CN109508453A - Across media information target component correlation analysis systems and its association analysis method - Google Patents
Across media information target component correlation analysis systems and its association analysis method Download PDFInfo
- Publication number
- CN109508453A CN109508453A CN201811134544.1A CN201811134544A CN109508453A CN 109508453 A CN109508453 A CN 109508453A CN 201811134544 A CN201811134544 A CN 201811134544A CN 109508453 A CN109508453 A CN 109508453A
- Authority
- CN
- China
- Prior art keywords
- information
- target
- target component
- personage
- extension
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention proposes across the media information target component correlation analysis systems of one kind and its association analysis methods, can significantly improve the analysis ability of target object using the present invention and analyze result to the applied value of user.The technical scheme is that: user's interactive operation module selects decision-tree model corresponding with intelligence analysis purpose, establish the target component model of goal-orientation, user demand information is sent into association extension process module, the target component model of user's initial retrieval information set is read according to index, obtain the component attributes item structural data that information data enters library module input target, the field priori knowledge and element extension rule provided using external data support module carries out element extension, obtain new extension target component and element extension information；All information is retrieved, the incidence relation result for obtaining object event is output to supporting module, output includes the target analysis result of target association relationship and goal activities rule.
Description
Technical field
The present invention relates to information association analysis fields, and in particular, to a kind of element information of object-oriented event
Information association analysis method.
Background technique
With multimedia information technology and the high speed development of Internet technology, more and more multimedia messages are full of
In whole network, also become the multimedia messages for increasingly be unableing to do without various kinds in daily life.People can be easy
The multimedia messages of these magnanimity are obtained, but to find the information that those most related, users are most interested in and often expend largely
Manpower and material resources, for information value resolution be also it is extremely difficult.In these media informations often included by image
Information content maximum and a kind of most widely used media.And image data is often isomery, its storage is also dispersion
, this just to how quickly and prepare find user needed for useful information cause very burden.
Across will be used wider and wider for medium technique (SIMA), each neck such as education, science and technology, management has been penetrated into
Domain will be that digital resource rolls the new transformant information retrieval medium saved across medium technique.Text may be implemented to figure in it
Picture, image handle the three-stage of text sound equipment, sound equipment, and play an important role in actual operation.Across media skills
Art refers to information spreading and interact between various media, includes two layers of meaning: one refer to mutual information different media it
Between cross transmission and integration；Secondly referring to the cooperation between media, symbiosis, interaction and coordination.One across media data aobvious
Writing characteristic is exactly its relevance, how more go deep into accurately retrieval being mesh using across the relevance between media data
Preceding research hotspot.Not only refer to cross integration of the information between various media across media data, further comprises various media
Coordination between data, so that the feature of a certain things be shown in terms of multi-angle, multi-level, multi-semantic meaning are several
Come.But it is also in the primary stage about the research across media data at present, many solutions are also only following multimedia
Treatment mechanism on the basis of be extended, do not form special analysis method and theory across media, it is much special across media
Category problem is also at Uniform semantic between stage of fumbling, such as each media, related information modeling, data source retrieval, attribute phase interaction
With the problems such as all need further to be furtherd investigate.Since different types of multi-medium data uses the low-level image feature of different attribute
It carries out data representation and allows media in identical data representation therefore, it is necessary to study the relativity problem across media
It is measured under frame.
Correlation across media includes convergence analysis and across media two parts of correlation model of building.Convergence analysis can solve
Multimedia application type is analysed, semantic gap is reduced, accelerates the accuracy rate of information output content, and the correlation of synchronous analogical object
Measurement.The similarity of media data influences the corresponding relationship that connection is concentrated in data characteristics selection, be it is a kind of hide it is similar
Relationship.Data acquisition system merged with information be resource description progressive form, no matter as a whole description or hierarchy unit
Description, the reflection that finally general character merges all in " isomorphism ", metadata map to semantic normalization and describe space identity, obtain useful
Interoperability situation search result.The building of cross-media correlation graph model is carried out by obtaining different media information low-level image features
Similarity distance calculates to construct single mode adjacent map, the incidence relation between different media is found out by link analysis, as base
Plinth constructs model.The superiority and inferiority of selected characteristic will directly affect the overall performance of whole system, at the same also with entire categorizing system
Accuracy rate and efficiency have very big correlation.In order to establish multimedia correlation model, it is necessary to first find out the pass that different media are asked
Then connection relationship resettles certain model.In fact, no matter which kind of method its core concept is all by comprehensively utilizing image
Various features help to understand multimedia in data, to realize the relativity measurement between different characteristic, but be difficult to solve across
The correlation calculations problem of media.The existing association letter between different types of medium information characteristic during information fusion
Breath is generally regarded as a kind of additional feature to be merged, these features is not made full use of in this process, so that not
It is not in full use in establishing across media correlation models with the incidence relation between media.
Across media information associations are one of intelligence analysis field key technologies, are associated with merge in across media informations at present
There is also certain shortcoming, pixel-based fusion, the feature-based fusions of lower layer although there are many theoretical methods for aspect, but not
There is practical large-scale use.The decision level fusion on upper layer lacks basic-level support again, therefore lacks a practical information and merge system
System is to improve its combat capabilities.But we are very urgent for the fusion joint demand of information, major information center, information centre,
There are many data to be associated with, and becomes information island and is unable to get effective use, thus we be badly in need of research one can be with
The utility system that information fusion association analysis is carried out using existing information, improves the utilization efficiency of information.
Most important source of the Text Intelligence as information source is one of the content that analysis is most difficult in information processing.Due to
Its data volume is huge, different formats, is difficult to find the related information of user's care in numerous Text Intelligences.Simultaneously as
The information such as image, video, situation map are usually associated with textual form Interpretation Report, the different formats of the Interpretation Report, description side
Method difference is huge, it is more difficult to merge with Text Intelligence association.Therefore research Text Intelligence information association technology, be solve it is all kinds of across
The key technology of media information data association convergence analysis.
Summary of the invention
For current cross-media retrieval algorithm do not make full use of the intrinsic semantic knowledge of event between various media and
Can not solve the problems, such as the semantic gap during cross-media retrieval, the purpose of the present invention is to provide it is a kind of analysis it is more accurate,
Across the media information target component information correlation analysis systems across media data correlation can be accurately measured, and can be preferably
The incidence relation between media is found out within the scope of high-level semantic, analyzes potential high-level semantic association letter between different media datas
Breath finds out intrinsic similitude between image, and can effectively promote the accuracy rate and efficiency of cross-media retrieval, object-oriented event
The information of element information is associated with and analysis method.
