Semantic mapping method and device for tasks

Abstract

The invention discloses a task semantic mapping method and device, and the method comprises the steps: obtaining a multi-source multi-mode data set of a target task, extracting the semantic features of the target task, and the semantic features related to the scene context of the target task, a target task semantic feature set and a scene context semantic feature set related to the scene context of the target task are obtained, and scene context semantic features are determined based on information provided by a user; fusing the scene context semantic feature set with the target task semantic feature set to obtain an influence space of the scene context semantic feature on the target task semantic feature; under the influence space, semantic features equivalent to or similar to core features in the semantic features of the target task are judged; and linking all the semantic features based on a preset rule. A solution is provided for semantic integration between heterogeneous models, and the cognition and discrimination result of a target task is achieved.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a task semantic mapping method and device. Background Art

[0002] The study of similarity theory originated from cognitive science. Similarity calculation is the core of the semantic mapping process. Research results show that similarity is established under a certain contextual premise. The same object may have different associations under different cognitions, and there are some concepts that are difficult to establish correspondence without limiting the target context. In addition, the best similarity match is often related to the context of user needs, focus, angle, etc., which is rarely reflected in existing cognitive computing methods, which easily makes the similar meaning as the calculation result ambiguous or even meaningless. Especially in multi-source and multi-modal data fusion across domains, industries, and time and space target tasks, the focus on similar information is usually different due to different scenarios and fields. Therefore, it is very meaningful to consider the influence of context when fusion similarity factors of multi-source and multi-modal data.

[0003] In addition, in terms of specific semantic mapping algorithms, various existing algorithms either use the similarity transitivity of adjacent concept nodes or set thresholds to determine the similarity between ontology concepts. They often do not consider the influence of scene contexts such as the user's focus, perspective, and needs, and cannot achieve the best similarity match. Summary of the invention

[0004] The main purpose of the present invention is to provide a task semantic mapping method and device to solve the technical problems existing in the related art.

[0005] According to a first aspect of the present invention, a semantic mapping method for a task is provided, comprising acquiring a multi-source multimodal dataset of a target task, extracting semantic features of the target task and semantic features related to the scene context of the target task, obtaining a target task semantic feature set and a scene context semantic feature set related to the scene context of the target task, wherein the scene context semantic features are determined based on information provided by a user; fusing the scene context semantic feature set with the target task semantic feature set to obtain an influence space of the scene context semantic features on the target task semantic features; in the influence space, determining semantic features that are equivalent to and similar to the core features in the target task semantic features; and connecting all semantic features based on preset rules.

[0006] Optionally, fusing the scene context semantic feature set with the target task semantic feature set includes: extracting standard semantic features related to the target task in the target task semantic feature set; and fusing the scene context semantic feature set with the target task semantic feature set based on a fuzzy C-means clustering algorithm.

[0007] Optionally, determining semantic features that are equivalent to and similar to core features in the semantic features of the target task includes: constructing a correspondence between conceptual entities of the target task and business conceptual entities of the target task under semantic features of different scenario contexts.

[0008] Optionally, the method further includes fusing the scene context semantics with the target task to determine the semantic features of the scene context when establishing a correspondence between the concept entity of the target task and the business concept entity of the target task.

[0009] According to a second aspect of the present invention, there is provided a semantic mapping device for a task, comprising a feature extraction unit, configured to obtain a multi-source multi-modal data set of a target task, extract semantic features of the target task and semantic features related to a scene context of the target task, and obtain a target task semantic feature set and a scene context semantic feature set related to the scene context of the target task, wherein the scene context semantic features are determined based on information provided by a user;

[0010] A fusion unit is configured to fuse the scene context semantic feature set with the target task semantic feature set to obtain an influence space of the scene context semantic features on the target task semantic features; a mapping unit is configured to determine semantic features that are equivalent to and similar to the core features in the target task semantic features in the influence space; and all semantic features are connected based on preset rules.

[0011] Optionally, fusing the scene context semantic feature set with the target task semantic feature set includes: extracting standard semantic features related to the target task in the target task semantic feature set; and fusing the scene context semantic feature set with the target task semantic feature set based on a fuzzy C-means clustering algorithm.

