JP2017191533A - Minute preparation system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform recognition with higher accuracy in preparing minutes on the basis of voice recognition.SOLUTION: If recognition cannot be performed when voice recognition is performed by using a dictionary set in advance, dictionaries on another network are used to perform recognition processing, and recognition results are described in minutes by differentiating the recognition results from other terms about voice recognition subjected to the recognition processing. Further, voice recognition which can be recognized by using the dictionaries on the network may be described by differentiating it from voice recognition which cannot be recognized for all that. Furthermore, a priority sequency of dictionaries to be used may be dynamically changed o the basis of recognition results.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a technique for creating meeting minutes.

  Currently, in companies and the like, meetings are frequently held, and minutes are often created at the meetings. Minutes are often created manually based on memory, etc., and errors may be mixed. As a technique for improving the accuracy of the minutes, there is a technique disclosed in Patent Document 1. In Patent Document 1, preparation of meeting minutes, distribution of drafts to attendees, proofreading, and distribution of final minutes are performed in real time, thereby reducing time and labor for proofreading and at the same time preventing falsification of the final minutes.

JP 2005-222118 A

  Here, since speech is spoken verbally in a meeting, efficient recognition of the minutes can be made if speech recognition is performed and converted into text. However, technical terms are often used in meetings, and if recognition is performed only with a general dictionary, misrecognition or unrecognizable words may occur.

  In order to solve the above problem, in the present invention, when speech recognition is performed using a preset dictionary, if recognition is not possible, recognition processing is performed using a dictionary on another network, and this recognition is performed. For those that have processed, the recognition results are shown in the minutes separately from other terms. In addition, what can be recognized using a dictionary on the network may be distinguished from what cannot be recognized yet.

  Furthermore, in the present invention, the priority order of dictionaries to be used may be dynamically changed based on the recognition result.

  According to the present invention, recognition accuracy can be improved. In addition, by distinguishing and describing those that have been recognized in other dictionaries, it is possible to identify words that should be noted in the confirmation and correction work by the minutes creator.

Conceptual diagram of minutes creation system 2 used in one embodiment of the present invention The figure which shows the flow of the preparation performed before the meeting in one Embodiment of this invention. The figure explaining the processing content of the minutes preparation in one Embodiment of this invention. The figure which shows the system configuration | structure figure and data relationship in one Embodiment of this invention. The figure which shows the detail of the process by S11 of one Embodiment of this invention The figure which shows the detail of the process by S12 of one Embodiment of this invention

  Hereinafter, an embodiment of the present invention will be described with reference to the drawings. First, FIG. 1 is a conceptual diagram of a minutes creation system 2 in the present embodiment. In the present embodiment, the minutes creation system 2 that records the audio during the meeting and creates the minutes after the meeting is linked with the schedule management server 3, and the person registered in the schedule management server 3 as the minutes manager Send the minutes you created. This eliminates the need for preparation of a small device for recording the minutes and the preparation of the minutes after the meeting, which are necessary in the past, and the minutes editing after the meeting proceeds smoothly. In more detail, after collaborating with a plurality of in-house DBs in advance, search priorities are set, and when technical terms etc. that are not registered in the dictionary DB 7 of the minutes creation system 2 appear, Search automatically in Network 1 and include in the minutes with annotations. Also, new words are registered in the dictionary DB7. Hereinafter, the contents of the above processing will be described with reference to FIG.

  FIG. 4 is a system configuration diagram of the present embodiment. Each device is connected via the network 1. The PC 8 is a personal PC for registering conference room reservations and minutes managers. The minutes creation system 2 is installed in each meeting room. Only one schedule management server 3 and minutes creation management server 5 are required. The schedule management server 3 includes a schedule management DB 4 having information on the minutes creation flag, minutes staff, reserved meeting room name, reservation date and time, reservation person, and meeting attendees. This makes it possible to identify the necessity of creating the minutes of each meeting and the person in charge of the minutes. In addition, the minutes creation management server 5 includes a meeting minutes manager, a reserved meeting room, reservation date and time, voice data, a minutes management DB 6 that stores information on meeting minutes in which voice data is converted into text, and an intranet The dictionary DB 7 is connected to the dictionary DB and the Internet, and has a dictionary DB 7 for storing voice information, its notation method, and intranet dictionary DB search order. The above minutes creation system 2, schedule management server 3, and minutes creation management server 5 are realized by a so-called computer, and the processing of the flowchart described below is performed by a CPU-like operation according to its own program. Executed by the device.

