Artificial intelligence-based intention mining method, device and equipment and storage medium

Abstract

The application discloses an artificial intelligence-based intention mining method and device, equipment and a storage medium, relates to the field of artificial intelligence, in particular to the fields of big data, knowledge graph, intelligent search and data mining, and can be applied to smart city, city management and application management scenarios. The specific implementation scheme is as follows: obtaining a user behavior data set to be processed, the user behavior data in the user behavior data set carrying an identifier and timestamp information of a user, dividing the user behavior data set based on the identifier and timestamp information of the user to obtain a first behavior data set of the user in a target time period, and determining a target intention data set existing in the first behavior data set according to a first knowledge base; wherein the first knowledge base comprises at least two types of keywords of a target intention, and intention keywords in the target intention data set cover the at least two types of keywords. The technical scheme can timely and accurately mine the intention of the user, and improves the intention mining efficiency.

Description

Technical Field

[0001] This disclosure relates to big data, knowledge graphs, intelligent search, and data mining technologies in the field of artificial intelligence, and particularly to an intent mining method, apparatus, device, and storage medium based on artificial intelligence. Background Technology

[0002] With the widespread adoption of mobile internet and smartphones, user intent can be reflected through online behavior, making the mining of online information for public opinion monitoring a popular and complex issue.

[0003] In today's world, all human behavior is inseparable from the Internet, and a large amount of information is output on the network every moment. However, the sensitive behaviors of users may only be the tip of the iceberg. Therefore, how to filter out the sensitive behaviors of users from the massive amount of user queries is a key factor in discovering whether users have a certain intention. Summary of the Invention

[0004] This disclosure provides an artificial intelligence-based intent mining method, apparatus, device, and storage medium.

[0005] According to a first aspect of this disclosure, an artificial intelligence-based intent mining method is provided, comprising:

[0006] Obtain the user behavior dataset to be processed, wherein the user behavior data in the user behavior dataset carries user identification and timestamp information;

[0007] Based on the user's identifier and the timestamp information, the user behavior dataset is divided to obtain the user's first behavior dataset within the target time period;

[0008] Based on the first knowledge base, the target intent dataset existing in the first behavior dataset is determined;

[0009] The first knowledge base includes at least two types of keywords related to the target intent, and the intent keywords in the target intent dataset cover the at least two types of keywords.

[0010] According to a second aspect of this disclosure, an artificial intelligence-based intent mining apparatus is provided, comprising:

[0011] The acquisition unit is used to acquire a user behavior dataset to be processed, wherein the user behavior data in the user behavior dataset carries the user's identifier and timestamp information;

[0012] The processing unit is configured to divide the user behavior dataset based on the user's identifier and the timestamp information to obtain the first behavior dataset of the user within a target time period.

[0013] The determining unit is used to determine the target intent dataset existing in the first behavior dataset based on the first knowledge base;

[0014] The first knowledge base includes at least two types of keywords related to the target intent, and the intent keywords in the target intent dataset cover the at least two types of keywords.

[0015] According to a third aspect of this disclosure, an electronic device is provided, comprising:

[0016] At least one processor; and

[0017] A memory communicatively connected to the at least one processor; wherein,

[0018] The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in the first aspect.

[0019] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are configured to cause the computer to perform the method described in the first aspect.

[0020] According to a fifth aspect of this disclosure, a computer program product is provided, the computer program product comprising: a computer program stored in a readable storage medium, wherein at least one processor of an electronic device can read the computer program from the readable storage medium, and the at least one processor executes the computer program to cause the electronic device to perform the method described in the first aspect.

[0021] According to the technical solution disclosed herein, by processing user behavior data within a target time period based on a pre-built first knowledge base, a target intent dataset can be determined. This enables timely and accurate mining of user intent, thereby improving the efficiency of intent mining.

[0022] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0023] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0024] Figure 1 This is a schematic diagram of an application scenario provided by an embodiment of this disclosure;

[0025] Figure 2This is a schematic diagram illustrating another application scenario provided by an embodiment of this disclosure;

[0026] Figure 3 This is a flowchart illustrating the intent mining method based on artificial intelligence provided in the first embodiment of this disclosure;

[0027] Figure 4 This is a schematic diagram of the composition of the first knowledge base corresponding to the preparation intention of the first compound;

[0028] Figure 5 This is a schematic diagram illustrating the intended composition of the first compound.

[0029] Figure 6 This is a flowchart illustrating the intent mining method based on artificial intelligence provided in the second embodiment of this disclosure;

[0030] Figure 7 This is a flowchart illustrating the intent mining method based on artificial intelligence provided in the third embodiment of this disclosure;

[0031] Figure 8 This is a schematic diagram of the implementation architecture of an embodiment of this disclosure;

[0032] Figure 9 This is a schematic diagram of the structure of an artificial intelligence-based intent mining device provided in an embodiment of this disclosure;

[0033] Figure 10 This is a schematic block diagram of an electronic device used to implement embodiments of the present disclosure. Detailed Implementation

[0034] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0035] Before introducing the application background and technical solutions of this disclosure, let's first introduce some terms that may be involved in the embodiments of this disclosure:

[0036] Artificial Intelligence (AI) is a comprehensive technology in computer science that studies the design principles and implementation methods of various intelligent machines, enabling them to possess the functions of perception, reasoning, and decision-making. AI technology is a multidisciplinary field, encompassing a wide range of areas, including natural language processing and machine learning / deep learning. With technological advancements, AI will be applied in more fields and play an increasingly important role.

[0037] Intent: refers to the true meaning that a user hopes to convey in a short and vague query.

[0038] With the development of the internet environment and artificial intelligence technology, most of people's behavior today is inseparable from the internet. More and more users obtain information and consultation through the internet, resulting in the internet outputting a large amount of information every moment. Therefore, mining information for public opinion monitoring from user behavior data generated by the internet has become a popular and complex problem.

[0039] Optionally, timely detection of sensitive behaviors can help prevent, combat, and verify potential sensitive intentions. Therefore, extracting sensitive behavior data from massive amounts of user behavior data is a key factor in uncovering user intent.

