CN115237967A

CN115237967A - Scene recommendation method, electronic device, storage medium and product

Info

Publication number: CN115237967A
Application number: CN202210828547.5A
Authority: CN
Inventors: 李禹肖; 宋韵瑾; 曹天元; 牟小峰; 唐剑
Original assignee: Midea Group Co Ltd; Midea Group Shanghai Co Ltd
Current assignee: Midea Group Co Ltd; Midea Group Shanghai Co Ltd
Priority date: 2022-07-13
Filing date: 2022-07-13
Publication date: 2022-10-25

Abstract

The application relates to the technical field of computers, and provides a scene recommendation method, which comprises the following steps: determining a scene query request; determining an entity to be associated based on the scene query request; inquiring a household appliance knowledge map based on the entity to be associated, and determining a scene recommendation result; the household appliance knowledge graph comprises a household appliance entity, a scene entity and a multi-dimensional incidence relation between the household appliance and the scene, the scene recommendation result comprises an incidence scene with the incidence relation of the entity to be associated, and the entity to be associated comprises the household appliance entity to be associated and the scene entity to be associated. According to the method and the device, the entity to be associated is obtained by analyzing the scene query request, the knowledge graph is used for identifying the existing scene linked with the entity to be associated, a potential new scene can be mined according to the entity to be associated, and the scene identification efficiency is improved.

Description

Scene recommendation method, electronic device, storage medium and product

Technical Field

The application relates to the technical field of computers, in particular to a scene recommendation method, electronic equipment, a storage medium and a product.

Background

The family scene mining is an important service in the intelligent home, and the service utilizes the family data of the user to learn and identify the use modes of the intelligent household appliances of the user in different life scenes through data analysis or mining algorithms. The accurate family scene mining can provide more comfortable, safe and energy-saving automatic control for the user based on the real use habit of the user, and recommend more interesting new scenes for the user.

However, the existing home scene mining needs to depend on manual predefining and pre-labeling of scenes, so that the new scene mining is difficult, and the predefined scene updating is also difficult.

Disclosure of Invention

The present application is directed to solving at least one of the technical problems occurring in the related art. Therefore, the scene recommendation method is provided, the knowledge graph is used for replacing manual labeling, after the query request is analyzed, the association relation of the body to be associated in the household knowledge graph is queried, the scene recommendation result under the query request is obtained, and meanwhile, a new scene can be automatically discovered according to the knowledge graph.

The application also provides a scene recommendation device.

The application also provides an electronic device.

The present application also proposes a non-transitory computer-readable storage medium.

The present application also proposes a computer program product.

A scene recommendation method according to an embodiment of a first aspect of the present application includes:

determining a scene query request;

determining an entity to be associated based on the scene query request;

inquiring a household appliance knowledge map based on the entity to be associated, and determining a scene recommendation result;

the household appliance knowledge graph comprises a household appliance entity, a scene entity and an incidence relation between the household appliance entity and the scene entity, the scene recommendation result comprises an incidence scene with the incidence relation of the entity to be associated, and the entity to be associated comprises the household appliance entity to be associated and the scene entity to be associated.

According to the scene recommendation method, after the scene query request is obtained, the household appliance knowledge graph is used for identifying and mining the entity to be associated, and finally the recommendation scene associated with the entity to be associated is obtained. The method can automatically discover a new scene by inquiring the scene which is expected to be pushed in the request and by inquiring the incidence relation of the body to be associated in the household electrical knowledge map. The limitation that the traditional method only can identify the preset scene and is not easy to identify a new scene is avoided, and the scene identification efficiency is effectively improved.

According to an embodiment of the present application, the scene recommendation result further includes: and (3) starting the associated household appliances and/or the associated household appliances in the associated scene.

According to an embodiment of the application, the querying the household appliance knowledge graph based on the entity to be associated to determine the scene recommendation result includes:

inquiring the household appliance knowledge graph based on the entity to be associated, and determining an alternative association scene having an association relation with the entity to be associated;

screening the alternative association scenes to determine the association scenes of the entities to be associated;

and determining the associated household appliances in the associated scene and the starting sequence of the associated household appliances based on the priority of the household appliances in the associated scene of the entity to be associated.

According to an embodiment of the present application, the screening the candidate association scenes to determine the association scene of the entity to be associated includes:

determining alternative association scenes in which conflict relationships exist between the alternative association scenes and the entity to be associated;

and deleting the alternative association scenes having conflict relations with the entities to be associated from the alternative association scenes, and determining the association scenes of the entities to be associated.

According to an embodiment of the application, the determining an entity to be associated based on the scenario query request includes:

extracting texts of the scene query request to obtain entity information, wherein the entity information comprises an entity and the state of the entity;

and screening the entity information based on the semantic contribution degree, and determining the entity to be associated.

According to an embodiment of the present application, the determining a scenario query request further includes:

acquiring user information;

associating user information with the household appliance knowledge graph, and determining a household appliance entity associated with the user information and a scene entity associated with the user information;

and determining an updated household appliance knowledge map based on the household appliance entity associated with the user information and the scene entity associated with the user information.

determining an entity to be associated for associating user information;

inquiring the updated household appliance knowledge map based on the entity to be associated of the associated user information to determine a scene recommendation result;

and the scene recommendation result comprises an association scene which has an association relation with the entity to be associated of the associated user information.

According to one embodiment of the application, the construction of the household appliance knowledge graph comprises the following steps:

collecting household appliance information;

performing entity extraction on the household appliance information, and determining a household appliance entity and a scene entity;

extracting the relation of the household appliance information and the incidence relation between the household appliance entity and the scene entity;

and matching the household electrical appliance entity and the scene entity by combining the relationship between the household electrical appliance entity and the scene entity, and determining the household electrical appliance knowledge graph.

