CN116028654A - Multi-mode fusion updating method for knowledge nodes - Google Patents
Multi-mode fusion updating method for knowledge nodes Download PDFInfo
- Publication number
- CN116028654A CN116028654A CN202310322871.4A CN202310322871A CN116028654A CN 116028654 A CN116028654 A CN 116028654A CN 202310322871 A CN202310322871 A CN 202310322871A CN 116028654 A CN116028654 A CN 116028654A
- Authority
- CN
- China
- Prior art keywords
- knowledge
- node
- nodes
- mode
- modal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 62
- 230000004927 fusion Effects 0.000 title claims abstract description 24
- 238000011156 evaluation Methods 0.000 claims abstract description 121
- 230000001174 ascending effect Effects 0.000 claims abstract description 11
- 238000012163 sequencing technique Methods 0.000 claims description 11
- 238000010606 normalization Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 9
- 238000004364 calculation method Methods 0.000 description 11
- 230000008569 process Effects 0.000 description 10
- 238000004590 computer program Methods 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 230000009193 crawling Effects 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides a multi-mode fusion updating method of knowledge nodes, which is used for determining knowledge mode types included in knowledge data; calculating according to the knowledge modal types of each knowledge node and the knowledge information quantity of the corresponding knowledge modal types to obtain modal evaluation sub-coefficients, and obtaining modal evaluation average coefficients according to the modal evaluation sub-coefficients of all knowledge nodes; ascending order sorting is carried out on the first knowledge nodes according to the modal evaluation sub-coefficients to obtain a knowledge node sequence; selecting a plurality of first knowledge nodes at the front part from the knowledge node sequence as second knowledge nodes, and generating recommended adding mode types corresponding to the second knowledge nodes according to the knowledge mode types of the second knowledge nodes at the current moment; and carrying out fusion updating processing on the knowledge data in the second knowledge node according to the configured new knowledge mode type and/or the new knowledge information.
Description
Technical Field
The invention relates to the technical field of data processing, in particular to a multi-mode fusion updating method of knowledge nodes.
Background
Knowledge graph is essentially a large-scale semantic network with entities, concepts as nodes and various semantic relationships between them as edges. However, existing knowledge maps are mostly represented in text form, which impairs the machine's ability to describe and understand the real world.
In the original single-mode knowledge graph (text mode), if new mode knowledge (images, videos, audios and the like) needs to be fused, manual matching is carried out by staff, so that corresponding nodes have multi-mode knowledge. However, in the prior art, along with the continuous update of the knowledge graph, which nodes are multi-mode and which nodes are single-mode staff cannot be mastered in time, and other mode contents cannot be configured for the nodes preferentially according to different states of the nodes, so that the knowledge graph is not targeted when updated and fused.
Disclosure of Invention
The embodiment of the invention provides a multi-mode fusion updating method of knowledge nodes, which can screen the knowledge nodes according to the different quantity and information quantity of modes of each knowledge node in a knowledge graph, and recommend users to add corresponding knowledge information for the corresponding knowledge nodes, so that the knowledge graph is more targeted when updated and fused.
In a first aspect of the embodiment of the present invention, a method for multi-modal fusion update of a knowledge node is provided, including:
acquiring knowledge data corresponding to each knowledge node in a knowledge graph, and determining knowledge mode types included in the knowledge data, wherein the knowledge data types comprise at least one of a text mode, an image mode, an audio mode and a video mode;
Counting the number of the mode types of the knowledge information files corresponding to all the knowledge mode types in the knowledge node and the knowledge information quantity corresponding to each knowledge information file, and determining a preset evaluation weight corresponding to each knowledge mode type;
comprehensively calculating according to the modal types of the knowledge information files, the knowledge information quantity corresponding to each knowledge information file and a preset evaluation weight to obtain modal evaluation sub-coefficients corresponding to each knowledge node;
obtaining a modal evaluation average coefficient according to the modal evaluation sub-coefficients of all knowledge nodes;
comparing the modal evaluation sub-coefficient of each knowledge node with the modal evaluation average coefficient, taking the knowledge node corresponding to the modal evaluation sub-coefficient smaller than the modal evaluation average coefficient as a first knowledge node, and carrying out ascending order sequencing on all the first knowledge nodes according to the modal evaluation sub-coefficient to obtain a knowledge node sequence;
selecting a plurality of first knowledge nodes at the front part from the knowledge node sequence as second knowledge nodes;
the knowledge mode type of the second knowledge node at the current moment is used as a first knowledge mode type, a second knowledge mode type which the second knowledge node does not have is generated according to the total knowledge mode type and the first knowledge mode type, and a recommended adding mode type is obtained according to the number of the second knowledge mode types;
And receiving a new knowledge information file corresponding to the new knowledge mode type configured by the staff according to the recommended adding mode type, correspondingly storing the new knowledge information file and the previous knowledge information file in a previous knowledge information file storage space, and updating the previous knowledge data in the second knowledge node.
Optionally, in one possible implementation manner of the first aspect, the obtaining knowledge data corresponding to each knowledge node in the knowledge graph, and determining a knowledge mode category included in the knowledge data, includes:
acquiring a storage space corresponding to each knowledge node in a knowledge graph, and calling a knowledge information file in the storage space;
and determining the format of the knowledge information file, and determining corresponding knowledge mode types according to the format of the knowledge information file, wherein each format has the corresponding knowledge mode type.
Optionally, in a possible implementation manner of the first aspect, the obtaining a modal evaluation average coefficient according to the modal evaluation sub-coefficients of all knowledge nodes includes:
obtaining a first total number of all knowledge nodes, summing the modal evaluation sub-coefficients of all knowledge nodes to obtain a sum of sub-coefficients, and calculating according to the first total number and the sum of the sub-coefficients to obtain a modal evaluation average coefficient;
The modal evaluation average coefficient is calculated by the following formula,
wherein ,
mean coefficient for mode evaluation ∈>
Is->
Modal evaluator coefficients of the knowledge nodes, +.>
For the upper limit value of knowledge node, +.>
A quantity value for a knowledge node;
Optionally, in one possible implementation manner of the first aspect, the comparing the modal evaluation sub-coefficient of each knowledge node with the modal evaluation average coefficient, taking the knowledge node corresponding to the modal evaluation sub-coefficient smaller than the modal evaluation average coefficient as the first knowledge node, and performing ascending order on all the first knowledge nodes according to the modal evaluation sub-coefficient to obtain the knowledge node sequence includes:
extracting node labels corresponding to each knowledge node respectively, and setting the node labels of each knowledge node in one-to-one correspondence with corresponding modal evaluation sub-coefficients;
counting node labels with all modal evaluation sub-coefficients smaller than modal evaluation average coefficients to generate a label set, and taking knowledge nodes corresponding to all node labels in the label set as first knowledge nodes;
and carrying out ascending order sequencing on all the node labels in the label set according to the modal evaluation sub-coefficients to obtain a knowledge node sequence based on the node label sequencing.
