem not only in the workplace but also in daily life. In such troubles, it is important to collect and analyze conversations to help prevent troubles and deal with them after the fact. Conventionally, aircraft cockpits are equipped with voice recorders that pick up and record crew conversations. In the unlikely event of an incident or accident, the voice recorded by this voice recorder will be played back and analyzed to help investigate the cause of the incident or accident. Call center operators often adopt a management method in which voice data recorded by a microphone device of the operator's response to a customer is played back and analyzed by the manager to assist in troubleshooting.

On the other hand, a portable voice recorder has been devised and used by combining a small microphone device and a recording device (Patent Document 1). According to the invention described in Patent Literature 1, conversations by employees while they are out and about are recorded by a voice recorder. Conversations can be played back and analyzed.

JP 2007-325392 A

However, in the management method described in Patent Literature 1, the recording time to the voice recorder becomes enormous. And this recording time increases as the number of subjects increases. Also, this recording time increases as the time to be managed increases. Therefore, it takes a lot of labor and time for the administrator to search for the voice data indicating the occurrence of the trouble from the recorded voice data. In addition, when dealing with troubles using recorded conversations, the administrator collects the recorded voice data, searches for the voice data corresponding to the trouble that occurred, and furthermore, plays back the found voice data. It is necessary to take steps to inspect the contents of the audio data using the Therefore, it takes time for the administrator to take action after the occurrence of the trouble. Therefore, the management of recording the subject's conversations with a voice recorder, etc. is insufficient in terms of the administrator's early response to troubles caused by conversations, and in terms of the time and effort required for analysis. There was a point.

In order for the manager to respond early to troubles caused by the conversation of the target person, it is important to find utterances that directly indicate the occurrence of trouble, such as "give me money" and "I'll kill you", from conversations. is. Furthermore, it is also important for the administrator to grasp the frequency of statements suspected of causing trouble. For example, the statement "mother" alone does not clearly indicate the occurrence of trouble. However, if the administrator can grasp the frequency of utterances such as "Mom", the administrator will be able to understand the situation where threats are being made by repeating phrases such as "I will tell your mother" and "Is it okay for your mother to know?" I can grasp it. Therefore, the administrator can take early action.

In management using collected and recorded conversations of subjects, the administrator grasps the frequency of utterances through the work of reproducing most of the recorded voice data and confirming the contents of the voice data. It is very difficult for an administrator to perform such confirmation work one by one for a large number of subjects under his charge, in terms of both labor and time. Therefore, in the management method for collecting and analyzing the subject's conversation, there is still room for further improvement in the conventional method in which the collected conversation is recorded and the content is confirmed by the manager.

The present invention was made based on such an idea, converts voice data received from a microphone device into text, compares it with predetermined prohibited word data, counts and displays the matching or similar frequency. Accordingly, it is an object of the present invention to provide a speech recognition apparatus, a management system, a management program, and a speech recognition method that enable an administrator to manage speech of a target person without checking the speech of the target person.

As a result of intensive studies to solve the above problems, the present inventors converted voice data received from a microphone device into text, compared it with predetermined prohibited word data, aggregated and displayed the frequency of coincidence or similarity. By doing so, the inventors have found that the above object can be achieved, and have completed the present invention. Specifically, the present invention provides the following.

A first feature of the invention is a speech recognition device connected to a microphone device capable of converting a subject's voice into voice data, and capable of converting the voice data received from the microphone device into text, wherein the microphone device A utterance data generation unit that converts the voice data received from the above into text and generates the utterance data of the subject, and a tally that compares the utterance data with predetermined prohibited word data and tallies the frequency of matching or similarity. and a display unit for displaying a result of aggregation by the aggregation unit to an administrator.

According to the first aspect of the invention, the frequency of prohibited words contained in the subject's voice is automatically totaled and displayed on the display unit. For example, by setting phrases that indicate the occurrence of trouble, such as "I'll kill you," as prohibited words, the frequency with which the target person made remarks containing such phrases is aggregated and displayed to the administrator through the display unit. . Through the display on the display unit, the administrator grasps the frequency of statements including prohibited words by the target person. The administrator uses the displayed frequency of utterances containing prohibited words to grasp the occurrence of troubles around the target, and performs management such as issuing warnings to the target and reporting to the police. The administrator can manage this series of utterances of the subject without checking the voice of the subject one by one.

Therefore, according to the invention according to the first characteristic, it is possible to provide a speech recognition apparatus capable of managing the utterances of the subject without the administrator having to check the speech of the subject one by one.

The invention according to a second characteristic is the invention according to the first characteristic, wherein as a result of the counting by the counting unit, the frequency of matching or similarity between the utterance data and the prohibited word data is equal to or greater than a predetermined frequency. Provided is a speech recognition device further comprising a warning unit for warning the administrator when .