According to an embodiment of the present application, matching the home appliance entity and the scene entity in combination with a relationship between the home appliance entity and the scene entity includes:

vectorizing the household appliance entity and the scene entity to determine a household appliance vector and a scene vector;

and matching the household appliances in the same scene by combining the relationship between the household appliance entity and the scene entity based on the similarity of the household appliance vector and the scene vector.

According to an embodiment of the application, the building of the household appliance knowledge graph further includes:

and creating a new scene ontology in the household appliance knowledge graph based on the change of the incidence relation of the household appliance entity in different time periods.

According to an embodiment of the present application, the creating a new scene ontology in the appliance knowledge graph based on a change of an association relationship of the appliance entity in different time periods includes:

visualizing the change of the association relationship of the household appliance entities in different time periods to obtain a household appliance relationship heat map;

determining, based on the appliance relationship heatmap, usage times and priorities of the appliance entities over different time periods;

and creating a new scene ontology in the household appliance knowledge graph based on the use time and the priority of the household appliance entity in different time periods.

A scene recommendation device according to an embodiment of a second aspect of the present application includes:

the input module is used for determining a scene query request;

the analysis module is used for determining an entity to be associated based on the scene query request;

the recommendation module is used for inquiring the household appliance knowledge map based on the entity to be associated and determining a scene recommendation result;

the household appliance knowledge graph comprises a household appliance entity, a scene entity and a multi-dimensional incidence relation between the household appliance entity and the scene entity, the scene recommendation result comprises an incidence scene with the incidence relation of the entity to be associated, and the entity to be associated comprises the household appliance entity to be associated and the scene entity to be associated.

According to the scene recommending device, after the scene query request is obtained, the household appliance knowledge graph is used for recognition and mining, and the recommended scene is finally obtained. The device can automatically discover a new scene by inquiring the incidence relation of the household appliance entity contained in the request in the household appliance knowledge graph. The limitation that the traditional method only can identify the preset scene and is not easy to identify a new scene is avoided, and the scene identification efficiency is effectively improved.

An electronic device according to an embodiment of the third aspect of the present application includes a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor implements the scene recommendation method or the appliance knowledge graph construction method when executing the program.

A non-transitory computer-readable storage medium according to an embodiment of a fourth aspect of the present application, has a computer program stored thereon, and when executed by a processor, implements the scene recommendation method or the appliance knowledge graph construction method.

A computer program product according to an embodiment of the fifth aspect of the present application includes a computer program, and when the computer program is executed by a processor, the computer program implements the scene recommendation method or the appliance knowledge graph construction method.

One or more technical solutions in the embodiments of the present application have at least one of the following technical effects: the scene recommendation result further comprises the starting sequence of the associated household appliances, so that the condition that the household appliances are interfered with each other and contradicted when a plurality of intelligent scenes are overlapped and linked is solved, and the household appliances can be controlled orderly under the condition that the intelligent scenes are linked.

Further, the embodiment of the application can prompt a user that the scene has the conflict relationship and can also prompt the user that the scene has the conflict relationship when the conflict relationship exists in the associated scene of the target household appliance, and the scene of the household appliance knowledge graph can be updated automatically.

Furthermore, according to the embodiment of the application, the user portrait and the user behavior data can be added to the established household appliance knowledge graph according to the behavior information of the user in different scenes, so that the household appliance knowledge graph is updated, and automatic scene recommendation can be performed according to the preference of the user.

Furthermore, the knowledge graph is used for replacing manual marking, manpower is saved, meanwhile entity marking quality is improved, and scene library updating, error correction and completion are conducted based on user information in reality. Compared with the traditional supervised and unsupervised learning methods, the method has the advantages that the automation degree, the missing judgment rate and the interpretability are greatly improved, and the cross-platform applicability is strong while the personalized features of the user are kept.

Additional aspects and advantages of the present application will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the present application.

Drawings

In order to more clearly illustrate the embodiments of the present application or technical solutions in related arts, the drawings used in the description of the embodiments or related arts will be briefly described below, it is obvious that the drawings in the description below are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative efforts.

FIG. 1 is a schematic flow chart of a scene recommendation method according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of a knowledge graph of a home appliance according to an embodiment of the present invention;

fig. 3 is a home appliance-scene relationship heat map of a home appliance knowledge graph provided by an embodiment of the present invention;

fig. 4 is a schematic flow chart of a method for building a knowledge graph of a home appliance according to an embodiment of the present application;

fig. 5 is a schematic structural diagram of a scene recommendation device according to an embodiment of the present application;

fig. 6 is a schematic structural diagram of an electronic device provided in an embodiment of the present application.

Detailed Description

To make the purpose, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be clearly and completely described below with reference to the drawings in the present application, and it is obvious that the described embodiments are some, but not all embodiments of the present application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

In the description of the present application, reference to the description of "one embodiment," "some embodiments," "an example," "a specific example," or "some examples" or the like is intended to mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the embodiments of the present application. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.

Embodiments of the present application will be described in further detail below with reference to the drawings and examples. The following examples are intended to illustrate the present application but are not intended to limit the scope of the present application.

In current scene recommendation methods based on family scene mining, high frequency scene mining is typically performed by manual-based scene recognition based on historical data analysis. And developing a customized system aiming at a high-frequency scene, and defining preset control programs and rules to repeatedly run, for example: cleaning at fixed time, opening in advance or closing scenes based on the timing on-off of the electric appliance; an abnormal intrusion alert mode, a gas leakage alert mode, etc. However, this method cannot provide more refined and personalized services according to the daily life habits of the user.