In order to obtain above-mentioned technical effect, the technical solution adopted by the present invention is that: a kind of across media information target component letters
Breath fusion correlation analysis system, comprising: user's interactive operation module, association extension process module, external data support module,
Information data enters library module and supporting module, it is characterised in that: user's interactive operation module passes through interactive selection point
Class model introduces the purpose factor of intelligence analysis, inputs the vocabulary to be retrieved, and select decision corresponding with intelligence analysis purpose
Tree-model establishes the target component model of goal-orientation for object event feature, and user demand information is sent into and is associated with
Extension process module；It is associated with the target component model that extension process module reads user's initial retrieval information set according to index, is obtained
Information data is taken to enter the component attributes item structural data of library module input target, the neck provided using external data support module
Domain priori knowledge and element extension rule carry out element extension, obtain new extension target component, and then retrieval extension target is wanted
Plain vocabulary obtains element extension information；According to the decision tree-based clustering for using user to select, all information is retrieved, target is obtained
The incidence relation of event is as a result, be output to supporting module for obtained information incidence relation result；Supporting module is defeated
It out include the target analysis result of target association relationship and goal activities rule.
The present invention has the following beneficial effects: compared with the prior art
The present invention inputs the vocabulary to be retrieved using user's interactive operation module, passes through decision template selection and intelligence analysis mesh
Corresponding decision-tree model, by user demand information be sent into association extension process module；Utilize user's interactive operation module
Characteristic matching rule finds out the interested image of user in the target component model built, and utilizes additional media information
The similarity of these images is ranked up, establishes across media association relation models based on this, and from similitude transmitting and
Two angles of user feedback are updated it.Then in the cross-media correlation graph model built, using based on cluster
Method establishes index for it, by calculating its similarity distance asked with object in class cluster after inquiring media information classification, and
It is added in association graph model as new media object, finally by the method for searching its neighbour in graph model, returns
Back to the interested media information of user, be finally reached it is more acurrate, more fully inquire effect by classification.Experimental result and the present invention
Expection it is almost the same.
The present invention is directed to the characteristics of object event, establishes the target component model of goal-orientation, and realize base
Different media informations can be preferably made to pass through the pass between object event element in the information association analysis of target component model
It is the information correlation analysis system merged；It is different for the demand of intelligence analysis, it is realized and is based on by interactive application
The information association analysis of user demand strengthens information user and analyzes object and analyze the specific aim of demand, improves target pair
Applied value of the analysis ability and analysis result of elephant to user.
User's interactive operation module introduces the purpose factor of intelligence analysis by interactive selection disaggregated model, so that
Information association process has definition, and analysis has specific aim.In online association analytic process, the retrieval vocabulary of user's input,
Obtain the basic intelligence comprising retrieval vocabulary；The target component that the information retrieved is read according to index, is known using based on field
Target component is extended to obtain new target component by the target component extension rule of knowledge, and retrieves fresh target element vocabulary,
Obtain the extension information comprising fresh target element vocabulary；End user selects information Clustering Decision-Making tree mould according to intelligence analysis purpose
Type is associated cluster to the above-mentioned basic intelligence retrieved and extension information, obtains the target association relational result of information, into
And the integration of information is carried out according to target association relational result, it realizes the intelligence analysis of object-oriented object, analyzes different media
Potential high-level semantic related information between data finds out the intrinsic similitude of object event in multimedia, and analysis is accurate, Neng Gouzhun
Correlation of the exactness amount across media data can preferably find out the incidence relation between media within the scope of high-level semantic, and can have
Effect ground promotes the accuracy rate and efficiency of cross-media retrieval.
External data support module of the present invention introduces domain expertise knowledge by element extension association, so that information
Association process has prior information, analyzes more accurate.Across media information pass through the text for being parsed into standardization, by pretreatment
Obtain the text segmentation sequence of information.And the target component of information is extracted, the index relative for establishing information and target component is put in storage.
Efficiently avoid image mismatch and can not find that there is the drawbacks of intrinsic similarity image.By establishing similitude inside media
Relational matrix and to link analysis is carried out between different media, acquires the correlation of high-level semantic between capable of finding out different media, from
The low-level visual feature for extracting image dynamicly avoids a series of problems caused by traditional images searching system, so that big rule
The application of the image database system of mould has more actuality.
The present invention is based on associations and analysis that the element information of object event realizes information, suitable for the pass across media information
Connection analysis, has stronger engineering practical value.
Detailed description of the invention
For a clearer understanding of the present invention, now will embodiment through the invention, referring concurrently to attached drawing, to describe this hair
It is bright, in which:
Fig. 1 is across the media information target component information fusion correlation analysis system schematic illustrations of the present invention.
Fig. 2 is Fig. 1 target component dictionary formwork structure schematic diagram.
Fig. 3 is that Fig. 1 external data support module utilizes a kind of associated schematic diagram of element extension rule progress element extension.
Fig. 4 is the decision-tree model schematic diagram of the training set training the present invention is based on intelligence analysis purpose mark.
Fig. 5 is across the media information association analysis flow charts based on target component model.
Specific embodiment
Refering to fig. 1.For a better understanding of the present invention, across media information target component information fusion associations under introducing first
The structure of analysis system.
In the embodiment described below, a kind of across media information target component information merge correlation analysis system, comprising:
User's interactive operation module, association extension process module, external data support module, information data enter library module and support clothes
Business module.Wherein, information data enters library module and needs to prepare offline, and in offline prepare, information data enters library module will collection
Across media information parsed, obtain unified semi-structured text data, and by text participle, remove stop words, extract
The pretreatments such as stem obtain text sequence of words；Information in information database is associated under external data support by information
Expansion module analysis processing is extracted target component information and is carried out target using name Entity recognition and element dictionary enquiring mode
Feature model filling, and establish the storage of the index between information and target component model.
Information association extension process module includes element extension submodule, element association cluster submodule, element extension
Module carries out element extension using the element extension rule that external data support module provides, and obtains new extension target component；
The decision tree-based clustering that element association cluster submodule is selected according to user, retrieval extension target component vocabulary, obtains element extension
Information.