According to the second aspect of the invention, the warning section warns the administrator when the frequency of matching or similarity between the utterance data and the prohibited word data is equal to or higher than a predetermined frequency. Since the warning unit issues a warning, the administrator can quickly deal with the problem. In addition, since the warning unit issues a warning when the frequency is equal to or higher than a predetermined frequency, mischievous warnings and false warnings due to aggregating words with similar pronunciations as matching or similar words can be prevented. Therefore, the utterance management device can quickly deliver a necessary warning to the administrator without being mistaken for a false warning. The manager can more appropriately and quickly grasp and manage remarks indicating the occurrence of trouble.

The invention according to a third feature is the invention according to the second feature, wherein the voice data is compared with pre-stored voiceprint data, and the source of the voice data is determined as the source of the voiceprint data. and the warning unit warns the administrator when the source of the voice data is the source of the voiceprint data. do.

According to the third aspect of the invention, the caller identification unit identifies the originator of the voice data, so that the warning unit notifies the administrator only when the voice data matches pre-stored voiceprint data. A warning action is taken. For example, by registering the target person's voiceprint and prohibited words such as "absolutely profitable" indicating unfair solicitation, the administrator warns the administrator that there has been an unfair solicitation by the target person. be done. As a result, prohibited words included in the voices of speakers other than the target person, especially those of third parties who were present around the target person who did not participate in the conversation with the target person, are aggregated and warned to the administrator. avoid false alarms.

The invention according to a fourth characteristic is the invention according to any one of the first to third characteristics, wherein the utterance data generated by the utterance data generation unit is classified according to the source of the voice data and stored. Provided is a speech recognition device, further comprising a utterance data storage unit that stores the speech data.

According to the fourth aspect of the invention, the generated utterance data is classified for each source, accumulated, and saved. This allows the administrator to read out the saved utterance data and scrutinize its content. In cases such as when fraudulent activity is suspected, there are often discrepancies between the parties regarding the presence or absence of remarks and the content of remarks. In such a situation where there is a discrepancy between the parties regarding the presence or absence of remarks and the content of the remarks, the administrator can objectively grasp the situation by examining the evidence of the saved remark data. Also, in a labor tribunal trial in which the subject is a party, the administrator can present the saved remark data as evidence.

Also, according to the fourth aspect of the invention, the utterance data is saved, so that the administrator can analyze and use the saved utterance data as big data. Through such analysis, the manager can analyze the utterance tendencies of the target person, find words that tend to appear before trouble occurs, and add them to the prohibited words.

The invention according to a fifth feature is the invention according to the fourth feature, wherein the display unit extracts utterance data of a specific caller from among the plurality of utterance data stored in the utterance data storage unit. to provide a speech recognition device that can be displayed as a

According to the fifth aspect of the invention, the display section can extract and display utterance data of a specific caller. As a result, the manager can analyze and use the utterance tendencies of senders who have favorable characteristics for the manager, such as good sales performance. Alternatively, the manager can analyze and use the utterance tendencies of senders with characteristics that are undesirable to the manager, such as poor sales performance and many complaints.

The invention according to a sixth feature is the invention according to any one of the first to fifth features, wherein the aggregating unit is predetermined and frequently used in an attribute to which the source of the voice data belongs. Provided is a speech recognition device that compares speech data with the speech data, and performs the aggregation for the speech data different from the specific speech data.

According to the sixth aspect of the invention, the manager defines loud sounds that frequently occur in the attribute to which the subject belongs as specific audio data. As a result, the loud sound is not processed by the tallying unit. For example, if the target person is a plumbing worker and the tool sounds frequently in the vicinity, the manager registers the tool sound as specific voice data. As a result, frequently occurring tool sounds are not processed in the totalizer. Therefore, the load on the counting unit is reduced.

The invention according to a seventh feature comprises a microphone device capable of converting a subject's voice into voice data, and the voice recognition device described above, wherein the microphone device measures the volume of sound contained in the voice data. A management system is provided, comprising: a volume measuring unit; and a transmitting unit that transmits voice data whose volume is equal to or greater than a predetermined volume to the voice recognition device.

According to the seventh aspect of the invention, the sound volume measurement unit measures the sound volume of the sound contained in the sound data, so that the sound data whose sound volume is equal to or greater than a predetermined amount is transmitted to the speech recognition device. As a result, it is possible to avoid processing speech data, which has a low volume and therefore does not contain speeches to be converted into text, compared and tabulated, by the speech data generation unit and the tabulation unit. Therefore, the amount of processing in the utterance data generation unit and totalization unit is reduced, and the load on the management system is reduced.