[0040] Currently, there are two main methods for mining user behavior data:

[0041] Method 1: Based on sensitive word matching. First, a sensitive word dictionary is established. Then, the keywords in each user's query are matched against the sensitive word dictionary. If a keyword match is found, the user's query is identified as a sensitive statement. However, this method only uncovers a limited range of sensitive behaviors and cannot detect statements where the keywords are not in the sensitive word dictionary, resulting in low accuracy.

[0042] Method 2: A mining scheme based on natural language processing (NLP) models. First, an NLP model is trained using user-labeled data. After training, the crawled user queries are input into the NLP model for sensitive statement identification, and then the expected sensitive statements are extracted. In this method, the accuracy of the NLP model is affected by the number of samples used during training, resulting in low mining accuracy.

[0043] As the above analysis shows, current sensitive statement mining solutions only analyze single query statements, resulting in low accuracy in behavior mining.

[0044] To address the aforementioned issues, the inventive concept of this disclosed technical solution is as follows: In practical applications, inventors discover that a person's purpose or intention, although not revealed by a single query statement, is often hidden in a series of inconspicuous behaviors over a period of time. Therefore, user behavior data over a period of time can be combined and analyzed to determine whether a user intends to perform a certain behavior within a certain time period, thereby improving the accuracy of intention discovery and laying the foundation for subsequent intention analysis.

[0045] Based on the above technical concept, this disclosure provides an artificial intelligence-based intent mining method. It involves acquiring a user behavior dataset to be processed, where the user behavior data carries user identifiers and timestamp information. Based on the user identifiers and timestamps, the user behavior dataset is divided to obtain a first behavior dataset for the user within a target time period. Finally, based on a first knowledge base, a target intent dataset is determined from the first behavior dataset. The first knowledge base includes at least two types of keywords representing the target intent, and the intent keywords in the target intent dataset cover at least two types of keywords. This technical solution, based on a pre-built first knowledge base, can determine the target intent dataset by processing user behavior data within a target time period, thus accurately mining user intent.

[0046] This disclosure provides an artificial intelligence-based intent mining method, apparatus, device, and storage medium, which are applied in the fields of big data, knowledge graphs, intelligent search, and data mining in the field of artificial intelligence. It can be applied in smart city, urban governance, and application management scenarios to accurately mine user intent and lay the foundation for subsequent analysis of whether a user will perform a certain behavior.

[0047] It should be noted that the first knowledge base in this embodiment is a collection of knowledge for a certain intention, including the knowledge required to realize the intention. However, this embodiment does not limit the knowledge content contained in the first knowledge base. It can be updated over time. Moreover, different intentions can correspond to knowledge bases containing different knowledge. This embodiment does not limit them.

[0048] The second knowledge base in this embodiment corresponds to the first knowledge base mentioned above. The second knowledge base is used to provide keywords for confirming interference intent. Through this second knowledge base, some irrelevant content in user data can be removed, laying the foundation for improving data processing efficiency.

[0049] It is understood that the user behavior datasets in this embodiment are all from publicly available datasets. The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solutions of this disclosure all comply with relevant laws and regulations and do not violate public order and good morals.

[0050] To facilitate understanding of the technical solutions provided in this disclosure, firstly, in conjunction with... Figure 1 The application scenarios of the embodiments of this disclosure will be described.

[0051] Figure 1 This is a schematic diagram illustrating an application scenario provided by an embodiment of this disclosure. For example... Figure 1 As shown, this application scenario involves two stages. Among them,

[0052] The first stage is the knowledge base construction stage. Optionally, in embodiments of this disclosure, the knowledge base may include a first knowledge base. The first knowledge base refers to the set of knowledge required to implement the target intent. The first knowledge base includes at least two types of keywords related to the target intent, and the number of keywords in each type may be multiple; this disclosure does not limit this.

[0053] Optionally, in embodiments of this disclosure, the knowledge base may further include a second knowledge base. The knowledge in the second knowledge base belongs to the same domain as the knowledge required to implement the target intent, but its relevance to the target intent is relatively small. By creating a second knowledge base, keywords in the second knowledge base can be removed from user behavior data during the intent recognition process, thereby improving the accuracy of intent recognition.

[0054] For example, see Figure 1 The device can analyze the target intent, determine the knowledge required to implement the target intent, and then, according to the steps or process of implementing the target intent, determine at least two types of keywords for the target intent. Furthermore, based on these at least two types of keywords, the device constructs a first knowledge base.

[0055] Optionally, the device can also analyze the knowledge required for the determined implementation target intent, calculate the relevance between the keywords corresponding to each knowledge and the target intent, and construct a second knowledge base based on keywords with a relevance less than the relevance threshold.

[0056] The second stage involves using a knowledge base for intent mining. See also... Figure 1 The knowledge base constructed in the first stage can be deployed to the execution device. For example, the first knowledge base can be deployed to the execution device, or both the first and second knowledge bases can be deployed to the execution device. In this way, the execution device can utilize the knowledge base to perform intent mining on the acquired user behavior data. It can be understood that the execution device can also be called an intelligent device.

[0057] In this embodiment, the aforementioned construction and execution devices can be terminal devices, servers, virtual machines, or distributed computer systems composed of one or more servers and / or computers. Terminal devices include, but are not limited to, smartphones, laptops, desktop computers, platform computers, in-vehicle devices, and smart wearable devices. Servers can be ordinary servers or cloud servers. Cloud servers, also known as cloud computing servers or cloud hosts, are a host product within the cloud computing service system. Servers can also be servers in distributed systems or servers integrated with blockchain technology.

[0058] It is worth noting that the product implementation disclosed herein is program code included in platform software and deployed on electronic devices (which can also be computing cloud or mobile terminal hardware with computing capabilities). Figure 1 In the system architecture diagram shown, the program code of this disclosure can be stored inside the execution device. At runtime, the program code runs in the host memory and / or GPU memory of the execution device.

[0059] For example, Figure 2 This is a schematic diagram illustrating another application scenario provided by an embodiment of this disclosure. For example... Figure 2 As shown, this application scenario may include: terminal device 21, network 22, and server 23. Both terminal device 21 and server 23 can communicate via network 22. Optionally, Figure 2 The application scenarios shown may also include data storage device 24 connected to server 23.