[0012] Optionally, determining semantic features that are equivalent to and similar to core features in the semantic features of the target task includes: constructing a correspondence between conceptual entities of the target task and business conceptual entities of the target task under semantic features of different scenario contexts.

[0013] According to a third aspect of the present invention, there is provided a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, wherein the computer instructions are used to cause the computer to execute the method described in any one of the implementations of the first aspect.

[0014] According to a fourth aspect of the present invention, there is provided an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor executes the method described in any one implementation of the first aspect.

[0015] According to a fifth aspect of the present invention, there is provided a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method according to any one of the implementation modes of the first aspect is implemented.

[0016] The semantic mapping method and device of the task of this embodiment include obtaining a multi-source multi-modal data set of the target task, extracting the semantic features of the target task and the semantic features related to the scene context of the target task, obtaining a target task semantic feature set and a scene context semantic feature set related to the scene context of the target task, wherein the scene context semantic features are determined based on information provided by the user; fusing the scene context semantic feature set with the target task semantic feature set to obtain the influence space of the scene context semantic features on the target task semantic features; under the influence space, judging the semantic features that are equivalent to and similar to the core features in the target task semantic features; and connecting all the semantic features based on preset rules. Provide a solution for semantic integration between heterogeneous models to achieve the results of cognition and discrimination of the target task. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 is a flow chart of a semantic mapping method for tasks according to an embodiment of the present invention;

[0019] Figure 2 is a schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0021] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so as to describe the embodiments of the present invention described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0022] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0023] According to an embodiment of the present invention, a semantic mapping method for a task is provided, such as Figure 1 As shown, it includes the following steps 101 to 103:

[0024] Step 101: Obtain a multi-source multimodal dataset of a target task, extract semantic features of the target task and semantic features related to the scene context of the target task, and obtain a target task semantic feature set and a scene context semantic feature set related to the scene context of the target task, wherein the scene context semantic features are determined based on information provided by a user.

[0025] In this step, a multi-source multi-modal dataset of the target task is obtained, the semantic features of the above target task dataset and the semantic features related to the target task scenario context are extracted, and a target task semantic feature set and a semantic feature set related to the target task scenario context are established. The target task class can be summarized as an overall semantic feature class and each specific business semantic feature class. Each feature class has its own member variables, which are mainly attribute objects; attributes are also abstracted into an object class, which also has member variables, including attribute name, attribute type, attribute value, etc.

[0026] For example, collect the index information and semantic feature information of nutrients and nutritional components contained in relevant agricultural products in a certain area, and simultaneously collect the index information and semantic feature information of nutrients and nutritional components and proportions contained in agricultural products in areas with large differences in altitude, climate, etc. Here, agricultural products include: cucumbers, tomatoes, potatoes (potatoes), sweet potatoes, etc.; nutrients include: such as protein, fat, energy, vitamins, carbohydrates, etc.; nutrients include: potassium, sodium, aluminum, etc. Based on the above semantic feature information, establish a comprehensive index information and semantic feature information set of unique quality of agricultural products including variety, soil, climate, nutritional components, etc. At the same time, obtain descriptive information of consumers' preferences for a specific type of agricultural products (including sensory and quality evaluation, etc.), so as to establish a semantic feature information set of consumers' preferred agricultural product quality preference evaluation (scenario context).

[0027] Step 102: Fusing the scene context semantic feature set with the target task semantic feature set to obtain an influence space of the scene context semantic feature on the target task semantic feature.

[0028] In this step, a scene context fusion model is established based on the fuzzy C-means clustering algorithm. The semantic features in the target task space are generally divided into two categories: scene context related and scene context irrelevant. Here, for the target task related to the scene context, a semantic influence domain representing the user's attention based on a specific task will be generated through fusion. The collection of these semantic influence domains constitutes a scene context influence space.

[0029] As an optional implementation method of this embodiment, the fusion of the scene context semantic feature set and the target task semantic feature set includes: extracting standard semantic features related to the target task in the target task semantic feature set; and fusing the scene context semantic feature set and the target task semantic feature set based on a fuzzy C-means clustering algorithm.

[0030] In this implementation, fusion can be performed in the following two ways:

[0031] 1) Utilizing the rich semantic relationships contained in the domain ontology, the fusion operations of different types of contexts are used to automatically determine, obtain or filter related entities that can affect the calculation of concept similarity, such as giving priority to a certain ancestor among multiple ancestor concepts, and excluding irrelevant items in concept description items.

