Dialogue Method and System Based on Complex Task Analysis

Abstract

A dialogue method based on complex task analysis, including analyzing the user's dialogue, extracting keywords from the dialogue, determining which complex tasks the keywords are associated with according to the knowledge graph and obtaining multiple target subtasks having a connection relationship therewith, searching for matching product service data for each target subtask, outputting information corresponding to the product service data matching each target subtask, receiving the next round of dialogue, and selectively adjusting one or more of the target complex tasks and target subtasks in the knowledge graph according to the next round of dialogue. The knowledge graph includes a plurality of pre-stored complex tasks and a plurality of pre-stored subtasks, there is a connection relationship between each pre-stored subtask and at least one pre-stored complex task, and the target complex task is one of the pre-stored complex tasks.

Description

Technical Field

[0001] The present invention relates to a dialogue method, and particularly to a dialogue method based on complex task analysis. Background Art

[0002] With the progress of technology and the prosperity of the economy, people can receive a vast amount of information every day, among which various news accounts for a quite high proportion. However, the pace of modern life is very fast, and modern people in the middle of it are difficult to efficiently obtain useful and beneficial information for themselves in their busy daily lives.

[0003] In recent years, chatbots have provided a way for modern people to obtain the information they want. Just like a virtual assistant, users can ask questions to the chatbot through voice or text, such as the weather, stock prices, etc., and the chatbot will search for the current weather, stock prices, etc. according to the question and respond to the user. However, this one-question-one-answer dialogue method still has deficiencies in terms of information provision efficiency and comprehensiveness. Summary of the Invention

[0004] In view of the above, the present invention provides a dialogue method and system based on complex task analysis.

[0005] According to an embodiment of the present invention, a dialogue method based on complex task analysis includes obtaining keyword vocabulary, determining, according to a knowledge graph, that the keyword vocabulary is associated with a target complex task and obtaining a plurality of target subtasks having a connection relationship with the target complex task, searching for matching product service data for each target subtask, outputting output information corresponding to the product service data matching each target subtask, receiving response information in response to the output information, and selectively adjusting one or more of the target complex task and the target subtasks in the knowledge graph according to the response information. The knowledge graph includes a plurality of pre-stored complex tasks and a plurality of pre-stored subtasks, each pre-stored subtask has a connection relationship with at least one pre-stored complex task, and the target complex task is one of the pre-stored complex tasks.

[0006] A dialogue system based on complex task analysis according to an embodiment of the present invention includes a keyword extraction device, a memory, a dialogue device, and a natural language processing device, wherein the natural language processing device is connected to the keyword extraction device, the memory, and the dialogue device. The keyword extraction device is used to obtain keywords. The memory stores a knowledge graph, wherein the knowledge graph includes a plurality of pre-stored complex tasks and a plurality of pre-stored subtasks, and there is a connection relationship between each pre-stored subtask and at least one pre-stored complex task, and the target complex task is one of the pre-stored complex tasks. The dialogue device is used to provide output information and receive response information in response to the output information. The natural language processing device is used to determine, according to the knowledge graph, that the keyword is associated with the target complex task and obtain a plurality of target subtasks having a connection relationship with the target complex task, find matching product service data for each target subtask, output, through the dialogue device, the output information corresponding to the product service data matching each target subtask, and selectively adjust one or more of the target complex task and the target subtasks in the knowledge graph according to the response information.

[0007] With the above structure, the dialogue method and system based on complex task analysis disclosed in this case can automatically find multiple associated subtasks based on a single semantic information and provide corresponding information, thereby enhancing the diversity of information provision and enabling users to efficiently obtain useful information.

[0008] The above description of the disclosure content and the following description of the embodiments are used to illustrate and explain the spirit and principle of the present invention and provide a further explanation of the claims of the present invention. Description of the Drawings

[0009] Figure 1 It is a functional block diagram of a dialogue system based on complex task analysis shown according to an embodiment of the present invention.

[0010] Figure 2A It is a schematic diagram of the knowledge graph of a dialogue system based on complex task analysis shown according to an embodiment of the present invention.

[0011] Figure 2B It is a schematic diagram of the knowledge graph of a dialogue system based on complex task analysis shown according to another embodiment of the present invention.