Optionally, in a possible implementation manner of the first aspect, the selecting a plurality of first knowledge nodes as second knowledge nodes in the first knowledge node sequence includes:
Obtaining the sum of knowledge information amounts corresponding to each first knowledge node in a knowledge node sequence, and obtaining the average information amount of all the first knowledge nodes according to the sum of the knowledge information amounts;
comparing the average information quantity with a preset information quantity to obtain a quantity offset coefficient, and offsetting the preset quantity according to the quantity offset coefficient to obtain a selected quantity;
and selecting a plurality of first knowledge nodes in front in the knowledge node sequence according to the selection quantity as second knowledge nodes.
Optionally, in one possible implementation manner of the first aspect, the obtaining a sum of knowledge information amounts corresponding to each first knowledge node in the knowledge node sequence, and obtaining an average information amount of all the first knowledge nodes according to the sum of knowledge information amounts includes:
obtaining the total number of all the first knowledge nodes in the knowledge node sequence to obtain a second total number, calculating according to the sum of the knowledge information amounts corresponding to all the first knowledge nodes and the second total number to obtain an average information amount, calculating the average information amount by the following formula,
wherein ,
for the average information quantity of all first knowledge nodes, < +.>
For knowledge node sequences Middle->
Sum of knowledge information amounts of the first knowledge nodes,/->
For the upper limit value of the first knowledge node in the sequence of knowledge nodes,/I>
For the number value of the first knowledge node in the sequence of knowledge nodes,/for the first knowledge node>
Is->
Sum of knowledge information amounts of the first knowledge nodes,/->
A text modality class of the first knowledge node +.>
Knowledge information amount corresponding to the knowledge information file, < >>
The upper limit value of the knowledge information file of the text modality class for the first knowledge node,/for>
The first knowledge node is the audio modality class +.>
Knowledge information amount corresponding to the knowledge information file, < >>
An upper limit value of a knowledge information file of the audio modality class for the first knowledge node,
video modality seed for a first knowledge nodeClass->
Knowledge information amount corresponding to the knowledge information file, < >>
The upper limit value of the knowledge information file of the video modality class for the first knowledge node,/for>
The first knowledge node is the first +.>
Knowledge information amount corresponding to the knowledge information file, < >>
The upper limit value of the knowledge information file of the image modality class of the first knowledge node.
Optionally, in one possible implementation manner of the first aspect, the comparing the average information amount with a preset information amount to obtain a number offset coefficient, and offsetting the preset number according to the number offset coefficient to obtain a selected number includes:
The pick-out number is calculated by the following formula,
wherein ,
to select the number +.>
For presetting information quantity->
Is information quantity weight value, +.>
Is a preset number;
and if the calculated selection quantity is judged not to be an integer, rounding the selection quantity.
Optionally, in one possible implementation manner of the first aspect, the generating, with the knowledge mode type of the second knowledge node at the current moment as the first knowledge mode type, a second knowledge mode type that the second knowledge node does not have according to the total knowledge mode type and the first knowledge mode type, and obtaining the recommended adding mode type according to the number of the second knowledge mode types includes:
acquiring the number of recommended categories preset by an administrator;
and if the number of the second knowledge mode types is smaller than or equal to the number of the recommended types, taking all the second knowledge mode types as recommended addition mode types.
Optionally, in one possible implementation manner of the first aspect, the method further includes:
if the number of the second knowledge mode types is larger than the number of the recommended types, extracting priority information corresponding to each second knowledge mode type, wherein the preset priority information corresponding to each knowledge mode type is different;
And sorting all the second knowledge mode types in a descending order according to the priority information, and selecting the second knowledge mode types corresponding to the recommended type number according to the order.
Optionally, in one possible implementation manner of the first aspect, if it is determined that the user actively adjusts the selection number, comparing the adjusted adjustment number with the selection number before adjustment to obtain an adjustment difference, where a value of the adjustment number is a value of the user after adjustment of the selection number;
if the adjustment difference value is greater than 0, increasing and adjusting the preset quantity according to the adjustment difference value and the selection quantity before adjustment;
if the adjustment difference value is smaller than 0, reducing and adjusting the preset quantity according to the adjustment difference value and the selection quantity before adjustment;
the preset number after the increase adjustment or the decrease adjustment is calculated by the following formula,
wherein ,
to increase the adjusted preset number, +.>
To increase the number of adjustments after adjustment, +.>
To increase the adjustment coefficient->
To reduce the number of adjustments after adjustment, +.>
To reduce the adjusted preset number +.>
To reduce the adjustment coefficient.
In a second aspect of the embodiments of the present invention, there is provided a storage medium having stored therein a computer program for implementing the method of the first aspect and the various possible designs of the first aspect when the computer program is executed by a processor.
The multi-mode fusion updating method of the knowledge nodes can comprehensively calculate the knowledge mode type dimension and the knowledge information quantity dimension of all knowledge nodes to obtain the mode evaluation average coefficient of all knowledge nodes. The second knowledge nodes are screened by combining the modal evaluation average coefficients, so that the knowledge nodes with fewer knowledge modal types and fewer knowledge information can be updated preferentially and pertinently when the knowledge graph needs to be updated, the knowledge modal types and the knowledge information quantity of the second knowledge nodes in the knowledge graph can be uniformly increased, and the knowledge of each knowledge node in the knowledge graph is relatively more balanced.
According to the method, when the modal evaluation average coefficient and the modal evaluation sub-coefficient are calculated, a plurality of kinds of dimensionalities such as the text modal kind, the audio modal kind, the video modal kind and the image modal kind and the knowledge information quantity under each dimensionality are comprehensively considered, so that the calculated modal evaluation average coefficient is referenced and calculated dimensionalities are more, and the modal evaluation average coefficient and the modal evaluation sub-coefficient relatively more accurately represent the modal condition and the data quantity condition of the knowledge node.
When the second total number of the second knowledge nodes is determined, calculation is performed according to the average information quantity of all the second knowledge nodes, when the average information quantity of the second knowledge nodes is large, the selection quantity is small, and when the average information quantity of the second knowledge nodes is small, the selection quantity is large. By means of the method, the average information quantity of the second knowledge nodes can be calculated differently, and workers can have proper workload when knowledge is added. In addition, after the staff actively adjusts the selection quantity, the invention continuously adjusts and trains the information quantity weight value, so that the modified information quantity weight value is more in line with the current calculation and application scene.