User's interactive operation module introduces the purpose factor of intelligence analysis by interactive selection disaggregated model, inputs
The vocabulary to be retrieved selects decision-tree model corresponding with intelligence analysis purpose to build for object event feature through decision template
User demand information is sent into association extension process module by the target component model for having found goal-orientation；It is associated at extension
The target component model that module reads user's initial retrieval information set according to index is managed, information data is obtained and enters library module input mesh
Target component attributes item structural data, the field priori knowledge provided using external data support module and element extension rule
Element extension is carried out, new extension target component is obtained, then retrieval extension target component vocabulary, obtains element extension information；
According to the decision tree-based clustering for using user to select, all information is retrieved, obtains the incidence relation of object event as a result, will obtain
Information incidence relation result be output to supporting module；The output of supporting module includes that target association relationship and target are living
Move the target analysis result of rule.
Refering to Fig. 2.The characteristics of paying close attention to target relevant information according to object event analysis devises external data support
Module is also required to prepare offline, and external data support module contains element extension rule based on field priori knowledge and by mesh
Several letter elements breath of mark event establishes characteristic model to characterize the target component dictionary template of information.
Target component dictionary template includes target component layer, object event layer and the information number positioned at top layer positioned at bottom
According to layer model, three-decker.Top layer is the regular information data of characterization；Object event layer is covered by event 1, event 2, event
The event of 3 ... event n information description takes out several class object events and forms the second layer；Target component layer is with object event
Center forms third layer to relevant target information item, and target information item has the activity time for respectively indicating target, actively
Point, related person, time, place, personage, equipment, tissue and the row of the affiliated organization of associated equipment and goal behavior trend
For.Each of object event layer event all corresponds to time, place, personage, equipment, tissue and the behavior of target component layer
Element.
Refering to Fig. 3.External data support module is expanded using the element that a kind of element extension rule of instantiation carries out information
Exhibition, to the information data collection that initial retrieval goes out, information data collection is classified as place, equipment, personage, place, equipment, people as required
Object is extended to downwards place 1, place 2, and equipment 1, equipment 2, personage 1, personage 2, personage 3 form target component item, and to
Under be extended to place 1- personage, place 2- personage, equipment 1- personage, equipment 2- personage, personage 1, personage 2, personage's 3- event are complete
It is extended at knowledge base；What personage continued to expand wants prime implicant: personage 1, personage 2, personage 3, personage 1, personage 2, personage 1, personage
2, personage 3, personage 1, personage 2, event are extended to event 1, event 2, until expanding information set.External data support module according to
The attribute classification of its target component model and element is read according to index, selects corresponding element extension rule in element extension knowledge base
Then, such as it is right to the rule of place generic attribute selection place element extension, such as Element_Expand (Relation)=E2→
E3Indicate the extension of " place element -> personage's element "；The rule for equipping generic attribute selection equipment element extension is right, such as
Element_Expand (Relation)=E4→E3Indicate the extension of " equipment element -> personage's element "；Figure kind's Attributions selection
The rule of personage's element extension is right, such as Element_Expand (Relation)=E3→E7Indicate " personage's element -> event name
The extension of title ".What is be expanded out wants prime implicant, is retrieved using extension element vocabulary, the information set being expanded out.
Refering to Fig. 4.The purpose of decision tree corresponding intelligence analysis is to need to count target in the crawler behavior in some area
Information.The training for the training set that system is marked now using different intelligence analyses obtains different decision classifying trees, in analysis mesh
It is lower by information training set by different target different location crawler behavior carry out classification mark, decision tree is expressed as leaf
Target component attribute of the node attribute list of node as information, equipment attribute of the root node as target；Attribute is equipped to make
For first layer node, site attribute is as second layer node, and behavior property is as third layer node；Node is equipped according to information number
Information is split into " equipment 1 ", " equipment 2 ", " equipment 3 " three classes according to the target component attribute in library, " equipment 1 " target class information according to
It is " place 1 ", " place 2 " two classes according to place node split；" equipment 2 " target class information is " place according to place node split
1 ", " place 2 ", " place 3 " three classes；" equipment 3 " target class information is " place 1 ", " place 2 " two according to place node split
Class, information set of the composition target in somewhere region；Place node forms target in the movable row in somewhere region according to behavior property
For information set, all kinds of middle information in upper layer continue to divide according to corresponding node, until sample class mark is homogeneous in the class of division
Together, the node attribute list of decision tree, the final classification of available sample, then using ID3 algorithm training decision are constructed
Classification tree；Wherein, decision tree as shown in the figure is obtained.
Based on the above method, the corresponding decision-tree model of the different intelligence analysis purposes of training.The purpose of intelligence analysis is foundation
The angle that intelligence analysis personnel often pay close attention to and analyze takes out the classification of information foundation for being conducive to the analysis purpose and typical feelings
It calls the score and analyses purpose, and mark the decision-tree model of collection training now using correspondence analysis.Constitute typical information as shown in Table 1
Analyze the hypothesis decision-tree model list of mark training white silk under purpose.
Table 1
In specific implementation process, open source web page news report, including video, image, situation and webpage are collected first, and solve
Analysis arranges and forms unified semi-structured text data, as intelligence information.First step processed offline preparation process has three parts,
First part's field target component dictionary is established: being segmented to all intelligence informations, stop words is gone to pre-process, obtain information
Segmentation sequence, the top layer as target component template: regular information data；Artificial statistical information data obtain a few class target things
Part, as the second layer of template: object event layer；It is artificial to extract for the information data collection under each object event
The date-time that object event occurs out, the place of generation, the personage of participation, relevant equipment, affiliated organization and hair
Raw crawler behavior respectively corresponds every component attributes of target: the activity time of target, activity venue, related person, dress
Standby, affiliated tissue and behavior trend, as the third layer of template: element layer.The target component word established as shown in table 1
Allusion quotation takes out four class object events to entire information data layer, event 1, event 2, event 3 and event 4, the mesh under every class event
Mark 6 component attributes typical values that element layer is target；
Second part is that element extension rule is established: user carries out statistical induction to target component in intelligence information first, in addition
Domain knowledge and expertise take out the intrinsic relationship between target componentSuch as The relationship of equipment 1 and personage 1,The pass in place 1 and personage 1
System,
The relationship of event 1 and personage 1, using the extension rule of these intrinsic relationship building target components, wherein according to above-mentioned pass
It is the element extension rule Element establishedExpand(Relation): Eresource_element→Edestination_elementHave: E4→E3
(equipment element expands to personage's element), E2→E3(place element expands to personage's element), E4→E2(equipment element expands to
Place element), E3→E7(personage's element expands to event title)；
Part III is intelligence information storage: system is pre-processed to obtain segmentation sequence to all intelligence informations, uses Chinese
Name Entity recognition and the mode of element dictionary enquiring extract the target component model of information, and establish information and feature model it
Between index storage.