An invention according to an eighth feature comprises a microphone device capable of converting a subject's voice into voice data, and the speech recognition device described above, wherein the microphone device includes a positioning unit for positioning the current position of the microphone device. wherein the prohibited word data is predetermined according to the location of the microphone device, and the tallying unit combines the utterance data and the prohibited word data based on the current location received from the microphone device. and tally the frequency of matches or similarities.

According to the invention according to the eighth characteristic, since the aggregating unit refers to the prohibited word data according to the location of the microphone device, the frequency of inappropriate remarks according to the facility including the location of the microphone is calculated. can be aggregated. Inappropriate remarks according to this facility include, for example, remarking confidential information in a public place such as a train station where many people are present. In addition, the aggregation unit can also aggregate the frequency of appropriate remarks according to the facility including the location. For example, the administrator registers words related to the subject's work as prohibited words. Then, the management system aggregates and displays the frequency of utterances including words related to work when the target person is away from the work place. As a result, the manager can determine whether the subject has left the place of work as part of the work, or whether the person has simply abandoned the work and left the place of work. can be determined using the amount of

The inventions according to the ninth and tenth features are different in category from the inventions according to the first feature.

According to the present invention, voice data received from a microphone device is converted into text, compared with predetermined prohibited word data, and the frequency of coincidence or similarity is aggregated and displayed so that the administrator can recognize the target person's voice. It is possible to provide a speech recognition device that can manage the utterances of a target person without checking them one by one.

FIG. 1 is a block diagram schematically showing the hardware configuration and software configuration of a management system 1 using a speech recognition device 20 according to this embodiment. FIG. 2 is a flow chart showing the flow of management using the management system 1 in this embodiment. FIG. 3 is a diagram showing an example of the prohibited word list 221 in this embodiment. FIG. 4 is a diagram showing an example of the tally list 222 in this embodiment. FIG. 5 is a diagram showing an example of the statement save list 223 in this embodiment. FIG. 6 is a diagram showing an example of when a management screen showing total results and the like is displayed on the display unit 24 using the management system 1 according to this embodiment. FIG. 7 is a diagram showing an example of the specific voice list 225 in Modification 2. As shown in FIG. FIG. 8 is a conceptual diagram showing an example of a process of referring to the specific sound list 225 and excluding frequently occurring loud sound data from the sound data picked up by the microphone device 10 and totalizing the data in the modification 2. .

An example of a preferred embodiment for carrying out the present invention will be described below with reference to the drawings. This is just an example, and the technical scope of the present invention is not limited to this.

<Management system 1>
FIG. 1 is a block diagram schematically showing the hardware configuration and software configuration of a management system 1 using a speech recognition device 20 according to this embodiment.

The management system 1 is connected to a microphone device 10 capable of converting a subject's voice into voice data, and is connected to the microphone device 10. In order to operate the management system 1, data management, data processing, screen display, and warning to the administrator are performed. and a speech recognition device 20 that performs the above.

[Microphone device 10]
The microphone device 10 includes a control unit 11 that controls the operation of the microphone device 10, a storage unit 12 that stores control programs and the like executed by the microcomputer of the control unit 11, and a speech recognition device 20 that communicates with other devices. and a positioning unit 15 for positioning the current position of the microphone device 10 .

Although not essential, it is preferable that the microphone device 10 have unique identification information and transmit this identification information to the speech recognition device 20 . If the unique identification information is transmitted, the speech recognition device 20 receives and uses this identification information, thereby easily linking the microphone device 10 and the caller. Thereby, the speech recognition device 20 can quickly identify the caller. As this unique identification information, for example, the IPv6 address of the microphone device 10, the MAC address of the network card included in the microphone device 10, or the like can be used.

The control unit 11 includes a CPU (Central Processing Unit), a RAM (Random Access Memory), a ROM (Read Only Memory), and the like.

The storage unit 12 is a device for storing data and files, and has a data storage unit such as a hard disk, a semiconductor memory, a recording medium, or a memory card. The storage unit 12 stores control programs and the like executed by the microcomputer of the control unit 11 .

The communication unit 13 has a device for enabling communication with the microphone device 10 and other devices, such as a network card compatible with the Ethernet (registered trademark) standard, a wireless device compatible with a mobile phone network, and the like.

The configuration of the positioning unit 15 is not particularly limited. Examples of the positioning unit 15 include a positioning system using a GPS receiver that receives signals from GPS satellites and performs positioning, and a positioning system that uses information from a mobile phone base station.