Drawings
FIG. 1 is a flow chart of a first embodiment of a multi-modal fusion update method of a knowledge node;
FIG. 2 is a flow chart of a second embodiment of a multi-modal fusion update method of a knowledge node.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The terms "first," "second," "third," "fourth" and the like in the description and in the claims and in the above drawings, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the invention described herein may be implemented in sequences other than those illustrated or otherwise described herein.
It should be understood that, in various embodiments of the present invention, the sequence number of each process does not mean that the execution sequence of each process should be determined by its functions and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.
It should be understood that in the present invention, "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements that are expressly listed or inherent to such process, method, article, or apparatus.
It should be understood that in the present invention, "plurality" means two or more. "and/or" is merely an association relationship describing an association object, meaning that there may be three relationships, e.g., a and/or B, may represent: a exists alone, A and B exist together, and B exists alone. The character "/" generally indicates that the context-dependent object is an "or" relationship. "comprising A, B and C", "comprising A, B, C" means that all three of A, B, C comprise, "comprising A, B or C" means that one of the three comprises A, B, C, and "comprising A, B and/or C" means that any 1 or any 2 or 3 of the three comprises A, B, C.
It should be understood that in the present invention, "B corresponding to a", "a corresponding to B", or "B corresponding to a" means that B is associated with a, from which B can be determined. Determining B from a does not mean determining B from a alone, but may also determine B from a and/or other information. The matching of A and B is that the similarity of A and B is larger than or equal to a preset threshold value.
As used herein, "if" may be interpreted as "at … …" or "at … …" or "in response to a determination" or "in response to detection" depending on the context.
The technical scheme of the invention is described in detail below by specific examples. The following embodiments may be combined with each other, and some embodiments may not be repeated for the same or similar concepts or processes.
The invention provides a multi-mode fusion updating method of knowledge nodes, which is shown in figure 1 and comprises the following steps:
step S110, knowledge data corresponding to each knowledge node in the knowledge graph is obtained, knowledge mode types included in the knowledge data are determined, and the knowledge data types include at least one of text modes, image modes, audio modes and video modes. The knowledge graph is composed of a plurality of nodes, each node may correspond to an entity, such as "XX mountain", "XX river" and "XX company", at this time, all knowledge data in the knowledge nodes may be actively added by staff, or may be actively crawled in a database based on a data crawling policy and a data crawling plug-in. There may be a variety of knowledge data types, for example, knowledge data of text modality may be a word introduction to "XX mountain", knowledge data of image modality may be an image of "XX mountain", knowledge data of video modality may be a video of "XX mountain", and so on. The invention is not limited in any way with respect to the specific content of knowledge data of text modality, image modality, audio modality, and video modality.
In one possible implementation manner, as shown in fig. 2, step S110 includes:
step 1101, obtaining a storage space corresponding to each knowledge node in the knowledge graph, and retrieving a knowledge information file in the storage space. Since each knowledge node may store knowledge information files of multiple modes correspondingly, and the modes corresponding to different knowledge nodes may be different, the invention establishes a storage space corresponding to each knowledge node. The invention can acquire the knowledge information files in all the storage spaces, and the modes of the knowledge information files at the moment can be only one or multiple.
Step S1102, determining a format of the knowledge information file, and determining a corresponding knowledge mode type according to the format of the knowledge information file, wherein each format has a knowledge mode type corresponding to the format. When each knowledge node is confirmed in knowledge mode type, the invention can confirm the knowledge mode type, for example, the format of the knowledge information file is txt, excel, word, and the corresponding knowledge information file has a high probability of text. When the format of the knowledge information file is jpg, png, etc., the corresponding knowledge information file is an image with a high probability. The invention can carry out corresponding setting on the format and the knowledge mode types in advance.
And step S120, calculating according to the knowledge mode types of each knowledge node and the knowledge information quantity of the corresponding knowledge mode types to obtain a mode evaluation sub-coefficient corresponding to each knowledge node, and obtaining a mode evaluation average coefficient according to the mode evaluation sub-coefficients of all knowledge nodes. The method and the system can calculate the knowledge information quantity of the knowledge mode type and the corresponding knowledge mode type of the knowledge node, and the knowledge information quantity can be regarded as the data quantity. For example, the knowledge mode category of one knowledge node has two kinds, that is, includes a text mode category and an image mode category, for example, the knowledge information amount of the text mode category may be 50KB, the knowledge information amount of the image mode category may be 120KB, and so on. According to the method, knowledge mode types and knowledge information amounts of corresponding knowledge mode types are combined to calculate, the mode evaluation sub-coefficient corresponding to each knowledge node is obtained, the mode evaluation average coefficient is obtained, and if the mode evaluation average coefficient is higher, the mode dimension number and the mode information amount of the corresponding knowledge map are proved to be relatively larger.
In one possible implementation manner, the step S120 includes:
And counting the number of the mode types of the knowledge information files corresponding to all the knowledge mode types in the knowledge node and the knowledge information quantity corresponding to each knowledge information file, and determining the preset evaluation weight corresponding to each knowledge mode type. The invention counts the number of the modal types of the knowledge information file, for example, one node in the knowledge graph respectively comprises a text modal type and an image modal type, and the number of the modal types at the moment is 2. Because knowledge information files of different modal types have different display modes, and therefore the corresponding information amounts of the knowledge information files of different modal types are different when the same data amount is obtained, the invention sets a preset evaluation weight for each knowledge modal type, and can understand that under the condition that the data amount is certain, the preset evaluation weight of the files which can provide more knowledge information can be larger. For example, a line of words, the possible data amount is smaller when displayed in a text form, and the data amount is larger when displayed in an image or video form, so that the preset evaluation weight of the text mode type is relatively larger.
And comprehensively calculating according to the modal types of the knowledge information files, the knowledge information quantity corresponding to each knowledge information file and the preset evaluation weight to obtain modal evaluation sub-coefficients corresponding to each knowledge node. The method and the system can comprehensively calculate by combining a plurality of dimensions such as the number of the modal types and the knowledge information quantity to obtain the modal evaluation sub-coefficient corresponding to each knowledge node, and if the number of the modal types is larger and the knowledge information quantity is larger, the finally obtained modal evaluation sub-coefficient is relatively larger.
And obtaining a modal evaluation average coefficient according to the modal evaluation sub-coefficients of all knowledge nodes. According to the method, the modal evaluation sub-coefficients of all knowledge nodes are combined to calculate to obtain the modal evaluation average coefficient, the average information quantity of the corresponding knowledge graph can be reflected through the modal evaluation average coefficient, and then the new modal knowledge information is added for the knowledge nodes with relatively less modal class dimension and information quantity dimension selected subsequently, so that the targeted knowledge addition can be carried out on the knowledge nodes with relatively less modal class dimension and information quantity dimension when the knowledge graph is subjected to multi-modal fusion.
