Language data processing system, language data processing method, and computer program

Abstract

To provide a language data processing system capable of improving at least one disadvantage of a conventional technique.SOLUTION: A language data processing system includes a processing unit and a storage unit electrically connected to the processing unit. The storage unit stores a language processing model achieved by a machine learning technique. The language processing unit includes a plurality of intention tags corresponding to a plurality of control procedures to be executed by the processing unit. The processing unit is configured so as to use a language processing model to acquire a plurality of designation intentions represented by text data and each corresponding to one intention by analyzing the text data, further determine whether each of the designation intentions is a clear designation intention or an ambiguous designation intention on the basis of the text data, determine at least one target designation intention from designation intentions determined to be clear designation intentions, and execute a control procedure to which an intention tag with the target designation intention corresponding thereto corresponds to.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to a data processing system, in particular to a language data processing system for processing content dictated by a user. The present invention further relates to a language data processing method for processing content dictated by a user and a computer program for causing an electronic device to carry out the language data processing method. [Background technology]

[0002] With the development of language processing technology, when operating a voice-operable electronic device, a user can issue commands to the electronic device in more everyday expressions, without being limited to specific voice commands.

[0003] In general speaking, multiple requirements are conveniently expressed in one sentence, but such an expression tends to be long and some requirements are not clearly expressed, which causes problems when processing using conventional natural language processing. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] China Patent Publication No. 114153956 Summary of the Invention [Problem to be solved by the invention]

[0005] SUMMARY OF THE PRESENTLY PREFERRED EMBODIMENTS It is therefore an object of the present invention to provide a language data processing system, a language data processing method, and a computer program that can improve at least one of the shortcomings of the prior art. [Means for solving the problem]

[0006] The language data processing system includes a processing unit and a storage unit electrically connected to the processing unit.

[0007] The storage unit stores a language processing model realized by machine learning technology. The language processing model includes a plurality of intention tags corresponding to a plurality of control procedures executed by the processing unit.

[0008] The processing unit is configured to analyze the text data using a language processing model to obtain a plurality of designated intentions represented by the text data and each of which corresponds to one intention tag, and further to determine whether each of the designated intentions is a clear designated intention or an ambiguous designated intention based on the text data, to determine at least one target designated intention from the designated intentions determined to be clear designated intentions, and to execute a control procedure corresponding to the intention tag to which the target designated intention corresponds.

[0009] The language data processing method is performed by a language data processing system.

[0010] The language data processing system includes a processing unit and a storage unit electrically connected to the processing unit. The storage unit stores a language processing model realized by a machine learning technique. The language processing model includes a plurality of intention tags corresponding to a plurality of control procedures executed by the processing unit.

[0011] The language data processing method includes a step in which a processing unit analyzes text data using a language processing model to obtain a plurality of designated intentions represented by the text data and each of which corresponds to one intention tag, and further determines whether each of the designated intentions is a clear designated intention or an ambiguous designated intention based on the text data; a step in which the processing unit determines at least one target designated intention from the designated intentions determined to be clear designated intentions; and a step in which the processing unit executes a control procedure corresponding to the intention tag to which the target designated intention corresponds.

[0012] The computer program is executed by the electronic device.

[0013] The computer program includes a language processing model implemented by machine learning techniques, the language processing model including a plurality of intent tags corresponding to a plurality of control procedures executed by the electronic device.

[0014] The computer program, when executed by an electronic device, causes the electronic device to perform the above-mentioned language data processing method. Effect of the Invention

[0015] The system analyzes the designated intention from the text data, judges for each designated intention whether the designated intention is clear or ambiguous, and executes a corresponding control procedure only for the designated intention that is judged to be clear. Therefore, even if some of the designated intentions expressed by the text data are unclear (i.e., even if there is ambiguous designated intention), the language data processing system executes a corresponding procedure for the clearly expressed designated intention (i.e., clear designated intention), thereby satisfying the user's needs.