[Voice recognition device 20]
The speech recognition device 20 includes a control unit 21 that controls the operation of the speech recognition device 20, a control program executed by a microcomputer of the control unit 21, various data used for processing speech data, and speech data. It includes a storage unit 22 that stores data generated by processing, etc., a communication unit 23 that communicates with the microphone device 10, and a display unit 24 that displays tallied results and the like. The data used for processing the utterance data includes prohibited word data and the like. The data generated by processing the utterance data includes the result of tallying utterances that match or are similar to prohibited words.

The type of display unit 24 is not particularly limited. Examples of the display unit 24 include a monitor and a touch panel.

The control unit 21 reads a predetermined program, and cooperates with the storage unit 22 and/or the communication unit 23 as necessary to generate an utterance data generation unit 211 and an aggregation unit 212, which are software configuration elements in the management system 1. , warning unit 213, utterance data storage unit 214, and the like.

The hardware configurations of the control unit 21, the storage unit 22, and the communication unit 23 are the same as the hardware configurations of the control unit 11, the storage unit 12, and the communication unit 13, respectively. The storage unit 22 also stores a prohibited word list 221 that lists prohibited word data, a total list 222 that tallies the frequencies of utterance data that match or are similar to prohibited words, and the like. In addition, the storage unit 22 is configured to be able to set the subject's voice data received from the microphone device 10 .

<Management procedure using management system 1>
FIG. 2 is an example of a flow chart showing a procedure for managing a subject's utterances using the management system 1 using the speech recognition device 20. As shown in FIG. A preferred software configuration of the management system 1 will be described in more detail below with reference to FIG.

[Step S10: Audio conversion]
First, the microphone device 10 picks up the voice of the subject. The control unit 11 of the microphone device 10 cooperates with the storage unit 12 to convert the collected sound into sound data.

Well-known formats such as WAV format, MP3 format, and FLAC format are used as the format of this audio data. In order to suppress communication band consumption, it is preferable to use a format using irreversible compression, such as the MP3 format.

[Step S11: Volume measurement]
Subsequently, the control unit 11 cooperates with the storage unit 12 to execute the volume measurement unit 111 to measure the volume of this audio data.

[Step S12: Determining whether the sound volume is equal to or greater than a predetermined amount]
If the volume measured in step S11 is equal to or greater than the predetermined amount, the process proceeds to step S13. If the volume is less than the predetermined amount, the processing is terminated and the device returns to the standby state for waiting for voice input.

According to the management system 1 according to the present embodiment, through steps S11 and S12, it is possible to transmit voice data to the voice recognition device 20 only when the volume is equal to or greater than a predetermined amount.

[Step S13: Measurement of current position]
The control unit 11 measures the current position of the microphone device 10 in cooperation with the positioning unit 15 and stores it in the storage unit 12 as current position data.

Although it is not an essential mode, the current position measured by the positioning unit 15 at a pre-scheduled timing is stored as current position data in the storage unit 12, and the current position data is read out to perform positioning of the current position in this step. It is also preferable to replace with As a result, it is possible to avoid positioning the current position each time the voice is converted and waiting for the completion of the positioning process.

[Step S14: Audio data transmission]
The control unit 11 cooperates with the storage unit 12 and the communication unit 13 to execute the transmission unit 112 and transmit the voice data and the current position data to the voice recognition device 20 .

According to the management system 1 according to the present embodiment, voice data is transmitted to the voice recognition device 20 through steps S10 to S14. Further, according to the management system 1 according to the present embodiment, the current position can be transmitted to the speech recognition device 20 through steps S13 to S14.

[Step S20: Receive voice data]
When the voice data and the current position data are transmitted in step S14, the communication unit 23 of the voice recognition device 20 receives them. The received voice data and current position data are stored in the storage unit 22 .

[Step S21: Generation of utterance data]
When the voice data is received in step S20, the control unit 21 cooperates with the storage unit 22 to execute the utterance data generation unit 211, convert the received voice data into text, and generate utterance data. The generated utterance data is stored in the storage unit 22 . A well-known speech recognition technique such as speech recognition using a statistical model is used to convert speech data into text.

[Step S23: Comparison with Prohibited Words]
When the utterance data is generated in step S21, the control unit 21 cooperates with the storage unit 22 to execute the counting unit 212, and compares the utterance data with each of the prohibited words stored in the prohibited word list 221. do. If the utterance data and the prohibited word match, the process proceeds to step S24. If the utterance data and the prohibited word do not match, the process proceeds to step S26.

FIG. 3 is a diagram showing an example of the prohibited word list 221 in this embodiment. As indicated by "source" and "prohibited word" in FIG. there is ID: 1 in FIG. 3 indicates that the phrase "mother" is defined as a prohibited word for the sender "Taro Tanaka" who is the target person.

