Decision-making methods for adaptive single-turn dialogue and task-oriented multi-turn dialogue

Abstract

This invention relates to the field of automated dialogue technology, specifically to a decision-making method for adaptive single-turn dialogue and task-oriented multi-turn dialogue. It includes: conducting dialogue within a user communication unit; a dialogue state memory unit storing the dialogue format, slot information, and dialogue state within the user communication unit to form memory data, which is then output to a dialogue input unit; and the dialogue input unit inputting a dialogue format into the user communication unit that differs from the format stored in the dialogue state memory unit. In this invention, the dialogue state memory unit stores the current dialogue format, facilitating the input of different dialogue formats and ensuring smooth switching between multiple different dialogue formats during the dialogue process, thus guaranteeing the overall fluency of the dialogue.

Description

Technical Field

[0001] This invention relates to the field of automated dialogue technology, and more specifically, to decision-making methods for adaptive single-turn dialogue and task-oriented multi-turn dialogue. Background Technology

[0002] With the advancement of intelligent technology, more and more artificial intelligence is replacing the way people work;

[0003] Artificial intelligence is a branch of computer science that attempts to understand the nature of intelligence and produce a new kind of intelligent machine that can react in a way similar to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing, and expert systems.

[0004] Artificial intelligence was also introduced into the dialogue process. In many cases, different dialogue formats are used to obtain information about user needs in the process of travel or purchasing goods, thereby improving work efficiency. However, there are many forms of dialogue, and multiple dialogue formats may be combined in the same communication process. Without a decision-making method, the information obtained from the communication cannot form a complete dialogue information.

[0005] In addition, current automated dialogues can only address one task topic at a time, and one task must be completed before another can begin; it is not possible to switch topics during the dialogue; when users want to address multiple types of tasks, they can only wait for the current dialogue to be resolved or for it to time out before restarting the topic, resulting in a large number of repetitive dialogues and operations. In view of this, we propose a decision-making method for adaptive single-turn dialogue and task-oriented multi-turn dialogue. Summary of the Invention

[0006] The purpose of this invention is to provide a decision-making method for adaptive single-turn dialogue and task-oriented multi-turn dialogue to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, an adaptive single-turn dialogue and task-oriented multi-turn dialogue decision-making method is provided, including the following method steps:

[0008] S1. Conduct dialogue within the user communication unit;

[0009] S2. The dialogue state memory unit remembers the dialogue format, slot information and dialogue state in the user communication unit, forms memory data, and outputs it to the dialogue input unit.

[0010] S3. The dialogue input unit inputs a dialogue format into the user communication unit that is different from the dialogue state memory unit outputting the memory format, and simultaneously inputs the corresponding memory data.

[0011] As a further improvement to this technical solution, the dialogue format in S2 includes single-turn dialogue and task-based dialogue.

[0012] As a further improvement to this technical solution, the single-turn dialogue format includes the following method steps:

[0013] S2.1.1 Identify the user's dialogue intent;

[0014] S2.1.2 Match the appropriate slot information template according to the user's dialogue intent;

[0015] S2.1.3 Obtain slot values ​​by communicating with the user;

[0016] S2.1.4. Fill the corresponding slot with the slot value obtained from the user dialogue to obtain a complete dialogue information.

[0017] As a further improvement to this technical solution, in S2.1.4, the slots that have not been filled with the corresponding slot value are formed as empty slots. If an empty slot appears, S2.1.3 is repeated to continue the dialogue with the user to obtain the slot value of the empty slot, and the obtained slot value is filled into the corresponding empty slot.

[0018] As a further improvement to this technical solution, the task-oriented dialogue format includes the following method steps:

[0019] S2.2.1 Identify the user's dialogue intent;

[0020] S2.2.2 Generate a finite state tree corresponding to the user's dialogue intent;

[0021] S2.2.3 The user communication unit conducts dialogue according to a finite state tree;

[0022] S2.2.4 During the dialogue, the dialogue state nodes are determined according to the finite state tree, and then the response information is matched according to the dialogue state nodes to obtain a complete dialogue information.