[0012] Figure 3 It is a flowchart of a dialogue method based on complex task analysis shown according to an embodiment of the present invention. Detailed Embodiments

[0013] The detailed features and advantages of the present invention are described in detail in the following embodiments. The content is sufficient for those skilled in the art to understand the technical content of the present invention and implement it accordingly. According to the content disclosed in this specification, the claims, and the drawings, those skilled in the art can easily understand the related purposes and advantages of the present invention. The following embodiments further illustrate the viewpoints of the present invention in detail, but do not limit the scope of the present invention in any way.

[0014] Please refer to Figure 1 , Figure 1 FIG. is a functional block diagram of a dialogue system based on complex task analysis according to an embodiment of the present invention. Among them, the so-called complex task refers to a task associated with multiple subtasks. For example, "preparing for marriage" can be associated with subtasks such as "venue rental", "purchasing wedding dresses in advance", and "wedding invitation design". Therefore, "preparing for marriage" is a type of complex task. As Figure 1 shown, the dialogue system 1 includes a keyword extraction device 11, a dialogue device 13, a memory 15, and a natural language processing device 17. The natural language processing device 17 can be connected to the keyword extraction device 11, the dialogue device 13, and the memory 15 in a wired or wireless manner.

[0015] The keyword extraction device 11 is used to obtain keywords and provide them to the natural language processing device 17. The natural language processing device 17 performs complex task analysis on the keywords obtained by the keyword extraction device 11 based on the data stored in the memory 15: determining the target complex task associated with the keywords, and obtaining multiple target subtasks having a connection relationship with this target complex task; for each target subtask, searching for one or more matching product / service data; and outputting, through the dialogue device 13, the output information corresponding to one or more product / service data matching each target subtask. After outputting the output information, the dialogue device 13 can receive, from the outside, response information in response to the output information, and the natural language processing device 17 can selectively adjust one or more of the target complex task and the target subtasks in the knowledge graph based on this response information.

[0016] The following is a description of one or more implementation aspects of each device in the system. In one implementation aspect, the keyword extraction device 11 includes a voice input component and a voice recognizer connected to each other. The voice input component is, for example, a sound receiver such as a microphone, and is used to receive a voice signal. Herein, the so-called voice signal is, for example, formed by a sound waveform emitted by a user. In this implementation aspect, the input of the voice signal can be used as a trigger action. When the voice input component receives the input of the voice signal, the voice recognizer is triggered to perform voice recognition on the voice signal to extract keywords from the voice signal. In another implementation aspect, the keyword extraction device 11 includes input components such as a keyboard, a mouse, or / and a touchpad, and a processor. The input components can be triggered by actions such as tapping, clicking, and swiping to determine the information corresponding to the received actions, and the processor then analyzes this information to obtain it. In yet another implementation aspect, when the time that the web page presenting an article (such as an online news article, a blog article, etc.) is in a browsed state exceeds a preset time, the keyword extraction device 11 obtains keywords from the article. The keywords are, for example, the title of the article, the hashtag of the article, etc. Specifically, the keyword extraction device 11 can be implemented by a processor in a computer for web page control. When the web page presenting the article is at the top layer of the display screen, the keyword extraction device 11 determines that the article is in a browsed state and starts timing. When the time that the web page presenting the article is at the top layer of the display screen exceeds the preset time, the extraction of keywords is performed.

[0017] As described above, the dialogue device 13 can output output information corresponding to the plurality of product service data, and can receive response information in response to the output information. Further, the dialogue device 13 includes an output component for outputting information and an input component for inputting information. The output component is, for example, a display, which can present information in the form of text or pictures on the screen of the display, or, for example, a speaker, which can output information in the form of sound. The input component is, for example, a keyboard, a mouse, a touchpad, and can be triggered by actions such as tapping, clicking, and swiping to determine the information corresponding to the received actions. For another example, the input component can be a sound receiver such as a microphone, which is used to receive a voice signal as response information. For yet another example, the input component can include a camera or a combination of a sound receiver and a camera, which can capture an image of the mouth to recognize the mouth shape, and then determine or assist in identifying the voice.