[0060] For example, in Figure 2 In the application scenario shown, when a user accesses the Internet through terminal device 21, for example, when using resources provided in network 22, user behavior data can be generated in network 22. At this time, server 23 can obtain the user behavior dataset to be processed from network 22 and store it in data storage device 24 so that it can be used directly in subsequent intent mining of user behavior data.

[0061] Understandably, the user behavior data stored in the data storage device 24 will carry the user's identifier and timestamp information, so that the dataset can be divided in a targeted manner in the future.

[0062] In this embodiment, the data storage device 24 can store a large amount of user behavior data, as well as the processing results of the server 23. The server 23 can execute program code of an artificial intelligence-based intent mining method based on the user behavior dataset in the data storage device 24 to determine the target intent dataset in the user behavior data.

[0063] It should be noted that the appendix Figure 2 This is merely a schematic diagram illustrating one application scenario provided by an embodiment of this disclosure; this embodiment does not necessarily represent an application scenario. Figure 2 The included equipment is not limited, nor is it restricted. Figure 2 The positional relationships between devices are defined, for example, in Figure 2 In this context, the data storage device 24 can be an external storage device relative to the server 23. In other cases, the data storage device 24 can also be placed inside the server 23.

[0064] In practical applications, since terminal devices are also processing devices with data processing capabilities, therefore, the above... Figure 2The server in the illustrated application scenario can also be implemented through a terminal device. In the embodiments of this disclosure, the server and terminal device with intent mining capabilities can be collectively referred to as electronic devices. Optionally, the embodiments of this disclosure are explained and described using electronic devices as the executing entity of the artificial intelligence-based intent mining method.

[0065] In this embodiment of the disclosure, "multiple" refers to two or more. "And / or" describes the relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following associated objects have an "or" relationship.

[0066] Below, in conjunction with the above Figure 1 and Figure 2 The application scenarios shown are illustrated in detail through specific embodiments to illustrate the technical solutions of this disclosure. It should be noted that the following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0067] For example, Figure 3 This is a flowchart illustrating the artificial intelligence-based intent mining method provided in the first embodiment of this disclosure. Figure 3 As shown, this AI-based intent mining method may include the following steps:

[0068] S301. Obtain the user behavior dataset to be processed. The user behavior data in the dataset carries the user's identifier and timestamp information.

[0069] In practical applications, if a user has a certain intention and wants to perform a certain action, they will usually search on the Internet to determine the way to perform the action or the knowledge required. Therefore, in order to uncover user intentions, we can determine them by analyzing the user behavior datasets generated by users on the Internet. That is, we can obtain the user behavior datasets to be processed from the Internet through web scraping tools.

[0070] In embodiments of this disclosure, the user behavior dataset may include internet user behavior data. Optionally, the user behavior data may include search data and order data, etc.

[0071] Optionally, in the user behavior dataset, the user behavior data typically carries a user identifier to mark the executor of the behavior, and the user behavior data also carries timestamp information to mark the time when the behavior occurred. It is understood that the user behavior data may also carry other information, such as device identifier, IP address, etc. The embodiments of this disclosure do not limit the information carried by the user behavior data, and it can be determined according to the actual scenario.

[0072] S302. Based on the user's identifier and timestamp information, the user behavior dataset is divided to obtain the first behavior dataset of the user within the target time period.

[0073] Optionally, in this embodiment, in order to analyze in a targeted manner whether a user has a certain intention during a certain time period, such as having a targeted intention, the user behavior dataset of that user during that time period can be analyzed. Therefore, based on the user's identifier and timestamp information carried in the user behavior data, the user behavior dataset can be divided to obtain the first behavior dataset of that user within the target time period.

[0074] In one optional embodiment, when dividing the user behavior dataset, the user behavior dataset can first be divided based on the user's identifier, and then the user behavior dataset corresponding to each user identifier can be divided based on the timestamp information, thereby obtaining the first behavior dataset of the user within the target time period.

[0075] In an optional embodiment, when dividing the user behavior dataset, the user behavior dataset can be divided first based on timestamp information, and then the user behavior dataset within each target time period can be divided based on the user's identifier, thereby obtaining the first behavior dataset of the user within the target time period.

[0076] It is understood that the embodiments disclosed herein do not limit the specific implementation of dividing the user behavior dataset to obtain the first behavior dataset of the user within the target time period. Moreover, the number of the first behavior datasets obtained by the division can be determined according to the user's identifier and / or the target time period covered by the user behavior dataset, which will not be elaborated here.

[0077] S303. Based on the first knowledge base, determine the target intent dataset that exists in the first behavior dataset.

[0078] The first knowledge base includes at least two types of keywords related to the target intent, and the intent keywords in the target intent dataset cover the at least two types of keywords included in the first knowledge base.

[0079] In this embodiment, the first knowledge base can be pre-stored in the device. After obtaining the first behavior dataset, the intent keywords included in the first behavior dataset can be queried in the first knowledge base to determine the intent keywords existing in the first knowledge base and the second behavior dataset corresponding to the intent keywords existing in the first knowledge base. Then, it is determined whether the keyword category corresponding to the intent keywords existing in the first knowledge base can cover the keyword categories included in the first knowledge base. If so, the second behavior dataset is the target intent dataset existing in the first behavior dataset.

[0080] In the embodiments of this disclosure, a user behavior dataset to be processed is acquired. Then, based on the user identifier and timestamp information carried in the user behavior data within the dataset, the dataset is divided to obtain a first behavior dataset for the user within a target time period. Finally, based on a first knowledge base, a target intent dataset is determined from the first behavior dataset. The first knowledge base includes at least two types of keywords representing the target intent, and the intent keywords in the target intent dataset cover at least two types of keywords. This technical solution, based on a pre-built first knowledge base, can determine the target intent dataset by processing user behavior data within a target time period, thus accurately mining the user's intent.

[0081] To help readers gain a deeper understanding of the implementation principles of this disclosure, the following will be discussed in conjunction with... Figures 4 to 8 The content shown is Figure 3 The illustrated embodiments are further refined.