Second step online processing process: user inputs retrieval vocabulary first, and intelligence analysis purpose is selected (to correspond to classification to determine
Plan tree-model), " place 1 " such as is inputted, system retrieval obtains the basic intelligence about place 1, then reads information according to index
Target component model, using initial retrieval go out information target component as source, user element extension rule is wanted by source target
Element extension obtains fresh target element, then retrieval extension target component vocabulary, obtains extension feelings relevant to extension element vocabulary
Report；The information Clustering Decision-Making tree-model finally selected using user is associated above-mentioned initial retrieval information and extension information
Cluster, obtains the target association relational result of information.And goal activities place, activity time, output are counted according to association results
The mechanics of the fresh targets such as the mechanics of target, and the related person, the equipment that are expanded by target.
Refering to Fig. 5.Across media isomeric datas include the image data of intelligence analysis concern, video data, situation data, text
Notebook data, wherein situation data include trend report, and text data includes XML webpage, WORD, PDF, TXT formatted data, image,
Across the media information datas such as video, situation, text.Across media information datas pass through Semiautomatic deconvolution at the text of standardization, so
It segmented afterwards by text, go the pretreatments such as stop words, extraction stem, obtain the text sequence of words of information.Its structuring step
Are as follows: across media information target component correlation analysis systems extract key images frame based on video content from video data, extract
Target relevant information in image carrys out the message that processing forms description target by image, and to XML open source webpage, WORD, PDF etc.
Half format data is parsed, and text abstract is extracted；
S1: across media information target component information correlation analysis systems (hereinafter referred to as system) by parsing and pre-process across
Media information obtains the text segmentation sequence of information；
S2: system extracts the target component in text segmentation sequence, establishes target component model；
System is for analysis in the object time, centered on object event, establishes the feature model of description object event, utilizes neck
Domain knowledge constructs target component dictionary；Target component includes the activity time of object event, participates in movable personage, is movable
Point, the equipment being related to, the organization being related to, the behavior trend of target and event 7 of title code name want prime implicant
Information and target component model；
2 target component model of table
| Want prime implicant | Time | Place | Personage | Equipment | Tissue | Behavior | Title |
| Component is described | E1 | E2 | E3 | E4 | E5 | E6 | E7 |
Assuming that m-th of information d of descriptionmTarget component model be X(m):
WhereinPrime implicant is wanted in v-th for indicating target m, is an attribute value set.
Text segmentation sequence d of the system to informationmText segmentation sequence carry out Chinese name, Entity recognition, entity word
Classification carries out assignment to target component model by classification results, such as time, date class entity are assigned to element of time E1, ground
Name entity is assigned to place element E2, name entity is assigned to personage's element E3.That does not find correspondent entity wants prime implicant, such as equips
Element E4, organizational factors E5, behavioral primitive E6The assignment by way of element dictionary enquiring.
System user element dictionary enquiring mode, to it is remaining do not find correspondent entity want prime implicant assignment, step includes:
System establishing element dictionary D={ D first1,D2,D3,D4,D5,D6,D7, then by matching entities vocabulary and element dictionary
Vocabulary to target component item carry out assignment, the specific implementation steps are as follows:
System constructs target component dictionary using domain knowledge, and dictionary includes three layers, and top layer to bottom is followed successively by information number
According to layer, object event layer, element layer.Object event refers to that the subject events of information description, element layer correspond to target component model and build
Vertical element dictionary D={ D1, D2, D3, D4, D5, D6, D7, refer to the element of time dictionary D of object event1, place element dictionary D2, people
Object element dictionary D3, equipment element dictionary D4, organizational factors dictionary D5, behavioral primitive dictionary D6With title code name element dictionary D7,
The value set of element dictionary carries out statistics filling using a large amount of information datas and obtains.
The text segmentation sequence d of system-computed informationmIn do not match the entity word w of corresponding elementiWith element dictionary DvIt is general
Similarity is read, wherein v=1,2,3,4,5,6,7, concept similarity presentation-entity vocabulary wiWhether element E is belonged tov.Usual concept
It is described with semantic, i-th of entity word wiWith the concept distance calculation formula of v class element are as follows:
In formula, NvRefer to element dictionary DvIn vocabulary sum, tjIt is dictionary DvIn j-th of vocabulary, pijFor two word wi,tj?
Semantic distance in semantic tree, a are constant.As text segmentation sequence dmIn vocabulary wiWith element dictionary DvConcept similarity
Dissim(wi,Dv) when meeting the given threshold requirement of system, it is believed that entity vocabulary wiBelong to target component Ev, then by the entity word
The assignment of such element as target.
S3: system establishes the index and storage of information Yu target component model.It establishes between information and target component model
Index, and configuration information title, information ID number, by information data index be put in storage.
S4: in online association treatment process, user inputs retrieval vocabulary, passes through matching retrieval vocabulary and information text vocabulary
Sequence retrieves the vocabulary of retrieval input, obtains the information comprising retrieval vocabulary.
S5: according to the index established in S3, the target component model of the information retrieved is read.