  The intranet and the Internet are shown as network 1.

  Next, a preparation flow to be executed before the meeting will be described with reference to FIG. The processing in FIG. 2 is executed by the schedule management server 3 unless otherwise specified.

  First, in S1, the minutes creation flag, the minutes staff, the meeting room name, and the date and time are registered in the schedule management server 3.

  In this registration, when making a reservation for a conference room, information used for creating the minutes is registered in the system as necessary. For this reason, (1) a meeting room along the desired meeting, (2) date and time, and (3) necessity of creating minutes are received from the user via the PC 8, and registration processing is performed accordingly. That is, (1) and (2) are registered in the schedule management server 3. As for (3), when the user's input indicates “necessary” to create the minutes (such as when the user presses a button), information indicating this is received, and the flag is recorded accordingly.

  When the flag is recorded, that is, when it is necessary to create the minutes, the PC 8 displays a message prompting the designation of the person in charge of the minutes. In response to this display, when the user inputs to the PC8 to identify the person in charge of the minutes, the personal identification number such as “LDAP” of the person in charge of the minutes and the name of the person in charge of the minutes are specified. This is registered in the schedule management server 3. The personal identification number and the name of the person in charge of the minutes are specified using the contents recorded in the employee DB (not shown) in response to the input.

  Next, in S2, it is determined whether the reservation status is complete before the reservation date and time. In this process, it is determined whether or not the reservation status of the conference reserved on the day is complete when it becomes scheduled every day. More specifically, for example, when the system is started, the schedule management DB 4 is searched using the reservation date and time and the reservation person as a key to identify the meeting of the day. Whether the reservation is complete for the identified meeting, that is, whether the minutes creation flag is registered, whether the minutes clerk is registered for those with the minutes creation flag, or whether the meeting room is reserved Make a decision. If it is determined that the reservation is complete, the process proceeds to S4, and an email indicating that the reservation is complete is sent to the schedule reservation person. On the other hand, if the reservation is incomplete, the process proceeds to S3, and an email indicating that the reservation is incomplete is transmitted to the schedule reservation person. The schedule reservation person who has received the email that the reservation is incomplete re-registers the incomplete part in S5 and completes the reservation. Finally, in S6, the schedule management server 3 transmits the reservation date / time, the conference room, and the information of the person in charge of the minutes to the minutes creation server 5. The above processing is executed for each conference reservation. In S2, the integrity is confirmed when the system is started. However, it may be any time before the date and time of the meeting or other time when S1 is input.

  Next, with reference to FIG. 3, the contents of the minutes creation process will be described. In S7, the names and voices of the conference participants are registered in the minutes creation management server 5 at the start of the conference. When the minutes are converted into text, the content of the speech and the name of the speaker are converted into text using this information.

  After the conference starts, in S8, the content of the speech in the conference is recorded as audio information. After the conference ends or in parallel with the conference, the voice information recorded in S8 is converted to text in S9. At this time, if there is a word that cannot be transcribed, the dictionary DB7 search of S10 is executed. If the word is already registered in the dictionary DB7, the process proceeds to S16. If the word is not registered in the dictionary DB 7, the word is searched on the intranet or the Internet as described in S11. As a result, in S12, an input as to whether a word has been searched, that is, whether it can be converted into a character is accepted from the intranet or the Internet. If an input indicating that the word has been searched is accepted, the process proceeds to S13.