[0032] 2) Return intermediate results at runtime to provide users with the ability to judge and correct the scene context fusion results.

[0033] The above two methods can be carried out alternately until a satisfactory scene context influence domain is obtained.

[0034] Specifically, based on the correlation probability, the fuzzy C-means clustering algorithm is used to fuse the above semantic feature information set to obtain a fusion result that meets the conditions. A correlation probability threshold can be set to retain the semantic feature information greater than or equal to the correlation probability threshold, and the fuzzy C-means clustering algorithm is used for feature fusion; the standard feature information less than the correlation probability threshold is discarded.

[0035] The specific steps include:

[0036] (1) Obtain the standard semantic feature information of the target task, expressed as X′={x′1,x′ 2 ,…,x′ n}, determine the number of clusters C, divide the standard semantic feature information into C groups, and initialize the cluster center c of each group j , determine the weighted index m, the termination error ε and the maximum number of iterations max; 1≤m≤∞, ε is a very small constant, representing the error threshold.

[0037] (2) Initialize the membership matrix determined by the membership function. The membership matrix should be an n*C matrix, which represents the degree to which each standard feature information belongs to each class.

[0038] (3), in each iteration, the membership values ​​of the membership matrix and the cluster center of each group are updated;

[0039] (4) When the membership value change of the membership matrix in two iterations is less than the termination error ε or reaches the maximum number of iterations max, the algorithm terminates, otherwise, repeat step (3).

[0040] Wherein, step (3) specifically includes:

[0041] During the tth iteration, the membership value calculation formula of the membership matrix is:

[0042]

[0043] Where i = 1, 2, ..., n, j = 1, 2, ..., C, c j t-1 is the cluster center of each group in the t-1th iteration, m is the weighted index, x′ i is the i-th standard feature information;

[0044] During the tth iteration, the calculation formula for each group of cluster centers is:

[0045]

[0046] Among them, U ij tIt represents the membership value of the membership matrix in the t-th iteration process. When the membership value gradually stabilizes, it is the global optimal state.

[0047] In this method, the correlation probability threshold is 0.75. When the absolute value of the correlation coefficient of the standard semantic feature information is greater than or equal to 0.75, the standard semantic feature information is saved; when the absolute value of the correlation coefficient of the standard semantic feature information is less than 0.75, the standard semantic feature information is discarded. The standard semantic feature information of the multi-source heterogeneous data of the target task is further screened by the event tree and the complex correlation coefficient, and the standard semantic feature information with low correlation with the corresponding task information is eliminated, which not only increases the accuracy of multi-source heterogeneous data fusion, but also reduces the computational complexity caused by too much data in the calculation process.

[0048] The standard semantic feature information is first extracted and then fused using the fuzzy C-means clustering algorithm, which achieves the effective fusion of multi-source heterogeneous data when the data type is complex and the data dimension is large, and improves the utilization rate of multi-source heterogeneous data fusion methods in practical applications.

[0049] Step 103: In the influence space, determine the semantic features that are equivalent to and similar to the core features in the semantic features of the target task; and connect all the semantic features based on preset rules.

[0050] In this step, based on the above set and the scene context fusion classification, a layer-by-layer semantic mapping mechanism and similarity mapping model are established to complete three different semantic correspondences of target task semantic equivalence, semantic similarity and semantic connection. Here, in the mapping, the scene context needs to be provided by the user according to the goal and task.

[0051] Acquire a multi-source and multi-modal dataset for the target task, extract the semantic features of the above target task dataset and the semantic features related to the target task scene context, establish a target task semantic feature set and a semantic feature set related to the target task scene context; establish a layer-by-layer in-depth semantic mapping mechanism and similarity mapping model based on the above set to complete three different semantic correspondences of semantic equivalence, semantic similarity and semantic connection for the target task; finally, use the scene context, semantic mapping mechanism and similarity mapping model to provide a solution for semantic integration between heterogeneous models and achieve the cognition and discrimination results of the target task.

[0052] As an optional implementation method of this embodiment, determining semantic features that are equivalent to and similar to the core features in the semantic features of the target task includes: constructing a correspondence between the conceptual entity of the target task and the business conceptual entity of the target task under the semantic features of different scenario contexts.

