CN109684625A - Entity handles method, apparatus and storage medium - Google Patents
Entity handles method, apparatus and storage medium Download PDFInfo
- Publication number
- CN109684625A CN109684625A CN201811290669.3A CN201811290669A CN109684625A CN 109684625 A CN109684625 A CN 109684625A CN 201811290669 A CN201811290669 A CN 201811290669A CN 109684625 A CN109684625 A CN 109684625A
- Authority
- CN
- China
- Prior art keywords
- entity
- entities
- target
- built
- similarity
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
- G06F40/295—Named entity recognition
Abstract
The present invention provides a kind of entity handles method, apparatus and storage medium, this method comprises: being grouped to the target entity in knowledge mapping, obtains multiple group of entities, includes multiple target entities pair in each group of entities;The similarity of two target entities of each target entity centering of each group of entities is obtained, and two target entities that similarity is greater than similarity threshold are merged, obtains multiple new group of entities;The associated entity of side object yet to be built is obtained in multiple new group of entities, and side object yet to be built is associated with associated entity;Object in triple yet to be built when object is yet to be built, the associated entity of side object yet to be built are as follows: be greater than the target entity of similarity threshold with the similarity of side object yet to be built.The present invention carries out entity in new entity set and builds side, so that entity fusion is built side with entity and is associated with, improves the degree of communication of knowledge mapping, and carry out entity fusion by the way of similarity, improves the accuracy of entity fusion.
Description
Technical field
The present invention relates to technical field of data processing more particularly to a kind of entity handles method, apparatus and storage mediums.
Background technique
Knowledge mapping (Knowledge Graph) is also known as mapping knowledge domains.Knowledge mapping either academia also
It is that industry suffers from very important status, it is the basis of artificial intelligence, and realizes the necessary of the applications such as intelligent answer
Road, efficiently and easily can provide information for user.It include multiple architectural entities, i.e., a large amount of Subject, Predicate and Object three in knowledge mapping
The set of tuple (SPO), S, that is, subject, P, that is, predict, O, that is, object.Generally comprise to the processing of entity: entity melts
It closes, entity builds side;Entity fusion refers to that novel entities are carried out include with normalizing, i.e., by the entity in novel entities and knowledge mapping
Match, novel entities and entity matched in knowledge mapping are merged;Entity is built side and is referred to the O progress entity disambiguation in SPO,
The entity of corresponding same concept is found in knowledge mapping.
In the prior art, without establishing connection between two steps handled entity, i.e., the two is isolated separates
Processing, entity cannot effectively be handled, and in the prior art to the fusion of entity only according to entity name and other
The mode of name is matched, and the accuracy of fusion is low.
Summary of the invention
The present invention provides a kind of entity handles method, apparatus and storage medium, and entity is carried out in new entity set and builds side,
So that entity fusion is built side with entity and is associated with, improve the degree of communication of knowledge mapping, and by the way of similarity into
The fusion of row entity, improves the accuracy of entity fusion.
The first aspect of the present invention provides a kind of entity handles method, comprising:
Target entity in knowledge mapping is grouped, multiple group of entities are obtained, includes more in each group of entities
A target entity pair, a target entity in the group of entities is to being by a target entity in the group of entities and its
What his target entity formed;
Obtain the similarity of two target entities of each of each described group of entities target entity centering, and by phase
It is merged like two target entities that degree is greater than similarity threshold, obtains multiple new group of entities, each new entity
The target entity that similarity is greater than similarity threshold is not included in group;
Obtain the associated entity of side object yet to be built in multiple new group of entities, and by the side object yet to be built and institute
Associated entity is stated to be associated;Object in the Subject, Predicate and Object triple yet to be built when object is described yet to be built, it is described yet to be built
The associated entity of side object are as follows: in the new group of entities, be greater than the similarity with the similarity of the side object yet to be built
The target entity of threshold value.
Optionally, the target entity in knowledge mapping is grouped, before obtaining multiple group of entities, comprising:
According to the semanteme of each entity in the knowledge mapping, the target entity is obtained.
Optionally, the phase for obtaining each of each group of entities two target entities of the target entity centering
Like degree, comprising:
Using preset attribute comparison method and Attribute Significance, each two target realities of the target entity centering are obtained
Attributes similarity between body;
Using preset iterative model, calculating is iterated to the attributes similarity, obtains each target entity
The similarity of two target entities of centering.
Optionally, using preset attribute comparison method and Attribute Significance, each target entity centering two is obtained
Attributes similarity between a target entity, comprising:
According to the comparison information in the Attribute Significance, determine each two target entities of the target entity centering it
Between multiple attributes pair to be compared;
According to the feature of each attribute pair to be compared, corresponding ratio is chosen from the preset attribute comparison method
Attribute is carried out to comparing compared with method, obtains the attributes similarity between each two target entities of the target entity centering.
Optionally, the associated entity that side object yet to be built is obtained in each new group of entities, comprising:
If in the new group of entities there are target entity be the side object yet to be built default associated entity when, by institute
Target entity is stated as the associated entity;And/or
If it exists the title of target entity it is identical as the side name of an object yet to be built or there are target entity with it is described
Build while entity have cooccurrence relation or there are the type of target entity with it is described yet to be built while the type of object it is identical when, by institute
Target entity is stated as the associated entity;And/or
In the new group of entities, multiple candidate target entities of each side object yet to be built are obtained;
The similarity for obtaining each the side object yet to be built and each candidate target entity, by the corresponding time of maximum similarity
Select target entity as candidate association entity;
It is if the corresponding entity of the side object yet to be built is present in the knowledge mapping, the candidate association entity is true
It is set to the associated entity.
Optionally, the similarity for obtaining each the side object yet to be built and each candidate target entity, comprising:
It is corresponding to feature and group feature to obtain each candidate target entity, described is that measurement is described yet to be built to feature
The feature of similitude between side object and the candidate target entity, described group of feature are to measure the visitor including the side yet to be built
The feature of similitude between the triple of body and the candidate target entity.
Optionally, if the corresponding entity of the side object yet to be built is present in the knowledge mapping, by the time
Associated entity is selected to be determined as before the associated entity, further includes:
It determines whether the candidate association entity exists with the entity on the side yet to be built to conflict;
If it is not, judging that the corresponding entity of the side object yet to be built whether there is in the knowledge mapping.
The second aspect of the present invention provides a kind of entity handles device, comprising:
Grouping module obtains multiple group of entities, each reality for being grouped to the target entity in knowledge mapping
It include multiple target entities pair in body group, a target entity in the group of entities is to being by one in the group of entities
What target entity and other target entities formed;
Fusion Module, for obtaining two target entities of each of each described group of entities target entity centering
Similarity, and two target entities that similarity is greater than similarity threshold are merged, multiple new group of entities are obtained, each
The target entity that similarity is greater than similarity threshold is not included in the new group of entities;
Relating module, for obtaining the associated entity of side object yet to be built in multiple new group of entities, and will be described
Side object yet to be built is associated with the associated entity;In the Subject, Predicate and Object triple yet to be built when object is described yet to be built
Object, the associated entity of the side object yet to be built are as follows: big with the similarity of the side object yet to be built in the new group of entities
In the target entity of the similarity threshold.