In one possible implementation manner, the method for obtaining the modal evaluation average coefficient according to the modal evaluation sub-coefficients of all knowledge nodes includes:
obtaining a first total number of all knowledge nodes, summing the modal evaluation sub-coefficients of all knowledge nodes to obtain a sum of sub-coefficients, and calculating according to the first total number and the sum of the sub-coefficients to obtain a modal evaluation average coefficient. For example, if a knowledge graph is composed of 1000 knowledge nodes, the corresponding first total number is 1000, and the method can sum the modal evaluation sub-coefficients to obtain the sum of the sub-coefficients, so as to obtain the modal evaluation average coefficient of the 1000 knowledge nodes.
The modal evaluation average coefficient is calculated by the following formula,
wherein ,
mean coefficient for mode evaluation ∈>
Is->
Modal evaluator coefficients of the knowledge nodes, +.>
For the upper limit value of knowledge node, +.>
Is the number value of the knowledge node.
By passing through
The sum of the sub-coefficients of all knowledge nodes can be obtained, will +.>
Dividing by the number of knowledge nodes>
The mode evaluation average coefficient can be obtained>
. By->
The calculation coefficient of the number of modality types can be obtained if the number of modality types +.>
The larger the relative, the corresponding modality evaluation sub-coefficient +.>
The larger will be. By->
The sum of knowledge information amounts of knowledge information files of the text mode types can be calculated, and in the calculation process, the method can be combined with the preset evaluation weight corresponding to the text mode types>
And (5) weighting processing is carried out. By->
The sum of knowledge information amounts of knowledge information files of audio mode types can be calculated, and in the calculation process, the method can be combined with preset evaluation weights corresponding to the audio mode types>
And (5) weighting processing is carried out. By->
The sum of the knowledge information amounts of the knowledge information files of the video mode types can be calculated, and in the calculation process, the method and the device can be combined with the preset evaluation weight corresponding to the video mode types>
And (5) weighting processing is carried out. By->
The sum of knowledge information amounts of knowledge information files of image mode types can be calculated, and in the calculation process, the method can be combined with the preset evaluation weight corresponding to the image mode types >
And (5) weighting processing is carried out. The invention can go through->
Obtaining the modal evaluation sub-coefficient of each knowledge node according to the calculation mode of (a), and finally
And obtaining a modal evaluation average coefficient.
And S130, comparing the modal evaluation sub-coefficient of each knowledge node with the modal evaluation average coefficient, taking the knowledge node corresponding to the modal evaluation sub-coefficient smaller than the modal evaluation average coefficient as a first knowledge node, and carrying out ascending order sequencing on all the first knowledge nodes according to the modal evaluation sub-coefficient to obtain a knowledge node sequence. When the modal evaluation sub-coefficient is smaller than the modal evaluation average coefficient, the knowledge mode and the knowledge information quantity in the knowledge node are proved to be relatively less compared with other knowledge nodes in the whole knowledge graph, and at the moment, the method and the device can screen the corresponding knowledge nodes and obtain the first knowledge node. The invention can perform ascending order sequencing on the first knowledge node to obtain a knowledge node sequence. The first knowledge node with less knowledge mode and knowledge information amount synthesis is positioned at the front part of the knowledge node sequence.
In one possible implementation manner, the step S130 includes:
and extracting node labels corresponding to each knowledge node respectively, and setting the node labels of each knowledge node in one-to-one correspondence with the corresponding modal evaluation sub-coefficients. The invention can obtain the node labels corresponding to all the knowledge nodes respectively, wherein the node labels can be the position labels, the identity labels and the like of the corresponding knowledge nodes in the knowledge graph, each knowledge node has a unique node label, and the node labels can be numbers, such as 00001. The invention sets the modal evaluation sub-coefficient of each knowledge node corresponding to the corresponding node label.
And counting node labels with all the modal evaluation sub-coefficients smaller than the modal evaluation average coefficient to generate a label set, and taking knowledge nodes corresponding to all the node labels in the label set as first knowledge nodes. The node labels smaller than the average coefficient of the modal evaluation are counted to obtain a final label set, and the first knowledge nodes corresponding to the label set are knowledge nodes with smaller modal dimension and information dimension.
And carrying out ascending order sequencing on all the node labels in the label set according to the modal evaluation sub-coefficients to obtain a knowledge node sequence based on the node label sequencing. According to the method, all the node labels are sequenced in an ascending order, so that the node labels corresponding to knowledge nodes with smaller modal dimension and information quantity dimension are arranged at the front part of the knowledge node sequence.
Step S140, selecting a plurality of first knowledge nodes at the front part in the knowledge node sequence as second knowledge nodes, and generating recommended adding mode types corresponding to the second knowledge nodes according to the knowledge mode types of the second knowledge nodes at the current moment. Because the number of the knowledge nodes in the knowledge graph is huge, a worker cannot add knowledge to all the first knowledge nodes at one time, the method can select the first knowledge nodes in the knowledge node sequence, determine the second knowledge nodes with corresponding number in a plurality of first knowledge nodes, and generate recommended adding mode types corresponding to the second knowledge nodes by combining the knowledge mode types of the second knowledge nodes at the current moment. Recommending the knowledge information addition of the corresponding knowledge nodes to the staff, so that the more the mode types and the knowledge information amount of the formed knowledge data are after the new knowledge information is added to the second knowledge node.
In one possible implementation manner, the step S140 includes:
and obtaining the sum of the knowledge information amounts corresponding to each first knowledge node in the knowledge node sequence, and obtaining the average information amount of all the first knowledge nodes according to the sum of the knowledge information amounts. The invention can obtain the average information quantity of all the second knowledge nodes, and if the average information quantity is larger, the relative information quantity of all the knowledge nodes in the knowledge graph is proved to be larger.