[0053] In this optional implementation, in the mapping, the scene context needs to be provided by the user according to the goals and tasks. In addition, the specified scene context acts on the definition of the candidate concept set in the mapping process, and participates in the adjustment of the similarity calculation method. Therefore, the mapping result is based on the scene context. The reason for the final similarity value can be understood from the scene context model content. That is, the scene context is used to define or clarify the premise for the generation or existence of semantic mapping between domain models. The context model content will help understand why and in what aspects the knowledge of different domain models can be shared and reused.

[0054] Specifically, the semantic mapping related to the scene context is the mapping Ms-t between the concepts of the target task, and its element is a four-tuple: (s, t, r, Ectx), where s is the concept entity in the target task, s∈S; t is the business concept entity in the target task, t∈T; r is the semantic correspondence between s and t, r∈R, R={equal, similarity, connection}; Ectx is the scene context, which represents the premise assumption for the occurrence of this semantic correspondence.

[0055] Here, given two target objects S and T, mapping requires finding the semantic feature information in T that corresponds to a semantic feature in S. Context is the premise for mapping, and fusion serves mapping. Since each target task shares core semantic feature information, the target task is merged with the core semantic feature during fusion, which is recorded as S∪CPO and T∪CPO.

[0056] As an optional implementation of this embodiment, the method also includes fusing the scene context semantics with the target task to determine the semantic features of the scene context when establishing a correspondence between the concept entity of the target task and the business concept entity of the target task.

[0057] In this optional implementation, the specific fusion process of the scene context and the target task before the semantic correspondence judgment is:

[0058] 1. For Si∈S, obtain the body context acting on Si through the has_context relationship;

[0059] 2. If Si is associated with the internal context imricRel, obtain the class-restricted context TRel;

[0060] 3. For t j ∈T, judge for each b in TRel ck , t j ~CRb ck If not satisfied, return;

[0061] 4. Obtain the attribute-restricted context PREl, the class-independent context TRelax, and the attribute-independent context PRelax for intricRel respectively;

[0062] 5. If TRel and PREl are not empty, call SelectFocusRegion(ζ, Trel, Prel, S∪CPO, s i ,s i , True) and SelectFocusRegion(ζ, Trel, Prel, T∪CPO, t j , t j , True).

[0063] 6. If TRelax and PRelax are not empty, call FilterOutRegion(ξ, TRelax, PRelax, S∪CPO, si, si, True) and FilterOutRegion(ζ, TRelax, PRelax, T∪CPO, t j , t j , True).

[0064] 7. Proceed to the next step of semantic correspondence judgment.

[0065] This optional implementation method is based on the target object semantic mapping technology and method of the scene context model. By utilizing the scene context model, a step-by-step in-depth semantic mapping mechanism and similarity mapping model are established to achieve three different semantic correspondences: semantic equivalence, semantic similarity, and semantic connection, providing a solution for semantic integration between heterogeneous models.

[0066] In the entire framework, first, the scene context semantic features are fused with the target task semantic features to determine their impact space on certain semantic features in the target task. Then, on this basis, corresponding judgments are made based on the semantic features of the core layer of the target task, and semantic equivalence and similarity are obtained based on the semantic similarity judgments of multi-level description features. Equivalence and similarity judgments can be made under the influence of the context influence space. On the basis of semantic equivalence and semantic similarity matching, combined with specific rules, semantic connection judgments are made through reasoning, and finally three semantic correspondences are generated: Fusion is to project the scene context into the target task space