Optionally, described device further include: target entity obtains module;
The target entity obtains module, for the semanteme according to each entity in the knowledge mapping, described in acquisition
Target entity.
Optionally, the Fusion Module is specifically used for using preset attribute comparison method and Attribute Significance, obtains every
Attributes similarity between a two target entities of the target entity centering;Using preset iterative model, to the attribute
Similarity is iterated calculating, obtains the similarity of two target entities of each target entity centering.
Optionally, the Fusion Module, specifically for determining each institute according to the comparison information in the Attribute Significance
State multiple attributes pair to be compared between two target entities of target entity centering;
According to the feature of each attribute pair to be compared, corresponding ratio is chosen from the preset attribute comparison method
Attribute is carried out to comparing compared with method, obtains the attributes similarity between each two target entities of the target entity centering.
Optionally, the relating module, if specifically in the new group of entities there are target entity be it is described to
When building the default associated entity of side object, using the target entity as the associated entity;And/or
If it exists the title of target entity it is identical as the side name of an object yet to be built or there are target entity with it is described
Build while entity have cooccurrence relation or there are the type of target entity with it is described yet to be built while the type of object it is identical when, by institute
Target entity is stated as the associated entity;And/or
In the new group of entities, multiple candidate target entities of each side object yet to be built are obtained;
The similarity for obtaining each the side object yet to be built and each candidate target entity, by the corresponding time of maximum similarity
Select target entity as candidate association entity;
It is if the corresponding entity of the side object yet to be built is present in the knowledge mapping, the candidate association entity is true
It is set to the associated entity.
Optionally, it is corresponding to feature and group to be specifically used for each candidate target entity of acquisition for the relating module
Feature, described is to measure the feature of the similitude between the side object yet to be built and the candidate target entity to feature, described
Group feature is to measure the feature of the similitude between the triple and the candidate target entity of the object including the side yet to be built.
Optionally, the relating module, specifically for the determination candidate association entity whether the reality with the side yet to be built
There is conflict in body;If it is not, judging that the corresponding entity of the side object yet to be built whether there is in the knowledge mapping.
The third aspect of the present invention provides a kind of entity handles device, comprising: at least one processor and memory;
The memory stores computer executed instructions;
At least one described processor executes the computer executed instructions of the memory storage, so that the entity handles
Device executes above-mentioned entity handles method.
The fourth aspect of the present invention provides a kind of computer readable storage medium, deposits on the computer readable storage medium
Computer executed instructions are contained, when the computer executed instructions are executed by processor, realize above-mentioned entity handles method.
The present invention provides a kind of entity handles method, apparatus and storage medium, this method comprises: to the mesh in knowledge mapping
Mark entity is grouped, and obtains multiple group of entities, includes multiple target entities pair in each group of entities;Obtain each group of entities
The similarity of two target entities of each target entity centering, and two target entities by similarity greater than similarity threshold
It is merged, obtains multiple new group of entities;Obtain the associated entity of side object yet to be built in multiple new group of entities, and will be to
Side object is built to be associated with associated entity;Object in triple yet to be built when object is yet to be built, the pass of side object yet to be built
Join entity are as follows: be greater than the target entity of similarity threshold with the similarity of side object yet to be built.The present invention in new entity set into
Row entity builds side, so that entity fusion is built side with entity and is associated with, improves the degree of communication of knowledge mapping, and using similar
The mode of degree carries out entity fusion, improves the accuracy of entity fusion.
Detailed description of the invention
Fig. 1 is the flow diagram one of entity handles method provided by the invention;
Fig. 2 is the flow diagram two of entity handles method provided by the invention;
Fig. 3 builds the process of another feasible embodiment on side for the entity in entity handles method provided by the invention
Schematic diagram;
Fig. 4 is the structural schematic diagram one of entity handles device provided by the invention;
Fig. 5 is the structural schematic diagram two of entity handles device provided by the invention;
Fig. 6 is the structural schematic diagram three of entity handles device provided by the invention.
Specific embodiment
To make the object, technical solutions and advantages of the present invention clearer, below in conjunction with the embodiment of the present invention, to this
Technical solution in inventive embodiments is clearly and completely described, it is clear that described embodiment is that a part of the invention is real
Example is applied, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not making creation
Property labour under the premise of every other embodiment obtained, shall fall within the protection scope of the present invention.
Technical term paraphrase of the present invention:
Triple: it when constructing knowledge mapping, needs to carry out Knowledge Extraction to document, using the form of triple to document
Content carry out corresponding expression, include entity, object in triple, and, the relation on attributes between subject and object;Wherein,
Entity S, that is, subject, relation on attributes P, that is, predict, object O, that is, object;Illustratively, knowledge such as is extracted to document content
For " daughter of A is B ", then entity is A, and object B, relation on attributes is daughter.
Entity: the subject in Subject, Predicate and Object triplet information.
Object: the object in Subject, Predicate and Object triplet information.
Fig. 1 is the flow diagram one of entity handles method provided by the invention, the executing subject of method flow shown in Fig. 1
It can be entity handles device, which can be by arbitrary software and or hardware realization.As shown in Figure 1, this implementation
Example provide entity handles method may include:
S101 is grouped the target entity in knowledge mapping, obtains multiple group of entities, includes more in each group of entities
A target entity pair, a target entity in group of entities is to being by a target entity and other target entity groups in group of entities
At.
Include multiple entities in knowledge mapping, may include repetition entity in multiple entity;As an entity has
Multiple alias, multiple corresponding entity of alias and the entity are to repeat entity;Alternatively, the semanteme of an entity and another entity
It is semantic identical, it can be using two entities as repetition entity.When in knowledge mapping including repetition entity, if will not repeat
Entity is merged, and illustratively, if A and B belong to repetition entity, knowledge mapping does not merge A and B, then user into
When the inquiry of row knowledge mapping, if the query entity of input is A, which can only return to use for the corresponding document of A
Family;But if knowledge mapping merges A and B, and which can return to user for the corresponding document of A and B, so that
The document that user obtains is more fully, accurately.
Target entity in the present embodiment in knowledge mapping be there may be duplicate entity, the target entity can be by
User carries out multiple entities of selection according to actual entities data.Specifically, being grouped to target entity, multiple entities are obtained
The concrete mode of group are as follows: semantic identical entity is divided into a group of entities;Alternatively, will include identical entity name
The entity of (including alias) is divided into a group of entities;Alternatively, will include that the entity of same alike result is divided into a group of entities, such as
The attribute of entity star A is respectively gender, nationality, and the attribute of entity star B is respectively birthday, nationality, wherein entity star A
It include attribute nationality with entity star B, it therefore, can be according to attribute having the same by entity star A and entity star B points
For a group of entities.
Specifically, each group of entities includes multiple target entities, wherein in a target entity and the group of entities
Other each target entities separately constitute target entity pair, include multiple target entities pair in each group of entities.Illustratively, such as
It include target entity 1, target entity 2 and target entity 3, the then entity pair for including are as follows: (mesh in group of entities A in group of entities A
Mark entity 1, target entity 2), (target entity 1, target entity 3) and (target entity 2, target entity 3).
It is envisioned that the mode for obtaining the target entity in knowledge mapping can also exist in advance other than user's selection
Entity handles dress centers target entity and obtains rule, such as according to the semanteme of entity or entity title or whether have
There is identical attribute, obtains that multiple there may be duplicate target entities in multiple entities.