In one possible implementation manner, the method for obtaining the sum of the knowledge information amounts corresponding to each first knowledge node in the knowledge node sequence, and obtaining the average information amount of all the first knowledge nodes according to the sum of the knowledge information amounts includes:
obtaining the total number of all the first knowledge nodes in the knowledge node sequence to obtain a second total number, calculating according to the sum of the knowledge information amounts corresponding to all the first knowledge nodes and the second total number to obtain an average information amount, calculating the average information amount by the following formula,
wherein ,
for the average information quantity of all first knowledge nodes, < +.>
Is the +.>
Sum of knowledge information amounts of the first knowledge nodes,/- >
For the upper limit value of the first knowledge node in the sequence of knowledge nodes,/I>
For the number value of the first knowledge node in the sequence of knowledge nodes,/for the first knowledge node>
Is->
Sum of knowledge information amounts of the first knowledge nodes,/->
A text modality class of the first knowledge node +.>
Knowledge information amount corresponding to the knowledge information file, < >>
The upper limit value of the knowledge information file of the text modality class for the first knowledge node,/for>
The first knowledge node is the audio modality class +.>
Knowledge information amount corresponding to the knowledge information file, < >>
An upper limit value of a knowledge information file of the audio modality class for the first knowledge node,
the video modality category of the first knowledge node +.>
Knowledge information amount corresponding to the knowledge information file, < >>
The upper limit value of the knowledge information file of the video modality class for the first knowledge node,/for>
The first knowledge node is the first +.>
Knowledge information amount corresponding to the knowledge information file, < >>
The upper limit value of the knowledge information file of the image modality class of the first knowledge node. The invention can perform average calculation on the knowledge information quantity of all the first knowledge nodes to obtain average information quantity. By->
The sum of the knowledge information amounts of all the first knowledge nodes can be obtained by +. >
The average information content of all the first knowledge nodes can be obtained. By->
The sum of knowledge information of the text modality types of one of the first knowledge nodes can be calculated by +.>
The sum of knowledge information of the audio modality types of one of the first knowledge nodes can be calculated by +.>
The sum of knowledge information of the video modality types of one of the first knowledge nodes can be calculated by +.>
The sum of knowledge information of the image modality category of one of the first knowledge nodes may be calculated. />
Is->
The first knowledge nodes sum the knowledge information amounts of all the modal types. />
Comparing the average information quantity with a preset information quantity to obtain a quantity offset coefficient, and offsetting the preset quantity according to the quantity offset coefficient to obtain a selected quantity. The invention can compare the average information quantity with the preset information quantity to obtain the quantity offset coefficient, if the average information quantity is larger, the corresponding quantity offset coefficient is smaller, because the information quantity in the knowledge node is larger, a large amount of content is contained in the knowledge node, and the information is required to be finely compared when being added, so that the time possibly consumed by staff when the knowledge is added is increased, and the selection quantity of the knowledge node is required to be relatively reduced. If the average information quantity is smaller, the corresponding quantity offset coefficient is larger at the moment, and because the information quantity in the knowledge node is smaller, the information quantity can have a small quantity of contents, and at the moment, a great amount of time is not consumed for comparing the information when the information is added (the new knowledge information is compared with the old knowledge information, the knowledge information is prevented from being repeated), so that the time possibly consumed by staff when the knowledge is added is reduced, and the selection quantity of the knowledge node can be relatively increased at the moment.
In one possible implementation manner, the comparing the average information amount with a preset information amount to obtain a number offset coefficient, and offsetting the preset number according to the number offset coefficient to obtain a selected number includes:
the pick-out number is calculated by the following formula,
wherein ,
to select the number +.>
For presetting information quantity->
Is information quantity weight value, +.>
Is a preset number. By->
The number offset coefficient, < > can be obtained>
The difference between the average information amount and the preset information amount can be obtained, the information amount weight value +.>
The information weight value +.A preset by the staff can be adopted, for example, when the number of staff is more, the working efficiency is higher, the working time period is longer, and the knowledge addition workload is more, the information weight value +.>
Can be reduced, corresponding selection quantity +.>
There will be a tendency to increase. Conversely, when the number of staff is smaller, the working efficiency is lower, the working time period is shorter, and the knowledge addition workload is smaller, the information weight value at the moment is +.>
Can be enlarged, corresponding selection quantity +.>
There will be a tendency to decrease.
And if the calculated selection quantity is judged not to be an integer, rounding the selection quantity. In the actual calculation scenario, the obtained number of choices may be an integer, so that the number of choices needs to be rounded, and the processing mode may be large, for example, the number of choices is 11.2, and the number of choices after the large is 12.
And selecting a plurality of first knowledge nodes in front in the knowledge node sequence according to the selection quantity as second knowledge nodes. The invention screens the knowledge node sequences according to the selected quantity to obtain a corresponding quantity of second knowledge nodes, and the second knowledge nodes at the moment can meet the current knowledge adding workload of the staff, wherein the knowledge adding workload can be one day, one week and the like.
And taking the knowledge mode type of the second knowledge node at the current moment as a first knowledge mode type, generating a second knowledge mode type which is not possessed by the second knowledge node according to the total knowledge mode type and the first knowledge mode type, and obtaining recommended addition mode types according to the number of the second knowledge mode types. In order to enable the mode types of knowledge information in the knowledge nodes to be more comprehensive, the mode types of the second knowledge nodes are used as the first knowledge mode types, and the second knowledge mode types which the second knowledge nodes do not have are determined. For example, the first knowledge mode type of the second knowledge node includes a text mode type and an image mode type, and the total knowledge mode type includes a text mode type, an image mode type, an audio mode type and a video mode type, and at this time, the second knowledge mode type that the second knowledge node does not have is the audio mode type and the video mode type.
In one possible implementation manner, the method for generating the second knowledge mode category not possessed by the second knowledge node according to the total knowledge mode category and the first knowledge mode category by using the knowledge mode category possessed by the second knowledge node at the current moment as the first knowledge mode category, and obtaining the recommended adding mode category according to the number of the second knowledge mode categories includes:
the recommended category number preconfigured by the administrator is obtained. The recommended category number may be 1, 2, 3, etc. Taking the example that the number of recommended categories is 1, the invention recommends 1 modal category at this time, namely when the second knowledge node is recommended and added with knowledge information.
And if the number of the second knowledge mode types is smaller than or equal to the number of the recommended types, taking all the second knowledge mode types as recommended addition mode types. For example, the number of the second knowledge mode types is 1, that is, the condition that the number of the second knowledge mode types is less than or equal to the number of the recommended types is achieved, and the present invention uses the second knowledge mode types as recommended addition mode types. For example, if the second knowledge mode type is a video mode type, the recommended adding mode type is the video mode type.
In one possible implementation manner, the technical scheme provided by the invention further comprises:
and if the number of the second knowledge mode types is greater than the number of the recommended types, extracting priority information corresponding to each second knowledge mode type, wherein the preset priority information corresponding to each knowledge mode type is different. In an actual application scenario, the number of the second knowledge mode types may be 2, and at this time, the number of the second knowledge mode types is greater than the number of the recommended types, so that selection needs to be performed among the 2 knowledge mode types at this time.
And sorting all the second knowledge mode types in a descending order according to the priority information, and selecting the second knowledge mode types corresponding to the recommended type number according to the order. If the second knowledge mode types are video knowledge mode types and audio knowledge mode types, the video knowledge mode types and the audio knowledge mode types are ranked according to priority information, and the front second knowledge mode types corresponding to the recommended type number are selected.
If the second knowledge mode type is 0, all the first knowledge mode types are directly ranked at the moment, and the first knowledge mode type corresponding to the recommended type number in the first knowledge mode types is selected as the recommended addition mode type.