[0067] By using the above mapping mechanism and mapping model, the cognitive and discrimination results of the target task can be achieved. For example, the target scene we set is various watermelons, and the cognitive task is to judge whether it is a good melon or a bad melon, or an unripe melon. Here, the semantic feature information of watermelons includes various colors of watermelons, the sound of watermelons being knocked, the specific gravity of watermelons, the size of watermelons, the elasticity of watermelons, the shape of watermelons, the shape description of the shape of watermelons, etc.; it also includes the semantic feature information of the watermelon scene context: production place, shelf life, storage place, storage time, etc. We turn these descriptions into a set of segmentation words, and the corresponding structures of these segmentation words are good melons and bad melons. By using the above mapping mechanism and mapping model, cognitive calculation can be performed to obtain whether a certain watermelon is a good melon or a bad melon, or an unripe melon. For example: 1. You can look at the color of the watermelon. If the melon skin is smooth, the pattern is clear, the lines are obvious, and the bottom is yellow, it is a ripe melon; if the surface is fuzzy, the luster is dull, the patterns and lines are unclear, it is an unripe melon; 2. You can look at the handle of the watermelon. Green ones are ripe melons; dark brown ones, with falling hairs, bent and brittle, and yellow and withered tips of curls, are melons picked before they are ripe; melons with dried stems are "dead vine melons" with poor quality; 3. You can weigh the watermelon. The ones that float up when put into the water are ripe melons; the ones that sink are raw melons; 4. You can look at the shape of the watermelon. The melons with neat and symmetrical bodies are growing normally and of good quality; the melons with deformed bodies are growing abnormally and of poor quality; 5. You can listen to the sound of the watermelon. If you tap the melon with your fingers and hear a "bang bang" sound, it is a ripe melon; if you hear a "dang dang" sound, it is not ripe yet; if you hear a "pu pu" sound, it is an overripe melon, and so on. These data come from different sources, including sound, shape, weight, color, and evaluation. Generally speaking, good and bad melons can be judged through a single source and modality, but it is not accurate enough. It is only a preliminary judgment, and the cognitive results have a large error. If the data from different sources and in different modes are integrated with cognitive computing, the cognitive ability to distinguish good melons from bad melons will be more scientific, more accurate and more explainable.

[0068] Advantages of the present invention:

[0069] 1) The scene context associates the two sides of the semantic mapping, and objects in the real world can often be associated in multiple aspects. Considering that when the target task knowledge is shared across domains, the domains often obtain the required knowledge based on their own goals. We believe that the association provided by the scene context should mainly reflect or characterize the focus related to user needs and target tasks, and at the same time associate the tasks or goals, as well as the different domain ontology entities that need to be semantically mapped. Some contextual information that affects the judgment of semantic association is related to human cognition and is difficult to quantify. Therefore, the content of the scene context model mainly considers those elements that can act on the mapping calculation, assist reasoning, and directly affect the calculation matching results, so that the mapping results can be dynamically adjusted as the tasks or focuses change, so as to achieve the purpose of meeting the model integration requirements to the greatest extent.

[0070] 2) For the three relationships of semantic equivalence, similarity and connection, the shared core is the target task ontology, the classification hierarchy of business concepts, sibling relationships, descriptive semantics and rules, and other semantic descriptions of different depths are used to obtain them layer by layer through three methods: core ontology-based, multi-level description feature-based, and domain rule-based.

[0071] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0072] According to an embodiment of the present invention, a semantic mapping device for a task is also provided, including a feature extraction unit, configured to obtain a multi-source multi-modal data set of a target task, extract semantic features of the target task and semantic features related to the scene context of the target task, and obtain a target task semantic feature set and a scene context semantic feature set related to the scene context of the target task, wherein the scene context semantic features are determined based on information provided by a user; a fusion unit, configured to fuse the scene context semantic feature set with the target task semantic feature set to obtain an influence space of the scene context semantic features on the target task semantic features; a mapping unit, configured to determine semantic features that are equivalent to and similar to the core features in the target task semantic features under the influence space; and connect all semantic features based on preset rules.

[0073] As an optional implementation method of this embodiment, the fusion of the scene context semantic feature set and the target task semantic feature set includes: extracting standard semantic features related to the target task in the target task semantic feature set; and fusing the scene context semantic feature set and the target task semantic feature set based on a fuzzy C-means clustering algorithm.

[0074] As an optional implementation method of this embodiment, determining semantic features that are equivalent to and similar to the core features in the semantic features of the target task includes: constructing a correspondence between the conceptual entity of the target task and the business conceptual entity of the target task under the semantic features of different scenario contexts.

[0075] As an optional implementation of this embodiment, the method also includes fusing the scene context semantics with the target task to determine the semantic features of the scene context when establishing a correspondence between the concept entity of the target task and the business concept entity of the target task.

[0076] According to an embodiment of the present invention, the present invention also provides an electronic device, which includes: 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, and the instructions are executed by the at least one processor so that the at least one processor can implement the method described in any of the above embodiments when executing.