In the present embodiment, the target entity in knowledge mapping is obtained in advance, delineation is right there may be duplicate entity in advance
The target entity of delineation carries out the calculating of similarity, avoids the meter that all entities in knowledge mapping are compared with generation
Calculate redundancy, it is possible to reduce the number that entity compares, and then reduce time complexity.
S102 obtains the similarity of two target entities of each target entity centering of each group of entities, and will be similar
Two target entities that degree is greater than similarity threshold are merged, the multiple new group of entities of acquisition, in each new group of entities not
It is greater than the target entity of similarity threshold comprising similarity.
Two target entities of each target entity centering of each group of entities have attribute information.In the present embodiment,
It can be according to Apriori property information, between two target entities that an entity centering is calculated using each entity attributes information
Similarity, wherein entity similarity is the similarity degree between two target entities, for determining that can two target entities
Fusion normalizing is carried out in knowledge mapping.
In the present embodiment, similarity threshold can be preset in entity handles device, wherein the similarity threshold can
The empirical value for thinking user is also possible to user according to the entity actually compared to the threshold value being configured;In each group of entities
In the middle similarity for obtaining multiple entities pair, when being greater than similarity threshold there are one or more entity similarities, it is determined that big
In the corresponding entity centering of the similarity of similarity threshold two target entities be repeat entity.
After entity handles device obtains repetition entity, two that entity, i.e. similarity is greater than similarity threshold will be repeated
Target entity is merged, and multiple new group of entities are obtained.Specifically, carrying out fusion to two target entities can be two
Target entity is associated, which is stored in a new group of entities, and will be real in the new group of entities
Body is indicated by a target entity;It is envisioned that may include in a group of entities it is multiple repeat entity, can be with
Multiple entity that repeats is stored in a new group of entities.
Wherein, the target entity that similarity is greater than similarity threshold is not included in each new group of entities;Specifically, each
Attribute in new group of entities is the adduction of the attribute of original target entity;Wherein, the attribute between entity is different if it exists
When, by the corresponding attribute of entity of the adduction of the attribute group of entities new as this.It is worth noting that, if there are phases between entity
Same attribute, then obtain true attribute according to the renewal time of the corresponding document of attribute, and true attribute refers to the renewal time of document
Newest attribute.
Illustratively, all target entities of the entity centering for including in group of entities A are repetition entity, therefore by target
Entity 1, target entity 2, target entity 3 are determined as repeating entity, and target entity 1, target entity 2, target entity 3 are stored
In a new entity sets A'.Wherein, target entity 1, target entity 2, target entity 3 are associated, alternatively, the new reality
The corresponding entity name of body set A' be target entity 1, target entity 2, in target entity 3 any one target entity name
Claim;The attribute of the new entity sets A' is the adduction of target entity 1, target entity 2, the corresponding attribute of target entity 3.
Such as: the corresponding attribute of target entity 1 is a, b and c, and the corresponding attribute of target entity 2 is d, and target entity 3 is corresponding
Attribute is a;Then the attribute of entity sets A' is (a, b, c, d), wherein target entity 1 and the corresponding attribute a of target entity 3 are
Same alike result then judges the renewal time of target entity 1 and the corresponding document of target entity 3, if target entity 1 is corresponding
The renewal time of document is newer than the renewal time of the corresponding document of target entity 1, then the corresponding attribute a of target entity 1 is true
The a being set in the attribute of entity sets A'.
S103, obtains the associated entity of side object yet to be built in multiple new group of entities, and by side object yet to be built be associated with
Entity is associated;Object in Subject, Predicate and Object triple yet to be built when object is yet to be built, the associated entity of side object yet to be built are as follows:
In new group of entities, the target entity of similarity threshold is greater than with the similarity of side object yet to be built.
Entity builds side, i.e., carries out entity disambiguation to the object O in SPO, corresponding same concept is found in knowledge mapping
Entity be associated, increase knowledge mapping in connection degree.Various entities are proposed in the prior art and build side mode, are such as compared
There are commonly: by machine learning method by knowledge base entity and relationship be converted to vector expression, between vector away from
From illustrating contacting between entity and entity, entity and relationship, but this mode needs to carry out all entities and attribute
Training, the training time is longer, and treatment effeciency is low.
In the present embodiment, after carrying out normalizing fusion to the entity in knowledge mapping, the association for obtaining side object yet to be built is real
Body, so that side object yet to be built is associated with the entity in the knowledge mapping.Wherein, the Subject, Predicate and Object three yet to be built when object is yet to be built
Object in tuple, the associated entity of side object yet to be built are as follows: in new group of entities, be greater than phase with the similarity of side object yet to be built
Like the target entity of degree threshold value.Specifically, the acquisition side of the similarity between entity in side object yet to be built and new group of entities
Formula, can be identical as the mode of similarity of two target entities of above-mentioned acquisition target entity centering, and this will not be repeated here.
Wherein, it after the associated entity that side object yet to be built is obtained in each new group of entities, by the side object yet to be built and closes
Connection entity is associated, so that the associated entity is associated with side object yet to be built.Specifically, being associated looking into for entity in user
When asking word, the corresponding document of the associated entity can be returned for user and the associated side object yet to be built of the associated entity is corresponding
Document so that user obtain search result it is more complete, accurately.
Entity handles method provided in this embodiment includes: to be grouped to the target entity in knowledge mapping, is obtained more
A group of entities includes multiple target entities pair in each group of entities;Obtain the two of each target entity centering of each group of entities
The similarity of a target entity, and two target entities that similarity is greater than similarity threshold are merged, it obtains multiple new
Group of entities;Obtain the associated entity of side object yet to be built in each new group of entities, and by side object yet to be built and associated entity
It is associated;Object in triple yet to be built when object is yet to be built, the associated entity of side object yet to be built are as follows: with side visitor yet to be built
The similarity of body is greater than the target entity of similarity threshold.The present invention carries out entity in new entity set and builds side, so that entity
Fusion is built side with entity and is associated with, and improves the degree of communication of knowledge mapping, and carry out entity by the way of similarity and melt
It closes, improves the accuracy of entity fusion.
On the basis of the above embodiments, the entity in entity handles method provided by the invention is melted below with reference to Fig. 2
Conjunction is described in detail, and Fig. 2 is the flow diagram two of entity handles method provided by the invention, as shown in Fig. 2, the present embodiment
The entity handles method of offer may include:
S201 obtains target entity according to the semanteme of each entity in knowledge mapping.
In the present embodiment, target entity is obtained in knowledge mapping, can will specifically be had according to the semanteme of each entity
The entity of identical semanteme is determined as target entity.It is envisioned that if in knowledge mapping, the semantic corresponding entity of A have it is multiple,
The semantic corresponding entity of B also has multiple, A semanteme and the semantic corresponding entity of B can be determined as target entity.Only exist
When being grouped to target entity, A semanteme and the semantic corresponding entity of B are respectively divided into different group of entities.
S202 is grouped the target entity in knowledge mapping, obtains multiple group of entities, includes more in each group of entities
A target entity pair.
S203 obtains two targets of each target entity centering using preset attribute comparison method and Attribute Significance
Attributes similarity between entity.