[0023] As a further improvement to this technical solution, the dialogue state memory unit in S2 adopts a detection-based memory algorithm, the steps of which are as follows:

[0024] S2.3.1 Real-time detection of dialogue formats within user communication units;

[0025] S3.3.2. Compare the current dialogue format with the current memory format. The results include:

[0026] The dialogue format is the same;

[0027] The dialogue formats are different;

[0028] S2.3.3 When the dialogue format is different, memorize it and replace the currently memorized dialogue format.

[0029] As a further improvement to this technical solution, in S2.3.1, while the dialogue form is detected in real time, the dialogue state memory unit also detects the completion status of the dialogue information, and then memorizes the completed dialogue information until the user communication unit ends the dialogue.

[0030] As a further improvement to this technical solution, the dialogue state memory unit automatically restores the remembered dialogue information when the dialogue is restarted after the user communication unit has abnormally ended the dialogue.

[0031] As a further improvement to this technical solution, the dialogue state memory unit automatically restores the dialogue state and generates a recovery list, on which the completed dialogue information is modified.

[0032] As a further improvement to this technical solution, the dialogue state memory unit directly fills the completed dialogue information into the corresponding slot or dialogue state node in both single-turn dialogue and task-based dialogue formats.

[0033] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0034] In this adaptive single-turn dialogue and task-oriented multi-turn dialogue decision-making method, the current dialogue form is memorized through the dialogue state memory unit, which facilitates the input of different dialogue forms, thereby ensuring the smoothness of the entire dialogue by switching between multiple different dialogue forms during the dialogue process. Attached Figure Description

[0035] Figure 1 This is a flowchart illustrating the overall method steps of the present invention.

[0036] Figure 2 This is a flowchart of the steps of the single-turn dialogue method of the present invention;

[0037] Figure 3 This is a flowchart of the steps of the task-oriented dialogue method of the present invention;

[0038] Figure 4 This is a flowchart of the detection-based memory algorithm method steps of the present invention. Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] Example 1

[0041] Please see Figure 1 As shown, the purpose of this embodiment is to provide a decision-making method for adaptive single-turn dialogue and task-oriented multi-turn dialogue, including the following method steps:

[0042] First, a dialogue takes place within the user communication unit; then, the dialogue state memory unit memorizes the dialogue format, slot information, and dialogue state within the user communication unit, forming memory data, and outputs it to the dialogue input unit; next, the dialogue input unit inputs a dialogue format into the user communication unit that is different from the format output by the dialogue state memory unit, and simultaneously inputs the corresponding memory data.

[0043] The dialogue formats specifically include single-turn dialogues and task-based dialogues.

[0044] Suppose that the user is interacting with the computer through a single-turn dialogue. The dialogue state memory unit remembers the dialogue format, slot information, and dialogue state within the user's communication unit, forming memory data. The remembered dialogue format is a single-turn dialogue. Then, during the interaction, a task-type dialogue format appears. Since the task-type dialogue format is different from the remembered dialogue format, the dialogue input unit inputs the task-type dialogue format into the user's communication unit. The single-turn dialogue format can only continue after the task-type dialogue format ends. In other words, the task-type dialogue ends. Alternatively, the user triggers a single-turn dialogue again. In this case, the task-type dialogue format changes to a single-turn dialogue format, and the information in the task-type dialogue is saved. If the information in the task-type dialogue updates the dialogue information of the single-turn dialogue, it returns to the single-turn dialogue format. These two dialogue formats can be switched at any time, and the slot information is interconnected.

[0045] If the communication process involves a single-turn dialogue, then normal communication is sufficient.

[0046] Example 2

[0047] This embodiment specifically discloses a single-turn dialogue format; please refer to [link / reference needed]. Figure 2 As shown, the single-turn dialogue format includes the following steps:

[0048] First, the user's dialogue intent is identified, and then a suitable slot information template is matched according to the user's dialogue intent. After the information template is obtained, the user communication unit obtains the slot value through dialogue with the user, and then fills the slot value obtained from the user dialogue into the corresponding slot to obtain a complete dialogue information.