S6: system utilizes the element extension rule established based on domain knowledge, and target component is extended, is obtained new
Target component；
Target component carries out statistical induction in the information that S61 system collects user, in conjunction with domain knowledge and expertise,
It determines the intrinsic relationship between target component, constructs the relation rule pair between target component, the structure of element relationship rule pair are as follows:Wherein, Relation indicates elementWithBetween relation rule,Indicate one
A element relationship rule is right.The element relationship rule that empirically knowledge is established has following a few classes: equipment element and personage's element
Relation rule:The relation rule of place element and personage's element:Equip element and place element
Relation rule:With the relation rule of personage's element and object event:
S62 system is to the target component model read in S5
Prime implicant is wanted with itAs extended source Eresource_element, using the relation rule for the correspondence v class element established in S61, into
Row extension obtains target component item Edestination_element,
Element extension is as follows: ElementExpand(Relation): Eresourceelement→Edestinationelement。
S7: user search extends element vocabulary, and be expanded information；
S8: the index that user establishes according to system reads the target component of all information retrieved of system in S4 and S7
Model.
S9: the Clustering Decision-Making tree that system uses user to select carries out the cluster association of target component.Consider intelligence analysis
Purpose and the association, angle pairs result of concern are affected, using the purpose of intelligence analysis as factor, by man-machine interactively mode,
It is introduced into association cluster decision-tree model, implementation step is as follows:
S91: user is according to intelligence analysis experience, concern target when extracting typical intelligence analysis purpose, i.e. intelligence analysis
Angle, it is specific as shown in table 2, according to different analysis purposes by information training set S={ d1, d2..., dMInto classification annotation, it obtains
Training set { S, C under corresponding K kind intelligence analysis purposek, k=1,2 ..., K, CkIt indicates under kth kind analysis purpose to training set
The classification of S marks, Ck={ y1,y2,…,yM, ymFor dmMark classification, value ym=1,2 ..., Nk, NkIndicate kth
Total classification number in the case of middle mark；
S92: decision tree learning training set { S, C are usedk, generate the hypothesis decision-tree model of corresponding kth kind analysis now
Treek, the process using ID3 algorithm building decision tree includes:
S921: initialization training set { S, CkAnd target component model 7 attributes composition decision tree node list List=
(E1,E2,E3,E4,E5,E6,E7)；
S922: a component attributes E is arbitrarily selected to initialize the root node of a decision tree from List, then computation attribute E
Information gain.Attribute E is defined as relative to the information gain Gain (S, E) of training set S:
Wherein, Value (E) is the set of all possible values of attribute E, SvThe value for being the attribute E in S is the subset of v, i.e. Sv
=s ∈ S | E (s)=v }.Entropy (S) indicates the entropy of training set S, and giving includes NkThe entropy of the training set S of a class are as follows:
Wherein piFor the ratio for belonging to class i in training set S, i.e.,
The information gain of attribute E indicates that the entropy due to S caused by using attribute E to divide training set S reduces, and usually has highest
The best segmentation S of the attribute energy of information gain, traverses all properties E in List, uses best attribute E* as decision tree
TreekRoot node；
S923: system update List=List-E*, repeat step (2), obtain decision tree TreekSecond layer node.
S924: it repeats step (2) and (3), obtains decision tree TreekThe node of each layer, untilOr Gain (S, E)
=0, the decision tree Tree finally obtainedkModel；
S93 user is directed to different labeled training set { S, Ck, k=1,2 ..., K repeat step S92, respectively obtain K kind analysis mesh
/ preference under Clustering Decision-Making tree Tree={ Tree1,Tree2,…,TreeK}.During online association, user can be according to feelings
The different decision-tree model Tree of purpose interactive selection of analysis of calling the score carries out information association cluster, realize for specific purposes with
The informational intelligence summary association analysis that multi-angular analysis combines.
S10: system carries out information integration using target association relationship.
The foregoing is merely presently preferred embodiments of the present invention, is merely illustrative for the purpose of the present invention, and not restrictive
's.Those skilled in the art understand that it can be carried out in the spirit and scope defined by the claims in the present invention it is many change,
It modifies, is even equivalent, but falling in protection scope of the present invention.
Claims (10)
1. a kind of across media information target component correlation analysis systems, comprising: user's interactive operation module, association extension process
Module, external data support module, information data enter library module and supporting module, it is characterised in that: user's interactive mode behaviour
Make module by interactive selection disaggregated model, introduce the purpose factor of intelligence analysis, input the vocabulary to be retrieved, and select with
The corresponding decision-tree model of intelligence analysis purpose establishes the target component model of goal-orientation for object event feature,
User demand information is sent into association extension process module；It is associated with extension process module and reads user's initial retrieval feelings according to index
The target component model of collection is reported, the component attributes item structural data that information data enters library module input target is obtained, using outer
The field priori knowledge and element extension rule that portion's data support module provides carry out element extension, obtain new extension target and want
Element, then retrieval extension target component vocabulary, obtains element extension information；According to the decision tree-based clustering for using user to select, inspection
The information of Suo Suoyou obtains the incidence relation of object event as a result, obtained information incidence relation result is output to support clothes
Business module；The output of supporting module includes the target analysis result of target association relationship and goal activities rule.
2. across media information target component information merge correlation analysis system as described in claim 1, it is characterised in that: information
Data loading module parses across the media information of collection, obtains unified semi-structured text data, and pass through text
It segments, stop words, extraction stem is gone to pre-process to obtain text sequence of words；Information in information database is supported in external data
Under, it extracts target using name Entity recognition and element dictionary enquiring mode by information association expansion module analysis processing and wants
Prime information carries out the filling of target component model, and establishes the storage of the index between information and target component model.
3. across media information target component information merge correlation analysis system as described in claim 1, it is characterised in that: external
Data support module contains the target component extension rule based on field priori knowledge and several elements by object event
Information establishes characteristic model to characterize the target component dictionary template of information；Target component dictionary template includes the mesh positioned at bottom
Mark element layer, object event layer and the information data layer model positioned at top layer, three-decker, wherein top layer is the rule of characterization
Information data；Object event layer is covered by event 1, event 2, the event of event 3 ... event n information description, if taking out Ganlei
Object event forms the second layer；Target component layer forms third layer, mesh centered on object event, to relevant target information item
Mark item of information has the activity time for respectively indicating target, activity venue, related person, the affiliated organization of associated equipment and target
Time, place, personage, equipment, tissue and the behavior of behavior trend, each of object event layer event all correspond to target
Time, place, personage, equipment, tissue and the behavioral primitive of element layer.