In the present embodiment, target entity in each group of entities is to being multiple, wherein two targets that entity centering includes
The corresponding attribute of entity is one or more.For each target entity pair, preset attribute comparison method and category can use
Property different degree calculates each attributes similarity probability between two entities of target entity centering.
Wherein, attribute comparison method refers to the method for how comparing similarity between attribute being arranged according to attributive character.
It generally may include: accurate comparison, editing distance compares, the time compares, text similarity compares, co-occurrence compares, number compares
Relatively etc. with type, particular user can select corresponding ratio according to the difference of attribute from preset attribute comparison method
Compared with method;Alternatively, be stored in entity handles device be solid data and attribute comparison method corresponding relationship, for including
The solid data of particular community selects corresponding attribute comparison method, obtains the similarity of attribute.
For example, the pre-set comparator accurately compared can be used and carried out to the attribute when attribute is character string
It is accurate to compare, if attribute is identical, 1 is returned, otherwise returns to 0;Alternatively, editing distance compares for returning to two character strings
Levinstein distance, a successive value of the result of return between 0-1.Year comparator is for comparing the time, two value differences
Absolute value be less than customized threshold value and then return to 1, otherwise return and return Plsa comparator for calculating the plsa similarity of two values,
Plsa model is obtained by Baidupedia training;Whether Coccur comparator is for judging first character string at second
Occurring in character string, appearance then returns to 1, and it otherwise returns and returns Phonenumber comparator whether identical for comparing two string numbers,
It is identical, 1 is returned, otherwise returns to 0;Float comparator is less than customized for comparing two floating numbers, two value absolute value of the difference
Threshold value then returns to 1, otherwise returns to 0.
In addition, it is contemplated that some attributes are monodromes, some attributes are multivalues, for example, this attribute of date of birth is monodrome,
For a cinematographic work, this attribute of performer is multivalue.Therefore, attribute comparison method further includes that monodrome compares and multivalue ratio
Compared with, wherein multivalue compares based on monodrome comparison result, needs to be arranged the pass of each monodrome comparison result Yu final multivalue result
System, that is to say, that multivalue compares the multiple values for needing to correspond to different entities under attribute and compares two-by-two, is then based on and compares two-by-two
Result obtain final comparison result.For example, performer A (entity) has m masterpiece (attribute representative work has m value), performer B
There is n masterpiece, the masterpiece of A and B is compared (this monodrome being equivalent under multivalue comparative approach compares) two-by-two, each monodrome ratio
It can be compared with result with the relationship of final multivalue result following any: having k (k is less than or equal to m) a masterpiece identical, it is believed that
The masterpiece attribute of A and B is identical, returns to 1, otherwise returns and returns all masterpieces all identical, just thinks that attribute is identical;K/m or
K/n is all larger than preset threshold, it is believed that attribute is identical;Directly using k/m or k/n as return value.
Attribute Significance is that user is preconfigured according to priori knowledge, indicate attribute comparison result for entity whether phase
Same significance level.Attribute Significance may include: that comparison information (whether being the attribute that must compare), entity confidence level are punished
Penalize information and entity confidence level award information.
Further, in the present embodiment, user can need to determine each target entity pair according to the different degree of attribute, selection
In multiple attributes pair to be compared between two target entities, the feature of each attribute pair to be compared compares from preset attribute
Corresponding comparative approach is chosen in method and carries out attribute to comparing, and is obtained between two target entities of each target entity centering
Attributes similarity.
S204 is iterated calculating to attributes similarity using preset iterative model, obtains each target entity centering
Two target entities similarity.
In the present embodiment, preset iterative model can be Bayesian inference model, wherein the principle of Bayesian inference is
Corresponding entity similarity probability value is calculated according to the prior probability that each attribute compares.Bayesian inference mould in the present embodiment
Type can be identical as model in the prior art, and this will not be repeated here.
Specifically, being iterated to calculate using Bayesian inference formula between two target entities in each target entity pair
Entity similarity probability when, the number of iterations is equal with the number of each attributes similarity probability, each attributes similarity probability
Corresponding an iteration, and the corresponding iteration sequence of each attributes similarity is unlimited need to be all over that is, when carrying out bayesian iterative
Each attributes similarity is gone through, without considering sequencing of each attributes similarity in iteration.When first iteration, it can incite somebody to action
The value of probcur is set as 0.5, and (when indicating primary iteration, the identical and different probability of the corresponding attribute of two entities is respectively
0.5), later in each iteration, the value of the probnext of all corresponding last iteration of the value of probcur.
S205 merges two target entities that similarity is greater than similarity threshold, obtains multiple new group of entities.
S206, obtains the associated entity of side object yet to be built in multiple new group of entities, and by side object yet to be built be associated with
Entity is associated.
Wherein, S202, S205-S206 in the present embodiment specifically can refer to S101, S102-S103 in above-described embodiment
In associated description, this will not be repeated here.
In the present embodiment, by utilizing preset attribute comparison method and Attribute Significance, each target entity pair is calculated
In each attributes similarity probability between two target entities, and using preset Bayesian model by priori knowledge and machine
Learning model effective integration is iterated calculating to each attributes similarity probability, can be improved the entity similarity between entity
The efficiency and accuracy of probability calculation.Bayes's scheme tune ginseng is convenient simultaneously and comes into force fastly, is applicable to quickly obtain result
Scene.
On the basis of the above embodiments, side is built to the entity in entity handles method provided by the invention below to carry out in detail
Describe in detail bright, the entity in entity handles method provided in this embodiment, which builds side S206, can specifically include three kinds of feasible embodiment party
Formula.
A kind of feasible embodiment are as follows:, can be using default associated entity for the side object yet to be built of negligible amounts
Mode.Specifically, default associated entity can be preset in entity handles device, if there are targets in new group of entities
When entity is the default associated entity of side object yet to be built, using target entity as associated entity.
Illustratively, such as in triple SPO, league matches, country etc. are to be able to carry out the P for building side, for it is each can
Corresponding default associated entity can be respectively configured in the P for build side, and can record in default associated entity has the corresponding side yet to be built P
Object and the corresponding entity of each side object yet to be built.Such as, P is constellation, corresponding side object yet to be built can include: Aries
Seat, Taurus, Gemini, Cancer, Leo, Virgo, Libra, day take off seat, Sagittarius, Capricorn, Aquarius and
Pisces, each side object yet to be built respectively correspond an entity.
In practical applications, it is possible to which identical name corresponds to different entities, for example constellation " Capricorn " and song " are rubbed
Castrated ram's seat ", can be distinguished by different entity ID.For side object yet to be built, corresponding mapping dictionary if it exists, then then
It can will be associated with recorded in corresponding mapping dictionary physically in side object yet to be built.Such as, side object yet to be built is " Capricorn ",
" Capricorn " is mapped to entity recorded in corresponding mapping dictionary i.e. constellation " Capricorn " physically.
Another feasible embodiment are as follows: if it exists the title of target entity it is identical as side name of an object yet to be built or
Person there are target entity with build while entity have cooccurrence relation or there are the type of target entity and it is yet to be built while object type
When identical, using target entity as associated entity.