[0077] According to an embodiment of the present invention, the present invention further provides a readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to implement the method described in any of the above embodiments when executed.

[0078] According to an embodiment of the present invention, the present invention further provides a computer program product, which can implement the method described in any of the above embodiments when executed by a processor.

[0079] Figure 2 A schematic block diagram of an example electronic device 300 that can be used to implement an embodiment of the present invention is shown. The electronic device 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 processing, cellular phones, smart phones, wearable devices, and other similar computing devices.

[0080] like Figure 2 As shown, the electronic device 300 includes a computing unit 301, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 302 or a computer program loaded from a storage unit 308 into a random access memory (RAM) 303. In the RAM 303, various programs and data required for the operation of the electronic device 300 can also be stored. The computing unit 301, the ROM 302, and the RAM 303 are connected to each other via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0081] A number of components in the electronic device 300 are connected to the I / O interface 305, including: an input unit 306, such as a keyboard, a mouse, etc.; an output unit 307, such as various types of displays, speakers, etc.; a storage unit 308, such as a disk, an optical disk, etc.; and a communication unit 309, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 309 allows the electronic device 300 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0082] The computing unit 301 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 301 performs the various methods and processes described above, such as the object matching method. For example, in some embodiments, the object matching method may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 308. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 300 via the ROM 302 and / or the communication unit 309. When the computer program is loaded into the RAM 303 and executed by the computing unit 301, one or more steps of the method described above may be performed.

[0083] Various implementations 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 chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0084] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0085] In the context of the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

Claims

1. A task semantic mapping method, characterized in that: include: Acquire a multi-source multi-modal data set of a target task, extract semantic features of the target task and semantic features related to the scene context of the target task, and obtain a target task semantic feature set and a scene context semantic feature set related to the scene context of the target task, wherein the scene context semantic features are determined based on information provided by a user; The scene context semantic feature set is merged with the target task semantic feature set to obtain an influence space of the scene context semantic feature on the target task semantic feature; In the influence space, the semantic features that are equivalent to and similar to the core features in the semantic features of the target task are determined; based on the preset rules, all the semantic features are connected.

2. The task semantic mapping method according to claim 1, characterized in that: The fusion of the scene context semantic feature set and the target task semantic feature set includes: Extracting standard semantic features related to the target task from the target task semantic feature set; The scene context semantic feature set and the target task semantic feature set are fused based on a fuzzy C-means clustering algorithm.

3. The task semantic mapping method according to claim 1, characterized in that: Semantic features that are judged to be equivalent and similar to the core features in the semantic features of the target task include: Under the semantic features of different scenario contexts, the correspondence between the conceptual entities of the target task and the business conceptual entities of the target task is constructed.

4. The task semantic mapping method according to claim 3, characterized in that: The method also includes fusing the scene context semantics with the target task to determine the semantic features of the scene context when establishing a corresponding relationship between the concept entity of the target task and the business concept entity of the target task.

5. A task semantic mapping device, characterized in that: include: a feature extraction unit configured to obtain a multi-source multi-modal data set of a target task, extract semantic features of the target task and semantic features related to the scene context of the target task, and obtain a target task semantic feature set and a scene context semantic feature set related to the scene context of the target task, wherein the scene context semantic features are determined based on information provided by a user; a fusion unit configured to fuse the scene context semantic feature set with the target task semantic feature set to obtain an influence space of the scene context semantic feature on the target task semantic feature; The mapping unit is configured to determine, in the influence space, semantic features that are equivalent to and similar to the core features in the semantic features of the target task; and to connect all the semantic features based on preset rules.

6. The task semantic mapping device according to claim 5, characterized in that: The fusion of the scene context semantic feature set and the target task semantic feature set includes: Extracting standard semantic features related to the target task from the target task semantic feature set; The scene context semantic feature set and the target task semantic feature set are fused based on a fuzzy C-means clustering algorithm.

7. The semantic mapping device according to claim 5, characterized in that: Determining semantic features that are equivalent to and similar to the core features in the semantic features of the target task includes: constructing a correspondence between the conceptual entity of the target task and the business conceptual entity of the target task under the semantic features of different scene contexts.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the method according to any one of claims 1 to 4.

9. An electronic device, characterized in that: include: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor executes the method described in any one of claims 1 to 4.

10. A computer program product, comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 4.

Images