[0018] In another embodiment, the output component and the input component can be implemented by a touch screen. In yet another embodiment, the output component and the input component can be implemented by a wired or wireless connection interface, which can be connected to an external device (such as a mobile phone, a tablet, a personal computer, etc.) to transmit text messages, e-mails, chat room information, etc. corresponding to the plurality of commodity service data. Specifically, in addition to the output component and the input component, the dialogue device 13 can further include a natural language generation module, which can be implemented by a central processing unit, a microcontroller, a programmable logic controller or other processors, or software run by the above processors. The natural language generation module can generate information such as text or sound conforming to natural language based on the commodity service data, and then output it through the output component.

[0019] The memory 15 can be composed of one or more non-volatile storage media (such as flash memory, read-only memory, magnetic memory, etc.) to store data required for complex task analysis. As Figure 1 shown, the memory 15 can include a knowledge graph 151 and a commodity service database 153. The knowledge graph 151 can include a plurality of pre-stored complex tasks and a plurality of pre-stored subtasks, and there is a connection relationship between each pre-stored subtask and at least one of the plurality of pre-stored complex tasks. The knowledge graph 151 can be constructed based on a plurality of news corpora and blog corpora. Further, the knowledge graph 151 can establish a connection relationship between a plurality of complex tasks and a plurality of subtasks by analyzing words in news articles, blog articles, etc. Among them, the construction method and examples of the knowledge graph 151 will be described later.

[0020] In addition to the knowledge graph 151, the memory 15 can also include a commodity service database 153. The commodity service database 153 stores a plurality of commodity service data, and each commodity service data can have a label of a pre-stored subtask that it matches. Further, in addition to the label of the pre-stored subtask that it matches, the commodity service data can also have a label of the consumption demand pattern to which it belongs (such as a privilege pattern, a reliable pattern, and an expectation pattern).

[0021] The natural language processing device 17 can include one or more central processing units, microcontrollers, programmable logic controllers or other processors. The natural language processing device 17 can perform complex task analysis and matching of commodity service data on the keyword vocabulary obtained by the keyword extraction device 11. Further, the complex task analysis and the matching of commodity service data can be respectively executed by different central processing units, microcontrollers, programmable logic controllers or other processors, or can be multiple software run by the processors.

[0022] In particular, the dialogue device 13 can share an input element with the aforementioned keyword extraction device 11. The keyword extraction device 11 can receive user dialogues from the input element, extract keywords from them, and then provide the keywords to the natural language processing module 17. The natural language processing module 17 performs complex task analysis on the keywords based on the data stored in the memory 15 to determine which complex tasks and matching commodity service data the keywords are associated with, and outputs information corresponding to the commodity service data through the output element of the dialogue device 13. After outputting the information, the input element can receive response information for the next round of dialogue from the outside, and the natural language processing device 17 can selectively adjust one or more of the target complex tasks and target subtasks in the knowledge graph based on the response information.

[0023] Further, the knowledge graph 151 stored in the memory 15 can be constructed by two submodules in the natural language processing device 17 or other processors. The two submodules include a related subtask detection submodule (relatedsubtask identification) and a complex task name generation submodule (complex task name generation), which can be implemented by two processors respectively, or software run by the same processor. The related subtask detection submodule can use syntactic analysis tools to cut sentences in news, microblogs, community question answering services (CQA) and other contents into several words and phrases, and then use simple template rules to extract candidate pairs of subtasks. The related subtask detection submodule continuously collects a large amount of blog corpus to train a word vector (Word2Vector) with semantic concepts, and then expands it into a subtask vector (Subtask2Vector). The complex task name generation submodule can use CQA and microblog containing related subtasks to generate complex task names, especially using the topic-event-based complex task model to generate a complex task structure containing subtasks, and form a knowledge graph with multiple complex task structures151.