[0016] Other features and advantages of the present invention will become apparent in the following detailed description of the embodiments which proceeds with reference to the accompanying drawings. [Brief description of the drawings]

[0017] [Figure 1] 1 is a block diagram showing an embodiment of a language data processing system of the present invention; [Diagram 2] 1 is a flowchart showing an embodiment of a language data processing method of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0018] As used herein, the terms "couple" or "connect" refer to a direct connection between two or more electrical devices / apparatus / facilities by a conductive material (e.g., wires) or an indirect connection between two electrical devices / apparatus / facilities by one or more other devices / apparatus / facilities or wireless communication.

[0019] 1, there is shown an embodiment of a language data processing system 1 of the present invention. The language data processing system 1 is electrically connected to a plurality of user side devices 5 via a network, and can communicate with each of the user side devices 5. For convenience of explanation, this specification will describe the language data processing system 1 connected to one user side device 5 as shown in FIG. 1.

[0020] The user-side device 5 includes a processing module 51, an input module 52 electrically connected to the processing module 51, and an output module 53 electrically connected to the processing module 51. The user-side device 5 may be, for example, but is not limited to, a voice-operated service device, a mobile electronic device (e.g., a smart phone, a tablet computer, or a notebook computer), a desktop computer, or an Internet-enabled television.

[0021] In this specification, the user-side device 5 is an example of a voice-operable service device that is installed in a guest room of a lodging facility and used by a user (guest). When the user-side device 5 is a voice-operable service device, the processing module 51 is a central processing unit, the input module 52 has at least a microphone electrically connected to the processing module 51, and the output module 53 has at least a speaker and a display electrically connected to the processing module 53. The user-side device 5, which is a voice-operable service device, picks up the voice of the guest through the microphone of the input module 52, outputs the corresponding voice through the speaker of the output module 53, and displays the corresponding information through the display of the output module 53. The processing module 51 also operates electronic devices such as an air conditioner, a telephone, a television, and lights in the guest room by outputting control commands through wireless communication (e.g., Wi-Fi (registered trademark), Bluetooth (registered trademark)).

[0022] In this embodiment, the language data processing system 1 is a server device, and includes a processing unit 11 and a storage unit 12 electrically connected to the processing unit 11. In this embodiment, the processing unit 11 is a central processing unit having data calculation and processing functions, and the storage unit 12 is a data storage device (e.g., a hard disk) that stores digital data, but in other embodiments, the processing unit 11 may be a combination of multiple central processing units, and the storage unit 12 may be a combination of different types of computer-readable storage media (e.g., flash memory) or multiple same or different types of computer-readable storage media, and is not limited to this embodiment. In other embodiments, the language data processing system 1 may be an electronic device such as a smartphone, a tablet computer, a notebook computer, or a desktop computer, or may be multiple server devices electrically connected to each other, and is not limited to this embodiment.

[0023] In this embodiment, the memory unit 12 of the language data processing system 1 includes a language processing model M realized by machine learning technology and used by the processing unit 11, and a plurality of control procedures P that are preset and executed by the processing unit 11 to control the user-side device 5.

[0024] The language processing model M is a neural network that realizes natural language understanding and natural language generation, and the language processing model M includes a plurality of intent tags L that are set in advance. Each of the intent tags L is a semantic feature that indicates a user intent, that is, it represents a user intent. Specifically, the language processing model M is trained by machine learning using at least a plurality of phrase data as training data after the intent tags L are set. Each of the phrase data is a phrase that expresses an intent in natural language, for example, "call room 303", "set an alarm for 8 o'clock tomorrow morning", "turn on the TV", etc. Preferably, a part of the phrase data is a phrase that expresses multiple intents simultaneously in natural language, for example, "turn on the TV and lights, and turn on the air conditioner at the same time", "turn off the music, then call room 501", "play Jay Chou's song, and tell me the weather in Taipei tomorrow", etc., but is not limited to these. After being trained, the language processing model M can perform an intent analysis process on the text data, and analyze at least one user intent represented by the text data from the text data based on the intent tags L.