In the present embodiment, target entity title, the type in new group of entities are inquired, and, corresponding relation on attributes is sentenced
Title in new group of entities of breaking with the presence or absence of target entity is identical as side name of an object yet to be built, or there are target entities
Type is identical as the type of side object yet to be built, title is identical or the identical target entity of type is as side object yet to be built
Associated entity.
Further, cooccurrence relation therein refers in side object yet to be built and new group of entities is between target entity
No there are identical relations on attributes, and if it exists, then determines associated entity of the target entity as side object yet to be built.
Illustratively, such as: the triplet information of the target entity in new group of entities is " wife of A is B ", and side yet to be built
The corresponding triplet information of object be " husband of B is A ", wherein A and B have corresponding cooccurrence relation, it is determined that A be it is yet to be built
The associated entity of side object B.
Another feasible embodiment are as follows: in new group of entities, according to multiple candidate mesh of each side object yet to be built
The similarity between entity is marked, the associated entity of each side object yet to be built is obtained.
Specifically, Fig. 3 builds another feasible embodiment party on side for the entity in entity handles method provided by the invention
The flow diagram of formula, as shown in figure 3, this method may include:
S2061 obtains multiple candidate target entities of each side object yet to be built in new group of entities.
In the present embodiment, is carried out in fused new group of entities to repetition entity, obtain each side object yet to be built
Multiple candidate target entities.When facing mass knowledge map, the acquisition of candidate target entity can substantially reduce subsequent processing
Data volume, by need to carry out subsequent processing entity limitation in a certain range, reduce most redundant computations, guarantee
The efficiency of processing mass data.
Specifically, the mode for obtaining candidate target entity can be with are as follows:
1, the acquisition for carrying out candidate target entity is mapped according to schema, i.e., inquiry is returned the result is limited for classification
Fixed, secondly classification is expanded again according to the obtained subclass of schema.For example, the object on side yet to be built belongs to " personage " classification, and
A certain entity belongs to " story " classification, then can not then recall the entity as candidate target entity.
2, have the target entity of identical semanteme as candidate target entity for the object on side yet to be built, even if entity name
It is completely inconsistent, it can also be returned.
3, using with the target entity of object alias having the same in side yet to be built as candidate target entity, such as " Zhou Jielun ",
Alias can be " Zhou Dong " etc..
It should be noted that the mode of above-mentioned acquisition candidate target entity is by way of example only, it is not limited to this hair
Bright technical solution, if using those skilled in the art it is conceivable that other way, and completely can with.
S2062 obtains the similarity of each side object yet to be built and each candidate target entity, and maximum similarity is corresponding
Candidate target entity is as candidate association entity.
After obtaining each side object yet to be built and multiple candidate target entities, it is corresponding that each candidate target entity can be obtained respectively
Default feature.Preferably, each candidate target entity can be directed to, it is corresponding to feature that the candidate target entity is obtained respectively
(pair features) and group feature (group features).
Wherein, it is to measure the feature of the similitude between side object and candidate target entity yet to be built to feature, such as may include
Title similarity (whether consistent, editing distance etc.), based on schema phase character (type of the type constraint of P and O whether one
Theme phase of the type of cause, the type of P and O with the presence or absence of the semantic text description with O of Chinese of schema father's subclass relation, P
Like degree, P Chinese it is semantic whether appear in the description of O, whether the text of the reciprocity of PP relation on attributes, O include P relevant
Keyword etc.), and, (whether S occurs relationship in the description text of O, whether S occurs in the SPO set of O between S and O
Deng) etc..
Group feature is to measure the feature of the similitude between the triple and candidate target entity of the object including side yet to be built.
Group feature can include: plsa similarity (the plsa similarity of SPO group, the SPO group of S and the O of the SPO group and O of S between text
Text description plsa similarity etc.), and, the various cooccurrence relations of S and O entity be (the text description of the SPO group and S of O
The co-occurrence number etc. that the text of the co-occurrence number of the SPO group of co-occurrence number, the SPO group of S and O, the SPO group of S and O describes) etc..
For each candidate target entity, it is assumed that 10 have been got respectively to feature and 5 group features, then can benefit
A feature vector is formed with this 15 features.It can be according to the feature vector of each candidate target entity come to each candidate target entity
It is ranked up.Preferably, can be according to the feature vector of each candidate target entity, using order models come to each candidate target entity
It is ranked up.Entity is built for side, sequence only needs to be concerned about that specifically ranksvm mould can be used in the entity of sequence first
Type is ranked up each candidate target entity.It, can will be wait locate after selecting sequence after the primary candidate target entity
O in the SPO of reason is associated with the candidate association selected physically, i.e., the O in SPO to be processed is mapped to the candidate selected
In associated entity.
S2063 determines whether candidate association entity exists with the entity on side yet to be built and conflicts;If it is not, S2064 is executed, if so,
Execute S2065.
In the present invention program also, it has been proposed that can further progress conflict resolution, sieve while fall have obviously conflict when building as a result, i.e.
It determines whether the candidate association entity selected exists with SPO to be processed to conflict, if it is not, being then associated with the O in SPO to be processed
Physically to the candidate association selected;If so, without association.
Specifically, using preset rule do not determine the candidate association entity do not selected whether with SPO to be processed
There are conflicts.For example, in rule can include: when the P in SPO is " wife ", the corresponding entity of O is necessary for female character.It is false
If the corresponding SPO of the object on side yet to be built is " wife Zhou Jielun elder brother icepro ", and candidate association entity is male character, then then can be true
There is conflict in the candidate association entity selected surely SPO corresponding with the object on side yet to be built, to the O in the SPO is not associated with
The candidate selected is physically.
Candidate association entity is determined as closing by S2064 if the corresponding entity of side object yet to be built is present in knowledge mapping
Join entity.
In practical applications, it is possible that following situations: based on introduction before it is found that candidate association entity be from
The target entity obtained in new group of entities, that is to say, that candidate association entity is oneself existing entity in new group of entities,
It is, however, possible to which the corresponding entity of O in SPO to be processed is not present in new group of entities, i.e., in new group of entities simultaneously
There is no correct entities, then O will be associated with to mistake physically after being handled in the manner described above.
It for the appearance for avoiding above situation, is proposed in the present invention program: in primary candidate mesh after selecting sequence
After marking entity, it can determine that the corresponding entity of O in the SPO on side yet to be built whether there is in new group of entities by decision model
In, if so, the O in the SPO on side yet to be built is associated in the associated entity selected, otherwise, without association.
S2065, without association.
In the present embodiment, without the operation such as being trained in the present invention program, so that treatment effeciency is improved, side of the present invention
Entity in case is built in side, and multidimensional technology can be used and carry out pulling for candidate target entity, and when facing mass knowledge library, it waits
It selects target entity to pull the data volume that can substantially reduce subsequent processing, the entity for carrying out subsequent processing will be needed to be limited in centainly
In range, reduce most redundant computations, ensure that the efficiency of processing mass data, meanwhile, multidimensional technology is in each dimension
It is effectively guaranteed the covering of candidate entity on degree, has ensured as far as possible that correct entity can be called back, has entered next
In the processing in stage;Also, by sequence, most possible entity output can be filtered out, the meter of next stage is greatly reduced
Calculation amount simultaneously provides distribution characteristics etc. for subsequent calculating, further, can also be handled by decision and conflict resolution, as far as possible
Ground avoids that side object yet to be built is associated with to mistake physically, to improve the accuracy etc. for building side result.