According to the research of the inventor, scene recommendation methods based on family scene mining are mainly divided into two categories: an artificial based method and an artificial intelligence algorithm based method. The manual method usually performs high-frequency scene mining according to historical data analysis, performs customized development on the high-frequency scene, and defines preset control programs and rules to run repeatedly. However, the method can only be operated based on the preset and cannot provide more refined and personalized services according to the daily life habits of the user. The method based on the artificial intelligence algorithm is based on the collection of the use information of the user intelligent household appliances by sensors such as a camera, voice, infrared and intelligent electric meters, and the like, models are built for the sensor data and the labeled scenes through the labeled scenes predefined manually by using supervised learning and unsupervised learning, and devices linked for many times are found out, so that the scenes are excavated and constructed. However, the method based on the artificial intelligence algorithm has practical problems of dependence on artificial labeling, poor interpretability, difficulty in acquiring user behavior information and the like.

However, the existing supervised or unsupervised learning method has the following problems: 1. relying on manual annotation: the supervised learning type invention relies on manual labeling of a large amount of intelligent equipment, but in the actual situation, many scenes have no instant information and lack effective labeling. 2. Poor interpretability: the unsupervised learning algorithm has low interpretability, cannot provide a specific scene recognition rule for a user, and is inconvenient for subsequent users to debug according to personal preference. The requirement on the number of household appliances is high, for example, when the number of household appliances is limited, the number of learned scenes is limited. 3. The user behavior information is difficult to collect: the data acquisition has high requirements on a network, delay may exist in transmission time, the method is not like high precision of a physical switch, and when supervised and unsupervised learning is used for carrying out short-term electrical appliance linkage mode prediction on a user using scene, the influence of a sensor transmission time interval is large, so that the model accuracy is reduced.

Therefore, on the basis of many problems existing in the conventional scene recommendation scheme based on family scene mining, the embodiment of the present application provides a scene recommendation method based on family scene mining, and the scene recommendation method of the embodiment of the present application is described below with reference to fig. 1 to 4.

The execution subject of the scene recommendation method can be a scene recommendation device, or a server, or also can be a terminal of a user, including but not limited to a mobile phone, a tablet computer, a pc terminal, a vehicle-mounted terminal, a household intelligent appliance and the like.

The scene or the family scene in the embodiment of the application refers to a typical scene when the geographical position of the terminal is at home, that is, a scene of a place where the user lives and lives for a long time. The household appliance is an appliance used in a family scene.

Fig. 1 illustrates a flowchart of a scene recommendation method provided in an embodiment of the present application, where the method at least includes the following steps:

step 101, determining a scene query request;

102, determining an entity to be associated based on a scene query request;

103, inquiring a household appliance knowledge graph based on the entity to be associated, and determining a scene recommendation result;

the household appliance knowledge graph comprises a household appliance entity, a scene entity and an incidence relation between the household appliance entity and the scene entity, the scene recommendation result comprises an incidence scene with the incidence relation with the entity to be associated, and the entity to be associated comprises the household appliance entity to be associated and the scene entity to be associated.

For step 101, it should be noted that, in this embodiment, the scene query request may be a piece of instruction information input by the user, where the instruction information includes a description of the scene to be queried by the user. The instruction information may be a description of the home appliance included in the scene that needs to be queried, for example: "what the corresponding scene is when the electric fan is turned on", "what scene the intelligent mattress and the intelligent lamp can compose"; the instruction information can also be the description of the scene needing to be queried according to the real-time state information. For example: "what scene the appliances operating in the kitchen correspond to".

For step 102, it should be noted that, for the scene query request, the entity to be associated is obtained after the analysis is needed and is used as the input of the next query. In addition, the state information or the implicit state information of the entity to be associated can be obtained after the analysis, and the entity to be associated in the embodiment of the invention comprises the state information of the entity. For example, taking the home appliance entity to be associated as an example, the scene query request is: the 'what scene corresponds to when the humidifier is started', and the entity to be associated is 'humidifier starting'. For another example, taking the scenario entity to be associated as an example, the scenario query request includes: "what the scene associated with the sleep scene is", and the entity to be associated is the "sleep scene".

In addition, it should be noted that, in general, if no status wording indicating the entity to be associated appears in the scenario query request, the default status is start or use. For example, "what scene the humidifier corresponds to", where there is no explicit state information, the entity to be associated after the analysis is "humidifier start", and the implicit state information is "start", then the instruction information should actually be "what scene the humidifier corresponds to when starting".

In step 103, it should be noted that the household appliance knowledge graph in the embodiment of the present application is pre-established, and meanwhile, due to the expandability of the knowledge graph, the household appliance knowledge graph may also adapt to environmental changes and user inputs for self-updating. The body of household electrical appliances knowledge map includes household electrical appliances body and scene body in this embodiment, and household electrical appliances are the electrical appliances of work under the scene at home, for example: televisions, refrigerators, air conditioners, treadmills, washing machines, and the like. The scene is a living scene of the electric appliance linkage working based on the characteristics of the household life, such as: a windowing scene, a sports scene, an away-from-home scene, a bathing scene, a sleeping scene, and the like. The multi-dimensional association relationship between the household appliance and the scene mainly refers to the relationship between the scene and attributes such as the state of the household appliance, the state skill of the household appliance and the state efficacy of the household appliance.

When the scene recommendation method provided by the embodiment of the application is applied, a scene query request needs to be obtained first, and an entity to be associated is determined based on the scene query request. And then the entity to be associated searches the scene of the household appliance knowledge graph to finally obtain a scene recommendation result. For example, after a scene query request of "what scene the intelligent mattress corresponds to" is acquired, a scene sub-graph related to the intelligent mattress in the household appliance knowledge graph is searched, and finally a sleep scene related to the intelligent mattress is recommended.

According to the scene recommendation method, after the scene query request is obtained, the household appliance knowledge graph is used for recognition and mining, and the recommended scene is finally obtained. The method can automatically discover the new scene through the incidence relation of the entity to be associated contained in the scene query request in the household knowledge graph. The limitation that the traditional method only can identify the preset scene and is not easy to identify a new scene is avoided, and the scene identification efficiency is effectively improved.