[0082] For example, taking a search as an example, by analyzing a user's multiple search combinations over a period of time, it can be determined whether the user intends to prepare the first compound. Optionally, the first compound can be a prohibited compound or a compound with a certain property; this disclosure does not limit it.

[0083] Optional, Figure 4 This is a schematic diagram illustrating the composition of the first knowledge base corresponding to the intended preparation of the first compound. For example... Figure 4 As shown, the preparation of the first compound requires three essential components: raw materials, process, and equipment.

[0084] Reference Figure 4 As shown, based on the contents of each component, it is difficult to discern the intention to prepare the first compound by analyzing each component in isolation. This is because the contents of each component, when viewed individually, are all commonly used knowledge. For example, the raw materials include chemicals such as hydrochloric acid, toluene, and sodium hydroxide; the processes include deoxidation and condensation reactions; and the equipment includes reaction vessels, agitators, condensers, and other chemical devices. Individually, these are all common chemical or industrial knowledge. However, if someone simultaneously seeks out these contents over a period of time, it is very likely that they have the intention to prepare the first compound.

[0085] For example, Figure 5 This is a schematic diagram illustrating the intended preparation of the first compound. (Example) Figure 5As shown, suppose a user searched for "hydrochloric acid synthesis" on November 30th, "deoxygenation process" on December 1st, and "where to buy a reaction vessel" on December 7th. Analyzing each of these user behavior data individually, it is difficult to find that the user intended to prepare the first compound. However, by combining and analyzing multiple user behavior data, it is found that they are all indispensable parts of the preparation of the first compound. Therefore, it can be inferred that the user intended to prepare the first compound, and the target intent dataset composed of {"hydrochloric acid synthesis", "deoxygenation process", "where to buy a reaction vessel"} can be output.

[0086] Understandable Figure 4 and Figure 5 The contents shown are all illustrative examples. The embodiments disclosed herein do not limit the specific composition and time range for preparing the first compound. These can be determined according to the actual scenario and will not be elaborated here.

[0087] Accordingly, regarding the intent to prepare the first compound, the user behavior dataset to be processed can be divided first to determine the first user behavior dataset of a certain user in the target time period. Then, it can be analyzed whether the user has searched for the above-mentioned raw materials, processes, and equipment in the target time period. If so, it can be determined that the user has the intent to prepare the first compound, and the user behavior dataset covering the above-mentioned raw materials, processes, and equipment can be output.

[0088] For example, Figure 6 This is a flowchart illustrating the artificial intelligence-based intent mining method provided in the second embodiment of this disclosure. Figure 6 As shown, in this embodiment, the above-mentioned S303 can be implemented through the following steps:

[0089] S601. Based on the keywords included in the first knowledge base, determine the candidate datasets and the keyword categories covered by the candidate datasets in the first row dataset.

[0090] In this embodiment, the first knowledge base is a collection of knowledge about how users achieve their target intent. The first knowledge base contains keywords for each component of achieving that target intent. Therefore, by querying the user behavior data included in the first behavior dataset based on the keywords in the first knowledge base, a set of behavior data containing the keywords from the first knowledge base can be determined, i.e., the candidate dataset. Simultaneously, based on the category to which each keyword in the first knowledge base belongs, the keyword categories covered by the candidate dataset can be determined.

[0091] For example, regarding the above Figure 4If certain behavioral data in the first user behavior dataset contains any of the keywords such as hydrochloric acid, toluene, sodium hydroxide, reaction vessel, mixer, and condenser, then the set of these behavioral data can be determined as a candidate dataset. Since hydrochloric acid, toluene, and sodium hydroxide belong to raw materials, and reaction vessel, mixer, and condenser belong to equipment, the keyword categories covered by the candidate dataset can be determined to be raw materials and equipment.

[0092] As an example, in the implementation of step S601, the behavioral data in the first behavioral dataset can first be split into phrases to obtain an intent phrase set, which includes intent keywords. Then, the intent keywords are searched in the first knowledge base. If the intent keywords are found, the behavioral dataset to which the intent keywords belong is determined to be a candidate dataset, and the keyword category to which the intent keywords belong is the keyword category covered by the candidate dataset.

[0093] In this embodiment, assuming that the behavioral data in the first user behavior dataset includes "alternatives to toluene", "uses of hydrochloric acid", "how to use a mixer", and "acid-base neutralization", the statements "alternatives to toluene", "uses of hydrochloric acid", "how to use a mixer", and "acid-base neutralization" are first split into intent word sets {toluene, alternatives, hydrochloric acid, uses, mixer, use, acid-base neutralization}. Then, for each intent keyword in the intent word set, it is checked whether it exists in the first knowledge base. For example, intent keywords such as "toluene", "hydrochloric acid", and "mixer" exist in the first knowledge base, while other intent keywords do not exist in the first knowledge base. At this time, the set of behavioral data containing intent keywords such as "toluene", "hydrochloric acid", and "mixer" can be determined as a candidate dataset, that is, the candidate dataset includes {"alternatives to toluene", "uses of hydrochloric acid", "how to use a mixer"}. Based on the keyword category of "toluene" and "hydrochloric acid" being raw materials, and the keyword category of "mixer" being equipment, it can be determined that the keyword categories covered by the candidate dataset include raw materials and equipment.

[0094] S602. In response to the fact that the keyword categories included in the first knowledge base are consistent with the keyword categories covered by the candidate dataset, the candidate dataset is determined to be the target intent dataset in the first row dataset.

[0095] For example, after determining the keyword categories covered by the candidate dataset, it can be determined whether the keyword categories covered by the candidate dataset are consistent with the keyword categories included in the first knowledge base. If they are consistent, the candidate dataset can be determined as the target intent dataset; if they are inconsistent, the candidate dataset is determined not to be the target intent dataset.

[0096] For example, if a candidate dataset {“alternatives to toluene”, “uses of hydrochloric acid”, “how to use a mixer”} covers keyword categories including raw materials and equipment, while the first knowledge base includes keyword categories including raw materials, processes and equipment, that is, the keyword categories covered by the candidate dataset are inconsistent with the keyword categories included in the first knowledge base, then it is determined that the candidate dataset {“alternatives to toluene”, “uses of hydrochloric acid”, “how to use a mixer”} is not the target intent dataset.