[0024] Please refer to Figure 1 , Figure 2A and Figure 2B ,in Figure 2A is a schematic diagram of a knowledge graph 151 of a dialog system 1 based on complex task analysis according to an embodiment of the present invention. Figure 2B It is a schematic diagram of a knowledge graph 151 of a dialog system 1 based on complex task analysis according to another embodiment of the present invention. Figure 2AExemplarily, one of the pre-stored complex tasks CT1 in the knowledge graph 151 and the pre-stored subtasks ST1, ST3, and ST5 having a connection relationship with this pre-stored complex task CT1 are illustrated. In Figure 2A the embodiment, the pre-stored complex task CT1 and its pre-stored subtasks ST1, ST3, and ST5 are in a tree structure, where each pre-stored subtask has a connection relationship with a single pre-stored complex task. Figure 2B The knowledge graph 151 in Figure 2B shows a more complex distribution structure, where the pre-stored subtasks can have connection relationships with multiple pre-stored complex tasks at the same time. As Figure 2B shown, the knowledge graph 151 stores the pre-stored complex tasks CT2 and CT4, where the pre-stored complex task CT2 has a connection relationship with the pre-stored subtasks ST2, ST4, ST6, and ST8, and the pre-stored complex task CT4 has a connection relationship with the pre-stored subtasks ST2 and ST4. In

[0025] In one embodiment, each pre-stored complex task in the knowledge graph 151 has a consumption demand pattern ratio; and in another embodiment, each pre-stored subtask in the knowledge graph 151 has a consumption demand pattern ratio. Among them, the consumption demand pattern ratio is stored in the form of tags, for example. Further, the consumption demand pattern ratio indicates the ratio between multiple consumption demand patterns, and the multiple consumption demand patterns include a Special-privilege pattern, a reliable pattern, and an expected pattern. Specifically, the consumption demand pattern is defined based on Maslow's hierarchy of needs. Further, the definition of the special-privilege pattern is the same as the physiological needs in Maslow's hierarchy of needs, the definition of the reliable pattern is the same as the safety needs in the hierarchy of needs, and the expected pattern combines the belongingness and love needs, esteem needs, and self-actualization needs in the hierarchy of needs. For example, for the pre-stored complex task of "preparing for marriage", its consumption demand pattern ratio can be: the special-privilege pattern accounts for 7.8%, the reliable pattern accounts for 26.7%, and the expected pattern accounts for 65.5%. Another example is that for the pre-stored subtask of "buying a mobile phone", its consumption demand pattern ratio can be: the special-privilege pattern accounts for 51.1%, the reliable pattern accounts for 38.5%, and the expected pattern accounts for 10.4%.

[0026] Please refer to Figure 3 , Figure 3 which is a flowchart of a dialogue method based on complex task analysis according to an embodiment of the present invention. As Figure 3 shown, the dialogue method based on complex task analysis includes step S11: obtaining keyword vocabulary; step S12: judging, according to the knowledge graph, that the keyword vocabulary is associated with a target complex task, and obtaining multiple target subtasks having a connection relationship with the target complex task; step S13: for each target subtask, searching for one or more matching commodity service data; step S14: outputting output information corresponding to the one or more commodity service data matching each target subtask; step S15: receiving response information in response to the output information; and step S16: selectively adjusting one or more of the target complex task and the target subtasks in the knowledge graph according to the response information.

[0027] Please refer to together Figure 1 and Figure 3 , Figure 3 The dialogue method shown is applicable to Figure 1 the dialogue system 1 shown. The following describes it exemplarilyFigure 1 The dialogue system 1 executes Figure 3 the dialogue method, however, the present invention does not limit Figure 3 the dialogue method to be only applicable to Figure 1 the shown dialogue system 1.

[0028] In step S11, the keyword extraction device 11 of the dialogue system 1 can obtain keywords. Further, the keyword extraction device 11 can be triggered by a triggering action associated with semantic information and extract keywords from the semantic information. In one embodiment, the semantic information is voice input and the triggering action is the input voice signal; in another embodiment, the semantic information is an article and the triggering action is the situation where the time for browsing the web page presenting the article exceeds a preset time; in yet another embodiment, the semantic information is a dialogue input by the user and the triggering action is the input action.