Step S150, a new knowledge information file corresponding to the new knowledge mode type configured by the staff is received, the knowledge information file and the previous knowledge information file are stored, and previous knowledge data in the second knowledge node are updated. The invention correspondingly stores the new knowledge information file and the previous knowledge information file in the previous knowledge information file storage space in the receiving of the new knowledge information file corresponding to the new knowledge mode type configured by the staff, and updates the previous knowledge data in the second knowledge node to realize multi-mode knowledge fusion updating of the knowledge graph.
In one possible implementation manner, the step S150 includes:
and receiving a new knowledge information file corresponding to the new knowledge mode type configured by the staff, storing the knowledge information file and the previous knowledge information file, and updating the previous knowledge data in the second knowledge node. The invention correspondingly stores the new knowledge information file and the previous knowledge information file in the previous knowledge information file storage space in the receiving of the new knowledge information file corresponding to the new knowledge mode type configured by the staff, and updates the previous knowledge data in the second knowledge node to realize multi-mode knowledge fusion updating of the knowledge graph.
In one possible implementation manner, if the user is judged to actively adjust the selection quantity, the adjusted adjustment quantity is compared with the selection quantity before adjustment to obtain an adjustment difference value, wherein the value of the adjustment quantity is the value of the user after adjustment of the selection quantity. In an actual application scene, if a user considers that the adjustment quantity has an error, the adjustment quantity is actively adjusted at the moment, and the adjustment difference value is obtained.
If the adjustment difference value is greater than 0, increasing and adjusting the preset quantity according to the adjustment difference value and the selection quantity before adjustment. At this time, the preset number is proved to be smaller, the preset number is required to be increased and adjusted by combining the adjustment difference value, and if the adjustment difference value is larger, the increase amplitude of the preset number is larger.
If the adjustment difference value is smaller than 0, reducing and adjusting the preset quantity according to the adjustment difference value and the selection quantity before adjustment. At this time, the preset number is proved to be larger, the preset number is required to be reduced and adjusted by combining the adjustment difference value, and if the adjustment difference value is smaller, the reduction amplitude of the preset number is smaller.
The preset number after the increase adjustment or the decrease adjustment is calculated by the following formula,
wherein ,
to increase the adjusted preset number, +.>
To increase the number of adjustments after adjustment, +.>
To increase the adjustment coefficient->
To reduce the number of adjustments after adjustment, +.>
To reduce the adjusted preset number +.>
To reduce the adjustment coefficient. By->
and />
The adjustment difference can be calculated in different situations according to +.>
An increased number of preset numbers is obtained according to +.>
A reduced number of the preset number is obtained. Through the mode, the preset quantity can be continuously updated according to different use scenes, so that the calculated adjustment quantity is more in line with the corresponding calculation scene and is relatively more accurate.
The present invention also provides a storage medium having stored therein a computer program for implementing the methods provided by the various embodiments described above when executed by a processor.
The storage medium may be a computer storage medium or a communication medium. Communication media includes any medium that facilitates transfer of a computer program from one place to another. Computer storage media can be any available media that can be accessed by a general purpose or special purpose computer. For example, a storage medium is coupled to the processor such the processor can read information from, and write information to, the storage medium. In the alternative, the storage medium may be integral to the processor. The processor and the storage medium may reside in an application specific integrated circuit (Application Specific Integrated Circuits, ASIC for short). In addition, the ASIC may reside in a user device. The processor and the storage medium may reside as discrete components in a communication device. The storage medium may be read-only memory (ROM), random-access memory (RAM), CD-ROMs, magnetic tape, floppy disk, optical data storage device, etc.
The present invention also provides a program product comprising execution instructions stored in a storage medium. The at least one processor of the device may read the execution instructions from the storage medium, the execution instructions being executed by the at least one processor to cause the device to implement the methods provided by the various embodiments described above.
In the above embodiments of the terminal or the server, it should be understood that the processor may be a central processing unit (english: central Processing Unit, abbreviated as CPU), or may be other general purpose processors, digital signal processors (english: digital Signal Processor, abbreviated as DSP), application specific integrated circuits (english: application Specific Integrated Circuit, abbreviated as ASIC), or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of a method disclosed in connection with the present invention may be embodied directly in a hardware processor for execution, or in a combination of hardware and software modules in a processor for execution.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the invention.
Claims (10)
1. The multi-mode fusion updating method of the knowledge node is characterized by comprising the following steps of:
acquiring knowledge data corresponding to each knowledge node in a knowledge graph, and determining knowledge mode types included in the knowledge data, wherein the knowledge data types comprise at least one of a text mode, an image mode, an audio mode and a video mode;
counting the number of the mode types of the knowledge information files corresponding to all the knowledge mode types in the knowledge node and the knowledge information quantity corresponding to each knowledge information file, and determining a preset evaluation weight corresponding to each knowledge mode type;
comprehensively calculating according to the modal types of the knowledge information files, the knowledge information quantity corresponding to each knowledge information file and a preset evaluation weight to obtain modal evaluation sub-coefficients corresponding to each knowledge node;
obtaining a modal evaluation average coefficient according to the modal evaluation sub-coefficients of all knowledge nodes;
comparing the modal evaluation sub-coefficient of each knowledge node with the modal evaluation average coefficient, taking the knowledge node corresponding to the modal evaluation sub-coefficient smaller than the modal evaluation average coefficient as a first knowledge node, and carrying out ascending order sequencing on all the first knowledge nodes according to the modal evaluation sub-coefficient to obtain a knowledge node sequence;
Selecting a plurality of first knowledge nodes at the front part from the knowledge node sequence as second knowledge nodes;
the knowledge mode type of the second knowledge node at the current moment is used as a first knowledge mode type, a second knowledge mode type which the second knowledge node does not have is generated according to the total knowledge mode type and the first knowledge mode type, and a recommended adding mode type is obtained according to the number of the second knowledge mode types;
and receiving a new knowledge information file corresponding to the new knowledge mode type configured by the staff according to the recommended adding mode type, correspondingly storing the new knowledge information file and the previous knowledge information file in a previous knowledge information file storage space, and updating the previous knowledge data in the second knowledge node.
2. The method of claim 1, wherein the multi-modal fusion update of knowledge nodes,
the obtaining the knowledge data corresponding to each knowledge node in the knowledge graph, and determining the knowledge mode category included in the knowledge data includes:
acquiring a storage space corresponding to each knowledge node in a knowledge graph, and calling a knowledge information file in the storage space;
And determining the format of the knowledge information file, and determining corresponding knowledge mode types according to the format of the knowledge information file, wherein each format has the corresponding knowledge mode type.