[0025] Regarding the intention tags L included in the language processing model M, the semantic features represented by each of the intention tags L represent the user's intention. In this embodiment, the user-side device 5 to which the language data processing system 1 is connected is a voice-operable service device installed in a guest room of a lodging facility, so the intention tags L are related to events occurring in the guest room and indicate intentions such as "make a phone call," "set an alarm," "turn on (a home appliance)," "turn off (a home appliance)," "look up information," and "play music." Note that since the user-side device 5 is not limited to a service device installed in a guest room, the intention tags L are not limited to the above examples.

[0026] In this embodiment, the intention tags L have an order, and the order of the intention tags L is set in advance. For example, the intention tag L indicating "turn on (home appliances)" and the intention tag L indicating "search for information" are prioritized over the intention tag L indicating "play music". Setting the order of the intention tags L means that, for example, when a user simultaneously expresses three user intentions, "turn on (home appliances)", "search for information", and "play music" in one phrase, the language data processing system 1 can respond to the user intentions of "turn on (home appliances)" and "search for information" before responding to the user intention of "play music". Note that the order of the intention tags L can be freely set and adjusted according to each user, and therefore the order is not the point of the present invention, and will not be further explained.

[0027] Moreover, in this embodiment, one of the multiple intention tags L is set as an exclusive intention tag L'. The user intention represented by the exclusive intention tag L' is the highest priority user intention that can be met first. For example, in this embodiment, the exclusive intention tag L' is the intention tag L indicating the user intention of "making a phone call". It should be noted that which of the intention tags L is set as the exclusive intention tag L' can be freely set according to needs. Moreover, without being limited to this embodiment, there may be multiple intention tags L set as the exclusive intention tag L', and the multiple exclusive intention tags L' have an order.

[0028] In this embodiment, the control procedures P stored in the storage unit 12 are used by the processing unit 11, which communicates with the processing module 51 of the user side device 5 via a network, to control the user side device 5. Specifically, in this embodiment, each of the control procedures P is used to control the user side device 5 to output specific data in a specific manner, for example, to play voice data, display character data or image data, and output control commands via wireless communication to electronic devices such as an air conditioner, telephone, television, and light in the guest room to operate those electronic devices. Furthermore, in this embodiment, the control procedures P correspond to the intention tags L included in the language processing model M. Specifically, each of the intention tags L corresponds to at least one of the control procedures P. For example, the intention tag L indicating "set an alarm" corresponds to the control procedure P for setting the alarm function of the user side device 5 itself, and the intention tag L indicating "make a phone call" corresponds to the control procedure P in which the user side device 5 sends a control command to a telephone to make a call to a specific person.

[0029] With reference also to FIG. 2, the language data processing method will be described below with respect to cooperation between the language data processing system 1 and the user side device 5.

[0030] In step S1, the processing unit 11 obtains text data corresponding to a speech input.

[0031] Specifically, in this embodiment, the voice input is a voice signal corresponding to the user's voice obtained by the processing module 51 of the user-side device 5 through the microphone of the input module 52, i.e., the voice input indicates a phrase spoken by the user. Upon obtaining the voice input, the processing module 51 of the user-side device 5 generates text data corresponding to the voice input by performing a voice recognition process on the voice input, and then transmits the text data to the language data processing system 1. In another embodiment, the processing module 51 of the user-side device 5 transmits the voice input as it is to the processing unit 11 of the language data processing system 1, and the processing unit 11 performs a voice recognition process on the voice input to generate corresponding text data. Therefore, the manner in which the processing unit 11 obtains the text data is not limited to this embodiment.

[0032] In step S2, the processing unit 11 loads the language processing model M from the storage unit 12, and executes an intention analysis process on the text data using the language processing model M. In this embodiment, by executing the intention analysis process, the processing unit 11 obtains multiple designated intentions represented by the text data, and further determines whether each of the designated intentions is a clear designated intention or an ambiguous designated intention based on the words included in the text data. A word here consists of at least one character. In addition, since the intention analysis process is executed based on the intention tag L, each of the designated intentions corresponds to one of the intention tags L.