The existing invention or technology has the difficulty of multi-scene linkage, including multi-scene implicit conflict. In particular, because intelligent devices have versatility, and labeled predefined scenes have no unified standard, business personnel can only focus on explicit conflicts, but are difficult to label implicit conflicts. Such as: the 'closing of the window' only marks scenes needing shading and resting, and omits the implicit scene which is air-conditioning refrigeration and can be influenced by the window.

According to the scene recommendation method, the multi-dimensional association relation between the scene and the household appliances can be marked through the built household appliance knowledge map, and the household appliance knowledge map is searched after the query request is obtained, so that the recommendation of the optimal scene is achieved, and the implicit conflict of multiple scenes is avoided.

It is understood that the scene recommendation result further includes: and associating the household appliances in the associated scene.

It should be noted that the associated household appliance in the associated scene is extra recommendation information, and with the recommendation of the associated household appliance, the user can more clearly know the relationship between the associated scene and the household appliance, which is beneficial to further controlling the household appliance.

For example, after a scene query request of "what scene the intelligent mattress corresponds to" is acquired, a scene subgraph related to the intelligent mattress in the household appliance knowledge graph is searched, and finally, a scene related to the intelligent mattress is recommended to be a sleep scene, and in addition, electrical appliances under the sleep scene are also recommended to be as follows: air-conditioning, intelligent mattress, humidifier, etc.

It is understood that the scene recommendation result further includes: and associating the association relation of the household appliances.

For example, when "what relationship the smart mattress and the door sensor have" is acquired, in addition to recommending a sleep scene and appliances in the scene, the relationship between the appliances and the sleep scene, such as the temperature and time set by the air conditioner in the sleep scene, whether the humidifier is on, and the like, is also output.

The difficulty of multi-scene linkage in the existing invention or technology also includes: multi-scenario device priority conflicts. Specifically, the existing method usually performs completely isolated independent analysis on multiple simultaneous activities, outputs the recognition result as a unit based on a single scene, and does not consider the details that the priority of the device possibly caused by multiple scenes simultaneously needs to be recognized, which causes mutual interference.

According to the scene recommendation method, the established household appliance knowledge graph comprises the multi-dimensional association relationship between the household appliances and the scene, wherein the multi-dimensional association relationship comprises the associated priorities of the household appliances in each scene, after the scene query request is obtained, the household appliance knowledge graph is used for identifying and mining, details which need to be identified according to the equipment priorities possibly caused by a plurality of scenes are considered, and the target associated household appliances in the associated scene are finally recommended by identifying different priorities of the household appliances in different scenes. According to the method, the predicted association scene can be obtained firstly through the association relation of the household appliance entity in the household appliance knowledge graph contained in the scene query request, and then the electric appliances which are possibly operated under the association scene are obtained.

It is understood that the scene recommendation result further includes: the start-up sequence of the associated appliance.

It should be noted that, when the scene recommendation is performed based on the scene query request in combination with the household appliance knowledge graph, the start sequence of the associated household appliances needs to be determined in combination with the priorities of the associated household appliances. The priority is generally determined by the combined use time length of the equipment and the execution sequence obtained from semantic information when the knowledge graph is constructed, and the starting sequence of the electric appliances with high priority is advanced.

The scene recommendation method can effectively solve the recommendation problem of the associated electrical appliances in the multi-scene linkage situation. For example, when an electrical appliance under a certain queried scene belongs to multiple scenes, the existing method usually performs completely isolated independent analysis on multiple simultaneous activity scenes, outputs an identification result based on a single scene as a unit, and does not consider the situation that the multiple scenes simultaneously cause the mutual interference of different scenes due to different execution sequences of the electrical appliance under each scene. The method of the embodiment of the application can acquire and output the optimal path among the associated household appliances based on the priority of the household appliances, and determines the starting sequence of the associated household appliances according to the priority.

It can be understood that, querying the household appliance knowledge graph based on the entity to be associated to determine the scene recommendation result includes:

step 201, inquiring a household appliance knowledge graph based on an entity to be associated, and determining an alternative association scene with an association relation with the entity to be associated;

it should be noted that a plurality of alternative association scenarios may be included, but all of the alternative association scenarios need not be output, and therefore, further filtering needs to be performed on the alternative association scenarios.

202, screening alternative association scenes to determine an association scene of an entity to be associated;

it should be noted that the screening of the alternative association scenes includes primary screening and fine screening, and the alternative association scenes corresponding to the required association relationship can be screened through the association relationship between the entity to be associated and the alternative association scenes during the primary screening. Then, when fine screening is carried out, a scene with the highest semantic contribution degree and the highest degree centrality with the demand information needs to be found according to the demand information obtained through semantic recognition and the degree centrality of the household appliance nodes. Meanwhile, only one determined associated scene is needed, and the number of the associated scenes is determined according to the screening result.

Step 203, determining the associated household appliances in the associated scene and the starting sequence of the associated household appliances based on the priorities of the household appliances in the associated scene of the entity to be associated.

It should be noted that, in the execution of step 203, it is first necessary to perform path screening and degree-centric screening on all the home appliance nodes in the determined association scene to obtain all the home appliances on the target path as associated home appliances, and then it is necessary to determine the start order of the associated home appliances according to the start priority marked on the selected associated home appliances, and finally output the start order. The optimal entity items of the local entity link and the global entity link under a specific scene are searched, and the problem of scene conflict in the same household appliance is solved.

According to the scene recommendation method, the most appropriate candidate associated scene can be determined by using the degree centrality of the knowledge graph, and then the multi-electric appliance combined use condition can be found in the structural level mining scene to find the target path between the electric appliances and the scene. Meanwhile, graph-based reasoning can be performed for self-discovery of new scenes. The degree centrality of the embodiment of the application is combined with a weight distribution strategy based on the semantic contribution degree, the control conflict can be judged according to the priority, and the condition that household appliances are interfered with each other and contradicted when a plurality of intelligent scenes are overlapped and linked is solved.