[0097] For example, if the analysis in S602 above determines that a candidate dataset includes {"alternatives to toluene", "uses of hydrochloric acid", "how to use a mixer", "deoxygenation reaction"}, and the keyword categories covered by the candidate dataset include raw materials, processes and equipment, which are consistent with the keyword categories (raw materials, processes and equipment) included in the first knowledge base, then it can be determined that the candidate dataset {"alternatives to toluene", "uses of hydrochloric acid", "how to use a mixer", "deoxygenation reaction"} is a target intent dataset in the first row dataset.

[0098] In one optional embodiment of this disclosure, in the first knowledge base, each keyword included in the at least two keyword categories has a weight coefficient. That is, the first knowledge base includes not only at least two keyword categories, but also the weight coefficients of each keyword in each category. Therefore, after determining that the keyword categories included in the first knowledge base are consistent with the keyword categories covered by the candidate dataset, and before determining the candidate dataset as the target intent dataset in the first row dataset, the following operations can be performed on the candidate dataset:

[0099] A1. Based on the keyword categories covered by the candidate dataset, the behavioral data in the candidate dataset are arranged and combined to obtain a subset of candidate data. The intent keywords in this subset of candidate data cover at least two categories of keywords.

[0100] A2. Based on the weight coefficients of the aforementioned intent keywords, calculate the intent weighting value of the candidate data subset.

[0101] Accordingly, in this embodiment, determining the candidate dataset as the target intent dataset in the first behavior dataset includes: in response to the intent weighting value being greater than a preset weighting threshold, determining the candidate data subset in the candidate dataset as the target intent dataset.

[0102] In practical applications, since candidate datasets can include multiple behavioral data points for a certain keyword category, for example, the candidate dataset {"alternatives to toluene", "uses of hydrochloric acid", "how to use a mixer", "deoxygenation reaction"} includes both "alternatives to toluene" and "uses of hydrochloric acid", which also correspond to raw material categories, the candidate dataset {"alternatives to toluene", "uses of hydrochloric acid", "how to use a mixer", "deoxygenation reaction"} can be split into two candidate data subsets {"alternatives to toluene", "how to use a mixer", "deoxygenation reaction"} and {"uses of hydrochloric acid", "how to use a mixer", "deoxygenation reaction"}, and the keyword categories covered by each candidate data subset include raw materials, processes, and equipment.

[0103] In practical applications, based on practical experience, it is known that among various keywords of a certain intent, different keywords contribute differently to the realization of the intent. Therefore, a weight coefficient can be assigned to each keyword in the first knowledge base to distinguish the probability of intent when a user performs different behaviors.

[0104] For example, in the preparation of the first compound, "toluene" is more suitable as a raw material than "hydrochloric acid." Therefore, in the first knowledge base, the weight coefficient of "toluene" can be set higher than that of "hydrochloric acid." Similarly, in the process and equipment, different weight coefficients can be set for each keyword to improve the output accuracy of the target intent dataset.

[0105] In this embodiment, for each candidate data subset, the intent keywords included in the candidate data subset can be weighted and summed based on the weight coefficient of each intent keyword to calculate the intent weight value of the candidate data subset.

[0106] Optionally, the device can also set a confirmation threshold for the target intent dataset, i.e., a preset weight threshold. When the intent weight of a candidate data subset is greater than the preset weight threshold, it is then confirmed as the target intent dataset, which can improve the accuracy of intent mining to a certain extent.

[0107] In the embodiments of this application, based on the keywords included in the first knowledge base, candidate datasets and keyword categories covered by the candidate datasets are determined in the first row dataset. In response to the consistency between the keyword categories included in the first knowledge base and the keyword categories covered by the candidate datasets, the candidate dataset is determined as the target intent dataset in the first row dataset. Moreover, by setting weight coefficients for different intent keywords, content can only be output when the intent weight value of the subset reaches a certain score, thus ultimately achieving the goal of mining rich content and high mining accuracy.

[0108] For example, Figure 7 This is a flowchart illustrating the artificial intelligence-based intent mining method provided in the third embodiment of this disclosure. Figure 7 As shown, in this embodiment, the above-mentioned S302 can be implemented through the following steps:

[0109] S701. Based on timestamp information, the user behavior dataset is divided into subsets with the duration of the target time period as the sliding window length and the preset time unit as the sliding step size.

[0110] In practical applications, in order to avoid others discovering their intentions, users usually do not reveal them through single behavioral data. However, they may hide in a series of inconspicuous behavioral combinations over a period of time. Therefore, by statistically analyzing the exact duration of users executing different intentions, a target time period can be set in the device to statistically analyze the user's behavioral combinations within that target time period.

[0111] Optionally, in order to effectively monitor the user's implementation of a certain intention and avoid information omission, in this embodiment, the user behavior dataset can be divided based on the timestamp information carried by each piece of user behavior data in the user behavior dataset. That is, the user behavior dataset is divided with the duration of the target time period as the sliding window length and the preset time unit as the sliding step size to obtain at least one subset of user behavior data within the target time period.

[0112] For example, assuming the target time period is 14 days and the preset time unit is 1 day, then day 1 to day 14 is one time period, day 2 to day 15 is another time period, and so on.

[0113] It is understood that the embodiments disclosed herein do not limit the specific values ​​of the target time period and the preset time unit, which can be set according to different scenarios and different intentions, and will not be elaborated here.

[0114] Based on the knowledge base system, the time range for data mining is determined. Taking drug manufacturing as an example, anti-drug departments often believe that combinations of behaviors within two weeks have a higher accuracy rate.

[0115] S702. Divide the user behavior data subset according to the user's identifier to obtain the first behavior dataset of the user within the target time period.

[0116] In today's world, everyone within network coverage can access information or obtain advice online. Therefore, to analyze user intent in a targeted manner, user behavior data can be segmented into subsets for each target time period based on the user identifiers carried within the user behavior data. This results in the first behavior dataset for each user within the target time period. In other words, the behavior data in the first behavior dataset represents the behavior data of a single user within a single target time period.