4. across media information target component correlation analysis systems as described in claim 1, it is characterised in that: external data is supported
Module using a kind of instantiation element extension rule carry out information element extend, to initial retrieval go out information data collection,
Information data collection is classified as place, equipment, personage as required, and place, equipment, personage are extended to downwards place 1, place 2, equipment
1,2 are equipped, personage 1, personage 2, personage 3 form target component item, and be extended to place 1- personage, place 2- people downwards again
Object, equipment 1- personage, equipment 2- personage, personage 1, personage 2, personage's 3- event complete knowledge base extension；Personage continues to expand
Want prime implicant: personage 1, personage 2, personage 3, personage 1, personage 2, personage 1, personage 2, personage 3, personage 1, personage 2, event extension
For event 1, event 2, until expanding information set.
5. across media information target component correlation analysis systems as described in claim 1, it is characterised in that: utilize different information
The training for analyzing the training set marked now obtains different decision classifying trees, in the case where analyzing purpose by information training set by difference
Crawler behavior of the target in different location carries out classification mark, using decision tree be expressed as the node attribute list of leaf node as
The target component attribute of information, equipment attribute of the root node as target；Attribute is equipped as first layer node, site attribute is made
For second layer node, behavior property is as third layer node；Node is equipped according to the target component attribute of information database by feelings
Report is split into " equipment 1 ", " equipment 2 ", " equipment 3 " three classes, and " equipment 1 " target class information is " place according to place node split
1 ", " place 2 " two classes；" equipment 2 " target class information is " place 1 ", " place 2 ", " place 3 " three according to place node split
Class；" equipment 3 " target class information is " place 1 ", " place 2 " two classes according to place node split, forms target in somewhere region
Information set；Place node forms crawler behavior information set of the target in somewhere region, all kinds of middle feelings in upper layer according to behavior property
Report continue to divide according to corresponding node, until in the class of division sample class mark it is all the same, construct the node of decision tree
Attribute list obtains the final classification of sample, then using ID3 algorithm training decision classifying tree.
6. across media information target component correlation analysis systems as described in claim 1, it is characterised in that: user inputs retrieval
Vocabulary, and intelligence analysis purpose is selected to correspond to categorised decision tree-model, the target component model of information is then read according to index,
As source, user element extension rule is extended to obtain new mesh by source target component the target component of the information gone out using initial retrieval
Element is marked, then retrieval extension target component vocabulary, obtains extension information relevant to extension element vocabulary；Finally use user
The information Clustering Decision-Making tree-model of selection is associated cluster to above-mentioned initial retrieval information and extension information, obtains information
Target association relational result, and goal activities place, activity time are counted according to association results, the mechanics of target is exported,
And the mechanics of the fresh targets such as related person, equipment expanded by target.
7. a kind of association analysis method for being carried out across media information target components based on system described in claim 1, is had as follows
Technical characteristic: across media information target component information correlation analysis systems (hereinafter referred to as system) are by parsing and pre-process across matchmaker
Body information obtains the text segmentation sequence of information；According to the target component extracted in text segmentation sequence, target component mould is established
Type；It is analyzed for object event, information and target component mould that description object event wants prime implicant is established centered on object event
Type constructs target component dictionary using domain knowledge, carries out Chinese name, Entity recognition, reality to the text segmentation sequence of information
Pronouns, general term for nouns, numerals and measure words classification carries out assignment to target component model by classification results；The index between information and target component model is established, with
And information data is indexed and is put in storage by configuration information title, information ID number, the index of building information and target component model is incorporated to
Library；In online association treatment process, input retrieval vocabulary, matching is retrieved vocabulary and information text sequence of words, is obtained comprising inspection
The information of rope vocabulary；According to the index of above-mentioned foundation, the target component model of the information retrieved is read；Then using based on neck
The element extension rule that domain knowledge is established, target component is extended, new target component is obtained；User search extends element
Vocabulary, be expanded information, according to the index that system is established, reads the target component model of all information retrieved；System
The Clustering Decision-Making tree selected using user, carries out the cluster association of target component, carries out information integration using target association relationship.
8. association analysis method as claimed in claim 7, it is characterised in that: using domain knowledge building top layer to bottom successively
For information data layer, object event layer, element layer target component dictionary, wherein element layer corresponds to target component model foundation
Element dictionary D={ D1,D2,D3,D4,D5,D6,D7, D1For the element of time dictionary of object event, D2For place element dictionary, D3
For personage's element dictionary, D4To equip element dictionary, D5For organizational factors dictionary, D6Behavioral primitive dictionary, and element dictionary
Value set carries out statistics filling using a large amount of information datas and obtains.
9. association analysis method as claimed in claim 7, it is characterised in that: according to concept similarity presentation-entity vocabulary wiBelong to
In element EvDegree calculate information d with semantic description conceptmText segmentation sequence in non-Matching Elements item entity word wi
With element dictionary DvConcept similarity, calculating formula of similarity are as follows:
As text segmentation sequence dmIn vocabulary
wiWith element dictionary DvConcept similarity Dissim(wi,Dv) when meeting threshold requirement, it is believed that entity vocabulary wiBelong to target component
Ev, then using the entity word as the element E of targetvAssignment；Wherein, v=1,2,3,4,5,6, DisSim(wi,Dv) presentation-entity
Word wiWith element dictionary DvConcept distance, NvRefer to element dictionary DvIn vocabulary sum, tjIt is element dictionary DvIn vocabulary, a
For constant, pijFor entity word wi, vocabulary tjSemantic distance in semantic tree.