[0033] Specifically, for each designated intention, the processing unit 11 determines whether or not the words included in the text data contain a key word that semantically matches the designated intention and can execute a corresponding control procedure P. If it is determined that a key word is present, the processing unit 11 determines that the designated intention is a clear designated intention, and if not, the processing unit 11 determines that the designated intention is an ambiguous designated intention.

[0034] For example, when one of the specified intentions corresponds to the intention tag L indicating “making a phone call”, the processing unit 11 determines whether the words contained in the text data include a word indicating a specific person to call (e.g., a room number in an accommodation facility, a specific store, organization or telephone number), and if a word indicating a specific person to call is included, the processing unit 11 determines that a key word is included by using the word as a key word, and determines that the specified intention is a clear specified intention; if not, the processing unit 11 determines that a key word is not included, and determines that the specified intention is an ambiguous specified intention.

[0035] When one of the specified intentions corresponds to the intention tag L indicating "set an alarm," the processing unit 11 determines whether the words contained in the text data include a word indicating a specific time to set the alarm (for example, "8 a.m.", "one hour later," etc.), and when a word indicating a specific time to set the alarm is included, the processing unit 11 uses the word as a key word, thereby determining that a key word is included and determining that the specified intention is a clear specified intention; when not, the processing unit 11 determines that a key word is not included and determining that the specified intention is an ambiguous specified intention.

[0036] In addition, in this embodiment, each of the intention tags L includes a necessary information feature. The necessary information feature is a semantic feature indicating the necessary information of the user intention represented by the intention tag L. For example, "specific telephone party" is necessary information of the user intention of "making a phone call", and therefore, the necessary information feature of the intention tag L indicating "making a phone call" is a semantic feature indicating "specific telephone party". By using the language processing model M, the processing unit 11 determines whether a key word matching the necessary information feature is present in the text data based on the necessary information feature of the intention tag L corresponding to each of the designated intentions, and thus determines whether the designated intention is a clear designated intention or an ambiguous designated intention. Note that in other embodiments, the determination of whether the designated intention is a clear designated intention or an ambiguous designated intention is performed by a language processing model M trained by machine learning, and is not limited to this embodiment.

[0037] In step S3, the processing unit 11 determines whether any of the designation intentions is an ambiguous designation intention. That is, it determines whether the designation intentions include at least one designation intention that is determined to be an ambiguous designation intention. If the processing unit 11 determines that at least one designation intention is an ambiguous designation intention, the processing proceeds to step S4, whereas if the processing unit 11 determines that none of the designation intentions is an ambiguous designation intention, that is, if there is no designation intention that is determined to be an ambiguous designation intention, the processing proceeds to step S6.

[0038] In step S4, for one of the designated intentions determined to be an ambiguous designated intention, the processing unit 11 uses a language processing model to generate a question message corresponding to the designated intention, and transmits the question message to the processing module 51 of the user-side device 5, which outputs the question message to the output module 53 (speaker). The question message is intended to allow the user to provide further explanation about the designated intention. For example, when the designated intention determined to be an ambiguous designated intention corresponds to the intention tag L indicating "make a phone call", the question message is "Who is the other party on the phone?", and when the designated intention corresponds to the intention tag L indicating "set an alarm", the question message is "What time do you want to set the alarm?", but is not limited thereto.

[0039] Then, in step S5, when the processing unit 11 obtains additional text data related to the question message and corresponding to other voice input within a certain response period after the question message is output (e.g., 5 seconds after the output), it determines whether the additional text data semantically matches the designated intention to which the question message corresponds, and determines whether the control procedure P corresponding to the designated intention to which the question message corresponds can be executed based on the additional text data. Specifically, the processing unit 11 determines whether there is a word that semantically matches the designated intention and can be used as a key word among the words included in the additional text data.