By associating the originator with the prohibited word, the administrator can set the prohibited word for each originator. As a result, it is possible to finely manage inappropriate remarks according to the source. For example, the phrase "My hair is beautiful" is a remark that could lead to sexual harassment, so we set it as a prohibited word and tabulate it, but if the sender is a hairdresser, it is normal for the sender's work. It is a remark, and it can be managed such that it is not set as a prohibited word and not counted.

Although not essential, the prohibited word list 221 preferably also includes area information associated with the prohibited words, as indicated by "Area" in FIG. By including the area in the prohibited word list 221, the tabulating unit 212 can compare this area with the current position data and refer to the prohibited word corresponding to the current position. As a result, it is possible to collect prohibited words more precisely according to the current position.

Prohibited words according to the current position will be described using IDs 1 to 3 and 101 to 104 in FIG. 3 as an example. In this example, 7 pairs of prohibited words and areas are defined for "Taro Tanaka" who is engaged in plumbing work. In IDs: 1 and 101, the phrase "mother" is defined as a prohibited word for the two areas of "station" and "residential area" for "Taro Tanaka", who is the target person and sender. The phrase "mother" is defined in both areas to indicate trouble, such as threats, regardless of the area in which the statement is made. In ID: 2, the phrase "water bill" is defined as a prohibited word for the "station" area. For Taro Tanaka, who is engaged in waterworks, the station is not a place where he does his usual work. However, since "water bill" is a phrase related to work, if Taro Tanaka repeats remarks including "water bill" at the station, the administrator can determine that Taro Tanaka has visited the station to do his work. IDs 102 to 104 define phrases referring to physical appearance as prohibited words for the area of "residential area", such as "beautiful woman", "baby face", and "beautiful hair". Mentioning appearance in the residential area where Taro Tanaka works indicates the possibility of sexual harassment by Taro Tanaka. The administrator can detect sexual harassment by Taro Tanaka from the frequency with which Taro Tanaka made remarks containing these prohibited words in the residential area.

Although not essential, the prohibited word list 221 preferably includes attribute information that bundles multiple senders, as indicated by "attribute" in FIG. Since the prohibited word list 221 has attribute information that bundles multiple senders, the administrator can perform more efficient management, such as collectively changing the prohibited words of multiple senders using attributes. .

[Step S24: Tabulation Processing]
Return to FIG. When step S 23 detects a match between the prohibited word and the utterance data, the control unit 21 cooperates with the storage unit 22 to increment the frequency corresponding to the prohibited word stored in the tally list 222 .

FIG. 4 is a diagram showing an example of the tally list 222 in this embodiment. As shown in FIG. 4, in the tally list 222, senders, prohibited words, and frequencies are associated with IDs and stored. ID: 1 in FIG. 4 stores that the originator "Taro Tanaka" has not made any remarks including the prohibited word "mother". ID: 2 in FIG. 4 stores that the caller "Taro Tanaka" made 886 remarks including the prohibited word "water bill".

By associating the originator, the prohibited word, and the frequency, the display unit 24 can display to the administrator the frequency of utterances containing the prohibited word for each originator. With this display, the administrator can finely judge the occurrence of troubles and the signs of troubles for each source.

Although not required, the summary list 222 also preferably stores the current location data and/or the area containing the current location data. By storing the current location data and areas in the tally list 222, the display unit 24 can display to the administrator the frequency of utterances containing prohibited words for each area including the stored current locations. With this display, the administrator can manage the frequency with which the target person made inappropriate remarks defined for each area.

[Step S26: Saving Speech Data]
Return to FIG. Control unit 21 cooperates with storage unit 22 to execute utterance data storage unit 214 and store utterance data in utterance storage list 223 .

FIG. 5 is a diagram showing an example of the statement save list 223 in this embodiment. As indicated by "source", "utterance data", and "voice data" in FIG. Saved with ID. In ID: 1 in Fig. 5, the originator "Taro Tanaka" made a statement in Chiyoda Ward, Tokyo, at exactly 12:00 on December 16, 2018, "This month's water bill is..." and the sound data [sound1] at that time are stored.

By associating the caller and the message data, the administrator can manage the message data for each caller. In addition, by linking the utterance data and the voice data, the administrator can reproduce the voice data and obtain more detailed information about the utterance, such as the subject's tone of voice and voice volume.

Although it is not essential, it is also preferable that position information indicating the current position of the target person at the time the statement was made and the statement data be associated with each other and stored in the statement storage list 223 . By storing the position information, the administrator can grasp the place where the statement was made. This allows the administrator to conduct additional research, such as asking around the location where the statement was made.

Although not essential, it is also preferable that the time information indicating the date and time when the statement was made is associated with the statement data and stored in the statement storage list 223 . By storing the time information, the administrator can grasp the date and time when the statement was made. This allows the administrator to determine whether the trouble that occurred on a specific date and time and the message data are related to each other. In addition, the administrator can search for statement data using this time information, and can efficiently find statement data related to trouble that occurred on a specific date and time. In addition, the administrator can conduct additional research, such as asking questions using the date and time the statement was made.