10. association analysis method as claimed in claim 7, it is characterised in that: using the purpose of intelligence analysis as factor, introduce
Into association cluster decision-tree model, the association of information is instructed, according to intelligence analysis experience, extracts typical intelligence analysis mesh
, i.e., intelligence analysis when pay close attention to the angle of target, according to different analysis purposes by information training set S={ d1,d2,…,dMInto point
Class mark obtains training set { S, the C under corresponding K kind intelligence analysis purposek, k=1,2 ..., K, wherein CkIndicate kth kind feelings
It calls the score to analyse and the classification of training set S is marked under purpose, Ck={ y1,y2,…,yM, ymFor dmMark classification, value ym=
{1,2,…,Nk, NkIndicate total classification number in the case of marking in kth.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811134544.1A CN109508453A (en) | 2018-09-28 | 2018-09-28 | Across media information target component correlation analysis systems and its association analysis method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811134544.1A CN109508453A (en) | 2018-09-28 | 2018-09-28 | Across media information target component correlation analysis systems and its association analysis method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109508453A true CN109508453A (en) | 2019-03-22 |
Family
ID=65746264
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811134544.1A Pending CN109508453A (en) | 2018-09-28 | 2018-09-28 | Across media information target component correlation analysis systems and its association analysis method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109508453A (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110287482A (en) * | 2019-05-29 | 2019-09-27 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Semi-automation participle corpus labeling training device |
| CN110287481A (en) * | 2019-05-29 | 2019-09-27 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Name entity corpus labeling training system |
| CN110288001A (en) * | 2019-05-28 | 2019-09-27 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Target identification method based on the training study of target data feature |
| CN110781348A (en) * | 2019-10-25 | 2020-02-11 | 北京威晟艾德尔科技有限公司 | Video file analysis method |
| CN110825839A (en) * | 2019-11-07 | 2020-02-21 | 成都国腾实业集团有限公司 | Incidence relation analysis method for targets in text information |
| CN111046130A (en) * | 2019-11-08 | 2020-04-21 | 杭州安恒信息技术股份有限公司 | Association retrieval method combining ElasticSearch and FSM |
| CN111191046A (en) * | 2019-12-31 | 2020-05-22 | 北京明略软件系统有限公司 | Method, device, computer storage medium and terminal for realizing information search |
| CN111339398A (en) * | 2019-12-19 | 2020-06-26 | 杭州安恒信息技术股份有限公司 | Diversified big data information analysis system and analysis method thereof |
| CN111680173A (en) * | 2020-05-31 | 2020-09-18 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | CMR model for uniformly retrieving cross-media information |
| CN111861830A (en) * | 2020-04-03 | 2020-10-30 | 深圳市天彦通信股份有限公司 | Information cloud platform |
| CN111859573A (en) * | 2020-07-22 | 2020-10-30 | 重庆大学 | Structural characterization method for technological process |
| CN111949696A (en) * | 2020-07-08 | 2020-11-17 | 国家计算机网络与信息安全管理中心 | Full-factor correlation analysis method and device |
| CN112307768A (en) * | 2019-07-25 | 2021-02-02 | 北京知元创通信息技术有限公司 | Artificial intelligence technology enterprise-oriented information monitoring method |
| CN112765992A (en) * | 2021-01-14 | 2021-05-07 | 深圳市人马互动科技有限公司 | Training data construction method and device, computer equipment and storage medium |
| CN113254641A (en) * | 2021-05-27 | 2021-08-13 | 中国电子科技集团公司第十五研究所 | Information data fusion method and device |
| CN113705415A (en) * | 2021-08-23 | 2021-11-26 | 中国电子科技集团公司第十五研究所 | Radar information-based air situation target feature extraction method and device |
| CN114282072A (en) * | 2022-03-01 | 2022-04-05 | 智器云南京信息科技有限公司 | Method and system for constructing intelligence teaching information network |
| CN115017336A (en) * | 2022-08-03 | 2022-09-06 | 中国电子科技集团公司第五十四研究所 | Deep learning model interpretation method based on task cognition |
| CN116628628A (en) * | 2023-04-17 | 2023-08-22 | 华北科技学院(中国煤矿安全技术培训中心) | User information literacy analysis method, system and storage medium based on search information |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105005578A (en) * | 2015-05-21 | 2015-10-28 | 中国电子科技集团公司第十研究所 | Multimedia target information visual analysis system |
| CN105677873A (en) * | 2016-01-11 | 2016-06-15 | 中国电子科技集团公司第十研究所 | Text information associating and clustering collecting processing method based on domain knowledge model |
| CN105740228A (en) * | 2016-01-25 | 2016-07-06 | 云南大学 | Internet public opinion analysis method |
| CN107194468A (en) * | 2017-04-19 | 2017-09-22 | 哈尔滨工程大学 | Towards the decision tree Increment Learning Algorithm of information big data |
| WO2017210949A1 (en) * | 2016-06-06 | 2017-12-14 | 北京大学深圳研究生院 | Cross-media retrieval method |
| US20180189691A1 (en) * | 2017-01-04 | 2018-07-05 | Richard Oehrle | Analytical system for assessing certain characteristics of organizations |
-
2018
- 2018-09-28 CN CN201811134544.1A patent/CN109508453A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105005578A (en) * | 2015-05-21 | 2015-10-28 | 中国电子科技集团公司第十研究所 | Multimedia target information visual analysis system |
| CN105677873A (en) * | 2016-01-11 | 2016-06-15 | 中国电子科技集团公司第十研究所 | Text information associating and clustering collecting processing method based on domain knowledge model |
| CN105740228A (en) * | 2016-01-25 | 2016-07-06 | 云南大学 | Internet public opinion analysis method |
| WO2017210949A1 (en) * | 2016-06-06 | 2017-12-14 | 北京大学深圳研究生院 | Cross-media retrieval method |
| US20180189691A1 (en) * | 2017-01-04 | 2018-07-05 | Richard Oehrle | Analytical system for assessing certain characteristics of organizations |
| CN107194468A (en) * | 2017-04-19 | 2017-09-22 | 哈尔滨工程大学 | Towards the decision tree Increment Learning Algorithm of information big data |
Non-Patent Citations (3)
| Title |
|---|
| EMMANUEL AHISHAKIYE 等: "Crime Prediction Using Decision Tree (J48)Classification Algorithm", 《INTERNATIONAL JOURNAL OF COMPUTER AND INFORMATION TECHNOLOGY》 * |
| 张伟民: "数字图书馆跨媒体语义检索模型研究", 《图书馆学刊》 * |
| 明均仁 等: "基于语义关联挖掘的数字图书馆跨媒体检索方法研究", 《图书馆情报工作》 * |
Cited By (30)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110288001B (en) * | 2019-05-28 | 2023-09-05 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Target recognition method based on target data feature training learning |