[0040] When the processing unit 11 determines that the additional text data semantically matches the specified intention, the processing unit 11 judges again that the specified intention is a clear specified intention. That is, the processing unit 11 judges again that the specified intention is a clear specified intention based on key words included in the additional text data. On the other hand, when the processing unit 11 determines that the additional text data does not semantically match the specified intention, the processing unit 11 generates, for example, another question message for prompting the user to further explain the specified intention, and transmits it to the user-side device 5 for output. When the processing unit 11 cannot judge again that the specified intention is a clear specified intention even if the number of times that a question message is generated for the same specified intention reaches a predetermined threshold, the processing unit 11 does not continue processing the specified intention, and ignores (discards) the specified intention.

[0041] For example, when the designated intention determined to be an ambiguous designated intention corresponds to the intention tag L indicating "make a phone call" and the additional text data related to the question message "Who is the other party on the phone?" is "Call room 602," the processing unit 11 determines that "room 602" included in the additional text data semantically matches the designated intention, and re-determines that the designated intention is a clear designated intention. Also, when the designated intention corresponds to the intention tag L indicating "set an alarm" and the additional text data related to the question message "What time would you like to set the alarm?" is "let's set it at half past five," the processing unit 11 determines that "half past five" included in the additional text data semantically matches the designated intention, and re-determines that the designated intention is a clear designated intention.

[0042] In step S6, the processing unit 11 executes a contradiction process and an exclusion process for the designation intention that is determined to be a clear designation intention, using the language processing model M, thereby determining at least one designation intention of the target from the designation intention that is determined to be a clear designation intention. In this embodiment, the contradiction process is executed first, and then the exclusion process is executed, but this is not limiting.

[0043] In this embodiment, the conflict processing is performed by the processing unit 11 using the language processing model M to determine whether a designated intention determined to be a clear designated intention includes multiple conflicting designated intentions, and if it is determined that multiple conflicting designated intentions exist, the processing unit 11 groups the multiple conflicting designated intentions into one conflicting intention group, selects one designated intention from each conflicting intention group as a candidate designated intention, and ignores designated intentions other than the selected designated intention. In this embodiment, the processing unit 11 sets a clear designated intention that is expressed first in text data in one conflicting intention group as a candidate designated intention, i.e., ignores clear designated intentions expressed thereafter, but this is not limited to this embodiment.

[0044] The above-mentioned "multiple mutually contradictory designation intentions" indicates that the designation intentions are semantically contradictory to each other, or that the control procedures P corresponding to the designation intentions cannot be executed simultaneously by the processing unit 11. Whether the designation intentions are semantically contradictory to each other can be determined using the natural language understanding of the language processing model M. The control procedures P that cannot be executed simultaneously are set in advance. For example, multiple control procedures P that cause the user side device 5 to output voice in the control procedure P are set as control procedures P that cannot be executed simultaneously.

[0045] For an example where the text data is “Turn up the volume, turn down the volume, stop the music”, the processing unit 11 analyzes the text data into three semantically contradictory clear specified intentions, “Turn up the volume”, “Turn down the volume”, and “Stop the music”, and sets the clear specified intention, “Turn up the volume”, which is expressed first in the text data, as a candidate specified intention, and ignores the two clear specified intentions, “Turn down the volume” and “Stop the music”, which are expressed subsequently.

[0046] For an example where the text data is "play news and music", the processing unit 11 analyzes two clear designation intentions, "play news" and "play music", from the text data. Since the two clear designation intentions are to cause the user side device 5 to output sound, in consideration of the viewing experience, it is inappropriate to simultaneously play news and music on the user side device 5, that is, it is inappropriate to simultaneously execute the control procedures P corresponding to the two clear designation intentions. Therefore, the processing unit 11 determines that the two clear designation intentions, "play news" and "play music", are mutually exclusive, and groups the two clear designation intentions into one mutually exclusive intention group, selects the first clear designation intention of "play news" from the mutually exclusive intention group as a candidate designation intention, and ignores the clear designation intention of "play music" expressed thereafter. Note that in this embodiment, if there is a clear designation intention that does not belong to any mutually exclusive intention group, the processing unit 11 sets it as a candidate designation intention as it is. In this embodiment, after selecting candidate designation intentions for each of the contradictory intention groups, an exclusion process is executed.