  Here, details of the processing in S11 are shown in FIG. In the dictionary DB7, when creating the minutes of the conference No. 1 that was held in advance before the conference concerned, the database was searched and the voice information of the newly registered word, the notation method, the DB search order, the DB ( 1 and record) are registered. Note that the number of DBs recorded as 1 is limited to one. The reason is that 1 is recorded in the first hit DB as a result of the search, the search is terminated at the time of the hit, and the search in other DBs is suppressed. In addition, since the DB that was initially recorded as 1 did not hit, “x” is recorded for the DB. In this example, when a new voice information “Shukkin” is to be converted into text, it is not registered in the dictionary DB 7 and thus searches on the intranet and the Internet are started. The order of priority for DB search in the intranet is set in the order of accounting DB, procurement DB, in-house SNS, and in-house dictionary. By default, search is started from the accounting DB. When searching for "Shukkin", it was not found in the accounting database, but it was hit in the second procurement database. In this case, since the word is searched in S12, the process proceeds to S13, and “withdrawal” is registered in the notation method column of the dictionary DB7. In addition, in the accounting DB that did not hit, the x mark and the ranking of the procurement DB are set to 1, and the other rankings are shifted one by one and registered. Further, in S15, when the searched word is described on the minutes, an annotation such as a meaning or a source is added in distinction from others, such as changing the color.

  Next, another example (when searched on the Internet) in S11 is shown. In S11, if there is an unregistered word "Zei" in the voice information, the voice information is first registered in the dictionary DB7, and the search target DB is the same as "Shukkin" searched before searching for "Zay" Set the search order and start the search according to this order. DBs with a cross are searched last. As a result of the search, there was no hit in the DB in the intranet, so when a search on the Internet was started, there was a page that hit on the Internet as a result. As a result, as shown in FIG. 6, in S12, the voice information “Zey” is registered, and in S13, the URL hit on the Internet is recorded in the URL item. Register a cross in the rank of the DB that was not found. Set 1 in the Internet column that can be searched, and set 2 in the accounting DB that was scheduled to be searched at the end (the x mark was written). In this way, the DB and the Internet in the search target intranet are searched, and the search result is accumulated in the dictionary DB 7.

  Here, the maintenance of the dictionary DB 7 performed after the minutes are created will be described. Of the search results stored in the dictionary DB 7, the words searched on the intranet are periodically searched in each DB in the intranet, and the rank is specified in the order of the number of hits. The specified order is overwritten in the dictionary DB7. This keeps the freshness of the information and at the same time makes the next search process efficient. Further, this processing may be performed by monthly processing or the like.

  Next, the process returns to S12 in FIG. If the word is not found as a result of the search, proceed to S14 and distinguish it from others, such as changing the color in the minutes, so that the voice information is written as it stands out and can be recognized as a word when it is transcribed Leave a note to the effect that it did not.

  The above operations are repeated until it is determined in S16 that the recording content has been transcribed and searched. After the characterization is completed, the minutes are sent to the minutes staff registered in the minutes creation management server 5 in S17.

  In addition, the characterization in this embodiment is voice recognition and characterizing the result. The above is the content of the processing in this embodiment.

1 Network 2 Minutes Creation System 3 Schedule Management Server 4 Schedule Management DB
5 Minutes creation management server 6 Minutes management DB
7 Dictionary DB

Claims (3)

In the minutes creation system to create the minutes at the meeting,
Means for performing speech recognition using a preset dictionary for the attendees'remarks;
Means for outputting a recognition request to the second dictionary for terms not recognized as a result of the speech recognition;
Means for receiving a recognition processing result by the second dictionary and creating a minutes for noting the term that is the received recognition processing result as distinct from other terms; Creation system. In the minutes preparing system according to claim 1,
The minutes creating system characterized in that the creating means displays a term recognized using the second dictionary and a term unrecognizable by distinguishing the difference. In the minutes creation system according to claim 1 or 2,
Further, when the second dictionary can be recognized, the minutes creation system is characterized in that the second dictionary is used as a preset dictionary instead of the dictionary.

Images