It can be understood that screening the alternative association scenarios to determine the association scenario of the entity to be associated includes:

determining alternative association scenes with conflict relations between the alternative association scenes and entities to be associated;

and deleting the alternative association scenes with conflict relations with the entities to be associated from the alternative association scenes, and determining the association scenes of the entities to be associated.

It should be noted that a conflict may occur between different alternative association scenarios, such as home and away scenarios. Therefore, when a scene with a conflict relationship is screened, the conflict situation needs to be reported to the user, and the conflict scene needs to be screened out from the alternatives. Further, since the determination of the existence of the conflict is not necessarily correct, it is necessary to obtain user feedback based on the report and determine whether or not the relationship in the home appliance knowledge map needs to be corrected or supplemented.

Meanwhile, in some cases, a user may need to query an associated scene and a conflicting scene through a scene, for example, "what is the associated scene and the conflicting scene of the query and the return scene, respectively", at this time, when screening candidate associated scenes of a scene query request, a candidate associated scene having a conflicting relationship with a target scene is retained, and an associated scene of the target scene is determined.

According to the scene recommendation method, the effectiveness of screening the final association scene can be improved by eliminating the alternative association scenes with conflict relations.

It is to be understood that, based on the scenario query request, determining the entity to be associated includes:

extracting a text of the scene query request to obtain entity information, wherein the entity information comprises an entity and the state of the entity;

and screening entity information based on the semantic contribution degree, and determining the entity to be associated.

It should be noted that the instruction input by the user is likely to be irregular, and therefore the acquired scene query request may be unclear, and therefore text extraction is often required to be performed on the instruction information in combination with the semantic contribution degree. After text extraction, entity information including the entity and the state information of the entity is obtained. At this time, an entity with a higher semantic contribution degree needs to be preferentially selected as an entity to be associated, and the entity to be associated also includes corresponding state information. Taking the entity to be associated as the scene entity as an example, the instruction information input by the user may be "close the window", the entity extracted by the text is the window at this time, the state information is closed, the entity to be associated is determined to be "close the window scene" at this time, and the scene query instruction may be: "what scene is associated with the scene of the window". Taking the entity to be associated as a home appliance entity as an example, the scene instruction information input by the user may be "air conditioner and humidifier", the entity information extracted from the text at this time is the air conditioner and humidifier, the state information is all on, the entity to be associated is determined to be "air conditioner on" and "humidifier on" at this time, and the scene query instruction may be: "what scenario the air conditioner is turned on and the humidifier is turned on".

It is understood that determining a scenario query request further includes:

acquiring user information;

associating the user information with the household appliance knowledge graph, and determining a household appliance entity associated with the user information and a scene entity associated with the user information;

It should be noted that the updated household appliance knowledge graph is also used for representing the association relationship between the user and the household appliance and between the user and the scene. By adding the information of the user into the household electrical knowledge map, the user information can be the portrait of the user, such as sex, age, height, occupation and the like, and more personalized scene recommendations can be made.

Besides, it should be noted that the user information further includes behavior information of the user in different scenarios, for example, usage frequency of the user on the electrical appliance and user label information in different scenarios.

It should be noted that, the use frequency of the electrical appliance by the user in different scenes determines the priority of the electrical appliance entity, and in the embodiment of the present application, the use frequency of different electrical appliances is detected, and the threshold of the common occurrence times of the electrical appliance and the scene is set to perform scene self-discovery. For example, if two electrical appliances are used at the same time in high frequency, whether a home scene exists in the two electrical appliances needs to be considered, and the knowledge graph is updated according to the situation. The user labeling information can be manually labeled and corrected aiming at problems in practical application, and the user labeling information can further update, correct and complement a scene library of the knowledge graph. In addition, for users with high similarity of the electric appliance use behavior data, the same electric appliance use prediction can be made moderately based on the map.

In the aspect of map updating, the household electrical information concept layer and the data layer are updated according to the new data flow about the user. And automatically adding new electric appliance and scene concepts to a concept layer of the electric appliance encyclopedia knowledge base. The user labeling information can consider the reliability and consistency of data sources, eliminate contradictions and redundancies of household appliance information, and select scenes and relations with high frequency in the user using information of each user to be added into the household appliance knowledge graph.

According to the scene recommendation method, the abnormal data can be conveniently marked and corrected by injecting new data streams such as behavior information of the user in different scenes, and new electric appliances and scenes are added into the knowledge graph. Other methods only relate to updating of the appliance-scene knowledge graph, inference of multi-appliance use conditions under the scene based on the knowledge graph, possible use scene prediction of a user and user interactive recommendation are all within the protection range of the application.

determining an entity to be associated for associating user information;

inquiring the updated household appliance knowledge map based on the entity to be associated of the associated user information, and determining a scene recommendation result;

It should be noted that, by identifying the entity to be associated of the associated user information in the scene query instruction, a specific scene with user preference may be queried, for example, the user information is added to the household electrical appliance knowledge graph as a female, and a plurality of associated scenes and associated household electrical appliances may be associated correspondingly. For another example, a recommended scene may be queried through a user tag by adding a home appliance knowledge graph of user information, where a scene query instruction is: "what the recommended scene is for the child".

The scene recommendation method can predict the scene and other possible operating electrical appliances in the scene according to the association under the condition that a few operating electrical appliances are known based on the household electrical appliance relation prediction of the knowledge graph. Based on the similarity of the users, the use habits among the users are used for reference, the prediction updating of the new scene is made, and similar recommendations are made for the users with the same use habits.