[Step S27: Comparison with Predetermined Frequency]
Return to FIG. Control unit 21 cooperates with storage unit 22 to execute warning unit 213, and compares the frequency of utterances containing prohibited words with a predetermined frequency. If the frequency of utterances is equal to or higher than the predetermined frequency, the process proceeds to step S28. If the frequency of utterances is less than the predetermined frequency, the process proceeds to step S29.

[Step S28: Warning processing]
In step S27, if the frequency of utterances is equal to or higher than the predetermined frequency, the control unit 21 cooperates with the storage unit 22 to warn the administrator. This warning preferably displays on the display unit 24 that the frequency of utterances has exceeded a predetermined frequency.

By being warned that the frequency of utterances exceeds a predetermined frequency, the administrator can quickly notice the occurrence of trouble or its precursor. As a result, the administrator can promptly issue a warning to the target person, report to the police, or the like. Further, by issuing a warning when the frequency is equal to or higher than a predetermined frequency, it is possible to prevent mischievous warnings and false warnings caused by aggregating words with similar pronunciations as matching or similar words.

Although not required, this alert also preferably includes the originator. By including the sender of the warning, the administrator can quickly determine which target person should be dealt with.

Although not required, this alert also preferably includes utterance data. By including the message data in the warning, the administrator can determine what kind of troubleshooting should be done without searching for the corresponding message data.

Although not an essential aspect, it is also preferable that the control unit 21 cooperate with the storage unit 22 and the communication unit 23 to send an email and/or an instant message to the administrator to warn the administrator. By sending an email and/or an instant message to the administrator, the administrator can be alerted even when the administrator is away from the speech recognition device 20 . As a result, the administrator can quickly know that the frequency of prohibited words exceeds a predetermined frequency, and can deal with the trouble.

Moreover, although not an essential mode, it is also preferable that the control section 21 cooperates with the storage section 22 and the communication section 23 to give an additional warning using a warning sound or a voice such as a telephone call. The administrator can receive the warning without always monitoring the display unit 24 by using the warning sound or the voice such as calling by telephone.

[Step S29: Display processing]
The control unit 21 cooperates with the storage unit 22 to display the tallied result on the display unit 24 . The aggregated results displayed include the source, the banned words, and the frequency of utterances containing the banned words.

By displaying the caller, the prohibited words, and the frequency of statements containing the prohibited words on the display unit 24, the administrator can grasp the frequency of the statements containing the prohibited words by the callers. The administrator uses the displayed frequency of utterances containing prohibited words to grasp the occurrence of troubles around the originator, and performs management such as issuing a warning to the target person and reporting to the police. The manager can manage this series of statements of the caller without checking the caller's voice one by one.

Although not essential, for the most recent utterance data stored in the utterance data storage unit 214, a set of outlines including caller, date and time, place, means of access to conversation data and/or voice data are stored in a predetermined number. It is also possible to enumerate only the sets and display them on the display unit 24 . By displaying a summary of the most recent speech data, the administrator can grasp the target person who is currently speaking. Since the summary includes means for accessing conversation data and/or voice data, the administrator can quickly check the speech data, voice data, etc. of the subject who is currently speaking.

<Usage example of management system 1>
Next, a usage example of the management system 1 using the speech recognition device 20 according to this embodiment will be described.

First, the administrator connects the microphone device 10 and the speech recognition device 20 . For this connection, for example, the administrator registers an IPv6 address indicating the voice recognition device 20 in the storage unit 12 of the microphone device 10, and further registers an IPv6 address indicating the microphone device 10 in the storage unit 22 of the voice recognition device 20. This is done by, for example, configuring them so that they can communicate with each other via the Internet.

Subsequently, the administrator causes the subject to carry the microphone device 10, installs the microphone device 10 on the subject's desk, or the like, so that the subject's voice can be converted into voice data. The volume of the converted voice data is measured, and the microphone device 10 transmits the voice data to the voice recognition device 20 when the volume is greater than or equal to a predetermined value.

The speech recognition device 20 receives speech data from the microphone device 10, converts the speech data into text to generate utterance data, compares it with predetermined prohibited words, counts and displays the frequency of matching or similar utterances. A series of operations to be displayed on the unit 24 are performed. Further, the speech recognition device 20 can classify the generated utterance data according to the originator and store them in the utterance storage list 223 .