[0049] This example illustrates the concept using a dialogue scenario of buying insurance:

[0050] First, the user communication unit sends "I want to buy insurance", and then matches the slot information template: "Age-Type of insurance-Whether you have social security";

[0051] Continue to ask through the user communication unit: "What is your age? Do you already have social security? Which type of insurance do you want to buy, health insurance or property insurance?"

[0052] The user replied, "I am 40 years old and already have social security."

[0053] Then, the user interaction unit can search for "40 years old - with social security". This retrieves the age and social security eligibility information, but the corresponding insurance slot is not yet filled, meaning the insurance slot is empty. The interaction then continues with the user, for example:

[0054] The user communication unit asked, "Which type of insurance do you want to buy: health insurance or property insurance?"

[0055] When a user answers "I want to buy health insurance," the system obtains the slot value "buy health insurance" for the insurance type. This slot value is then filled into the corresponding empty slot (i.e., the insurance type slot). The user's communication unit can then search for a list of insurance options for "40 years old - buy health insurance - have social security" for the user to choose from. This ensures the completeness of the dialogue information, narrows down the search scope later, and improves search accuracy.

[0056] Example 3

[0057] This embodiment specifically discloses the task-based dialogue format; please refer to [link / reference]. Figure 3 As shown, the task-based dialogue format includes the following steps:

[0058] First, the user's dialogue intent is identified, and a finite state tree corresponding to the user's dialogue intent is generated. Then, the user communication unit conducts dialogue according to the finite state tree. During the dialogue, the dialogue state nodes are determined according to the finite state tree, and then the response information is matched through the dialogue state nodes to obtain a complete dialogue information.

[0059] Using insurance as an example again, the specific principle is as follows:

[0060] First, obtain the dialogue intent. For example, if a user asks, "Can I buy insurance A if I have high blood pressure?", the first step is to obtain the intent "health disclosure during purchase consultation". The first step is to determine if the user has the following diseases or symptoms: high blood pressure (specific categories include past medical history, family medical history, current medical history, physical examination records, etc.). Then, determine the specific status. If the user has been examined and diagnosed with high blood pressure in the past two years, then determine the level. The high blood pressure level is 1.

[0061] So: Hypertension - Examination and Diagnosis - Level 1 is the information in the current dialogue, which further narrows down the list of insurance products that can be purchased. The dialogue information formed in the task-based dialogue format will not be incomplete, because the whole process is carried out step by step, and the next step can only be carried out after the previous step is completed.

[0062] Example 4

[0063] Considering that if the dialogue state memory unit were to remember all dialogue formats during the communication process, a high amount of memory would be required. To reduce the storage memory required during the communication process, please refer to [link to relevant documentation]. Figure 4 As shown, the dialogue state memory unit uses a detection-based memory algorithm, and its algorithm steps are as follows:

[0064] First, the dialogue format within the user's communication unit is detected in real time. Then, the current dialogue format is compared with the currently remembered format. The results include: the dialogue format is the same.

[0065] The dialogue formats are different;

[0066] Finally, when the dialogue format is different, it is memorized and the currently memorized dialogue format is replaced.

[0067] Assumption: The current dialogue format is a single-turn dialogue. During real-time detection, the dialogue format is only remembered and replaced when a task-oriented dialogue format appears. At this time, the current dialogue format is a task-oriented dialogue format. The above steps are repeated to continue detection. In this way, the dialogue format is remembered when it changes, which greatly reduces the number of times it needs to be remembered and reduces the storage burden caused by remembering.

[0068] Example 5

[0069] Considering that the ongoing dialogue may be canceled due to special circumstances, in order to avoid the clearing of previously completed dialogue information, this embodiment detects the completion status of dialogue information in real time while detecting the dialogue form, and then remembers the completed dialogue information until the user communication unit ends the dialogue, thereby preventing the completed dialogue information from being cleared.

[0070] Furthermore, if the user's communication unit abnormally ends the conversation (e.g., the phone is turned off, or the user accidentally returns everything while modifying information), the conversation state memory unit will automatically restore the remembered conversation information when the conversation is restarted, and then the conversation can continue. This completes the subsequent information replenishment, reduces the time spent manually re-entering information, and greatly improves the efficiency of the conversation.