It can be understood that, the method for constructing the knowledge graph of the household appliance provided by the embodiment of the present application at least includes the following steps:

step 301, collecting household appliance information;

step 302, performing entity extraction on the household appliance information, and determining a household appliance entity and a scene entity;

step 303, extracting the relationship of the household appliance information and the incidence relationship between the household appliance entity and the scene entity;

and 304, matching the household appliance entity and the scene entity by combining the relationship between the household appliance entity and the scene entity, and determining the household appliance knowledge graph.

In step 301, it should be noted that the home appliance information is obtained by extracting structured, semi-structured, and unstructured data from a relational database or a graph database such as an internet corpus, a public knowledge graph, and manually marked home scene data, and in an embodiment, the database is an RDF ontology.

Specifically, the structured data is derived from scenes and household appliance data predefined by service personnel according to the use condition of the user. The semi-structured data includes but is not limited to XML, JSON, encyclopedia, and open source graph databases such as Freebase, etc., the unstructured data can be obtained from internet text corpus, and in addition, pictures, audio, video, etc. are all used as protected household appliance knowledge information sources, and other methods are also within the protection scope of the patent as long as the methods relate to constructing household appliance related graph databases.

With respect to step 302 and step 303, it should be noted that step 302 introduces a schema layer on the basis of the RDF ontology library obtained in step 301. The method comprises the steps of performing entity-based and relation-based combined extraction on data of household appliance information, screening existing information based on a coarse-grained and fine-grained method, obtaining the most probable entity items, extracting basic attribute knowledge, association knowledge, event knowledge, time sequence knowledge, resource knowledge and the like related to the household appliance entity in a scene, storing the basic attribute knowledge, the association knowledge, the event knowledge, the time sequence knowledge, the resource knowledge and the like in a knowledge base mode layer, constructing an RDFs triple mode base, and establishing an OWL household appliance-scene base after completing equivalence and mutual exclusion relations in the RDFs triple mode base scene.

The entity extraction in the embodiment of the application refers to performing coarse-grained entity identification on semi-structured data and unstructured data, and aims to obtain potential entity candidates and then obtain the most possible entity items based on a fine-grained method. And the relation extraction is to finish automatic household appliance-scene relation extraction by using a machine learning method according to the vocabulary relation template and the syntax relation template after the entity is determined. Extracting the relation between the electric appliance entity and the electric appliance-scene, relating to the action of the electric appliance entity, setting a unique ID number and forming an RDFs triple pattern library.

Specifically, the coarse-grained method comprises a dictionary-based, fuzzy matching-based and word vector-based extraction method, the entity extraction method comprises a rule-based, dictionary-based and machine learning method, such as a hidden Markov model, a conditional random field, a support vector machine, a deep learning method, a transfer learning method and other methods, to extract fine-grained household appliance-scene entity information, and according to the selected household appliance entity and context information, information of time types, such as using switch time, date, digital types, such as model and appliance price, and the like, which can imply the relationship of hierarchy, mutual exclusion and the like between the household appliance-scene entities, so as to form an OWL household appliance-scene library. Other methods are within the scope of the embodiments of the present application as long as they relate to the extraction of the home appliance-scene entity.

For step 304, it should be noted that, the matching performed in combination with the relationship between the household appliance entity and the scene entity refers to performing entity matching and structure matching on the schema layer data. The entity matching means that the entity matching is aligned at the level of the household appliance entity in pairs and in coordination, so that the unified and reference resolution of the household appliance entity is realized. The structure matching refers to mining the data rule in the ontology base at the structure level and grouping and aggregating the household appliances operating in the same scene by combining the time sequence knowledge.

Because the extracted household appliance entity information may have redundancy, the household appliance-entity relationship presents the layering problems of flattening and lack of priority. And reasoning and calculating the relation information in the existing OWL library on the basis of the logic, graph and deep learning of the entity level and the structure level of the household appliance-scene. And performing similarity calculation according to the relation between the entity vectors, achieving the effect of entity unification and reference resolution, and removing redundant household appliance entities. And establishing a structured and networked knowledge system, quantizing the credibility of the household appliance-scene, and abandoning the household appliance-scene relation with lower confidence coefficient.

Specifically, entity matching can be achieved by embedding a knowledge information graph into vectorization through the generated OWL library information, disambiguation and coreference resolution of household appliances and scene entities are achieved through triple confidence and entity similarity calculation, correct household appliance entities are confirmed to be linked to corresponding entities of a knowledge base, and hierarchical construction based on data clustering, pattern tree mining and hierarchical loss standardization is achieved.

The structure matching can be realized by combining an external knowledge base, checking all entities connected under a certain scene by using methods such as integer programming and the like, performing conflict checking by using a voting method, an AFET model and the like, processing conflicts between a data layer and a mode layer, and cleaning redundant and wrong entity relations. Other methods are within the scope of the embodiments of the present application as long as the methods relate to entity linking, data clustering mining, and denoising in the building of the home appliance-scene graph.

The household appliance knowledge graph constructed in the embodiment of the application is shown in fig. 3, and in the knowledge graph, the household appliance comprises: lampblack absorber, gas-cooker, intelligent mattress, door magnetic sensor, air conditioner, air quality sensor, intelligent lock, treadmill, water heater and gas alarm. The scene comprises the following steps: cooking, sleeping, windowing, leaving home, returning home, bathing, exercising, and speaking. C1, C2 and C3 represent priority relations, the priorities of the priority relations decrease from C1 to C3, C1, C2 and C3 represent a high priority relation, a medium priority relation and a low priority relation respectively, and ME represents that mutual exclusion relation (mutual exclusive) exists between two nodes.

It can be understood that matching the home appliance entity and the scene entity in combination with the relationship between the home appliance entity and the scene entity includes:

vectorizing the household appliance entity and the scene entity, and determining a household appliance vector and a scene vector;

It should be noted that, aiming at the generated OWL household appliance-scene library, household appliance entities and scene entities are embedded into a vectorization mode, the disambiguation and coreference resolution of the household appliances and the scene entities are carried out through triple confidence and entity similarity calculation, the problems of homonymy ambiguity and homonymy multi-index are solved, the correct household appliance entities are confirmed to be linked to the corresponding entities of the knowledge base, and hierarchical construction based on data clustering, pattern tree mining and hierarchical loss standardization is carried out.