[0047] In this embodiment, the exclusive processing is performed by the processing unit 11 determining whether the intention tag L corresponding to the designated intention of the candidate includes an intention tag L that is an exclusive intention tag L', and if it is determined that the intention tag L that is an exclusive intention tag L' is included, the designated intention of the candidate corresponding to the intention tag L that is the exclusive intention tag L' is set as the only target designated intention, and the designated intentions of the other candidates are ignored, and otherwise, the designated intentions of all the candidates are set as the target designated intention. For example, if the exclusive intention tag L' is the intention tag L indicating "make a phone call" and the text data is "call the front desk, and turn on the TV and air conditioner at the same time," the processing unit 11 analyzes three clear designated intentions, "make a phone call," "turn on the TV," and "turn on the air conditioner," from the text data, and performs exclusive processing on the three clear designated intentions, thereby setting the clear designated intention of "make a phone call" as the only target designated intention, and ignoring the two clear designated intentions, "turn on the TV" and "turn on the air conditioner."

[0048] In another embodiment, in step S6, the processing unit 11 may execute only the contradictory process without executing the exclusive process for the designation intention that is determined to be a clear designation intention.

[0049] In step S7, the processing unit 11 executes the control procedures P corresponding to the intention tags L corresponding to the respective intentions to specify the target. In this embodiment, since the intention tags L have an order, when there are multiple intentions to specify the target, the processing unit 11 executes the corresponding control procedures P one by one in accordance with the order of the intention tags L corresponding to the intentions to specify the target, but is not limited to this embodiment.

[0050] For example, in this embodiment, in the case where the text data is “Play the Titanic theme song, check the weather forecast for tomorrow, and turn on the lights”, the processing unit 11 determines the three specified intentions of the goals “Play music”, “Check the weather information”, and “Turn on the lights” from the text data, and then executes the control procedures P corresponding to the intention tags L corresponding to the three specified intentions of the goals according to the order of the intention tags L corresponding to the three specified intentions of the goals. For example, first, the control procedure P corresponding to the specified intention of the goal “Turn on the lights” is executed, and the lights in the cabin are controlled and turned on via the processing module 51 of the user-side device 5; then, the control procedure P corresponding to the specified intention of the goal “Check the weather information” is executed, and the weather information is displayed on the display of the output module 53 of the user-side device 5; and finally, the control procedure P corresponding to the specified intention of the goal “Play music” is executed, and the music file of the Titanic theme song is played on the speaker of the output module 53 of the user-side device 5.

[0051] It should be noted that the target designation intention is always a clear designation intention. In steps S4 and S5 of this embodiment, one of the designation intentions determined to be ambiguous designation intentions is processed (such as generating a question message), and the designation intention is determined to be a clear designation intention again or the designation intention is ignored. In this embodiment, if there is one or more designation intentions determined to be ambiguous designation intentions, the processing unit 11 processes only the ambiguous designation intention that is expressed first in the text data, and ignores the other ambiguous designation intentions. In other embodiments, the processing unit 11 may process the ambiguous designation intention that has the highest priority according to the order of the intention tag L to which the ambiguous designation intention corresponds. In some embodiments, the processing unit 11 may execute steps S4 and S5 multiple times, for example, processing all or some of the designation intentions determined to be ambiguous designation intentions (ignoring the designation intentions that are not processed as they are).

[0052] 2 are merely illustrative of the language data processing method of the present invention, and are not intended to limit the present invention. It should be noted that even if steps S1 to S7 are combined, divided, or rearranged, they are still within the scope of the present invention as long as the technical effect is the same as that of the present embodiment.

[0053] The computer program of the present invention includes the above-mentioned language processing model M, and when executed by an electronic device (e.g., a smartphone, a tablet computer, a notebook computer, a desktop computer, etc.), the electronic device becomes the language data processing system 1 of the present invention and executes the above-mentioned language data processing method.