| CN110288001A (en) * | 2019-05-28 | 2019-09-27 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Target identification method based on the training study of target data feature |
| CN110287481A (en) * | 2019-05-29 | 2019-09-27 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Name entity corpus labeling training system |
| CN110287482A (en) * | 2019-05-29 | 2019-09-27 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Semi-automation participle corpus labeling training device |
| CN110287481B (en) * | 2019-05-29 | 2022-06-14 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Named entity corpus labeling training system |
| CN110287482B (en) * | 2019-05-29 | 2022-07-08 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | Semi-automatic participle corpus labeling training device |
| CN112307768A (en) * | 2019-07-25 | 2021-02-02 | 北京知元创通信息技术有限公司 | Artificial intelligence technology enterprise-oriented information monitoring method |
| CN110781348A (en) * | 2019-10-25 | 2020-02-11 | 北京威晟艾德尔科技有限公司 | Video file analysis method |
| CN110825839A (en) * | 2019-11-07 | 2020-02-21 | 成都国腾实业集团有限公司 | Incidence relation analysis method for targets in text information |
| CN111046130A (en) * | 2019-11-08 | 2020-04-21 | 杭州安恒信息技术股份有限公司 | Association retrieval method combining ElasticSearch and FSM |
| CN111046130B (en) * | 2019-11-08 | 2023-05-23 | 杭州安恒信息技术股份有限公司 | Correlation retrieval method combining elastic search and FSM |
| CN111339398A (en) * | 2019-12-19 | 2020-06-26 | 杭州安恒信息技术股份有限公司 | Diversified big data information analysis system and analysis method thereof |
| CN111191046A (en) * | 2019-12-31 | 2020-05-22 | 北京明略软件系统有限公司 | Method, device, computer storage medium and terminal for realizing information search |
| CN111861830B (en) * | 2020-04-03 | 2024-04-26 | 深圳市天彦通信股份有限公司 | Information cloud platform |
| CN111861830A (en) * | 2020-04-03 | 2020-10-30 | 深圳市天彦通信股份有限公司 | Information cloud platform |
| CN111680173B (en) * | 2020-05-31 | 2024-02-23 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | CMR model for unified searching cross-media information |
| CN111680173A (en) * | 2020-05-31 | 2020-09-18 | 西南电子技术研究所（中国电子科技集团公司第十研究所） | CMR model for uniformly retrieving cross-media information |
| CN111949696A (en) * | 2020-07-08 | 2020-11-17 | 国家计算机网络与信息安全管理中心 | Full-factor correlation analysis method and device |
| CN111859573B (en) * | 2020-07-22 | 2023-11-03 | 重庆大学 | Structural characterization method of technological process |
| CN111859573A (en) * | 2020-07-22 | 2020-10-30 | 重庆大学 | Structural characterization method for technological process |
| CN112765992A (en) * | 2021-01-14 | 2021-05-07 | 深圳市人马互动科技有限公司 | Training data construction method and device, computer equipment and storage medium |
| CN113254641A (en) * | 2021-05-27 | 2021-08-13 | 中国电子科技集团公司第十五研究所 | Information data fusion method and device |
| CN113254641B (en) * | 2021-05-27 | 2021-11-16 | 中国电子科技集团公司第十五研究所 | Information data fusion method and device |
| CN113705415B (en) * | 2021-08-23 | 2023-10-27 | 中国电子科技集团公司第十五研究所 | Air condition target feature extraction method and device based on radar information |
| CN113705415A (en) * | 2021-08-23 | 2021-11-26 | 中国电子科技集团公司第十五研究所 | Radar information-based air situation target feature extraction method and device |
| CN114282072A (en) * | 2022-03-01 | 2022-04-05 | 智器云南京信息科技有限公司 | Method and system for constructing intelligence teaching information network |
| CN115017336B (en) * | 2022-08-03 | 2022-11-01 | 中国电子科技集团公司第五十四研究所 | Deep learning model interpretation method based on task cognition |
| CN115017336A (en) * | 2022-08-03 | 2022-09-06 | 中国电子科技集团公司第五十四研究所 | Deep learning model interpretation method based on task cognition |
| CN116628628A (en) * | 2023-04-17 | 2023-08-22 | 华北科技学院(中国煤矿安全技术培训中心) | User information literacy analysis method, system and storage medium based on search information |
| CN116628628B (en) * | 2023-04-17 | 2024-01-30 | 华北科技学院(中国煤矿安全技术培训中心) | User information literacy analysis method, system and storage medium based on search information |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109508453A (en) | Across media information target component correlation analysis systems and its association analysis method | |
| CN109948911B (en) | Evaluation method for calculating network product information security risk | |
| CN107609052B (en) | A kind of generation method and device of the domain knowledge map based on semantic triangle | |
| CN104834686B (en) | A kind of video recommendation method based on mixing semantic matrix | |
| CN108920544A (en) | A kind of personalized position recommended method of knowledge based map | |
| CN105045875B (en) | Personalized search and device | |
| CN106815369A (en) | A kind of file classification method based on Xgboost sorting algorithms | |
| CN111382276B (en) | Event development context graph generation method | |
| CN108427714A (en) | The source of houses based on machine learning repeats record recognition methods and system | |
| CN106815310A (en) | A kind of hierarchy clustering method and system to magnanimity document sets | |
| CN110110225A (en) | Online education recommended models and construction method based on user behavior data analysis | |
| CN104778224A (en) | Target object social relation identification method based on video semantics | |
| CN117290489B (en) | Method and system for quickly constructing industry question-answer knowledge base | |
| CN107577782A (en) | A kind of people-similarity depicting method based on heterogeneous data | |
| CN110688549A (en) | Artificial intelligence classification method and system based on knowledge system map construction | |
| CN105869058B (en) | A kind of method that multilayer latent variable model user portrait extracts | |
| CN107908749B (en) | Character retrieval system and method based on search engine | |
| CN110309355A (en) | Generation method, device, equipment and the storage medium of content tab | |
| Feng et al. | ModelsKG: A Design and Research on Knowledge Graph of Multimodal Curriculum Based on PaddleOCR and DeepKE | |
| CN115712511A (en) | Web service classification method based on bilinear graph neural network | |
| Xu et al. | Estimating similarity of rich internet pages using visual information | |
| Shen et al. | Predicting named entity location using Twitter | |
| CN114003803A (en) | Method and system for discovering media account in specific region on social platform | |
| Zheng | Individualized Recommendation Method of Multimedia Network Teaching Resources Based on Classification Algorithm in a Smart University | |
| Stoica et al. | A Local Structure-Based Method for Nodes Clustering: Application to a Large Mobile Phone Social Network |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190322 |
|
| WD01 | Invention patent application deemed withdrawn after publication |