[0029] In step S12, the natural language processing device 17 of the dialogue system 1 can, according to the knowledge graph 151, determine that the keyword is associated with a target complex task and obtain multiple target subtasks having a connection relationship with the target complex task. Further, the natural language processing device 17 can determine whether the keyword conforms to one of the multiple pre-stored complex tasks and multiple pre-stored subtasks stored in the knowledge graph 151, which is called a conformer, and thereby determine the target complex task, and then use the pre-stored subtasks having a connection relationship with the target complex task as the target subtasks. Even further, the natural language processing device 17 can determine whether there are words identical to the pre-stored complex tasks or pre-stored subtasks in the keyword. When the conformer is one of the pre-stored complex tasks, the natural language processing device 17 takes this conformer as the target complex task; and when the conformer is one of the pre-stored subtasks, the natural language processing device 17 takes the pre-stored complex task having a connection relationship with this conformer as the target complex task. After determining the target complex task, the natural language processing device 17 will, according to the knowledge graph 151, use the pre-stored subtasks having a connection relationship with the target complex task as the target subtasks.

[0030] For Figure 2ATaking the knowledge graph 151 shown as an example, when the keyword contains the word "preparing to get married", the natural language processing device 17 will determine that the keyword is associated with the pre-stored complex task CT1, take it as the target complex task, and obtain the pre-stored subtasks ST1, ST3, and ST5 that have a connection relationship with this target complex task, and use these three pre-stored subtasks ST1, ST3, and ST5 as the target subtasks; and when the keyword contains the word "venue rental", the natural language processing device 17 will determine that the keyword is associated with the pre-stored subtask ST1, and take the pre-stored complex task CT1 that has a connection relationship with the pre-stored subtask ST1 as the target complex task, and then use the pre-stored subtasks ST1, ST3, and ST5 that have a connection relationship with it as the target subtasks.

[0031] In addition, when the conformer is a pre-stored subtask and there are multiple pre-stored complex tasks in the knowledge graph that have a connection relationship with this conformer, the natural language processing device 17 can select the one with the highest historical usage ratio among these pre-stored complex tasks (candidate complex tasks) as the target complex task. Taking Figure 2B the knowledge graph 151 shown as an example, when the keyword contains the word "air ticket", the natural language processing device 17 will determine that the keyword is associated with the pre-stored subtask ST2, and those having a connection relationship with the pre-stored subtask ST2 include the pre-stored complex tasks CT2 and CT4. Therefore, the natural language processing device 17 can determine which of the pre-stored complex tasks CT2 and CT4 has a higher historical usage ratio and take it as the target complex task. In another implementation manner, the natural language processing device 17 can also output options of the pre-stored complex tasks CT2 and CT4 through the dialogue device 13 (for example, output button icons for the two options), and take the triggered one (for example, the clicked button icon) as the target complex task. Or, the natural language processing device 17 can randomly select the pre-stored complex task CT2 or CT4 as the target complex task.

[0032] In step S13, the natural language processing device 17 searches for one or more pieces of product / service data that match each target subtask. Further, the natural language processing device 17 can search for the product / service data that matches the target subtask from the product / service database 153 in the memory 15. As described above, each piece of product / service data in the product / service database 153 can have the label of the pre-stored subtask it matches. The natural language processing device 17 can obtain the product / service data that matches the target subtask by searching for the label. In addition, when there is no match in the product / service database 153, the natural language processing device 17 can also search for the matching product / service data from the network search result snippet or blog article.

[0033] In step S14, the natural language processing device 17 outputs, through the dialogue device 13, output information corresponding to one or more product service data items that match each target subtask. The output information may be in the form of text, sound, etc. For example, the natural language processing device 17 may output information such as the brand, style, price, etc. of the product service data item or the ordering website in the form of chat room information, text messages, or emails. In steps S15 and S16, the natural language processing device 17 may receive, through the dialogue device 13, response information in response to the output information, and selectively adjust one or more of the target complex task and the target subtasks in the knowledge graph according to the response information. Specifically, the user may input, through the dialogue device 13, response information on whether the selected product service data item is appropriate.