[0054] In summary, the language data processing system 1, the language data processing method, and the computer program of the present invention analyze the designated intention from the text data, judge whether each of the designated intentions is clear or ambiguous, and execute the corresponding control procedure P only for the designated intentions that are judged to be clear. Therefore, even if some of the designated intentions expressed by the text data are unclear (i.e., even if there is an ambiguous designated intention), the language data processing system 1 executes the corresponding procedure for the clearly expressed designated intention (i.e., clear designated intention) to satisfy the needs of the user. In addition, for the ambiguous designated intention, the language data processing system 1 can generate and output a question message for the user to provide additional explanation. Furthermore, even if the user expresses multiple mutually contradictory needs in one phrase, the language data processing system 1 can deal with them by executing a contradictory process and can respond to at least one of the needs, and will not be unable to operate due to the needs that are semantically contradictory to each other. In this way, the language data processing system 1 improves the shortcomings of the prior art and reliably achieves the object of the present invention.

[0055] In the above description, for the purpose of explanation, numerous specific details are set forth to provide a thorough understanding of the embodiments. However, it will be apparent to one skilled in the art that one or more other embodiments may be implemented without the specific details. In addition, in the description of "one embodiment" or "one embodiment" in this specification, it should be understood that all descriptions accompanied by ordinal numbers or other indications may be included in the specific implementation of the present invention having specific aspects, structures, and features. Furthermore, in this specification, multiple variations are sometimes incorporated into one embodiment, drawing, or description thereof, but this is for the purpose of streamlining the specification and for the purpose of understanding the multiple aspects of the present invention, and one or more features or specific examples in one embodiment may be implemented together with one or more features or specific examples in other embodiments, where appropriate, in the implementation of the present invention.

[0056] Although the embodiments and modified examples of the present invention have been described above, the present invention is not limited to these, and is intended to encompass all modifications and equivalent configurations as various configurations falling within the spirit and scope of the broadest interpretation. [Explanation of symbols]

[0057] 1 Language Data Processing System 11 Processing Unit 12 Storage Unit M Language Processing Model L Intent Tag L' exclusive intent tag P Control Procedure 5 User side equipment 51 Processing Module 52 Input Module 53 Output Module S1-S7 Steps

Claims

1. A processing unit; a storage unit electrically connected to the processing unit; a language processing model realized by a machine learning technique is stored in the storage unit, the language processing model including a plurality of intention tags corresponding to a plurality of control procedures executed by the processing unit; The processing unit includes: Using the language processing model, text data is analyzed to obtain a plurality of designated intentions represented by the text data and each of the designated intentions corresponding to one of the intention tags, and further, based on the text data, it is determined whether each of the designated intentions is a clear designated intention or an ambiguous designated intention; When at least one of the designated intentions is determined to be the ambiguous designated intention, for one of the designated intentions determined to be the ambiguous designated intention, a question message corresponding to the designated intention is generated using the language processing model, the question message is output, and additional text data related to the question message is obtained, then it is determined whether the additional text data semantically matches the designated intention corresponding to the question message, and whether the control procedure corresponding to the designated intention to which the question message corresponds can be executed based on the additional text data, and when it is determined that the additional text data semantically matches the designated intention and it is determined that the control procedure can be executed based on the additional text data, the designated intention is determined to be the clear designated intention, determining a designation intent of at least one target from the designation intents determined to be the clear designation intent; The control procedure corresponding to the intention tag corresponding to the specified intention of the target is executed. Language data processing system.

2. The language data processing system of claim 1, wherein the processing unit is configured to determine whether each of the designated intentions is the clear designated intention or the ambiguous designated intention by determining, for each of the designated intentions, whether a key word that can execute the corresponding control procedure is present in the text data, and if it is determined that the key word is present, determining that the designated intention is the clear designated intention, and if it is determined that the key word is not present, determining that the designated intention is the ambiguous designated intention.

3. The language data processing system of claim 1, wherein the processing unit is configured to determine whether the designated intention judged to be the clear designated intention includes multiple mutually contradictory designated intentions, and when it is judged that there are multiple mutually contradictory designated intentions, to determine the designated intention of the target by selecting one of the multiple mutually contradictory designated intentions to be one of the designated intentions of the target.