FIG. 6 is a display example of the totalization result displayed on the display unit 24 . Under the "number of detected keywords (Taro Tanaka)", for each prohibited word, a set of information consisting of the prohibited word and the frequency of statements that match or are similar to the prohibited word is displayed for a predetermined number. . "(Taro Tanaka)" in FIG. 6 indicates that prohibited words and frequency of utterances are displayed for "source: Taro Tanaka". By specifying the sender, the administrator can similarly display the prohibited words and the frequency of utterances for arbitrary senders. By operating "See more", the administrator can display more combinations of prohibited words and frequency of remarks.

As shown in FIG. 6, it is also possible to display a bar graph whose size corresponds to the frequency of utterances. By displaying a bar graph whose size corresponds to the frequency of utterances, the administrator can more intuitively grasp the frequency of utterances that match or are similar to prohibited words.

As shown in FIG. 6 as a "latest usage list" and a table below it, for the most recent saved utterance data, the sender, date and time, place, and means of access to the utterance data and/or voice data of the target person are displayed. It is also possible to display a set of summaries including a predetermined number of sets listed. By displaying a summary of the most recent speech data, the administrator can grasp the target person who is currently speaking. By displaying the summary, the administrator can quickly check the utterance data, voice data, etc. of the target person who is currently speaking. The administrator can access the saved statement data by operating "details" in the "conversation content (text)" column. Also, the administrator can access the saved voice data by operating "details" in the "conversation content (voice)" column. The administrator can display more overviews by operating "See more".

The administrator looks at the tallied result displayed on the display unit 24 and judges whether or not trouble has occurred or whether there is a sign of trouble based on the frequency of utterances. Then, for a target person who is determined to be in trouble or to have a sign of trouble, management including trouble handling such as warning the target person or reporting to the police is performed. In addition, the administrator can refer to the current location linked to the displayed voice data, go to the site where the trouble is occurring, contact the person in charge near the site where the trouble is occurring, etc. Management including response.

In addition, the speech recognition device 20 can warn the administrator through the display unit 24 or the like when the frequency of statements that match or are similar to prohibited words is equal to or higher than a predetermined frequency. The administrator who received the warning can perform management including troubleshooting in the same manner as above.

This series of operations of the management system 1 is automatically performed without the administrator checking the target person's voice one by one. Therefore, with the management system 1 of this embodiment, the administrator can manage the utterances of the subject without checking the voice of the subject one by one. As a result, the administrator can manage the utterances of the target person with less effort. By managing utterances with less effort, the manager can simultaneously manage utterances of more subjects without increasing the effort required to manage utterances. In addition, the warning allows the administrator to more quickly deal with troubles that occur around the target person.

<Can be provided as a management program>
So far, the present invention has been described as being provided as the management system 1, but it is not limited to this. The invention described in this embodiment can also be provided as a management program that can be executed by a computer. The management program includes a pronunciation data generation step in which the utterance data generation unit 211 converts the voice data received from the microphone device 10 into text and generates utterance data of the target person, and Comparing the prohibited word data with the prohibited word data obtained, totaling the matching or similar frequencies, and a displaying step of displaying the result of totaling by the totalizing unit 212 to the administrator.

<Modification>
Various modifications of the invention described in this embodiment will be exemplified below.

[Modification 1] Determining Caller The storage unit 22 has a voiceprint data list 224 that stores voiceprint data in advance, and the control unit 21 compares the voice data with voiceprint data stored in the voiceprint data list 224. However, it is also preferable to include a caller determination unit 215 for determining whether or not the caller of the voice data is the caller of the voiceprint data. At this time, it is preferable that the utterance data is classified according to the originator and stored in the utterance storage list 223 . Furthermore, the warning unit 213 may warn the administrator when the source of the voice data is the source of the voiceprint data.

By storing the voiceprint data of the target person in the voiceprint data list 224 and determining the caller by the caller determination unit 215, the voice recognition device 20 can count only the utterances of the target person and warn the administrator. In other words, the prohibited words contained in the voices of third parties other than the target, especially the voices of third parties who were present around the target who did not participate in the conversation with the target, are aggregated and sent to the administrator. You will no longer be warned. The administrator can get a more accurate tally of the forbidden word frequency required.

Further, by identifying the originator, the utterance data storage unit 214 can more appropriately associate the utterance data with the originator, classify the utterance data, and store the data in the utterance storage list 223 . Thereby, the administrator can specify a caller, read out the saved message data corresponding to the specified caller, and examine the contents of the caller's message data. In a situation where there is a discrepancy between the parties regarding the presence or absence of statements and the content of the statements, the administrator can objectively grasp the situation by carefully examining the evidence of the saved statement data by the originator. Also, in a labor tribunal trial in which the subject is a party, the administrator can present the saved utterance data of the sender as evidence.