[0117] It is understood that the implementation of this application does not limit the execution order of S701 and S702. In practical applications, the user behavior dataset can be divided into user behavior data subsets based on the user's identifier first, and then the user behavior data subsets can be divided based on the timestamp information, with the duration of the target time period as the sliding window length and the preset time unit as the sliding step size, to obtain the first behavior dataset of each user in each target time period.

[0118] In the embodiments of this disclosure, the user behavior dataset is divided based on timestamp information and user identifiers to obtain the first behavior dataset of each user within each target time period, laying the foundation for subsequent periodic routines of intent mining.

[0119] In the above embodiments of this disclosure, by analyzing historical intent mining information, it can be found that some intent keywords in the user behavior dataset may not be closely related to the intent. For example, when mining the intent to prepare the first compound, the user behavior dataset may analyze behavioral data with keywords such as "chemical bond" and "pH value". These keywords are general knowledge in chemistry and are often not closely related to the preparation of the first compound. Therefore, in order to reduce the workload of subsequent processing, behavioral data with such keywords can be intervened (removed) at a certain stage of data processing, thereby optimizing the output target intent dataset.

[0120] Optionally, in this embodiment, based on the target intent, a second knowledge base can be formed by identifying keywords frequently used in the compound preparation process but whose relevance to the target intent is less than a relevance threshold, according to actual applications. In this embodiment, it is assumed that the second knowledge base includes a first category of keywords, which have a relevance to the target intent less than a relevance threshold. Therefore, during intent analysis, behavioral data containing the first category of keywords can be removed based on the second knowledge base.

[0121] In one optional embodiment of this disclosure, before the above-mentioned S302 (based on the user's identifier and timestamp information, the user behavior dataset is divided to obtain the first behavior dataset of the user within the target time period), behavioral data with the first type of keywords in the user behavior dataset can be removed based on the second knowledge base.

[0122] Optionally, after obtaining the user behavior dataset to be processed, some behavioral data in the user behavior dataset can be removed based on the first type of keywords contained in the second knowledge base. This can minimize the amount of data to be processed in the subsequent process and improve the efficiency of intent mining.

[0123] In one optional embodiment of this disclosure, before the above-mentioned S303 (determining the target intent dataset existing in the first behavior dataset according to the first knowledge base), behavior data with the first type of keywords in the first behavior dataset can be removed based on the second knowledge base.

[0124] Optionally, after dividing the user behavior dataset and before mining the target intent dataset from the first behavior dataset based on the first knowledge base, some behavior data in the first behavior dataset can be removed based on the first type of keywords contained in the second knowledge base. This can reduce the workload of comparing keywords, reduce the workload of the data filtering process, and improve the efficiency of intent mining.

[0125] In one optional embodiment of this disclosure, after the above-described S303 (determining the target intent dataset existing in the first behavior dataset according to the first knowledge base), the target output dataset is obtained by removing the behavior data with the first type of keywords from the target intent dataset based on the second knowledge base.

[0126] In this optional embodiment, when the target intent dataset is determined but not yet output, irrelevant behavioral data in the intent dataset can be removed to improve the accuracy of the target intent dataset and reduce the workload of subsequent analysis and processing by relevant personnel.

[0127] Based on the above embodiments, Figure 8 This is a schematic diagram of the implementation architecture of an embodiment of this disclosure. For example... Figure 8 As shown, the current process of this embodiment includes:

[0128] S801, Construction of the knowledge base system.

[0129] In this embodiment, the knowledge base system may include a first knowledge base and a second knowledge base. The first knowledge base is a collection of knowledge reflecting the target intent, and the second knowledge base is a collection of irrelevant knowledge reflecting the target intent. It is understood that in practical applications, the first knowledge base is essential for target intent mining, while the second knowledge base, acting as an intervention aid to the first knowledge base, can improve the accuracy of target intent mining and reduce the workload of data processing.

[0130] Understandably, a knowledge base system for each intent in the target scenario can be built according to different scenario requirements, in order to discover datasets that realize different intents.

[0131] S802, Mining Intent Datasets.

[0132] Optionally, the mining of intent datasets can also be understood as the combined recall of target intent datasets. That is, by setting rules for combined recall, it is possible to output target intent datasets that meet the accuracy requirements.

[0133] S803, Periodic Mining of Intent Datasets.

[0134] In this embodiment, the periodic mining of the intent dataset can also be called data periodic routine. It can be explained as taking a set target time period as the processing period and a preset time unit as the step size, and performing intent mining sequentially within each target time period. Optimization strategies can be drawn from this process to update the knowledge base system, realize the mining closed loop, improve the mining accuracy, and reduce the user's processing workload.

[0135] Figure 9 This is a schematic diagram of the structure of an artificial intelligence-based intent mining device provided in an embodiment of this disclosure. The artificial intelligence-based intent mining device provided in this embodiment can be an electronic device or a device within an electronic device. Figure 9 As shown, the artificial intelligence-based intent mining device 900 provided in this embodiment may include:

[0136] The acquisition unit 901 is used to acquire a user behavior dataset to be processed, wherein the user behavior data in the user behavior dataset carries the user's identifier and timestamp information;

[0137] Processing unit 902 is used to divide the user behavior dataset based on the user's identifier and the timestamp information to obtain the first behavior dataset of the user within the target time period;

[0138] The determining unit 903 is used to determine the target intent dataset existing in the first behavior dataset based on the first knowledge base;

[0139] The first knowledge base includes at least two types of keywords related to the target intent, and the intent keywords in the target intent dataset cover the at least two types of keywords.

[0140] In an optional embodiment, the determining unit 903 includes:

[0141] The filtering module is used to determine candidate datasets and keyword categories covered by the candidate datasets in the first behavioral dataset based on the keywords included in the first knowledge base.

[0142] The determination module is configured to determine the candidate dataset as the target intent dataset in the first behavior dataset in response to the fact that the keyword categories included in the first knowledge base are consistent with the keyword categories covered by the candidate dataset.

[0143] In one optional embodiment, the filtering module includes:

[0144] The splitting submodule is used to split the behavioral data in the first behavioral dataset into phrases to obtain an intent phrase set, which includes intent keywords.