3. The method for multi-modal fusion update of a knowledge node as claimed in claim 2, wherein,
the obtaining the modal evaluation average coefficient according to the modal evaluation sub-coefficients of all knowledge nodes comprises the following steps:
obtaining a first total number of all knowledge nodes, summing the modal evaluation sub-coefficients of all knowledge nodes to obtain a sum of sub-coefficients, and calculating according to the first total number and the sum of the sub-coefficients to obtain a modal evaluation average coefficient;
the modal evaluation average coefficient is calculated by the following formula,
wherein ,
mean coefficient for mode evaluation ∈>
Is->
Modal evaluator coefficients of the knowledge nodes, +.>
For the upper limit value of knowledge node, +.>
A quantity value for a knowledge node;
4. The method for multi-modal fusion update of a knowledge node as claimed in claim 3,
comparing the modal evaluation sub-coefficient of each knowledge node with the modal evaluation average coefficient, taking the knowledge node corresponding to the modal evaluation sub-coefficient smaller than the modal evaluation average coefficient as a first knowledge node, and carrying out ascending order sequencing on all the first knowledge nodes according to the modal evaluation sub-coefficient to obtain a knowledge node sequence, wherein the method comprises the following steps:
Extracting node labels corresponding to each knowledge node respectively, and setting the node labels of each knowledge node in one-to-one correspondence with corresponding modal evaluation sub-coefficients;
counting node labels with all modal evaluation sub-coefficients smaller than modal evaluation average coefficients to generate a label set, and taking knowledge nodes corresponding to all node labels in the label set as first knowledge nodes;
and carrying out ascending order sequencing on all the node labels in the label set according to the modal evaluation sub-coefficients to obtain a knowledge node sequence based on the node label sequencing.
5. The method of multimodal fusion update of a knowledge node of claim 4,
the selecting the first knowledge nodes as the second knowledge nodes in the front part in the knowledge node sequence comprises the following steps:
obtaining the sum of knowledge information amounts corresponding to each first knowledge node in a knowledge node sequence, and obtaining the average information amount of all the first knowledge nodes according to the sum of the knowledge information amounts;
comparing the average information quantity with a preset information quantity to obtain a quantity offset coefficient, and offsetting the preset quantity according to the quantity offset coefficient to obtain a selected quantity;
and selecting a plurality of first knowledge nodes in front in the knowledge node sequence according to the selection quantity as second knowledge nodes.
6. The method of multimodal fusion update of a knowledge node of claim 5,
the obtaining the sum of the knowledge information amounts corresponding to each first knowledge node in the knowledge node sequence, and obtaining the average information amount of all the first knowledge nodes according to the sum of the knowledge information amounts, includes:
obtaining the total number of all the first knowledge nodes in the knowledge node sequence to obtain a second total number, calculating according to the sum of the knowledge information amounts corresponding to all the first knowledge nodes and the second total number to obtain an average information amount, calculating the average information amount by the following formula,
wherein ,
for the average information quantity of all first knowledge nodes, < +.>
Is the +.>
Sum of knowledge information amounts of the first knowledge nodes,/->
For the upper limit value of the first knowledge node in the sequence of knowledge nodes,/I>
For the number value of the first knowledge node in the sequence of knowledge nodes,/for the first knowledge node>
Is->
Sum of knowledge information amounts of the first knowledge nodes,/->
A text modality class of the first knowledge node +.>
Knowledge information amount corresponding to the knowledge information file, < >>
The upper limit value of the knowledge information file of the text modality class for the first knowledge node,/for >
The first knowledge node is the audio modality class +.>
Knowledge information amount corresponding to the knowledge information file, < >>
The upper limit value of the knowledge information file for the audio modality class of the first knowledge node,/for>
The video modality category of the first knowledge node +.>
Knowledge information amount corresponding to the knowledge information file, < >>
The upper limit value of the knowledge information file of the video modality class for the first knowledge node,/for>
The first knowledge node is the first +.>
Knowledge information amount corresponding to the knowledge information file, < >>
The upper limit value of the knowledge information file of the image modality class of the first knowledge node.
7. The method of multimodal fusion update of a knowledge node of claim 6, wherein,
comparing the average information amount with a preset information amount to obtain a quantity offset coefficient, and offsetting the preset quantity according to the quantity offset coefficient to obtain a selected quantity, wherein the method comprises the following steps:
the pick-out number is calculated by the following formula,
wherein ,
to select the number +.>
For presetting information quantity->
Is information quantity weight value, +.>
Is a preset number;
and if the calculated selection quantity is judged not to be an integer, rounding the selection quantity.
8. The method of claim 1, wherein the multi-modal fusion update of knowledge nodes,
the step of using the knowledge mode type of the second knowledge node at the current moment as the first knowledge mode type, generating a second knowledge mode type which the second knowledge node does not have according to the total knowledge mode type and the first knowledge mode type, and obtaining recommended addition mode types according to the number of the second knowledge mode types, includes:
acquiring the number of recommended categories preset by an administrator;
and if the number of the second knowledge mode types is smaller than or equal to the number of the recommended types, taking all the second knowledge mode types as recommended addition mode types.
9. The method for multimodal fusion update of a knowledge node of claim 8, further comprising:
if the number of the second knowledge mode types is larger than the number of the recommended types, extracting priority information corresponding to each second knowledge mode type, wherein the preset priority information corresponding to each knowledge mode type is different;
and sorting all the second knowledge mode types in a descending order according to the priority information, and selecting the second knowledge mode types corresponding to the recommended type number according to the order.