[0034] For example, the output information corresponding to the product service data item may include, in addition to information such as the brand, style, price, etc. of the product service data item or the ordering website, an option on whether the recommendation for the target subtask corresponding to this product service is needed. When the natural language processing device 17 receives a response information indicating "not needed", the natural language processing device 17 may save and delete the connection relationship between this target subtask and the target complex task in the knowledge graph as a dedicated knowledge graph belonging to the user who provides the "not needed" response information, thereby providing the function of customizing the knowledge graph. When the number of "not needed" response information received by the natural language processing device 17 for a specific target subtask reaches a specific quantity, the connection relationship between the target subtask and the target complex task in the knowledge graph 151 will be deleted. Additionally, when the natural language processing device 17 receives "not needed" information corresponding to multiple target subtasks simultaneously, the natural language processing device 17 may determine that the target complex task judged in the previous step S12 is incorrect, re-judge the target complex task, and lower the historical usage ratio of the originally selected target complex task.

[0035] In addition to searching for matching product and service data by the tags of the pre-stored subtasks to which it belongs, the natural language processing device 17 can also search for product and service data that matches the target subtask based on the consumption demand pattern ratio, so that the provided product and service data can be closer to the consumer psychology. As mentioned above, the pre-stored complex tasks or pre-stored subtasks in the knowledge graph 151 can have consumption demand pattern ratios. In an embodiment where each pre-stored subtask has a consumption demand pattern ratio, after the natural language processing device 17 finds the product and service data with the target subtask tag, it can further screen the product and service data according to the consumption demand pattern with the highest ratio in the consumption demand pattern ratio of the target subtask. For example, when the ratio of the reliable pattern in the consumption demand pattern of the target subtask is the highest, the natural language processing device 17 will select the product and service data that has the target subtask tag and the reliable pattern tag. Alternatively, the natural language processing device 17 can determine the order of presenting the product and service data according to the ratio of each consumption demand pattern in the consumption demand pattern ratio.

[0036] In an embodiment where each pre-stored complex task has a consumption demand pattern ratio, the natural language processing device 17 can set each target subtask to have the consumption demand pattern ratio of the target complex task, and output the output information corresponding to the consumption demand pattern ratio of each target subtask through the dialogue device 13. The natural language processing device 17 can receive, through the dialogue device 13, a response information in response to the output information including the consumption demand pattern ratio, and selectively adjust the consumption demand pattern ratio of one or more of the target subtasks based on this response information. Specifically, the natural language processing device 17 can use the consumption demand pattern ratio of the target complex task as the default consumption demand pattern ratio of each target subtask, and then provide a channel for the user to adjust the consumption demand pattern ratio of each target subtask through the dialogue device 13. Then, the natural language processing device 17 screens the product and service data according to the consumption demand pattern ratio of the target subtask after selective adjustment. The method of screening or presenting the product and service data is as described in the above embodiment where each pre-stored subtask has a consumption demand pattern ratio, and will not be elaborated here. Alternatively, the natural language processing device 17 can directly search for product and service data that has the target subtask tag and the tag of the pattern with the highest ratio in this consumption demand pattern ratio based on the consumption demand pattern ratio of the target complex task, or determine the presentation order of the product and service data according to this consumption demand pattern ratio. By the above method of providing appropriate product and service information based on psychological needs, the opportunity for successful advertising recommendation and product and service sales can be increased.

[0037] With the above structure, the dialogue method and system based on complex task analysis disclosed in this case can automatically find multiple associated subtasks based on a single semantic information and provide corresponding information, thereby enhancing the diversity of information provision and enabling users to efficiently obtain useful information.

[0038]

Symbol Explanation

[0039] 1 Dialogue system

[0040] 11 Keyword extraction device

[0041] 13 Dialogue device

[0042] 15 Memory

[0043] 151 Knowledge graph

[0044] 153 Product service database

[0045] 17 Natural language processing device

[0046] CT1, CT2, CT4 Pre-stored complex tasks

[0047] ST1~ST6, ST8 Pre-stored subtasks.

Claims

1. A dialogue method based on complex task analysis, applicable to a dialogue system, comprising: Obtaining keyword vocabulary; Based on a knowledge graph, determining that the keyword vocabulary is associated with a target complex task, and obtaining a plurality of target subtasks having a connection relationship with the target complex task; For each of the target subtasks, searching for one or more matching product service data; Outputting output information corresponding to the one or more product service data that match each of the target subtasks; Receiving response information in response to the output information; And Selectively adjusting one or more of the target complex task and the target subtasks in the knowledge graph according to the response information; Wherein the knowledge graph includes a plurality of pre-stored complex tasks and a plurality of pre-stored subtasks, each of the pre-stored subtasks has a connection relationship with at least one of the pre-stored complex tasks, and the target complex task is one of the pre-stored complex tasks, Wherein each of the pre-stored subtasks in the knowledge graph has a consumption demand pattern ratio, and for each of the target subtasks, the step of searching for one or more matching product service data is based on the consumption demand pattern ratio of the target subtask.