According to the scene recommendation method, the entities and the relations thereof are converted into vectors through the graph embedding operation, multi-attribute embedded coding can be performed, and the method has expandability. The method can obtain qualitative relationships among various electric appliances based on content similarity, and has more applications in the aspects of follow-up intelligent question answering, multi-turn conversation and the like of the household appliances.

The construction process schematic diagram of the household appliance knowledge graph in the embodiment of the application is shown in fig. 2, and the construction process schematic diagram comprises four steps of (1) household appliance encyclopedia construction, (2) household appliance-scene knowledge extraction, (3) household appliance-scene information fusion, and (4) new scene self-discovery and updating, wherein multiple methods may exist for realizing each step.

For step (1), the embodiment of the application constructs the household appliance encyclopedia knowledge by collecting the structured data, the semi-structured data and the unstructured data, and inputs the household appliance encyclopedia knowledge into step (2). The structured data is derived from scenes and household appliance data predefined by service personnel according to the use condition of a user. The semi-structured data comprises but is not limited to XML, JSON, encyclopedia and open source graph databases such as Freebase and the like, the unstructured data can be obtained from an Internet text corpus, and in addition, pictures, audio, video and the like are used as protected household appliance knowledge information sources.

As for the step (2), coarse-grained entity identification is performed on the semi-structured data and the unstructured data in the step (1) to extract the household appliance entity, the scene entity, and the association relationship between the household appliance entity and the scene entity. The association relationship comprises equivalent and exclusive relationships between the household appliance and the scene entity. Thereby constituting an OWL household appliance-scene library.

As for the step (3), the embodiment of the application aims at the OWL library information generated in the step (2), embeds the knowledge information graph into vectorization, unifies entities and resolves references, solves the problems of homonymy ambiguity and multi-reference of the same entity, and realizes the entity level matching of the household appliances. In addition, combining the structured data generated in the step (2), merging an external knowledge base, grouping and aggregating the household electrical appliance entities in the same scene according to time sequence, and then connecting the global entities with the optimal entity items based on the voice contribution degree and the weight distribution, thereby realizing the matching of the household electrical appliance structure level. And (4) constructing a household appliance knowledge graph through the two-step matching in the step (3).

And (4) for the knowledge graph constructed in the step (3), injecting new data streams such as real behavior information and user portrait, and performing semantic visualization on the knowledge information of the graph through human-computer interaction. And marking and correcting abnormal data, adding a new electric appliance and a new scene into the knowledge graph, and setting a threshold value of the co-occurrence times of the electric appliance and the scene to perform self-discovery of the scene. In addition, for users with high similarity of the appliance use behavior data, the same appliance use prediction can be made moderately based on the map.

According to an embodiment of the application, the construction of the household appliance knowledge graph further comprises:

and creating a new scene ontology in the household appliance knowledge graph based on the change of the incidence relation of the household appliance entities in different time periods.

It should be noted that, the change of the association relationship of the household appliance entities in different time periods is monitored, the latest state of the household appliance entity can be known in real time, and the existing household appliance knowledge graph is further subjected to scene mining according to the latest state, so that the self-discovery of the scene is realized. Meanwhile, the inference of the use condition of multiple electric appliances under a new scene can be realized based on the household appliance knowledge graph, or a new scene body is created according to the user preference by adding the use information of the user. The change of the incidence relation can be a parameter reduction process of the incidence relation between the household appliances, and through observation, the household appliances which densely appear in the same time period can be divided into new scenes and added into a household appliance knowledge map, and mining can be carried out according to the priority of the household appliances to obtain new step scenes. For example: the game on the computer is firstly opened, then the air conditioner is opened, the curtain is closed in sequence, the intelligent lamp is turned on, and the game scene is created.

According to one embodiment of the application, a new scene ontology is created in a household appliance knowledge graph based on changes of incidence relations of household appliance entities in different time periods, and the method comprises the following steps:

determining the use time and priority of the household appliance entity in different time periods based on the household appliance relation heat map;

It should be noted that, in the embodiments of the present application, semantic visualization of a knowledge graph of a home appliance may also be performed, and a home appliance relationship heat map is constructed, where the home appliance relationship heat map represents the use time and priority of each home appliance at different time periods in a day. And searching common electric appliances and common behavior patterns of the user by combining the distribution of the using time of each point in each time period every day. For example, step-type scenes can be mined based on the priority of the electric appliances, common electric appliances and behavior patterns of users can be mined based on the use time, linkage-type scenes can be mined through the co-occurrence relation in the same time, and the like.

For example, in the embodiment of the present application, the knowledge information of the household appliance knowledge graph shown in fig. 3 may be semantically visualized, a heat map of dense relationships is created through a parameter restoration process of the electrical appliance relationships, and a plurality of electrical appliance-scene relationships are visualized as the heat map, as shown in fig. 4, priorities of an air conditioner, an intelligent door lock, a door magnetic sensor, an air quality sensor, a water heater, and a voice household appliance at different time periods in a day are displayed, and as shown in the figure, a darker color indicates a higher priority. The abnormal data can be conveniently marked and corrected through visual display, new electric appliances and scenes are added into the knowledge graph, and the threshold value of the co-occurrence times of the electric appliances and the scenes is set to perform self-discovery of the scenes. The embodiment of the application can realize the visual expression map of the entity relationship of the household appliances in the scene, automatically detect the new scene, update the concept layer and the data layer after actual application and complete further scene discovery.