[0071] In addition, the dialogue state memory unit automatically restores the dialogue state and generates a restoration list. The restored dialogue information can be modified on the restoration list, which makes it convenient for users to modify the previous information.

[0072] Example 6

[0073] Considering that the same information requirement may occur in both single-turn and task-based dialogue formats during the dialogue process, the dialogue state memory unit can directly fill the completed dialogue information into the corresponding slot or dialogue state node in both single-turn and task-based dialogue formats, thereby further improving the efficiency of the dialogue.

[0074] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A decision-making method for adaptive single-turn dialogue and task-oriented multi-turn dialogue, characterized in that, The methods and steps include the following: S1. Conduct dialogue within the user communication unit; S2. The dialogue state memory unit remembers the dialogue format, slot information and dialogue state in the user communication unit, forms memory data, and outputs it to the dialogue input unit. S3. The dialogue input unit inputs a dialogue format into the user communication unit that is different from the output memory format of the dialogue state memory unit, and simultaneously inputs the corresponding memory data. The dialogue formats in S2 include single-turn dialogue and task-based dialogue. Users interact with the computer through single-turn dialogue. The dialogue state memory unit remembers the dialogue format, slot information, and dialogue state within the user's communication unit, forming memory data. At this point, the remembered dialogue format is single-turn dialogue. Then, during the interaction, a task-type dialogue format appears. The dialogue input unit inputs the task-type dialogue format into the user's communication unit. Single-turn dialogue can only continue after the task-type dialogue ends. If the user triggers a single-turn dialogue again, the task-type dialogue format changes to single-turn dialogue, and the information in the task-type dialogue is saved. If the information in the task-type dialogue updates the dialogue information in the single-turn dialogue, it returns to single-turn dialogue. These two dialogue formats can be switched at any time, and the slot information is interconnected. The task-oriented dialogue format includes the following steps: S2.2.1 Identify the user's dialogue intent; S2.2.2 Generate a finite state tree corresponding to the user's dialogue intent; S2.2.3 The user communication unit conducts dialogue according to a finite state tree; S2.2.4 During the dialogue, the dialogue state nodes are determined according to the finite state tree, and then the response information is matched according to the dialogue state nodes to obtain a complete dialogue information; The dialogue state memory unit in S2 adopts a detection-based memory algorithm, the steps of which are as follows: S2.3.1 Real-time detection of dialogue formats within user communication units; S3.3.

2. Compare the current dialogue format with the current memory format. The results include: The dialogue format is the same; The dialogue formats are different; S2.3.

3. When the dialogue format is different, memorize it and replace the currently memorized dialogue format; When a user's communication unit abnormally ends a conversation and then restarts the conversation, the conversation state memory unit automatically restores the stored conversation information. Simultaneously, the conversation state memory unit generates a recovery list and modifies the completed conversation information on this list. In both single-turn and task-based conversation formats, the conversation state memory unit directly fills the completed conversation information into the corresponding slot or conversation state node.

2. The decision-making method for adaptive single-turn dialogue and task-oriented multi-turn dialogue according to claim 1, characterized in that: The single-turn dialogue format includes the following steps: S2.1.1 Identify the user's dialogue intent; S2.1.2 Match the appropriate slot information template according to the user's dialogue intent; S2.1.3 Obtain slot values ​​by communicating with the user; S2.1.

4. Fill the corresponding slot with the slot value obtained from the user dialogue to obtain a complete dialogue information.

3. The decision-making method for adaptive single-turn dialogue and task-oriented multi-turn dialogue according to claim 2, characterized in that: In step S2.1.4, slots that have not been filled with the corresponding slot value are empty slots. If an empty slot appears, step S2.1.3 is repeated to continue the dialogue with the user to obtain the slot value of the empty slot and fill the corresponding empty slot with the obtained slot value.

4. The decision-making method for adaptive single-turn dialogue and task-oriented multi-turn dialogue according to any one of claims 1-3, characterized in that: In S2.3.1, while the dialogue format is detected in real time, the dialogue state memory unit also detects the completion status of the dialogue information, and then remembers the completed dialogue information until the user communication unit ends the dialogue.

Images