[0145] The search submodule is used to search for the intent keywords in the first knowledge base;

[0146] The filtering submodule is used to, in response to finding the intent keyword, determine the behavior dataset to which the intent keyword belongs as the candidate dataset, and the keyword category to which the intent keyword belongs as the keyword category covered by the candidate dataset.

[0147] In one optional embodiment, in the first knowledge base, each keyword included in the at least two types of keywords has a weight coefficient;

[0148] The determining unit 903 further includes: a combination module and a calculation module;

[0149] The combination module is used to arrange and combine the behavioral data in the candidate dataset according to the keyword categories covered by the candidate dataset to obtain a subset of candidate data, wherein the intent keywords in the subset of candidate data cover the at least two categories of keywords;

[0150] The calculation module is used to calculate the intent weighting value of the candidate data subset based on the weight coefficient of the intent keyword;

[0151] The determining module is specifically used to determine the candidate data subset in the candidate dataset as the target intent dataset in response to the intent weighting value being greater than a preset weighting threshold.

[0152] In one optional embodiment, the processing unit 902 includes:

[0153] The first processing module is used to divide the user behavior dataset based on the timestamp information, with the duration of the target time period as the sliding window length and the preset time unit as the sliding step size, to obtain a subset of user behavior data within the target time period.

[0154] The second processing module is used to divide the user behavior data subset according to the user's identifier to obtain the first behavior dataset of the user within the target time period.

[0155] In an optional embodiment, the processing unit 902 is further configured to remove behavioral data containing a first type of keyword from the user behavior dataset based on a second knowledge base; the second knowledge base includes the first type of keyword, wherein the relevance of the first type of keyword to the target intent is less than a relevance threshold.

[0156] In an optional embodiment, the processing unit 902 is further configured to remove behavioral data with a first type of keyword from the first behavioral dataset based on a second knowledge base; the second knowledge base includes the first type of keyword, wherein the relevance of the first type of keyword to the target intent is less than a relevance threshold.

[0157] In an optional embodiment, the processing unit 902 is further configured to, based on a second knowledge base, remove behavioral data containing first-type keywords from the target intent dataset to obtain a target output dataset; the second knowledge base includes the first-type keywords, wherein the relevance of the first-type keywords to the target intent is less than a relevance threshold.

[0158] The AI-based intent mining device provided in this embodiment can be used to execute the AI-based intent mining method in any of the above method embodiments. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0159] According to embodiments of this disclosure, this disclosure also provides a readable storage medium, which is a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause the computer to execute the scheme provided in any of the above embodiments.

[0160] According to embodiments of this disclosure, this disclosure also provides a computer program product, which includes: a computer program stored in a readable storage medium, at least one processor of an electronic device can read the computer program from the readable storage medium, and the at least one processor executes the computer program to cause the electronic device to perform the solution provided in any of the above embodiments, that is, when the computer program is executed by the processor, it implements the solution provided in any of the above embodiments.

[0161] According to embodiments of this disclosure, this disclosure also provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the scheme provided in any of the above embodiments.

[0162] Figure 10 This is a schematic block diagram of an electronic device used to implement embodiments of the present disclosure. Figure 10 As shown, the electronic device 1000 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0163] like Figure 10 As shown, device 1000 includes a computing unit 1001, which can perform various appropriate actions and processes according to a computer program stored in read-only memory (ROM) 1002 or a computer program loaded from storage unit 1008 into random access memory (RAM) 1003. The RAM 1003 may also store various programs and data required for the operation of device 1000. The computing unit 1001, ROM 1002, and RAM 1003 are interconnected via bus 1004. Input / output (I / O) interface 1005 is also connected to bus 1004.

[0164] Multiple components in device 1000 are connected to I / O interface 1005, including: input unit 1006, such as keyboard, mouse, etc.; output unit 1007, such as various types of monitors, speakers, etc.; storage unit 1008, such as disk, optical disk, etc.; and communication unit 1009, such as network card, modem, wireless transceiver, etc. Communication unit 1009 allows device 1000 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0165] The computing unit 1001 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 1001 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 1001 performs the various methods and processes described above, such as an AI-based intent mining method. For example, in some embodiments, the AI-based intent mining method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 1008. In some embodiments, part or all of the computer program may be loaded and / or installed on device 1000 via ROM 1002 and / or communication unit 1009. When the computer program is loaded into RAM 1003 and executed by the computing unit 1001, one or more steps of the AI-based intent mining method described above may be performed. Alternatively, in other embodiments, the computing unit 1001 may be configured to perform an artificial intelligence-based intent mining method by any other suitable means (e.g., by means of firmware).

[0166] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0167] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0168] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0169] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0170] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0171] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service ecosystem, addressing the shortcomings of traditional physical hosts and VPS (Virtual Private Server, or simply "VPS") services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0172] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0173] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. An intent mining method based on artificial intelligence, comprising: Obtain the user behavior dataset to be processed, wherein the user behavior data in the user behavior dataset carries user identification and timestamp information; Based on the user's identifier and the timestamp information, the user behavior dataset is divided to obtain the user's first behavior dataset within the target time period; Based on the first knowledge base, a target intent dataset existing in the first behavior dataset is determined, which includes: based on the keywords included in the first knowledge base, a candidate dataset and the keyword categories covered by the candidate dataset are determined in the first behavior dataset; wherein, the first knowledge base is constructed by analyzing the target intent, determining the knowledge required to implement the target intent, and then determining at least two types of keywords of the target intent according to the process of implementing the target intent; In response to the fact that the keyword categories included in the first knowledge base are consistent with the keyword categories covered by the candidate dataset, the candidate dataset is determined to be the target intent dataset in the first behavior dataset; wherein, the first knowledge base refers to the knowledge set required to implement the target intent; the intent keywords in the target intent dataset cover the at least two categories of keywords; The user behavior dataset is divided based on the user's identifier and the timestamp information to obtain the user's first behavior dataset within a target time period, including: Based on the timestamp information, the user behavior dataset is divided with the duration of the target time period as the sliding window length and the preset time unit as the sliding step size to obtain a subset of user behavior data within the target time period. The user behavior data subset is divided according to the user's identifier to obtain the first behavior dataset of the user within the target time period.