Fig. 4 is the structural schematic diagram one of entity handles device provided by the invention, as shown in figure 4, the entity handles device
400 include: grouping module 401, Fusion Module 402 and relating module 403.
Grouping module 401 obtains multiple group of entities, Mei Geshi for being grouped to the target entity in knowledge mapping
It include multiple target entities pair in body group, a target entity in group of entities is to being by a target entity in group of entities and its
What his target entity formed.
Fusion Module 402, two target entities of each target entity centering for obtaining each group of entities it is similar
Degree, and two target entities that similarity is greater than similarity threshold are merged, multiple new group of entities are obtained, it is each new
The target entity that similarity is greater than similarity threshold is not included in group of entities.
Relating module 403, for obtaining the associated entity of side object yet to be built in multiple new group of entities, and by side yet to be built
Object is associated with associated entity;Object in Subject, Predicate and Object triple yet to be built when object is yet to be built, side object yet to be built
Associated entity are as follows: in new group of entities, the target entity of similarity threshold is greater than with the similarity of side object yet to be built.
Entity handles device provided in this embodiment is similar with principle and technical effect that above-mentioned entity handles method is realized,
Therefore not to repeat here.
Optionally, Fig. 5 is the structural schematic diagram two of entity handles device provided by the invention, as shown in figure 5, at the entity
Manage device 400 further include: target entity obtains module 404.
Target entity obtains module 404, for the semanteme according to each entity in knowledge mapping, obtains target entity.
Optionally, Fusion Module 402 are specifically used for using preset attribute comparison method and Attribute Significance, obtain every
Attributes similarity between two target entities of a target entity centering;Using preset iterative model, to attributes similarity into
Row iteration calculates, and obtains the similarity of two target entities of each target entity centering.
Optionally, Fusion Module 402, specifically for determining that each target is real according to the comparison information in Attribute Significance
Multiple attributes pair to be compared between two target entities of body centering;
According to the feature of each attribute pair to be compared, chosen from preset attribute comparison method corresponding comparative approach into
Row attribute obtains the attributes similarity between two target entities of each target entity centering to comparing.
Optionally, relating module 403, if specifically for there are target entity being side object yet to be built in new group of entities
When default associated entity, using target entity as associated entity;And/or
The title of target entity is identical as side name of an object yet to be built if it exists or there are target entity and build Bian Shiti
When identical as the type of side object yet to be built with cooccurrence relation or there are the type of target entity, using target entity as closing
Join entity;And/or
In new group of entities, multiple candidate target entities of each side object yet to be built are obtained;
The similarity for obtaining each side object yet to be built and each candidate target entity, by the corresponding candidate mesh of maximum similarity
Entity is marked as candidate association entity;
If the corresponding entity of side object yet to be built is present in knowledge mapping, it is real that candidate association entity is determined as association
Body.
Optionally, relating module 403, it is corresponding to feature and group feature to be specifically used for obtaining each candidate target entity,
It is to measure the feature of the similitude between side object and candidate target entity yet to be built to feature, group feature is to measure including side yet to be built
Object triple and candidate target entity between similitude feature.
Optionally, relating module 403 are rushed specifically for determining whether candidate association entity exists with the entity on side yet to be built
It is prominent;If it is not, judging that the corresponding entity of side object yet to be built whether there is in knowledge mapping.
Fig. 6 is the structural schematic diagram three of entity handles device provided by the invention, which for example can be
Terminal device, such as smart phone, tablet computer, computer etc..As shown in fig. 6, the entity handles device 600 includes: storage
Device 601 and at least one processor 602.
Memory 601, for storing program instruction.
Processor 602, for being performed the entity handles method realized in the present embodiment, specific implementation in program instruction
Principle can be found in above-described embodiment, and details are not described herein again for the present embodiment.
The entity handles device 600 can also include and input/output interface 603.
Input/output interface 603 may include independent output interface and input interface, or integrated input and defeated
Integrated interface out.Wherein, output interface is used for output data, and input interface is used to obtain the data of input, above-mentioned output
Data are the general designation exported in above method embodiment, and the data of input are the general designation inputted in above method embodiment.
The present invention also provides a kind of readable storage medium storing program for executing, it is stored with and executes instruction in readable storage medium storing program for executing, work as entity handles
When at least one processor of device executes this and executes instruction, when computer executed instructions are executed by processor, realize above-mentioned
Entity handles method in embodiment.
The present invention also provides a kind of program product, the program product include execute instruction, this execute instruction be stored in it is readable
In storage medium.At least one processor of entity handles device can read this from readable storage medium storing program for executing and execute instruction, at least
One processor executes this and executes instruction so that entity handles device implements the entity handles that above-mentioned various embodiments provide
Method.
In several embodiments provided by the present invention, it should be understood that disclosed device and method can pass through it
Its mode is realized.For example, the apparatus embodiments described above are merely exemplary, for example, the division of the unit, only
Only a kind of logical function partition, there may be another division manner in actual implementation, such as multiple units or components can be tied
Another system is closed or is desirably integrated into, or some features can be ignored or not executed.Another point, it is shown or discussed
Mutual coupling, direct-coupling or communication connection can be through some interfaces, the INDIRECT COUPLING or logical of device or unit
Letter connection can be electrical property, mechanical or other forms.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple
In network unit.It can select some or all of unit therein according to the actual needs to realize the mesh of this embodiment scheme
's.
It, can also be in addition, the functional units in various embodiments of the present invention may be integrated into one processing unit
It is that each unit physically exists alone, can also be integrated in one unit with two or more units.Above-mentioned integrated list
Member both can take the form of hardware realization, can also realize in the form of hardware adds SFU software functional unit.
The above-mentioned integrated unit being realized in the form of SFU software functional unit can store and computer-readable deposit at one
In storage media.Above-mentioned SFU software functional unit is stored in a storage medium, including some instructions are used so that a computer
Equipment (can be personal computer, server or the network equipment etc.) or processor (English: processor) execute this hair
The part steps of bright each embodiment the method.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory
(English: Read-Only Memory, abbreviation: ROM), random access memory (English: Random Access Memory, letter
Claim: RAM), the various media that can store program code such as magnetic or disk.
In the embodiment of the above-mentioned network equipment or terminal device, it should be appreciated that processor can be central processing unit
(English: Central Processing Unit, referred to as: CPU), it can also be other general processors, digital signal processor
(English: Digital Signal Processor, abbreviation: DSP), specific integrated circuit (English: Application
Specific Integrated Circuit, referred to as: ASIC) etc..General processor can be microprocessor or the processor
It is also possible to any conventional processor etc..Hardware handles can be embodied directly in conjunction with the step of method disclosed in the present application
Device executes completion, or in processor hardware and software module combination execute completion.
Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention., rather than its limitations;To the greatest extent
Pipe present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: its according to
So be possible to modify the technical solutions described in the foregoing embodiments, or to some or all of the technical features into
Row equivalent replacement;And these are modified or replaceed, various embodiments of the present invention technology that it does not separate the essence of the corresponding technical solution
The range of scheme.
Claims (10)
1. a kind of entity handles method characterized by comprising
Target entity in knowledge mapping is grouped, multiple group of entities are obtained, includes multiple mesh in each group of entities
Entity pair is marked, a target entity in the group of entities is to being by a target entity and other mesh in the group of entities
Mark entity composition;
Obtain the similarity of two target entities of each of each described group of entities target entity centering, and by similarity
Two target entities greater than similarity threshold are merged, the multiple new group of entities of acquisition, in each new group of entities
It is greater than the target entity of similarity threshold not comprising similarity;
Obtain the associated entity of side object yet to be built in multiple new group of entities, and by the side object yet to be built and the pass
Connection entity is associated;Object in the Subject, Predicate and Object triple yet to be built when object is described yet to be built, the side visitor yet to be built
The associated entity of body are as follows: in the new group of entities, be greater than the similarity threshold with the similarity of the side object yet to be built
Target entity.
2. the method according to claim 1, wherein the target entity in knowledge mapping is grouped,
Before obtaining multiple group of entities, comprising:
According to the semanteme of each entity in the knowledge mapping, the target entity is obtained.
3. method according to claim 1 or 2, which is characterized in that described to obtain described in each of each described group of entities
The similarity of two target entities of target entity centering, comprising:
Using preset attribute comparison method and Attribute Significance, obtain each two target entities of the target entity centering it
Between attributes similarity;
Using preset iterative model, calculating is iterated to the attributes similarity, obtains each target entity centering
Two target entities similarity.
4. according to the method described in claim 3, it is characterized in that, using preset attribute comparison method and Attribute Significance,
Obtain the attributes similarity between each two target entities of the target entity centering, comprising:
According to the comparison information in the Attribute Significance, determine between each two target entities of the target entity centering
Multiple attributes pair to be compared;
According to the feature of each attribute pair to be compared, corresponding comparison side is chosen from the preset attribute comparison method
Method carries out attribute to comparing, and obtains the attributes similarity between each two target entities of the target entity centering.
5. according to the method described in claim 4, it is characterized in that, described obtain side yet to be built in multiple new group of entities
The associated entity of object, comprising:
If in the new group of entities there are target entity be the side object yet to be built default associated entity when, by the mesh
Entity is marked as the associated entity;And/or
If or presence identical as the side name of an object yet to be built that there are the titles of target entity in the new group of entities
Target entity and it is described build while entity have cooccurrence relation or there are the type of target entity and it is described yet to be built while object class
When type is identical, using the target entity as the associated entity;And/or
In the new group of entities, multiple candidate target entities of each side object yet to be built are obtained;
The similarity for obtaining each the side object yet to be built and each candidate target entity, by the corresponding candidate mesh of maximum similarity
Entity is marked as candidate association entity;
If the corresponding entity of the side object yet to be built is present in the knowledge mapping, the candidate association entity is determined as
The associated entity.
6. according to the method described in claim 5, it is characterized in that, described obtain each side object yet to be built and each candidate
The similarity of target entity, comprising:
It is corresponding to feature and group feature to obtain each candidate target entity, described is to measure the side visitor yet to be built to feature
The feature of similitude between body and the candidate target entity, described group of feature are to measure the object including the side yet to be built
The feature of similitude between triple and the candidate target entity.
7. if according to the method described in claim 6, it is characterized in that, the corresponding entity of the side object yet to be built is present in
When in the knowledge mapping, the candidate association entity is determined as before the associated entity, further includes:
It determines whether the candidate association entity exists with the entity on the side yet to be built to conflict;
If it is not, judging that the corresponding entity of the side object yet to be built whether there is in the knowledge mapping.
8. a kind of entity handles device characterized by comprising
Grouping module obtains multiple group of entities, each group of entities for being grouped to the target entity in knowledge mapping
In include multiple target entities pair, a target entity in the group of entities is to being by a target in the group of entities
What entity and other target entities formed;
Fusion Module, for obtain the target entity centering of each of each described group of entities two target entities it is similar
Degree, and two target entities that similarity is greater than similarity threshold are merged, multiple new group of entities are obtained, it is each described
The target entity that similarity is greater than similarity threshold is not included in new group of entities;
Relating module, for obtaining the associated entity of side object yet to be built in multiple new group of entities, and will be described yet to be built
Side object is associated with the associated entity;Visitor in the Subject, Predicate and Object triple yet to be built when object is described yet to be built
Body, the associated entity of the side object yet to be built are as follows: in the new group of entities, be greater than with the similarity of the side object yet to be built
The target entity of the similarity threshold.
9. a kind of entity handles device characterized by comprising at least one processor and memory;
The memory stores computer executed instructions;
At least one described processor executes the computer executed instructions of the memory storage, so that the entity handles device
Perform claim requires the described in any item methods of 1-7.
10. a kind of computer readable storage medium, which is characterized in that be stored with computer on the computer readable storage medium
It executes instruction, when the computer executed instructions are executed by processor, realizes the described in any item methods of claim 1-7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811290669.3A CN109684625B (en) | 2018-10-31 | 2018-10-31 | Entity processing method, device and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811290669.3A CN109684625B (en) | 2018-10-31 | 2018-10-31 | Entity processing method, device and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109684625A true CN109684625A (en) | 2019-04-26 |
CN109684625B CN109684625B (en) | 2021-01-12 |
Family
ID=66185666
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811290669.3A Active CN109684625B (en) | 2018-10-31 | 2018-10-31 | Entity processing method, device and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109684625B (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111259659A (en) * | 2020-01-14 | 2020-06-09 | 北京百度网讯科技有限公司 | Information processing method and device |
CN111459999A (en) * | 2020-03-27 | 2020-07-28 | 北京百度网讯科技有限公司 | Identity information processing method and device, electronic equipment and storage medium |
CN111522968A (en) * | 2020-06-22 | 2020-08-11 | 中国银行股份有限公司 | Knowledge graph fusion method and device |
CN111563133A (en) * | 2020-05-06 | 2020-08-21 | 支付宝(杭州)信息技术有限公司 | Method and system for data fusion based on entity relationship |
CN111597788A (en) * | 2020-05-18 | 2020-08-28 | 腾讯科技(深圳)有限公司 | Attribute fusion method, device and equipment based on entity alignment and storage medium |
CN111680498A (en) * | 2020-05-18 | 2020-09-18 | 国家基础地理信息中心 | Entity disambiguation method, device, storage medium and computer equipment |
CN112579770A (en) * | 2019-09-30 | 2021-03-30 | 北京国双科技有限公司 | Knowledge graph generation method, device, storage medium and equipment |
CN112699667A (en) * | 2020-12-29 | 2021-04-23 | 京东数字科技控股股份有限公司 | Entity similarity determination method, device, equipment and storage medium |
CN113392220A (en) * | 2020-10-23 | 2021-09-14 | 腾讯科技(深圳)有限公司 | Knowledge graph generation method and device, computer equipment and storage medium |
CN115130435A (en) * | 2022-06-27 | 2022-09-30 | 北京百度网讯科技有限公司 | Document processing method and device, electronic equipment and storage medium |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7970592B2 (en) * | 2004-09-01 | 2011-06-28 | Behrens Clifford A | System and method for consensus-based knowledge validation, analysis and collaboration |
CN104035917A (en) * | 2014-06-10 | 2014-09-10 | 复旦大学 | Knowledge graph management method and system based on semantic space mapping |
CN106777331A (en) * | 2017-01-11 | 2017-05-31 | 北京航空航天大学 | Knowledge mapping generation method and device |
CN108038183A (en) * | 2017-12-08 | 2018-05-15 | 北京百度网讯科技有限公司 | Architectural entities recording method, device, server and storage medium |
CN108154198A (en) * | 2018-01-25 | 2018-06-12 | 北京百度网讯科技有限公司 | Knowledge base entity normalizing method, system, terminal and computer readable storage medium |
CN108304381A (en) * | 2018-01-25 | 2018-07-20 | 北京百度网讯科技有限公司 | Entity based on artificial intelligence builds side method, apparatus, equipment and storage medium |
CN108647318A (en) * | 2018-05-10 | 2018-10-12 | 北京航空航天大学 | A kind of knowledge fusion method based on multi-source data |
-
2018
- 2018-10-31 CN CN201811290669.3A patent/CN109684625B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US7970592B2 (en) * | 2004-09-01 | 2011-06-28 | Behrens Clifford A | System and method for consensus-based knowledge validation, analysis and collaboration |
CN104035917A (en) * | 2014-06-10 | 2014-09-10 | 复旦大学 | Knowledge graph management method and system based on semantic space mapping |
CN106777331A (en) * | 2017-01-11 | 2017-05-31 | 北京航空航天大学 | Knowledge mapping generation method and device |
CN108038183A (en) * | 2017-12-08 | 2018-05-15 | 北京百度网讯科技有限公司 | Architectural entities recording method, device, server and storage medium |
CN108154198A (en) * | 2018-01-25 | 2018-06-12 | 北京百度网讯科技有限公司 | Knowledge base entity normalizing method, system, terminal and computer readable storage medium |
CN108304381A (en) * | 2018-01-25 | 2018-07-20 | 北京百度网讯科技有限公司 | Entity based on artificial intelligence builds side method, apparatus, equipment and storage medium |
CN108647318A (en) * | 2018-05-10 | 2018-10-12 | 北京航空航天大学 | A kind of knowledge fusion method based on multi-source data |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112579770A (en) * | 2019-09-30 | 2021-03-30 | 北京国双科技有限公司 | Knowledge graph generation method, device, storage medium and equipment |
CN111259659A (en) * | 2020-01-14 | 2020-06-09 | 北京百度网讯科技有限公司 | Information processing method and device |
CN111259659B (en) * | 2020-01-14 | 2023-07-04 | 北京百度网讯科技有限公司 | Information processing method and device |
CN111459999B (en) * | 2020-03-27 | 2023-08-18 | 北京百度网讯科技有限公司 | Identity information processing method, device, electronic equipment and storage medium |
CN111459999A (en) * | 2020-03-27 | 2020-07-28 | 北京百度网讯科技有限公司 | Identity information processing method and device, electronic equipment and storage medium |
CN111563133A (en) * | 2020-05-06 | 2020-08-21 | 支付宝(杭州)信息技术有限公司 | Method and system for data fusion based on entity relationship |
CN111597788A (en) * | 2020-05-18 | 2020-08-28 | 腾讯科技(深圳)有限公司 | Attribute fusion method, device and equipment based on entity alignment and storage medium |
CN111680498A (en) * | 2020-05-18 | 2020-09-18 | 国家基础地理信息中心 | Entity disambiguation method, device, storage medium and computer equipment |
CN111597788B (en) * | 2020-05-18 | 2023-11-14 | 腾讯科技(深圳)有限公司 | Attribute fusion method, device, equipment and storage medium based on entity alignment |
CN111680498B (en) * | 2020-05-18 | 2023-04-07 | 国家基础地理信息中心 | Entity disambiguation method, device, storage medium and computer equipment |
CN111522968A (en) * | 2020-06-22 | 2020-08-11 | 中国银行股份有限公司 | Knowledge graph fusion method and device |
CN111522968B (en) * | 2020-06-22 | 2023-09-08 | 中国银行股份有限公司 | Knowledge graph fusion method and device |
CN113392220A (en) * | 2020-10-23 | 2021-09-14 | 腾讯科技(深圳)有限公司 | Knowledge graph generation method and device, computer equipment and storage medium |
CN113392220B (en) * | 2020-10-23 | 2024-03-26 | 腾讯科技(深圳)有限公司 | Knowledge graph generation method and device, computer equipment and storage medium |
CN112699667A (en) * | 2020-12-29 | 2021-04-23 | 京东数字科技控股股份有限公司 | Entity similarity determination method, device, equipment and storage medium |
CN115130435B (en) * | 2022-06-27 | 2023-08-11 | 北京百度网讯科技有限公司 | Document processing method, device, electronic equipment and storage medium |
CN115130435A (en) * | 2022-06-27 | 2022-09-30 | 北京百度网讯科技有限公司 | Document processing method and device, electronic equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN109684625B (en) | 2021-01-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109684625A (en) | Entity handles method, apparatus and storage medium | |
JP6975377B2 (en) | Computer implementation methods, devices, electronic devices, computer-readable storage media and computer programs for determining the degree of text correlation. | |
CN110609902B (en) | Text processing method and device based on fusion knowledge graph | |
CN110377903B (en) | Sentence-level entity and relation combined extraction method | |
CN109885660A (en) | A kind of question answering system and method based on information retrieval that knowledge mapping is energized | |
CN110019732B (en) | Intelligent question answering method and related device | |
CN110427623A (en) | Semi-structured document Knowledge Extraction Method, device, electronic equipment and storage medium | |
CN109284397A (en) | A kind of construction method of domain lexicon, device, equipment and storage medium | |
CN111444320A (en) | Text retrieval method and device, computer equipment and storage medium | |
CN111858940B (en) | Multi-head attention-based legal case similarity calculation method and system | |
US20190340503A1 (en) | Search system for providing free-text problem-solution searching | |
CN110929498B (en) | Method and device for calculating similarity of short text and readable storage medium | |
CN111966793B (en) | Intelligent question-answering method and system based on knowledge graph and knowledge graph updating system | |
CN113987155B (en) | Conversational retrieval method integrating knowledge graph and large-scale user log | |
CN115062134B (en) | Knowledge question-answering model training and knowledge question-answering method, device and computer equipment | |
CN115438674A (en) | Entity data processing method, entity linking method, entity data processing device, entity linking device and computer equipment | |
CN115600593A (en) | Method and device for acquiring key content of literature | |
CN113297355A (en) | Method, device, equipment and medium for enhancing labeled data based on countermeasure interpolation sequence | |
CN117194616A (en) | Knowledge query method and device for vertical domain knowledge graph, computer equipment and storage medium | |
CN117435685A (en) | Document retrieval method, document retrieval device, computer equipment, storage medium and product | |
CN113807102B (en) | Method, device, equipment and computer storage medium for establishing semantic representation model | |
CN112416754B (en) | Model evaluation method, terminal, system and storage medium | |
CN115344698A (en) | Label processing method, label processing device, computer equipment, storage medium and program product | |
CN113705692A (en) | Emotion classification method and device based on artificial intelligence, electronic equipment and medium | |
WO2024099037A1 (en) | Data processing method and apparatus, entity linking method and apparatus, and computer device |
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 |