2. The dialogue method according to claim 1, wherein each of the pre-stored complex tasks in the knowledge graph has the consumption demand pattern ratio, the output information is first output information, the response information is first response information, and the dialogue method further comprises: Setting that each of the target subtasks has the consumption demand pattern ratio of the target complex task, and outputting second output information corresponding to the consumption demand pattern ratio of each of the target subtasks; and Receiving second response information in response to the second output information, and selectively adjusting the consumption demand pattern ratio of one or more of the target subtasks according to the second response information.

3. The dialogue method according to claim 1 or 2, wherein the consumption demand pattern ratio indicates a ratio between a plurality of consumption demand patterns, and the consumption demand patterns include a privilege pattern, a reliable pattern, and an expectation pattern.

4. The dialogue method according to claim 1, wherein determining that the keyword vocabulary is associated with the target complex task based on the knowledge graph includes: Determining that the keyword vocabulary conforms to those that conform to the pre-stored complex tasks and the pre-stored subtasks in the knowledge graph; When the conformer is one of the pre-stored complex tasks, taking the conformer as the target complex task; and When the conformer is one of the pre-stored subtasks, selecting the target complex task from those of the pre-stored complex tasks that have a connection relationship with the conformer.

5. The dialogue method according to claim 4, wherein those of the pre-stored complex tasks that have a connection relationship with the one of the pre-stored subtasks include a plurality of candidate complex tasks, and each of the candidate complex tasks has a historical usage ratio. The step of selecting the target complex task from those of the pre-stored complex tasks that have a connection relationship with the conformer of the pre-stored subtasks includes: Selecting the one with the highest historical usage ratio among the candidate complex tasks as the target complex task.

6. The conversation method according to claim 1, wherein obtaining the keyword includes: Receiving a voice signal and obtaining the keyword from the voice signal.

7. The conversation method according to claim 1, wherein obtaining the keyword includes: When it is determined that the time for the article to be in the viewed state exceeds a preset time, being triggered to obtain the keyword from the article.

8. A conversation system based on complex task analysis, comprising: A keyword extraction device for obtaining keywords; A memory storing a knowledge graph, wherein the knowledge graph includes a plurality of pre-stored complex tasks and a plurality of pre-stored subtasks, and each of the pre-stored subtasks has a connection relationship with at least one of the pre-stored complex tasks; A conversation device for providing output information and receiving response information in response to the output information; And A natural language processing device connected to the keyword extraction device, the memory, and the conversation device, and configured to, according to the knowledge graph, determine that the keyword is associated with a target complex task, obtain a plurality of target subtasks having a connection relationship with the target complex task, find one or more matching product data or service data for each of the target subtasks, output, through the conversation device, the output information corresponding to the one or more product / service data matching each of the target subtasks, and selectively adjust, according to the response information, one or more of the target complex task and the target subtasks in the knowledge graph, and the target complex task is one of the pre-stored complex tasks included in the knowledge graph, wherein each of the pre-stored subtasks in the knowledge graph has a consumption demand pattern ratio, and the finding of the one or more matching product / service data for each of the target subtasks performed by the natural language processing device is based on the consumption demand pattern ratio of the target subtask.

9. The conversation system according to claim 8, wherein each of the pre-stored complex tasks in the knowledge graph has the consumption demand pattern ratio, the output information is the first output information, the response information is the first response information, wherein the natural language processing device further sets that each of the target subtasks has the consumption demand pattern ratio of the target complex task, and outputs the second output information corresponding to the consumption demand pattern ratio of each of the target subtasks, and receives the second response information in response to the second output information, and selectively adjusts the consumption demand pattern ratio of one or more of the target subtasks according to the second response information.

Images