The following describes the scene recommendation device provided in the present application, and the scene recommendation device described below and the scene recommendation method described above may be referred to in correspondence with each other. As shown in fig. 5, the embodiment of the present application further discloses a fall detection device, including:

an input module 501, configured to determine a scene query request;

the analysis module 502 is configured to determine an entity to be associated based on the scene query request;

the recommending module 503 is configured to query the household appliance knowledge graph based on the entity to be associated, and determine a scene recommending result;

The scene recommendation device provided by the embodiment of the application utilizes the knowledge map to replace manual labeling, saves manpower, improves the quality of entity labeling, and is convenient for updating, correcting and completing the scene library. Compared with the traditional supervised and unsupervised learning methods, the method has the advantages that the automation degree, the missing judgment rate and the interpretability are greatly improved, and the cross-platform applicability is strong while the personalized features of the user are kept.

It is understood that the scene recommendation result further includes: and (3) starting the associated household appliances and/or the associated household appliances in the associated scene.

inquiring the household appliance knowledge graph based on the entity to be associated, and determining an alternative association scene with association relation with the entity to be associated;

and determining the associated household appliances and the starting sequence of the associated household appliances in the associated scene based on the priority of the household appliances in the associated scene of the entity to be associated.

It can be understood that, screening the alternative association scenarios to determine the association scenario of the entity to be associated includes:

determining alternative association scenes with conflict relations with entities to be associated in the alternative association scenes;

It is understood that determining a scenario query request further includes:

acquiring user information;

determining an entity to be associated for associating user information;

It can be understood that the construction of the knowledge graph of the household appliance comprises the following steps:

collecting household appliance information;

and matching the household appliance entity and the scene entity by combining the relationship between the household appliance entity and the scene entity to determine a household appliance knowledge graph.

It can be understood that the construction of the knowledge graph of the household appliance further comprises the following steps:

It can be understood that, based on the change of the association relationship of the home appliance entities in different time periods, a new scene ontology is created in the home appliance knowledge graph, which includes:

visualizing the change of the association relationship of the household appliance entity in different time periods to obtain a household appliance relationship heat map;

Fig. 6 illustrates a physical structure diagram of an electronic device, which may include, as shown in fig. 6: a processor (processor) 610, a communication Interface (Communications Interface) 620, a memory (memory) 630 and a communication bus 640, wherein the processor 610, the communication Interface 620 and the memory 630 communicate with each other via the communication bus 640. The processor 610 may call logic instructions in the memory 630 to perform the following method:

determining a scene query request;

determining an entity to be associated based on the scene query request;

In addition, the logic instructions in the memory 630 may be implemented in software functional units and stored in a computer readable storage medium when the logic instructions are sold or used as independent products. Based on such understanding, the technical solution of the present application or portions thereof that contribute to the related art in essence may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.

In another aspect, embodiments of the present application disclose a computer program product, the computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program includes program instructions, and when the program instructions are executed by a computer, the computer can perform the methods provided by the above-mentioned method embodiments, for example, the methods include:

determining a scene query request;

determining an entity to be associated based on the scene query request;

inquiring a household appliance knowledge graph based on the entity to be associated, and determining a scene recommendation result;

In another aspect, the present application further provides a non-transitory computer-readable storage medium, on which a computer program is stored, where the computer program is implemented by a processor to execute the transmission method provided in the foregoing embodiments, for example, the method includes:

determining a scene query request;

determining an entity to be associated based on the scene query request;

The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.

Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment may be implemented by software plus a necessary general hardware platform, and may also be implemented by hardware. With this in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods of the various embodiments or some parts of the embodiments.

Finally, it should be noted that the above embodiments are only for illustrating the present application, and do not limit the present application. Although the present application has been described in detail with reference to the embodiments, it should be understood by those skilled in the art that various combinations, modifications, or equivalents may be made to the technical solutions of the present application without departing from the spirit and scope of the technical solutions of the present application, and the technical solutions of the present application should be covered by the scope of the present application.

Claims

1. A method for scene recommendation, comprising:

determining a scene query request;

determining an entity to be associated based on the scene query request;

2. The scene recommendation method according to claim 1, wherein said scene recommendation result further comprises: and (3) starting the associated household appliances and/or the associated household appliances in the associated scene.

3. The scene recommendation method according to claim 2, wherein the querying a household appliance knowledge graph based on the entity to be associated to determine a scene recommendation result comprises:

4. The scene recommendation method according to claim 3, wherein the screening the candidate association scenes to determine the association scene of the entity to be associated comprises:

determining alternative association scenes with conflict relations with the entities to be associated in the alternative association scenes;

5. The scene recommendation method according to claim 1, wherein said determining the entity to be associated based on the scene query request comprises:

6. The scenario recommendation method according to any of claims 1 to 5, wherein the determining of the scenario query request further comprises:

acquiring user information;

7. The scene recommendation method according to claim 6, wherein the querying a household appliance knowledge graph based on the entity to be associated to determine a scene recommendation result comprises:

determining an entity to be associated for associating user information;

8. The scene recommendation method according to any one of claims 1 to 5, wherein the building of the appliance knowledge graph comprises:

collecting household appliance information;

extracting the relationship of the household appliance information and the incidence relationship between the household appliance entity and the scene entity;

9. The scene recommendation method according to claim 8, wherein matching the home appliance entity and the scene entity in combination with a relationship between the home appliance entity and the scene entity comprises:

10. The scene recommendation method according to claim 8, wherein the building of the appliance knowledge graph further comprises:

11. The scene recommendation method according to claim 10, wherein the creating a new scene ontology in the home appliance knowledge graph based on a change of an association relationship of the home appliance entity in different time periods comprises:

12. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the scene recommendation method according to any one of claims 1 to 11 when executing the program.

13. A non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the scene recommendation method according to any one of claims 1 to 11.

14. A computer program product comprising a computer program, characterized in that the computer program, when being executed by a processor, implements the scenario recommendation method according to any of the claims 1 to 11.