2. The method of claim 1, wherein, The step of determining candidate datasets and keyword categories covered by the candidate datasets in the first behavioral dataset based on the keywords included in the first knowledge base includes: The behavioral data in the first behavioral dataset is split into phrases to obtain an intent phrase set, which includes intent keywords. Search for the intent keywords in the first knowledge base; In response to finding the intent keyword, the behavioral dataset to which the intent keyword belongs is determined as the candidate dataset, and the keyword category to which the intent keyword belongs is the keyword category covered by the candidate dataset.

3. The method according to claim 2, wherein in the first knowledge base, each keyword included in the at least two types of keywords has a weight coefficient; Before determining the candidate dataset as the target intent dataset in the first behavior dataset, the method further includes: Based on the keyword categories covered by the candidate dataset, the behavioral data in the candidate dataset are arranged and combined to obtain a subset of candidate data, wherein the intent keywords in the subset of candidate data cover at least two categories of keywords; Based on the weight coefficients of the intent keywords, calculate the intent weighting value of the candidate data subset; Determining the candidate dataset as the target intent dataset in the first behavior dataset includes: In response to the intent weighting value being greater than a preset weighting threshold, the candidate data subset in the candidate dataset is determined as the target intent dataset.

4. The method according to any one of claims 1 to 3, before dividing the user behavior dataset based on the user's identifier and the timestamp information to obtain the first behavior dataset of the user within a target time period, the method further includes: Based on the second knowledge base, remove behavioral data containing the first type of keywords from the user behavior dataset; The second knowledge base includes the first type of keywords, and the relevance of the first type of keywords to the target intent is less than a relevance threshold.

5. The method according to any one of claims 1 to 3, wherein before determining the target intent dataset existing in the first behavior dataset based on the first knowledge base, the method further comprises: Based on the second knowledge base, remove behavioral data with the first type of keywords from the first behavioral dataset; The second knowledge base includes the first type of keywords, where the relevance of the first type of keywords to the target intent is less than a relevance threshold.

6. The method according to any one of claims 1 to 3, wherein after determining the target intent dataset existing in the first behavior dataset based on the first knowledge base, the method further comprises: Based on the second knowledge base, behavioral data containing the first type of keywords are removed from the target intent dataset to obtain the target output dataset; The second knowledge base includes the first type of keywords, where the relevance of the first type of keywords to the target intent is less than a relevance threshold.

7. An artificial intelligence-based intent mining device, comprising: The acquisition unit is used to acquire a user behavior dataset to be processed, wherein the user behavior data in the user behavior dataset carries the user's identifier and timestamp information; The processing unit is configured to divide the user behavior dataset based on the user's identifier and the timestamp information to obtain the first behavior dataset of the user within a target time period. The determining unit is used to determine the target intent dataset existing in the first behavior dataset based on the first knowledge base; wherein, the first knowledge base is constructed by analyzing the target intent, determining the knowledge required to implement the target intent, and then determining at least two types of keywords of the target intent according to the process of implementing the target intent; Wherein, the first knowledge base refers to the set of knowledge required to implement the target intent; The determining unit includes: The filtering module is used to determine candidate datasets and keyword categories covered by the candidate datasets in the first behavioral dataset based on the keywords included in the first knowledge base. The determining module is configured to determine the candidate dataset as the target intent dataset in the first behavior dataset in response to the fact that the keyword categories included in the first knowledge base are consistent with the keyword categories covered by the candidate dataset; the intent keywords in the target intent dataset cover the at least two categories of keywords; The processing unit includes: The first processing module is used to divide the user behavior dataset based on the timestamp information, with the duration of the target time period as the sliding window length and the preset time unit as the sliding step size, to obtain a subset of user behavior data within the target time period. The second processing module is used to divide the user behavior data subset according to the user's identifier to obtain the first behavior dataset of the user within the target time period.

8. The apparatus of claim 7, wherein, The filtering module includes: The splitting submodule is used to split the behavioral data in the first behavioral dataset into phrases to obtain an intent phrase set, which includes intent keywords. The search submodule is used to search for the intent keywords in the first knowledge base; The filtering submodule is used to, in response to finding the intent keyword, determine the behavior dataset to which the intent keyword belongs as the candidate dataset, and the keyword category to which the intent keyword belongs as the keyword category covered by the candidate dataset.

9. The apparatus according to claim 8, wherein in the first knowledge base, each keyword included in the at least two types of keywords has a weight coefficient; The determining unit further comprises: Combination module and calculation module; The combination module is used to arrange and combine the behavioral data in the candidate dataset according to the keyword categories covered by the candidate dataset to obtain a subset of candidate data, wherein the intent keywords in the subset of candidate data cover the at least two categories of keywords; The calculation module is used to calculate the intent weighting value of the candidate data subset based on the weight coefficient of the intent keyword; The determining module is specifically used to determine the candidate data subset in the candidate dataset as the target intent dataset in response to the intent weighting value being greater than a preset weighting threshold.

10. The apparatus according to any one of claims 7 to 9, wherein the processing unit is further configured to, based on a second knowledge base, remove behavioral data containing a first type of keyword from the user behavior dataset; the second knowledge base includes the first type of keyword, wherein the relevance of the first type of keyword to the target intent is less than a relevance threshold.

11. The apparatus according to any one of claims 7 to 9, wherein the processing unit is further configured to, based on a second knowledge base, remove behavioral data containing a first type of keyword from the first behavioral dataset; the second knowledge base includes the first type of keyword, wherein the relevance of the first type of keyword to the target intent is less than a relevance threshold.

12. The apparatus according to any one of claims 7 to 9, wherein the processing unit is further configured to, based on a second knowledge base, remove behavioral data containing first-type keywords from the target intent dataset to obtain a target output dataset; the second knowledge base includes the first-type keywords, wherein the relevance of the first-type keywords to the target intent is less than a relevance threshold.

13. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 6.

14. A non-transitory computer readable storage medium having stored thereon computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1 to 6.

15. A computer program product comprising a computer program that, when executed by a processor, implements the steps of the method according to any one of claims 1 to 6.

Images