Furthermore, since the utterance data is stored in a more appropriate correspondence with the originator, the administrator can more effectively analyze and use the stored utterance data as big data. Through such analysis, the manager can specify a source and analyze the utterance tendency of the source, find words that tend to appear before trouble occurs, and add them to the prohibited word list 221. .

[Modification 2] Reduction of load using specific voice data The storage unit 22 preferably includes a specific voice list 225 that stores specific voice data determined according to the attributes of the subject. By providing the specific audio list 225, the counting unit 212 compares the specific audio data corresponding to the attributes of the subject stored in the specific audio list 225 with the audio data, and totals the audio data different from the specific audio data. can do

FIG. 7 is an example of the specific voice list 225. As shown in FIG. In the specific voice list 225, the attribute of the caller and the specific voice data corresponding to the attribute are linked and stored. FIG. 7 shows that audio data [noise1] (knocking sound of "gangan") is defined as specific audio data for the attribute "waterworks". Although not essential, as shown as "remarks" in FIG. 7, the specific sound list 225 can include, as text, explanations of the specific sound data associated with the specific sound data, and the like. Since the specific audio list 225 includes explanations and other information as text, the administrator can grasp and easily manage the specified specific audio data without reproducing the specific audio data one by one.

Aggregation using this specific audio data will be described by way of an example using the conceptual diagram of FIG. In this example, the target person is a worker who engages in waterworks, and the knocking sound of the tap water frequently occurs near the target person.

The a) voice data in the uppermost row is an example of voice data transmitted from the microphone device 10, which is an utterance (indicated by "this water supply", "wet", and "bad" in FIG. 8) that can be counted. In addition, it is shown to include the knocking sound of tap water (indicated by "banging" in FIG. 8).

The b) specific voice data on the second row indicates a knocking sound of "gan-gan" corresponding to the attribute of plumbing, which is stored as specific voice data in the specific voice list 225 of FIG.

Return to FIG. The process c) in the third row shows how the control unit 21 executes the counting unit 212 and compares the specific audio data shown in b) with the audio data shown as an example. In c), the portion where the background color of the audio data is changed is the portion that matches the specific audio data as a result of this comparison.

As a result of this comparison, as shown in d) after processing at the bottom, the part matching the specific voice data was removed from the exemplified voice data, and the subject indicated by "this water", "wet", and "bad". Only the contents of the utterances by are aggregated as utterance data. In this way, only the parts different from the specific audio data are tallied, so that the processing load due to the execution of the tallying unit 212 is reduced.

Although the embodiments of the present invention have been described above, the present invention is not limited to these embodiments described above. Moreover, the effects described in the embodiments of the present invention are merely a list of the most suitable effects resulting from the present invention, and the effects of the present invention are limited to those described in the embodiments of the present invention. not a thing

Moreover, the above-described embodiments are described in detail for easy understanding of the present invention, and are not necessarily limited to those having all the described configurations. Also, part of the configuration of one embodiment can be replaced with the configuration of another embodiment, and the configuration of another embodiment can be added to the configuration of one embodiment. .

Further, each configuration, function, and processing unit described above may be implemented partially or entirely by hardware (for example, an integrated circuit). Further, each of the configurations, functions, and processing units described above may be realized by software by a processor interpreting and executing a program for realizing each function. Information such as programs, tables, and files that implement each function can be stored in recording devices such as memories, hard disks, SSDs (Solid State Drives), or recording media such as IC cards, SD cards, and DVDs.

1 management system 10 microphone device 11 control unit 111 volume measurement unit 112 transmission unit 12 storage unit 13 communication unit 15 positioning unit 20 speech recognition unit 21 control unit 211 utterance data generation unit 212 aggregation unit 213 warning unit 214 utterance data storage unit 215 transmission Source discrimination unit 22 Storage unit 221 Prohibited word list 222 Aggregation list 223 Statement storage list 224 Voiceprint data list 225 Specific voice list 23 Communication unit 24 Display unit

Claims (1)

A speech recognition device connected to a microphone device capable of converting a subject's voice into voice data, and capable of converting the voice data received from the microphone device into text,
a utterance data generation unit that converts the voice data received from the microphone device into text and generates utterance data of the subject person;
an aggregation unit that compares the utterance data with predetermined prohibited word data and aggregates the frequency of matching or similarity;
a display unit for displaying a result of counting by the counting unit to an administrator;
an utterance data storage unit for classifying, accumulating, and storing the utterance data generated by the utterance data generation unit for each source of the voice data;
has
The aggregating unit is predetermined and includes a target person who is a source of the voice data.
As corresponding to frequently occurring sounds in the vicinity of people who have the same attribute as the attribute,
A speech recognition device that compares specific speech data stored in association with an attribute with the speech data, and performs the aggregation for the speech data that is different from the specific speech data.

Images