4. one of the intent tags is an exclusive intent tag; The language data processing system of claim 1, wherein the processing unit is configured to determine whether the intention tag corresponding to the designated intention judged to be the clear designated intention includes the intention tag which is the exclusive intention tag, and when it is judged that the intention tag which is the exclusive intention tag is included, determine the designated intention of the target by making the designated intention corresponding to the intention tag which is the exclusive intention tag the only designated intention of the target.

5. The plurality of intention tags have an order, The language data processing system of claim 1, wherein the processing unit is configured to execute the control procedure by executing the control procedure corresponding to the intention tag to which the target's designated intention corresponds in accordance with the order of the intention tags to which the target's designated intention corresponds when there are multiple designated intentions determined to be the target's designated intention.

6. A language data processing method executed by a language data processing system, comprising: the language data processing system includes a processing unit and a storage unit electrically connected to the processing unit; a language processing model realized by a machine learning technique is stored in the storage unit, the language processing model including a plurality of intention tags corresponding to a plurality of control procedures executed by the processing unit; The language data processing method includes: a step of the processing unit analyzing text data using the language processing model to obtain a plurality of designated intentions represented by the text data and each of the designated intentions corresponding to one of the intention tags, and further determining whether each of the designated intentions is a clear designated intention or an ambiguous designated intention based on the text data; a step of the processing unit generating a question message corresponding to one of the designated intentions determined to be the ambiguous designated intention using the language processing model when at least one of the designated intentions is determined to be the ambiguous designated intention, outputting the question message, and obtaining additional text data related to the question message, determining whether the additional text data semantically matches the designated intention to which the question message corresponds, and determining whether the control procedure corresponding to the designated intention to which the question message corresponds can be executed based on the additional text data, and determining that the designated intention is the clear designated intention when it is determined that the additional text data semantically matches the designated intention and when it is determined that the control procedure can be executed based on the additional text data; determining, by the processing unit, a designated intent of at least one target from the designated intents determined to be the distinct designated intent; The processing unit executes the control procedure corresponding to the intention tag corresponding to the specified intention of the target. Language data processing methods.

7. The language data processing method of claim 6, wherein the processing unit determines, for each of the designated intentions, whether or not the text data contains a key word that can execute the corresponding control procedure, and if it is determined that the key word is present, determines that the designated intention is the clear designated intention, and if it is determined that the key word is not present, determines that the designated intention is the ambiguous designated intention, thereby determining whether each of the designated intentions is a clear designated intention or an ambiguous designated intention.

8. The language data processing method of claim 6, wherein the step of determining a designated intention of at least one target includes the processing unit determining whether the designated intention judged to be the clear designated intention includes a plurality of mutually contradictory designated intentions, and when it is judged that there are a plurality of mutually contradictory designated intentions, selecting one of the plurality of mutually contradictory designated intentions to be one of the designated intentions of the target.

9. one of the intent tags is an exclusive intent tag; The language data processing method of claim 6, wherein the step of determining a designated intention of at least one target includes the processing unit determining whether the intention tag corresponding to the designated intention determined to be the clear designated intention includes the intention tag that is the exclusive intention tag, and when it is determined that the intention tag that is the exclusive intention tag is included, the designated intention corresponding to the intention tag that is the exclusive intention tag is set as the only designated intention of the target.

10. The plurality of intention tags have an order, The language data processing method of claim 6, wherein the step of executing the control procedure includes, when there are multiple designation intentions determined to be the designation intention of the target, the processing unit executing the control procedure corresponding to the intention tag to which the designation intention of the target corresponds in accordance with the order of the intention tags to which the designation intention of the target corresponds.

11. A computer program executed by an electronic device, comprising: the computer program includes a language processing model implemented by machine learning technology, the language processing model including a plurality of intent tags corresponding to a plurality of control procedures executed by the electronic device; A computer program which, when executed by the electronic device, causes the electronic device to carry out the method according to any one of claims 6 to 10.

Images