10. The method of claim 7, wherein the multi-modal fusion update of knowledge nodes,
if the user is judged to actively adjust the selection quantity, comparing the adjusted adjustment quantity with the selection quantity before adjustment to obtain an adjustment difference value, wherein the value of the adjustment quantity is the value of the user after adjustment of the selection quantity;
if the adjustment difference value is greater than 0, increasing and adjusting the preset quantity according to the adjustment difference value and the selection quantity before adjustment;
if the adjustment difference value is smaller than 0, reducing and adjusting the preset quantity according to the adjustment difference value and the selection quantity before adjustment;
the preset number after the increase adjustment or the decrease adjustment is calculated by the following formula,
wherein ,
to increase the adjusted preset number, +.>
To increase the number of adjustments after adjustment, +.>
To increase the adjustment coefficient->
To reduce the number of adjustments after adjustment, +.>
To reduce the adjusted preset number +.>
To reduce the adjustment coefficient. />
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310322871.4A CN116028654B (en) | 2023-03-30 | 2023-03-30 | Multi-mode fusion updating method for knowledge nodes |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310322871.4A CN116028654B (en) | 2023-03-30 | 2023-03-30 | Multi-mode fusion updating method for knowledge nodes |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116028654A true CN116028654A (en) | 2023-04-28 |
| CN116028654B CN116028654B (en) | 2023-06-13 |
Family
ID=86089683
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310322871.4A Active CN116028654B (en) | 2023-03-30 | 2023-03-30 | Multi-mode fusion updating method for knowledge nodes |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116028654B (en) |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180165352A1 (en) * | 2016-12-09 | 2018-06-14 | Microsoft Technology Licensing, Llc | Reducing churn in knowledge graphs |
| CN112200317A (en) * | 2020-09-28 | 2021-01-08 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Multi-modal knowledge graph construction method |
| US20210217323A1 (en) * | 2019-07-03 | 2021-07-15 | Obrizum Group Ltd. | Cheating detection in remote assessment environments |
| CN113486190A (en) * | 2021-06-21 | 2021-10-08 | 北京邮电大学 | Multi-mode knowledge representation method integrating entity image information and entity category information |
| CN113792768A (en) * | 2021-08-27 | 2021-12-14 | 清华大学 | Hypergraph neural network classification method and device |
| CN113849575A (en) * | 2020-12-10 | 2021-12-28 | 阿里巴巴集团控股有限公司 | Data processing method, device and system |
| CN114218400A (en) * | 2021-12-13 | 2022-03-22 | 上海交通大学 | Semantic-based data lake query system and method |
| US20220198358A1 (en) * | 2021-04-27 | 2022-06-23 | Baidu International Technology (Shenzhen) Co., Ltd. | Method for generating user interest profile, electronic device and storage medium |
| CN114880527A (en) * | 2022-06-09 | 2022-08-09 | 哈尔滨工业大学（威海） | Multi-modal knowledge graph representation method based on multi-prediction task |
| CN115114443A (en) * | 2022-04-27 | 2022-09-27 | 腾讯科技（深圳）有限公司 | Training method and device of multi-modal coding model, electronic equipment and storage medium |
| CN115331769A (en) * | 2022-07-15 | 2022-11-11 | 北京大学 | Medical image report generation method and device based on multi-modal fusion |
| CN115700516A (en) * | 2022-09-23 | 2023-02-07 | 中国电子科技集团公司第七研究所 | Short-wave communication frequency-selecting data fusion system and frequency-selecting method based on knowledge graph |
-
2023
- 2023-03-30 CN CN202310322871.4A patent/CN116028654B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180165352A1 (en) * | 2016-12-09 | 2018-06-14 | Microsoft Technology Licensing, Llc | Reducing churn in knowledge graphs |
| US20210217323A1 (en) * | 2019-07-03 | 2021-07-15 | Obrizum Group Ltd. | Cheating detection in remote assessment environments |
| CN112200317A (en) * | 2020-09-28 | 2021-01-08 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Multi-modal knowledge graph construction method |
| CN113849575A (en) * | 2020-12-10 | 2021-12-28 | 阿里巴巴集团控股有限公司 | Data processing method, device and system |
| US20220198358A1 (en) * | 2021-04-27 | 2022-06-23 | Baidu International Technology (Shenzhen) Co., Ltd. | Method for generating user interest profile, electronic device and storage medium |
| CN113486190A (en) * | 2021-06-21 | 2021-10-08 | 北京邮电大学 | Multi-mode knowledge representation method integrating entity image information and entity category information |
| CN113792768A (en) * | 2021-08-27 | 2021-12-14 | 清华大学 | Hypergraph neural network classification method and device |
| CN114218400A (en) * | 2021-12-13 | 2022-03-22 | 上海交通大学 | Semantic-based data lake query system and method |
| CN115114443A (en) * | 2022-04-27 | 2022-09-27 | 腾讯科技（深圳）有限公司 | Training method and device of multi-modal coding model, electronic equipment and storage medium |
| CN114880527A (en) * | 2022-06-09 | 2022-08-09 | 哈尔滨工业大学（威海） | Multi-modal knowledge graph representation method based on multi-prediction task |
| CN115331769A (en) * | 2022-07-15 | 2022-11-11 | 北京大学 | Medical image report generation method and device based on multi-modal fusion |
| CN115700516A (en) * | 2022-09-23 | 2023-02-07 | 中国电子科技集团公司第七研究所 | Short-wave communication frequency-selecting data fusion system and frequency-selecting method based on knowledge graph |
Non-Patent Citations (2)
| Title |
|---|
| RUOTIAN HOU 等: ""Global Relation Auxiliary Graph Attention Network for Knowledge Graph Completion"", 《2022 IEEE THE 5TH INTERNATIONAL CONFERENCE ON ARTIFICIAL INTELLIGENCE AND BIG DATA》, pages 532 - 538 * |
| 马昂 等: ""基于强化学习的知识图谱综述"", 《计算机研究与发展》, pages 1694 - 1722 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116028654B (en) | 2023-06-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8385660B2 (en) | Mixed media reality indexing and retrieval for repeated content | |
| US8676810B2 (en) | Multiple index mixed media reality recognition using unequal priority indexes | |
| US8369655B2 (en) | Mixed media reality recognition using multiple specialized indexes | |
| US8073263B2 (en) | Multi-classifier selection and monitoring for MMR-based image recognition | |
| US8965145B2 (en) | Mixed media reality recognition using multiple specialized indexes | |
| CN110909222B (en) | User portrait establishing method and device based on clustering, medium and electronic equipment | |
| WO2016155493A1 (en) | Data processing method and apparatus | |
| KR20100052896A (en) | Mehod and system for recommending intelligent contents according to user | |
| CN111797320B (en) | Data processing method, device, equipment and storage medium | |
| CN111125528A (en) | Information recommendation method and device | |
| KR20120049551A (en) | Method for recommending content based on user preference and profile of content | |
| KR101725510B1 (en) | Method and apparatus for recommendation of social event based on users preference | |
| CN112765484A (en) | Short video pushing method and device, electronic equipment and storage medium | |
| KR20120049550A (en) | Method for recommending content based on user preference with time flow | |
| CN116028654B (en) | Multi-mode fusion updating method for knowledge nodes | |
| CN111177564B (en) | Product recommendation method and device | |
| CN114296775B (en) | Intelligent operation and maintenance method and system based on big data | |
| CN111339448B (en) | Query rewriting method, system, device and storage medium | |
| CN112613775A (en) | Resource quality evaluation method and device, electronic device and storage medium | |
| CN111026913A (en) | Video distribution method and device, electronic equipment and storage medium | |
| CN107491462B (en) | Method and system for providing search results | |
| KR102064709B1 (en) | Method for measuring contents level of difficulty and apparatus using the method | |
| CN116304303B (en) | Asset recommendation method and system based on knowledge graph | |
| CN110162648B (en) | Picture processing method, device and recording medium | |
| CN112231545B (en) | Method, device, equipment and storage medium for ordering aggregation set |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |