CN108536702A - A kind of related entities determine method, apparatus and computing device - Google Patents
A kind of related entities determine method, apparatus and computing device Download PDFInfo
- Publication number
- CN108536702A CN108536702A CN201710120836.9A CN201710120836A CN108536702A CN 108536702 A CN108536702 A CN 108536702A CN 201710120836 A CN201710120836 A CN 201710120836A CN 108536702 A CN108536702 A CN 108536702A
- Authority
- CN
- China
- Prior art keywords
- related entities
- entity
- edges
- target entity
- candidate
- 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
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
A kind of related entities of offer of the embodiment of the present invention determine that method, apparatus and computing device, this method include:Object knowledge collection of illustrative plates is obtained, the object knowledge collection of illustrative plates at least has target entity;It determines in the object knowledge collection of illustrative plates, the candidate entity sets of the target entity;It is described candidate entity sets include:The corresponding candidate entity of each number of edges up to the target entity can be touched;According to the candidate entity sets, the related entities of the target entity are determined.The embodiment of the present invention can promote the recall rate of related entities definitive result.
Description
Technical field
The present invention relates to technical field of data processing, and in particular to a kind of related entities determine that method, apparatus and calculating are set
It is standby.
Background technology
Related entities may be considered other entities with the target entity co-occurrence inquired in the same query, for
The relevant information that family obtains the target entity inquired is of great significance;For example user, after input inquiry sentence, search is drawn
It holds up and is presented to use open air in addition to target entity (such as web page interlinkage) corresponding with the query statement will be searched, can will also look into
Related entities during inquiry with the target entity co-occurrence recommend user, to guide user to search again for, promote user
Obtain the convenience of relevant information;A kind of typical scene is, search engine is to search target corresponding with query statement real
After body, except searched target entity is shown in result of page searching, can also result of page searching setting regions (such as
Left area) the recommended related entities of display, so that user searches again for.
It was found by the inventors of the present invention that being presently mainly to be counted and a target by open text (such as newsletter archive)
Other entities of entity co-occurrence, to determine the related entities of a target entity;However, the content of open Characters has centainly
Limitation and timeliness, this makes the related entities definitive result by open text statistics uncontrollable, leads to related entities
The recall rate of definitive result is relatively low, and (recall rate indicates the ratio of determining related entities quantity and related entities total quantity, is true
Determine a kind of comprehensive embodiment of result).
Invention content
In view of this, a kind of related entities of offer of the embodiment of the present invention determine method, apparatus and computing device, to promote phase
Close the recall rate of entity definitive result.
To achieve the above object, the embodiment of the present invention provides the following technical solutions:
A kind of related entities determine method, including:
Object knowledge collection of illustrative plates is obtained, the object knowledge collection of illustrative plates at least has target entity;
It determines in the object knowledge collection of illustrative plates, the candidate entity sets of the target entity;Candidate's entity sets packet
It includes:The corresponding candidate entity of each number of edges up to the target entity can be touched;
According to the candidate entity sets, the related entities of the target entity are determined.
The embodiment of the present invention also provides a kind of related entities determining device, including:
Object knowledge collection of illustrative plates acquisition module, for obtaining object knowledge collection of illustrative plates, the object knowledge collection of illustrative plates at least has mesh
Mark entity;
Candidate entity sets determining module, for determining in the object knowledge collection of illustrative plates, the candidate of the target entity is real
Body set;It is described candidate entity sets include:The corresponding candidate entity of each number of edges up to the target entity can be touched;
Related entities determining module, for according to the candidate entity sets, determining the related entities of the target entity.
The embodiment of the present invention also provides a kind of computing device, including related entities determining device described above.
Based on the above-mentioned technical proposal, related entities provided in an embodiment of the present invention determine that method includes:Obtain object knowledge
Collection of illustrative plates, the object knowledge collection of illustrative plates at least have target entity;It determines in the object knowledge collection of illustrative plates, the time of the target entity
Select entity sets;It is described candidate entity sets include:The corresponding candidate entity of each number of edges up to the target entity can be touched;According to
Candidate's entity sets, determine the related entities of the target entity.At least have as can be seen that the embodiment of the present invention uses
The object knowledge collection of illustrative plates of target entity excavates the candidate entity sets that can be touched in object knowledge collection of illustrative plates up to the target entity, into
And according to the candidate entity sets, determine the related entities of the target entity, the target included due to object knowledge collection of illustrative plates
The relevant information of entity more fully, therefore can be believed with the excavation of maximum probability to the previous comprehensive correlation of target entity history
Breath so that the related entities result for the target entity excavated is more comprehensive, and the correlation for promoting identified target entity is real
The recall rate of body result.
Description of the drawings
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
The embodiment of invention for those of ordinary skill in the art without creative efforts, can also basis
The attached drawing of offer obtains other attached drawings.
Fig. 1 is the flow chart that related entities provided in an embodiment of the present invention determine method;
Fig. 2 is the method flow diagram provided in an embodiment of the present invention for obtaining object knowledge collection of illustrative plates;
Fig. 3 is another flow chart that related entities provided in an embodiment of the present invention determine method;
Fig. 4 is the schematic diagram of relationship between entity in object knowledge collection of illustrative plates;
Fig. 5 is to determine the method flow diagram of the related entities of target entity according to candidate entity sets;
Fig. 6 is another flow chart that related entities provided in an embodiment of the present invention determine method;
Fig. 7 is the method flow diagram of the recommendation sequence of determining related entities provided in an embodiment of the present invention;
Fig. 8 is the another method flow chart of the recommendation sequence of determining related entities provided in an embodiment of the present invention;
Fig. 9 is the another method flow chart of the recommendation sequence of determining related entities provided in an embodiment of the present invention;
Figure 10 is the structure diagram of related entities determining device provided in an embodiment of the present invention;
Figure 11 is another structure diagram of related entities determining device provided in an embodiment of the present invention;
Figure 12 is the hardware block diagram of computing device provided in an embodiment of the present invention.
Specific implementation mode
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation describes, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Fig. 1 is the flow chart that related entities provided in an embodiment of the present invention determine method, and this method can be applied to have number
According to the computing device of operational capability, the corresponding program of method shown in Fig. 1 executed by the computing device, it can be achieved that related entities
It determines;The computing device can select the server of network side, can also select the electronic equipments such as the computer of user side;
Referring to Fig.1, related entities provided in an embodiment of the present invention determine that method may include:
Step S100, object knowledge collection of illustrative plates is obtained, the object knowledge collection of illustrative plates at least has target entity.
Target entity be the embodiment of the present invention it is to be determined go out related entities entity, the embodiment of the present invention may specify that needs are true
The target entity of related entities is made, and there is the target entity in object knowledge collection of illustrative plates.
Knowledge mapping is intended to various entities or concept present in description real world;Each entity or concept can use one
The ID (identity number) of globally unique determination is identified, and each attribute-value can be used to (attribute-value pair)
The intrinsic characteristic of entity is portrayed, and relationship (relation) is used for connecting two entities, portrays the association between them;Therefore,
Knowledge mapping is mainly made of the side between node and connecting node, wherein a node can indicate an entity or general
Read, the Bian Zeke of connecting node by between institute's connecting node attribute or relationship constitute;
In the embodiment of the present invention, the data source of knowledge mapping can come from encyclopaedia class website and various vertical websites by collection
Structural data, realize that the universal quality of these data is higher with the most of common sense knowledge of covering, but update slow;And
On the other hand, the data source of knowledge mapping also can be related real by being extracted from various semi-structured data (shaped like html table)
The attribute-value of body is realized, the description of entity is enriched with this;In addition, by search for daily record (query log) find new entity or
New entity attribute also can constantly extend the coverage rate of knowledge mapping;
In a kind of possible realization, the embodiment of the present invention can construct target by the data source comprising target entity and know
Know collection of illustrative plates.
To promote the comprehensive of follow-up related entities definitive result, the embodiment of the present invention also can be by including target entity
Knowledge mapping constructed by data source understands the meaning of the input text comprising target entity so that the related letter of target entity
The understanding of breath is more fully;In realization, the embodiment of the present invention can obtain the input text comprising target entity, by comprising
After the data source of target entity constructs knowledge mapping, the name entity given in text will be inputted, is mapped to and constructed knows
On the target entity for knowing collection of illustrative plates, object knowledge collection of illustrative plates is obtained.
Step S110, it determines in the object knowledge collection of illustrative plates, the candidate entity sets of the target entity;It is described candidate real
Body set includes:The corresponding candidate entity of each number of edges up to the target entity can be touched.
In object knowledge collection of illustrative plates, entity may be considered a node, can be connected by side between entity;Target entity can
It can be touched by a line up to a candidate entity, it is also possible to be touched up to a candidate entity by multiple summits, the embodiment of the present invention can be from institute
It states target entity to set out, determines that target entity touches the entity reached by a line, the corresponding candidate entity of number of edges one is obtained, from mesh
Mark entity sets out, and determines that target entity touches the entity reached by two sides, the corresponding candidate entity of number of edges two is obtained, with such
It pushes away, obtains the corresponding candidate entity of each number of edges.
Optionally, in a kind of realization, the embodiment of the present invention can set number of edges range, which may include multiple
Number of edges, then for each number of edges in number of edges range, the embodiment of the present invention can determine from the target entity with corresponding number of edges
The candidate entity reached is touched, obtains that candidate entity corresponding up to each number of edges of the target entity can be touched;
For example, setting number of edges range includes number of edges one to number of edges three, then for number of edges one, the embodiment of the present invention can determine with
A line touches the candidate entity up to target entity, the corresponding candidate entity of number of edges one is obtained, for number of edges two, the embodiment of the present invention
It can determine the candidate entity for being touched with two sides and reaching target entity, the corresponding candidate entity of number of edges two obtained, for number of edges three, this hair
Bright embodiment can determine the candidate entity for being touched with three sides and reaching target entity, obtain the corresponding candidate entity of number of edges three, to
To the corresponding candidate entity of each number of edges in the number of edges range.
It should be noted that setting number of edges range is only to determine the corresponding candidate entity of each number of edges that can be touched up to target entity
Optional mode, the embodiment of the present invention also can determine in object knowledge collection of illustrative plates, other entities are touched up to the institute involved by target entity
There is number of edges, to determine the corresponding candidate entity of each number of edges that can be touched up to target entity with this.
Step S120, according to the candidate entity sets, the related entities of the target entity are determined.
Optionally, on the one hand, the embodiment of the present invention can be by the identified candidate entity sets, as target reality
The related entities of body.
Optionally, on the other hand, in candidate's entity sets, it is understood that there may be the candidate of the repetition of corresponding different numbers edge is real
Body, the embodiment of the present invention can carry out duplicate removal processing to the candidate entity repeated in candidate entity sets, to retain the candidate repeated
The candidate entity of number of edges minimum in entity, thus by the candidate entity included by the candidate entity sets after duplicate removal processing, as
The related entities of the target entity.
Using the object knowledge collection of illustrative plates at least with target entity, excavating can touch in object knowledge collection of illustrative plates the embodiment of the present invention
Up to the candidate entity sets of the target entity, and then according to the candidate entity sets, determine the correlation of the target entity
Entity, the relevant information for the target entity included due to object knowledge collection of illustrative plates more fully, can be with the excavations of maximum probability
To target entity history comprehensive relevant information in the past so that the related entities result for the target entity excavated is more complete
Face promotes the recall rate of the related entities result of identified target entity.
It has been described above when obtaining object knowledge collection of illustrative plates, the data source comprising target entity can have been obtained, by including mesh
The data source for marking entity builds object knowledge collection of illustrative plates, the acquisition modes mesh that is relatively simple, and including of this object knowledge collection of illustrative plates
The relevant information of entity is marked more comprehensively, it can be achieved that the related entities result for the target entity finally excavated has higher call together
The rate of returning.
And in another implementation, the embodiment of the present invention can pass through the knowledge of data source (including target entity) structure
Collection of illustrative plates, understands the meaning of the input text comprising target entity, and then gets object knowledge collection of illustrative plates so that object knowledge collection of illustrative plates
In about target entity relevant information understanding more fully;Correspondingly, Fig. 2 shows acquisitions provided in an embodiment of the present invention
A kind of optional implementation method flow of object knowledge collection of illustrative plates, with reference to Fig. 2, this method may include:
Step S200, it obtains and inputs text, it is previously given in the input text to have multiple name entities, the name real
Body includes at least the target entity.
Optionally, input text can be one kind of open text, which, which at least records, target entity, simultaneously
Other entities may also have been recorded;The embodiment of the present invention can in inputting text the previously given life having including at least target entity
Name entity, name entity may be considered in input text given name, mechanism name, place name and other with entitled mark
The entity of knowledge.
Step S210, the name entity given in text will be inputted, is mapped on the target entity of knowledge mapping, obtains mesh
Mark knowledge mapping;The knowledge mapping is built by the data source comprising target entity.
Text is inputted obtaining, and after the determining knowledge mapping by the data source structure comprising target entity, the present invention is real
The name entity given in text can will be inputted by applying example, be mapped on the target entity of the knowledge mapping, obtained object knowledge figure
Spectrum;
The name entity given in text will be inputted, is mapped on the target entity of knowledge mapping, may recognize that it is one
The name entity given in text will be inputted, is linked to the process of the target entity in knowledge mapping unambiguously, this process can
With processing such as the disambiguations of merging, ambiguity entity including synonymous entity;
Optionally, in specific implementation, the embodiment of the present invention can use named entity linking (name entities
Link) technology will input the name entity given in text, it is mapped on the target entity of knowledge mapping, so that text will be inputted
In in given name entity link to knowledge mapping on target entity unambiguously;Name entity link technology can mainly be promoted
The information filtering ability of the systems such as online commending system, internet search engine.
Optionally, the embodiment of the present invention can be set used in the candidate entity for excavating target entity in object knowledge collection of illustrative plates
Number of edges range, after getting object knowledge collection of illustrative plates, by each number of edges corresponding to the number of edges range, to excavate object knowledge figure
The corresponding candidate entity of each number of edges up to the target entity can be touched in spectrum;Optionally, Fig. 3 shows that the embodiment of the present invention provides
Related entities determine another flow chart of method, with reference to Fig. 3, this method may include:
Step S300, object knowledge collection of illustrative plates is obtained, the object knowledge collection of illustrative plates at least has target entity.
Optionally, step S300 can be realized by method shown in Fig. 2, can also pass through the data source comprising target entity
Construct object knowledge collection of illustrative plates.
Step S310, presetting number of edges range is obtained, the number of edges range includes multiple number of edges.
Step S320, each number of edges included by the number of edges range determines in the object knowledge collection of illustrative plates, can touch and reach
The corresponding candidate entity of each number of edges of the target entity, obtains the candidate entity sets of the target entity.
Optionally, after setting number of edges range, for each number of edges in number of edges range, the embodiment of the present invention can determine from institute
Target entity is stated to set out with the tactile candidate entity reached of corresponding number of edges, so that it is determined that going out each number of edges pair that can be touched up to the target entity
The candidate entity answered obtains the candidate entity sets of the target entity.
For ease of understanding, as shown in figure 4, in object knowledge collection of illustrative plates, " Xiao Ming " is target entity, " small with target entity
It is bright " it sets out and is touched up to multiple candidate entities by the way that a variety of relationships are extensible, and there is certain relationship between the entity connected;
For setting number of edges range and include number of edges one to three, as shown in figure 4, from target entity " Xiao Ming's ", number of edges
One corresponding candidate entity includes:" small red ", " small strong " and " film A ";From target entity, " Xiao Ming's ", number of edges two are corresponding
Candidate entity includes:" small strong " and " little Rong ";From target entity, " Xiao Ming's ", the corresponding candidate entity of number of edges three include:It is " small
Hold ";Accordingly, it may be determined that go out the corresponding candidate entity of each number of edges that can be touched up to target entity " Xiao Ming ", obtain object knowledge collection of illustrative plates
In, the candidate entity sets up to the target entity can be touched;Candidate's entity sets can specifically include:
Number of edges one:" small red ", " small strong " and " film A ";
Number of edges two:" small strong " and " little Rong ";
Number of edges three:" little Rong ".
It should be noted that setting number of edges range can be not limited to the number of edges one to three of foregoing description, but can be according to reality
Number of edges included by the situation setting number of edges range of border.
Step S330, according to the candidate entity sets, the related entities of the target entity are determined.
Optionally, after obtaining the candidate entity sets, the embodiment of the present invention will directly can wrap in candidate entity sets
The candidate entity included realizes the determination of the related entities of target entity as the related entities of target entity.
Optionally, on the other hand, there may be the candidate entities of the repetition of corresponding different numbers edge in candidate entity sets, i.e.,
One candidate entity is likely to be present in different number of edges, and such as one candidate entity is likely to be present in the corresponding candidate entity of number of edges one
In, it is also possible to it is present in the corresponding candidate entity of number of edges two;In this case, a candidate entity can with the actual relationship of target
Can be a variety of;Based on this, to promote the relationship precision of the related entities and target entity excavated, the embodiment of the present invention can be to waiting
The candidate entity repeated in entity sets is selected to carry out duplicate removal processing, it is real to retain the minimum candidate of number of edges in the candidate entity repeated
Body;
Optionally, Fig. 5 shows the optional method stream for the related entities that target entity is determined according to candidate entity sets
Journey, with reference to Fig. 5, this method may include:
If in step S400, the described candidate entity sets, there are the candidate entities of the repetition of corresponding different numbers edge, will be described
The candidate entity repeated in candidate entity sets carries out duplicate removal processing, to retain the candidate of number of edges minimum in the candidate entity repeated
Entity.
After determining the candidate entity sets of target entity, the embodiment of the present invention can determine each number of edges of target entity
Corresponding candidate's entity, if wherein corresponding to different number of edges in the presence of the candidate entity of the candidate entity and repetition that repeat, according to
The principle for retaining the candidate entity of number of edges minimum carries out duplicate removal processing to the candidate entity repeated in candidate entity sets, to
Obtain the candidate entity sets after duplicate removal processing;
With one corresponding candidate's entity of number of edges candidate entity corresponding with number of edges two, for the candidate entity of repetition,
Then can be by the candidate entity corresponding with number of edges two of number of edges one, the candidate entity positioned at the repetition of number of edges two is removed so that
The candidate entity corresponding from number of edges two of number of edges one is different, realizes and carries out duplicate removal to the candidate entity repeated in candidate entity sets;
For shown in Fig. 4, candidate entity sets include:
Number of edges one:" small red ", " small strong " and " film A ";
Number of edges two:" small strong " and " little Rong ";
Number of edges three:" little Rong ";
As can be seen that the candidate entity corresponding with number of edges two of number of edges one includes the candidate entity of " small strong " this repetition,
Number of edges two and the corresponding candidate entity of number of edges three include the candidate entity of " little Rong " this repetition, then minimum according to number of edges is retained
Candidate entity principle, duplicate removal processing is carried out to the candidate entity repeated in candidate entity sets, can remove in number of edges two
Candidate entity " small strong " retains the candidate entity " small strong " in number of edges one, removes the candidate entity " little Rong " in number of edges three, retain
Candidate entity " little Rong " in number of edges two, realize in candidate entity sets repeat candidate entity duplicate removal processing, obtain as
Lower duplicate removal treated candidate entity sets:
Number of edges one:" small red ", " small strong " and " film A ";
Number of edges two:" little Rong ".
Step S410, by the candidate entity included by the candidate entity sets after duplicate removal processing, as the target entity
Related entities.
Candidate entity sets after duplicate removal processing include:Each number of edges that can be touched up to the target entity is corresponding unduplicated
Candidate entity.
Using the object knowledge collection of illustrative plates at least with target entity, excavating can touch in object knowledge collection of illustrative plates the embodiment of the present invention
Up to the candidate entity sets of the target entity, the relevant information for the target entity included due to object knowledge collection of illustrative plates is more complete
Face, therefore can be with the excavation of maximum probability to target entity history comprehensive relevant information in the past so that the target excavated
The candidate entity sets of entity are more comprehensive;And then pass through the repetition to corresponding to different numbers edge in the candidate entity sets again
Candidate entity carries out duplicate removal processing, obtains the related entities of target entity, can promote the related entities excavated and target entity
Relationship precision, finally obtain that recall rate is higher, and with the higher related entities definitive result of the relationship precision of target entity.
Preferably, Fig. 6 shows that related entities provided in an embodiment of the present invention determine another flow chart of method, with reference to figure
6, this method may include:
Step S500, it obtains and inputs text, it is previously given in the input text to have multiple name entities, the name real
Body includes at least the target entity.
Step S510, the name entity given in text will be inputted, is mapped on the target entity of knowledge mapping, obtains mesh
Mark knowledge mapping;The knowledge mapping is built by the data source comprising target entity.
Step S520, presetting number of edges range is obtained, the number of edges range includes multiple number of edges.
Step S530, each number of edges included by the number of edges range determines in the object knowledge collection of illustrative plates, can touch and reach
The corresponding candidate entity of each number of edges of the target entity, obtains the candidate entity sets of the target entity.
If in step S540, the described candidate entity sets, there are the candidate entities of the repetition of corresponding different numbers edge, will be described
The candidate entity repeated in candidate entity sets carries out duplicate removal processing, to retain the candidate of number of edges minimum in the candidate entity repeated
Entity.
Step S550, by the candidate entity included by the candidate entity sets after duplicate removal processing, as the target entity
Related entities.
In a possible implement scene, the related entities provided through the embodiment of the present invention determine method, it may be determined that
The related entities of film star " Xiao Ming ", i.e., with film star " Xiao Ming " for target entity, the embodiment of the present invention determines its correlation
The implementation process of entity can be as follows:
Server can from encyclopaedia class website, the structural data of various vertical websites and various semi-structured data and
Crawl includes the data source of target entity " Xiao Ming " in search daily record;
Server builds knowledge mapping by the data source comprising target entity " Xiao Ming ";It, can be by data when specific structure
As node, the relationship between entity, by the relationship between entity, each reality is connected with corresponding edge as side for each entity in source
Body;
Server obtains the input text for including target entity " Xiao Ming ", and being removed in the input text has target entity " small
It is bright " outside, other entities can also have been recorded;Specifically, previously given in the input text have multiple name entities, these lives
Target entity " Xiao Ming " is included at least in name entity;
Server will be inputted the name entity given in text, be mapped to by named entity linking technologies
On the target entity of knowledge mapping, object knowledge collection of illustrative plates is obtained;Specifically can be by named entity linking technologies, it will be defeated
Enter the name entity given in text, carries out the merging of synonymous entity with the entity in object knowledge collection of illustrative plates, ambiguity entity disappears
The processing such as discrimination;
Server transfers preset number of edges range, determines that the target corresponding to each number of edges in the number of edges range is real
The candidate entity of body obtains the candidate entity sets of the target entity;I.e. for each number of edges in the number of edges range, service
Device can determine in the object knowledge collection of illustrative plates, touches the candidate entity reached from the target entity with corresponding number of edges, obtains
Candidate's entity sets;As shown in figure 4, including number of edges one to number of edges three with number of edges range, then server can determine logical respectively
Cross a line, two while and three whiles touch candidate entity up to target entity, determine that each side up to the target entity can be touched
The corresponding candidate entity of number, obtains candidate entity sets;
If in candidate entity sets, there are the candidate entity of the repetition of corresponding different numbers edge, then server can will be candidate
The candidate entity repeated in entity sets carries out duplicate removal processing, hence for the candidate entity repeated in different numbers edge, only protects
Stay the candidate entity of number of edges minimum therein;In turn, server can be by the time included by the candidate entity sets after duplicate removal processing
Entity is selected, the related entities as target entity;
If, can be by the candidate included by candidate entity sets there is no the candidate entity repeated in candidate entity sets
Entity, the related entities as target entity.
After the related entities for determining target entity by above-described scheme, the embodiment of the present invention can recommend in search
Etc. need recommend related entities scene under, the related entities of target entity are recommended;Such as when user searches target reality
When body, the search entrance of the related entities of target entity can be recommended, to guide user to search again for, user is promoted and obtain mesh
Mark the convenience of the relevant information of entity;Correspondingly, the embodiment of the present invention can determine the recommendation sequence of each related entities, with root
The recommendation of related entities is carried out according to the recommendation sequence of each related entities, which will be described below.
A kind of relatively simple recommendation sortord is the recommendation sequence of random definition related entities, with what is defined at random
Sequence is recommended to carry out the recommendation of corresponding related entities;This mode is although relatively simple, but recommends the precision of sequence may be relatively low,
It in scene is recommended in some search and is not suitable for, is based on this, preferably, the embodiment of the present invention provides at least following three kinds
The recommendation sequence of related entities determines scheme.
One, the degree of correlation score of related entities and target entity is calculated on open text, it is true with degree of correlation score
Determine the recommendation sequence of related entities, and degree of correlation is higher, recommends sequence more forward;Optional realization process can be as shown in Figure 7;
Fig. 7 is the method flow of the recommendation sequence of determining related entities provided in an embodiment of the present invention, with reference to Fig. 7, the party
Method may include:
Step S600, the degree of correlation score of each related entities and target entity is counted in open text.
The degree of correlation score of statistical correlation entity and target entity is that the embodiment of the present invention exists offline in open text
A kind of application of co-occurrence semantic network is calculated on open text, it is considered that if (such as target entity is related to one real for 2 entities
Body) it frequently occurs in the same sentence, chapter, it is judged that this 2 entities are strong correlations.
The degree of correlation score of related entities and target entity, the mutual information that related entities and target entity can be used weigh
Amount, mutual information (Mutual Information) is a kind of useful measure information in information theory, it can regard as one with
The information content about another stochastic variable for including in machine variable, or perhaps a stochastic variable due to it is known another with
Machine variable and the uncertainty of reduction;
For the degree of correlation score of a related entities and target entity determines, the embodiment of the present invention can determine the phase
The mutual information for closing entity and target entity, the degree of correlation score of the related entities and target entity is determined with the mutual information;Tool
On body is realized, the embodiment of the present invention can determine while occurring the related entities and the amount of text and text total quantity of target entity
The first ratio, determine the second ratio of amount of text and text total quantity that the related entities occur, determine that target occur real
The amount of text of body and the third ratio of text total quantity, to according to first ratio, the second ratio and third ratio, determine
The mutual information of the related entities and target entity indicates related entities journey related to target entity with identified mutual information
Spend score;
On specific calculate, following formula may be used and realize:
Wherein, big X may be considered a set, and small x is interpreted as the specific data obtained in set, big Y and small y's
It defines similar;P (x, y) is indicated while the ratio of the quantity and text total quantity of the text of entity x and y is occurred, and p (x) is represented
The ratio of the quantity and text total quantity of the text of existing x, p (y) indicate the ratio for the quantity and text total quantity of the text of y occur
Value.
Step S610, according to the degree of correlation score of each related entities and target entity, the recommendation of each related entities is determined
Sequence, wherein related journey point is higher several times, recommends sequence more forward.
Two, according to the relationship weight in object knowledge collection of illustrative plates, determine that each related entities are corresponding in object knowledge collection of illustrative plates
Weight score is determined with the weight score of related entities and recommends sequence;
It was found by the inventors of the present invention that relationship between some related entities is due to very fixed (becoming common sense), so
Open the smaller such as well known film star man and wife etc. of the probability referred in text;But these and target entity
The very fixed related entities of relationship, it is again very high with the degree of correlation of target entity, it should be pushed away again when related entities are recommended
It recommends out, this is unapproachable in such a way that above the first is by co-occurrence semantic network;Therefore the invention of the present invention
People considers that larger weight is arranged in important setting relationship between entity so that the pass with target entity by knowledge mapping
System is more important, but the less related entities referred in open text can be recommended out;
Optionally, Fig. 8 shows the another method stream of the recommendation sequence of determining related entities provided in an embodiment of the present invention
Journey, with reference to Fig. 8, this method may include:
Step S700, using after duplicate removal processing candidate entity sets and the target entity as range, determine each correlation
Entity can touch the nearest entity reached.
With the corresponding related entities of target entity (the candidate entity sets after duplicate removal processing) and target entity sheet as
Range, the embodiment of the present invention is it needs to be determined that each related entities can touch the nearest entity reached in the range;
Optionally, a related entities can touch the nearest entity reached in object knowledge collection of illustrative plates, it may be possible to target entity (phase
One) it is to close entity number of edges corresponding with target entity, it is also possible to other related entities (such as related entities and target entities
Corresponding number of edges is more than one, needs related entities closer to target entity by other, is transitioned into target entity);
As shown in figure 4, after being removed processing to candidate entity sets, the related entities of target entity include:
Number of edges one:" small red ", " small strong " and " film A ";
Number of edges two:" little Rong ".
Wherein, related entities " small red ", " small strong " and " film A " can be touched directly and be reached target entity " Xiao Ming ", therefore can be touched
The nearest entity reached is target entity,
And related entities " little Rong " need to touch by related entities " small strong " and reach target entity, therefore related entities " little Rong "
It is " small strong " that the nearest entity reached, which can be touched,.
Step S710, it according to the corresponding relationship weight of each relationship in presetting object knowledge collection of illustrative plates, determines each related real
Body relationship weight corresponding with that can touch the relationship of nearest entity that reaches, obtains the corresponding relationship weight of each related entities.
Optionally, the embodiment of the present invention can utilize Heuristics, and needle is arranged in the different relationships between entity in knowledge mapping
To the relationship weight of property so that the more important entity of relationship has higher relationship weight;
Such as the recommendation for sciemtifec and technical sphere entity, it can be by position, affiliated company, public affairs where entity in corresponding knowledge mapping
Take charge of the larger relationship weight of the setting such as shareholder;The entity of sports field is recommended, it can will be in corresponding knowledge mapping where entity
Larger relationship weight is arranged in team, teammate etc.;
In presetting object knowledge collection of illustrative plates after each corresponding relationship weight of relationship, for each related real of target entity
Body, the embodiment of the present invention can determine the corresponding pass of each related entities according to related entities and the relationship that can touch the nearest entity reached
It is weight;
As above for example, related entities " small red ", " small strong " and " film A " can directly touch " small up to target entity
It is bright ", therefore it is target entity that can touch the nearest entity that reaches, then the corresponding relationship weight of related entities " small red " be " small red " with
The corresponding relationship weight of relationship of " Xiao Ming ", the corresponding relationship weight of related entities " small strong " are the relationship of " small strong " with " Xiao Ming "
Corresponding relationship weight, the corresponding relationship weight of related entities " film A " are " film A " relationship corresponding with the relationship of " Xiao Ming "
Weight;
And related entities " little Rong " need to touch by related entities " small strong " and reach target entity, therefore related entities " little Rong "
It is " small strong " that the nearest entity reached, which can be touched, then the corresponding relationship weight of related entities " little Rong " is the relationship of " little Rong " and " small strong "
Corresponding relationship weight;
I.e. for a related entities, the embodiment of the present invention can determine that the related entities can touch the nearest entity reached, by this
Related entities relationship weight corresponding with that can touch the relationship of nearest entity that reaches, determines the corresponding relationship weight of the related entities.
Step S720, each related entities are weighed the number of edges weight of the corresponding number of edges of related entities with corresponding relationship
Heavy phase combines, and obtains the corresponding weight score of each related entities;Wherein, number of edges is bigger, and number of edges weight is smaller.
After determining the corresponding relationship weight of each related entities, the embodiment of the present invention is real in combination with each related entities and target
The number of edges weight of the corresponding number of edges of body determines the weight score of each related entities;It is generally believed that number of edges is bigger, number of edges weight
Smaller, this is to carry out drop power with the related entities of the number of edges of target entity farther out, may make some that extend
Related entities can be removed;
As related entities number of edges corresponding with target entity for one (in the candidate entity sets after such as duplicate removal processing, or
In person's candidate's entity sets, which is touched by a line reaches target entity), then it is assumed that the corresponding number of edges of the related entities
Number of edges weight be one, such as related entities number of edges corresponding with target entity then needs the number of edges to the related entities more than one
Weight carries out drop power so that the number of edges weight of the related entities is less than 1;
Optionally, the embodiment of the present invention can be arranged in adjacent number of edges, and the number of edges weight of small number of edges is the number of edges power of big number of edges
One times of weight, it is one that the corresponding number of edges of related entities, which can such as be arranged, then corresponding number of edges weight is one, the corresponding side of related entities
Number is two, then corresponding number of edges weight is 1/2=0.5, and the corresponding number of edges of related entities is three, then corresponding number of edges weight is
0.5/2=0.25, and so on.
After determining the number of edges weight of each number of edges, for a related entities, the embodiment of the present invention can be by the related entities pair
The number of edges weight for the number of edges answered obtains the corresponding weight score of the related entities, for each phase with corresponding relationship multiplied by weight
It closes entity to be handled with this, then the corresponding weight score of each related entities can be obtained.
As shown in figure 4, after being removed processing to candidate entity sets, the related entities of target entity include:
Number of edges one:" small red ", " small strong " and " film A ";
Number of edges two:" little Rong ".
For example, " small red " is wife with the relationship that can touch the nearest entity " Xiao Ming " reached, and corresponding relationship can be arranged
Weight is 1;" small strong " and the relationship that can touch the nearest entity " Xiao Ming " reached are to work together, and it is 0.5 that corresponding relationship weight, which can be arranged,;
" film A " and the relationship that can touch the nearest entity " Xiao Ming " reached are to act the leading role, and it is 0.7 that corresponding relationship weight, which can be arranged,;" little Rong "
It is wife with the relationship of the nearest entity " small strong " reached can be touched, it is 1 that corresponding relationship weight, which can be arranged,;
And " small red ", " small strong " and " film A " corresponding number of edges are one, and it is 1 that corresponding number of edges weight, which can be arranged,
" little Rong " corresponding number of edges is two, and it is 0.5 that corresponding number of edges weight, which can be arranged,;
Correspondingly, the weight score of related entities " small red " is calculated as:Relationship weight is multiplied by number of edges weight, i.e. 1*1=
1;The weight score of related entities " small strong " is calculated as:Relationship weight is multiplied by number of edges weight, i.e. 0.5*1=0.5;Related entities
The weight score of " film A " is calculated as:Relationship weight is multiplied by number of edges weight, i.e. 0.7*1=0.7;Related entities " little Rong "
Weight score is calculated as:Relationship weight is multiplied by number of edges weight, i.e. 1*0.5=0.5;
Correspondingly, the weight score signal of each related entities can be as shown in table 1 below
Related entities | Weight score |
It is small red | 1 |
It is small strong | 0.5 |
Film A | 0.7 |
Little Rong | 0.5 |
Table 1
Farther out due to the number of edges with target entity, whole as can be seen that although small container has higher relationship weight
The weight score of body is weighed by drop.
Step S730, according to the corresponding weight score of each related entities, the recommendation sequence of each related entities is determined, wherein
Weight score is higher, recommends sequence more forward.
The embodiment of the present invention, can be by some and target entity by the constraint of object knowledge collection of illustrative plates relationship weight itself
Relationship is more important, but the recommendation sequence of the related entities seldom referred to due to known is promoted so that recommends the phase
The sequence for closing entity has higher precision.
Three, above-mentioned recommendation sequence is determined that scheme one is combined with two, i.e., is calculated on open text by scheme one
The degree of correlation score of related entities and target entity is determined by scheme two according to the relationship weight in object knowledge collection of illustrative plates
After the weight score of related entities, the degree of correlation score of same related entities is added with weight score, it is real to obtain the correlation
The ranking score of body determines the recommendation sequence of related entities with the ranking score of related entities;
Optionally, Fig. 9 shows the another method stream of the recommendation sequence of determining related entities provided in an embodiment of the present invention
Journey, with reference to Fig. 9, this method may include:
Step S800, the degree of correlation score of each related entities and target entity is counted in open text.
Step S810, using after duplicate removal processing candidate entity sets and the target entity as range, determine each correlation
Entity can touch the nearest entity reached.
Step S820, it according to the corresponding relationship weight of each relationship in presetting object knowledge collection of illustrative plates, determines each related real
Body relationship weight corresponding with that can touch the relationship of nearest entity that reaches, obtains the corresponding relationship weight of each related entities.
Step S830, each related entities are weighed the number of edges weight of the corresponding number of edges of related entities with corresponding relationship
Heavy phase combines, and obtains the corresponding weight score of each related entities;Wherein, number of edges is bigger, and number of edges weight is smaller.
Step S840, for each related entities, the corresponding degree of correlation score of related entities is added with weight score, is obtained
To the corresponding ranking score of each related entities.
Step S850, according to the corresponding ranking score of each related entities, the recommendation sequence of each related entities is determined, wherein
Ranking score is higher, recommends sequence more forward.
The combination that may be considered Fig. 7 and Fig. 9 schemes shown in Fig. 9, for each related entities, the embodiment of the present invention can
The degree of correlation score and the corresponding weight score of the related entities of the related entities and target entity are determined, to the phase
The degree of correlation score for closing entity is added with weight score, obtains the corresponding ranking score of the related entities, correlation is carried out with this
The recommendation of entity is sorted.
The embodiment of the present invention realizes the digging of the related entities of target entity based on the object knowledge collection of illustrative plates comprising target entity
The relevant information of pick, the target entity included due to object knowledge collection of illustrative plates can more fully be arrived with the excavation of maximum probability
Target entity history comprehensive relevant information in the past so that the related entities result for the target entity excavated is more comprehensive,
The recall rate of the related entities result of target entity determined by being promoted;
Further, according to co-occurrence semantic network, and/or, object knowledge collection of illustrative plates relationship weight itself determines excavated phase
The recommendation sequence for closing entity, may make the related entities recommended when carrying out the recommendation of related entities to have degree of precision
Sequence, promotes the probability that the relevant information of target entity is widely-available for users, and the relevant information for promoting target entity obtains just
Profit.
Related entities determining device provided in an embodiment of the present invention is introduced below, related entities described below are true
Determining device can determine that method corresponds reference with above-described related entities.Related entities determining device described below can
To be considered that the computing device related entities that embodiment provides to realize the present invention determine method, the function module frame of required setting
Structure.
Figure 10 is the structure diagram of related entities determining device provided in an embodiment of the present invention, which can be applied to have
The computing device of data operation ability, the computing device can select the server of network side, can also select the electricity of user side
The electronic equipments such as brain;
Referring to Fig.1 0, related entities determining device provided in an embodiment of the present invention may include:
Object knowledge collection of illustrative plates acquisition module 100, for obtaining object knowledge collection of illustrative plates, the object knowledge collection of illustrative plates at least has
Target entity;
Candidate entity sets determining module 200, for determining in the object knowledge collection of illustrative plates, the candidate of the target entity
Entity sets;It is described candidate entity sets include:The corresponding candidate entity of each number of edges up to the target entity can be touched;
Related entities determining module 300, for according to the candidate entity sets, determining that the correlation of the target entity is real
Body.
Optionally, object knowledge collection of illustrative plates acquisition module 100 is specifically included for obtaining object knowledge collection of illustrative plates:
It obtains and inputs text, it is previously given in the input text to there are multiple name entities, the name entity at least to wrap
Include the target entity;
The name entity given in text will be inputted, is mapped on the target entity of knowledge mapping, obtains object knowledge figure
Spectrum;The knowledge mapping is built by the data source comprising target entity.
Optionally, candidate entity sets determining module 200, for determining in the object knowledge collection of illustrative plates, the target is real
The candidate entity sets of body, specifically include:
Presetting number of edges range is obtained, the number of edges range includes multiple number of edges;
It according to each number of edges included by the number of edges range, determines in the object knowledge collection of illustrative plates, can touch up to the target
The corresponding candidate entity of each number of edges of entity, obtains the candidate entity sets of the target entity.
Optionally, related entities determining module 300, for according to the candidate entity sets, determining the target entity
Related entities, specifically include:
If in candidate's entity sets, there are the candidate entities of the repetition of corresponding different numbers edge, by the candidate entity
The candidate entity repeated in set carries out duplicate removal processing, to retain the candidate entity of number of edges minimum in the candidate entity repeated;
By the candidate entity included by the candidate entity sets after duplicate removal processing, the correlation as the target entity is real
Body.
Optionally, Figure 11 shows another structure diagram of related entities determining device provided in an embodiment of the present invention, knot
It closes shown in Figure 10 and Figure 11, which can also include:
Recommend sequence determining module 400, for determining that the recommendation of each related entities is sorted, with pushing away according to each related entities
Recommend the recommendation that sequence carries out related entities.
Optionally, on the one hand, recommend sequence determining module 400, it is specific to wrap for determining that the recommendation of each related entities is sorted
It includes:
The degree of correlation score of each related entities and target entity is counted in open text;
Determine each related entities corresponding weight score in the object knowledge collection of illustrative plates;
For each related entities, the corresponding degree of correlation score of related entities is added with weight score, obtains each correlation
The corresponding ranking score of entity;
According to the corresponding ranking score of each related entities, the recommendation sequence of each related entities is determined, wherein ranking score is got over
Height recommends sequence more forward.
Specifically, recommending sequence determining module 400, for determining that each related entities are corresponding in the object knowledge collection of illustrative plates
Weight score, specifically include:
Using after duplicate removal processing candidate entity sets and the target entity as range, determine that each related entities can be touched and reach
Nearest entity;
According to the corresponding relationship weight of each relationship in presetting object knowledge collection of illustrative plates, determine that each related entities are reached with can touch
Nearest entity the corresponding relationship weight of relationship, obtain the corresponding relationship weight of each related entities;
The number of edges weight of the corresponding number of edges of related entities is combined by each related entities with corresponding relationship weight,
Obtain the corresponding weight score of each related entities;Wherein, number of edges is bigger, and number of edges weight is smaller.
And recommend sequence determining module 400, it is related to target entity for counting each related entities in open text
Degree score, specifically includes:
For a related entities, the related entities and the amount of text and text total quantity of target entity are determined while occurred
The first ratio, there is the amount of text of the related entities and the second ratio of text total quantity, the text of target entity occur
The third ratio of quantity and text total quantity;
According to first ratio, the second ratio and third ratio determine the mutual information of the related entities and target entity,
The degree of correlation score of the related entities and target entity is indicated with identified mutual information.
On the other hand, recommend sequence determining module 400, for determining that the recommendation of each related entities is sorted, specifically include:
The degree of correlation score that each related entities and target entity are counted in open text, according to each related entities and mesh
The degree of correlation score for marking entity determines the recommendation sequence of each related entities, wherein related journey point is higher several times, recommends sequence
It is more forward.
In another aspect, recommending sequence determining module 400, for determining that the recommendation of each related entities is sorted, specifically include:
Determine each related entities corresponding weight score in the object knowledge collection of illustrative plates, it is corresponding according to each related entities
Weight score determines the recommendation sequence of each related entities, wherein weight score is higher, recommends sequence more forward.
Related entities determining device provided in an embodiment of the present invention can promote the related entities of identified target entity
As a result recall rate, and the related entities recommended have the sequence of degree of precision, can promote the relevant information quilt of target entity
The probability that user utilizes.
Optionally, the embodiment of the present invention also provides a kind of computing device, which may include phase described above
Close entity determining device.
Optionally, Figure 12 shows the hardware block diagram of the computing device, referring to Fig.1 2, which can wrap
It includes:Processor 1, communication interface 2, memory 3 and communication bus 4;
Wherein processor 1, communication interface 2, memory 3 complete mutual communication by communication bus 4;
Optionally, communication interface 2 can be the interface of communication module, such as the interface of gsm module;
Processor 1 may be a central processor CPU or specific integrated circuit ASIC (Application
Specific Integrated Circuit), or be arranged to implement the integrated electricity of one or more of the embodiment of the present invention
Road.
Memory 3 may include high-speed RAM memory, it is also possible to further include nonvolatile memory (non-volatile
Memory), a for example, at least magnetic disk storage.
Wherein, processor 1 is specifically used for:
Object knowledge collection of illustrative plates is obtained, the object knowledge collection of illustrative plates at least has target entity;
It determines in the object knowledge collection of illustrative plates, the candidate entity sets of the target entity;Candidate's entity sets packet
It includes:The corresponding candidate entity of each number of edges up to the target entity can be touched;
According to the candidate entity sets, the related entities of the target entity are determined.
Each embodiment is described by the way of progressive in this specification, the highlights of each of the examples are with other
The difference of embodiment, just to refer each other for identical similar portion between each embodiment.For device disclosed in embodiment
For, since it is corresponded to the methods disclosed in the examples, so description is fairly simple, related place is said referring to method part
It is bright.
Professional further appreciates that, unit described in conjunction with the examples disclosed in the embodiments of the present disclosure
And algorithm steps, can be realized with electronic hardware, computer software, or a combination of the two, in order to clearly demonstrate hardware and
The interchangeability of software generally describes each exemplary composition and step according to function in the above description.These
Function is implemented in hardware or software actually, depends on the specific application and design constraint of technical solution.Profession
Technical staff can use different methods to achieve the described function each specific application, but this realization is not answered
Think beyond the scope of this invention.
The step of method described in conjunction with the examples disclosed in this document or algorithm, can directly be held with hardware, processor
The combination of capable software module or the two is implemented.Software module can be placed in random access memory (RAM), memory, read-only deposit
Reservoir (ROM), electrically programmable ROM, electrically erasable ROM, register, hard disk, moveable magnetic disc, CD-ROM or technology
In any other form of storage medium well known in field.
The foregoing description of the disclosed embodiments enables those skilled in the art to implement or use the present invention.
Various modifications to these embodiments will be apparent to those skilled in the art, as defined herein
General Principle can in other embodiments be realized in the case where not departing from core of the invention thought or scope.Therefore, originally
Invention is not intended to be limited to the embodiments shown herein, and is to fit to and the principles and novel features disclosed herein
Consistent widest range.
Claims (16)
1. a kind of related entities determine method, which is characterized in that including:
Object knowledge collection of illustrative plates is obtained, the object knowledge collection of illustrative plates at least has target entity;
It determines in the object knowledge collection of illustrative plates, the candidate entity sets of the target entity;It is described candidate entity sets include:It can
Touch the corresponding candidate entity of each number of edges up to the target entity;
According to the candidate entity sets, the related entities of the target entity are determined.
2. related entities according to claim 1 determine method, which is characterized in that the acquisition object knowledge collection of illustrative plates packet
It includes:
It obtains and inputs text, it is previously given in the input text to there are multiple name entities, the name entity to include at least institute
State target entity;
The name entity given in text will be inputted, is mapped on the target entity of knowledge mapping, obtains object knowledge collection of illustrative plates;Institute
Knowledge mapping is stated to be built by the data source comprising target entity.
3. related entities according to claim 1 or 2 determine method, which is characterized in that the determination object knowledge
In collection of illustrative plates, the candidate entity sets of the target entity include:
Presetting number of edges range is obtained, the number of edges range includes multiple number of edges;
It according to each number of edges included by the number of edges range, determines in the object knowledge collection of illustrative plates, can touch up to the target entity
The corresponding candidate entity of each number of edges, obtain the candidate entity sets of the target entity.
4. related entities according to claim 1 determine method, which is characterized in that described according to the candidate entity set
It closes, determines that the related entities of the target entity include:
If in candidate's entity sets, there are the candidate entities of the repetition of corresponding different numbers edge, by the candidate entity sets
The candidate entity of middle repetition carries out duplicate removal processing, to retain the candidate entity of number of edges minimum in the candidate entity repeated;
By the candidate entity included by the candidate entity sets after duplicate removal processing, the related entities as the target entity.
5. related entities according to claim 4 determine method, which is characterized in that further include:
The recommendation sequence for determining each related entities, to carry out the recommendation of related entities according to the recommendation sequence of each related entities.
6. related entities according to claim 5 determine method, which is characterized in that the recommendation of each related entities of determination
Sequence includes:
The degree of correlation score of each related entities and target entity is counted in open text;
Determine each related entities corresponding weight score in the object knowledge collection of illustrative plates;
For each related entities, the corresponding degree of correlation score of related entities is added with weight score, obtains each related entities
Corresponding ranking score;
According to the corresponding ranking score of each related entities, the recommendation sequence of each related entities is determined, wherein ranking score is higher,
Recommend sequence more forward.
7. related entities according to claim 5 determine method, which is characterized in that the recommendation of each related entities of determination
Sequence includes:
The degree of correlation score that each related entities and target entity are counted in open text, it is real according to each related entities and target
The degree of correlation score of body determines the recommendation sequence of each related entities, wherein related journey point is higher several times, and sequence is recommended more to lean on
Before;
Or, determine each related entities corresponding weight score in the object knowledge collection of illustrative plates, it is corresponding according to each related entities
Weight score determines the recommendation sequence of each related entities, wherein weight score is higher, recommends sequence more forward.
8. the related entities described according to claim 6 or 7 determine method, which is characterized in that each related entities of determination exist
Corresponding weight score includes in the object knowledge collection of illustrative plates:
Using after duplicate removal processing candidate entity sets and the target entity as range, determine that each related entities can be touched and reach most
Nearly entity;
According to the corresponding relationship weight of each relationship in presetting object knowledge collection of illustrative plates, determines each related entities and can touch and reach most
The corresponding relationship weight of relationship of nearly entity, obtains the corresponding relationship weight of each related entities;
The number of edges weight of the corresponding number of edges of related entities is combined with corresponding relationship weight, is obtained by each related entities
The corresponding weight score of each related entities;Wherein, number of edges is bigger, and number of edges weight is smaller.
9. the related entities described according to claim 6 or 7 determine method, which is characterized in that described to be counted in open text
The degree of correlation score of each related entities and target entity includes:
For a related entities, determine occur the of the amount of text of the related entities and target entity and text total quantity simultaneously
There is the amount of text of the related entities and the second ratio of text total quantity, the amount of text of target entity occurs in one ratio
With the third ratio of text total quantity;
According to first ratio, the second ratio and third ratio determine the mutual information of the related entities and target entity, with institute
Determining mutual information indicates the degree of correlation score of the related entities and target entity.
10. a kind of related entities determining device, which is characterized in that including:
Object knowledge collection of illustrative plates acquisition module, for obtaining object knowledge collection of illustrative plates, the object knowledge collection of illustrative plates at least has target real
Body;
Candidate entity sets determining module, for determining in the object knowledge collection of illustrative plates, the candidate entity set of the target entity
It closes;It is described candidate entity sets include:The corresponding candidate entity of each number of edges up to the target entity can be touched;
Related entities determining module, for according to the candidate entity sets, determining the related entities of the target entity.
11. related entities determining device according to claim 10, which is characterized in that the object knowledge collection of illustrative plates obtains mould
Block is specifically included for obtaining object knowledge collection of illustrative plates:
It obtains and inputs text, it is previously given in the input text to there are multiple name entities, the name entity to include at least institute
State target entity;
The name entity given in text will be inputted, is mapped on the target entity of knowledge mapping, obtains object knowledge collection of illustrative plates;Institute
Knowledge mapping is stated to be built by the data source comprising target entity.
12. related entities determining device according to claim 10, which is characterized in that the related entities determining module,
For according to the candidate entity sets, determining the related entities of the target entity, specifically including:
If in candidate's entity sets, there are the candidate entities of the repetition of corresponding different numbers edge, by the candidate entity sets
The candidate entity of middle repetition carries out duplicate removal processing, to retain the candidate entity of number of edges minimum in the candidate entity repeated;
By the candidate entity included by the candidate entity sets after duplicate removal processing, the related entities as the target entity.
13. related entities determining device according to claim 12, which is characterized in that further include:
Recommend sequence determining module, for determining that the recommendation of each related entities is sorted, to sort according to the recommendation of each related entities
Carry out the recommendation of related entities.
14. related entities determining device according to claim 13, which is characterized in that the recommendation sequence determining module,
For determining that the recommendation of each related entities is sorted, specifically include:
The degree of correlation score of each related entities and target entity is counted in open text;
Determine each related entities corresponding weight score in the object knowledge collection of illustrative plates;
For each related entities, the corresponding degree of correlation score of related entities is added with weight score, obtains each related entities
Corresponding ranking score;
According to the corresponding ranking score of each related entities, the recommendation sequence of each related entities is determined, wherein ranking score is higher,
Recommend sequence more forward.
15. related entities determining device according to claim 14, which is characterized in that the recommendation sequence determining module,
For determining each related entities corresponding weight score in the object knowledge collection of illustrative plates, specifically include:
Using after duplicate removal processing candidate entity sets and the target entity as range, determine that each related entities can be touched and reach most
Nearly entity;
According to the corresponding relationship weight of each relationship in presetting object knowledge collection of illustrative plates, determines each related entities and can touch and reach most
The corresponding relationship weight of relationship of nearly entity, obtains the corresponding relationship weight of each related entities;
The number of edges weight of the corresponding number of edges of related entities is combined with corresponding relationship weight, is obtained by each related entities
The corresponding weight score of each related entities;Wherein, number of edges is bigger, and number of edges weight is smaller;
The recommendation sequence determining module, for counting each related entities and the degree of correlation point of target entity in open text
Number, specifically includes:
For a related entities, determine occur the of the amount of text of the related entities and target entity and text total quantity simultaneously
There is the amount of text of the related entities and the second ratio of text total quantity, the amount of text of target entity occurs in one ratio
With the third ratio of text total quantity;
According to first ratio, the second ratio and third ratio determine the mutual information of the related entities and target entity, with institute
Determining mutual information indicates the degree of correlation score of the related entities and target entity.
16. a kind of computing device, which is characterized in that including claim 10-15 any one of them related entities determining devices.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710120836.9A CN108536702B (en) | 2017-03-02 | 2017-03-02 | Method and device for determining related entities and computing equipment |
PCT/CN2018/077416 WO2018157790A1 (en) | 2017-03-02 | 2018-02-27 | Method and device for determining related entity, computing device and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710120836.9A CN108536702B (en) | 2017-03-02 | 2017-03-02 | Method and device for determining related entities and computing equipment |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108536702A true CN108536702A (en) | 2018-09-14 |
CN108536702B CN108536702B (en) | 2022-12-02 |
Family
ID=63369790
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710120836.9A Active CN108536702B (en) | 2017-03-02 | 2017-03-02 | Method and device for determining related entities and computing equipment |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN108536702B (en) |
WO (1) | WO2018157790A1 (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110008352A (en) * | 2019-03-28 | 2019-07-12 | 腾讯科技(深圳)有限公司 | Entity finds method and device |
CN110825821A (en) * | 2019-09-30 | 2020-02-21 | 深圳云天励飞技术有限公司 | Personnel relationship query method and device, electronic equipment and storage medium |
CN112069323A (en) * | 2020-08-04 | 2020-12-11 | 扬州制汇互联信息技术有限公司 | Recommendation method based on industrial knowledge graph |
CN113010769A (en) * | 2019-12-19 | 2021-06-22 | 京东方科技集团股份有限公司 | Knowledge graph-based article recommendation method and device, electronic equipment and medium |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110134796B (en) * | 2019-04-19 | 2023-06-02 | 平安科技(深圳)有限公司 | Knowledge graph-based clinical trial retrieval method, device, computer equipment and storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130346421A1 (en) * | 2012-06-22 | 2013-12-26 | Microsoft Corporation | Targeted disambiguation of named entities |
CN103593792A (en) * | 2013-11-13 | 2014-02-19 | 复旦大学 | Individual recommendation method and system based on Chinese knowledge mapping |
CN104199872A (en) * | 2014-08-19 | 2014-12-10 | 北京搜狗科技发展有限公司 | Information recommendation method and device |
CN105095433A (en) * | 2015-07-22 | 2015-11-25 | 百度在线网络技术(北京)有限公司 | Recommendation method and device for entities |
CN106372118A (en) * | 2016-08-24 | 2017-02-01 | 武汉烽火普天信息技术有限公司 | Large-scale media text data-oriented online semantic comprehension search system and method |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102591862A (en) * | 2011-01-05 | 2012-07-18 | 华东师范大学 | Control method and device of Chinese entity relationship extraction based on word co-occurrence |
US9390174B2 (en) * | 2012-08-08 | 2016-07-12 | Google Inc. | Search result ranking and presentation |
CN104102713B (en) * | 2014-07-16 | 2018-01-19 | 百度在线网络技术(北京)有限公司 | Recommendation results show method and apparatus |
CN104537065A (en) * | 2014-12-29 | 2015-04-22 | 北京奇虎科技有限公司 | Search result pushing method and system |
US9547823B2 (en) * | 2014-12-31 | 2017-01-17 | Verizon Patent And Licensing Inc. | Systems and methods of using a knowledge graph to provide a media content recommendation |
CN106326211B (en) * | 2016-08-17 | 2019-09-20 | 海信集团有限公司 | Determination of distance method and apparatus between the keyword of alternate statement |
-
2017
- 2017-03-02 CN CN201710120836.9A patent/CN108536702B/en active Active
-
2018
- 2018-02-27 WO PCT/CN2018/077416 patent/WO2018157790A1/en active Application Filing
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130346421A1 (en) * | 2012-06-22 | 2013-12-26 | Microsoft Corporation | Targeted disambiguation of named entities |
CN103593792A (en) * | 2013-11-13 | 2014-02-19 | 复旦大学 | Individual recommendation method and system based on Chinese knowledge mapping |
CN104199872A (en) * | 2014-08-19 | 2014-12-10 | 北京搜狗科技发展有限公司 | Information recommendation method and device |
CN105095433A (en) * | 2015-07-22 | 2015-11-25 | 百度在线网络技术(北京)有限公司 | Recommendation method and device for entities |
CN106372118A (en) * | 2016-08-24 | 2017-02-01 | 武汉烽火普天信息技术有限公司 | Large-scale media text data-oriented online semantic comprehension search system and method |
Non-Patent Citations (1)
Title |
---|
刘峤等: "基于图的中文集成实体链接算法", 《计算机研究与发展》, vol. 53, no. 02, 15 February 2016 (2016-02-15), pages 1 - 4 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110008352A (en) * | 2019-03-28 | 2019-07-12 | 腾讯科技(深圳)有限公司 | Entity finds method and device |
CN110008352B (en) * | 2019-03-28 | 2022-12-20 | 腾讯科技(深圳)有限公司 | Entity discovery method and device |
CN110825821A (en) * | 2019-09-30 | 2020-02-21 | 深圳云天励飞技术有限公司 | Personnel relationship query method and device, electronic equipment and storage medium |
CN110825821B (en) * | 2019-09-30 | 2022-11-22 | 深圳云天励飞技术有限公司 | Personnel relationship query method and device, electronic equipment and storage medium |
CN113010769A (en) * | 2019-12-19 | 2021-06-22 | 京东方科技集团股份有限公司 | Knowledge graph-based article recommendation method and device, electronic equipment and medium |
CN112069323A (en) * | 2020-08-04 | 2020-12-11 | 扬州制汇互联信息技术有限公司 | Recommendation method based on industrial knowledge graph |
CN112069323B (en) * | 2020-08-04 | 2024-04-26 | 扬州制汇互联信息技术有限公司 | Recommendation method based on industrial knowledge graph |
Also Published As
Publication number | Publication date |
---|---|
CN108536702B (en) | 2022-12-02 |
WO2018157790A1 (en) | 2018-09-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108536702A (en) | A kind of related entities determine method, apparatus and computing device | |
US7194454B2 (en) | Method for organizing records of database search activity by topical relevance | |
CN109189867A (en) | Relationship discovery method, apparatus and storage medium based on Corporate Intellectual map | |
US7343551B1 (en) | Autocompleting form fields based on previously entered values | |
US7647306B2 (en) | Using community annotations as anchortext | |
US20080313137A1 (en) | Behavioral WEB Graph | |
US20100241647A1 (en) | Context-Aware Query Recommendations | |
EP1225517A2 (en) | System and methods for computer based searching for relevant texts | |
CN106126521A (en) | The social account method for digging of destination object and server | |
US20140040371A1 (en) | Systems and methods for identifying geographic locations of social media content collected over social networks | |
US20130117287A1 (en) | Methods and systems for constructing personal profiles from contact data | |
JP2014513826A (en) | Computer systems, databases and their use | |
CN104572889A (en) | Method, device and system for recommending search terms | |
CN107690637B (en) | Connecting semantically related data using large-table corpus | |
JP2005535039A (en) | Interact with desktop clients with geographic text search systems | |
JP2008507792A (en) | A search engine that uses the background situation placed on the network | |
CN101911065A (en) | Access subject information retrieval device | |
EP4109293A1 (en) | Data query method and apparatus, electronic device, storage medium, and program product | |
CN110096646A (en) | The generation of category related information and its video pushing method and relevant device | |
JP2003016094A (en) | Method for profile management used for information filtering and program therefor | |
CN103262079B (en) | Search device and search method | |
Ozdikis et al. | Ontology-based recommendation for points of interest retrieved from multiple data sources | |
CN101425981A (en) | Information publishing system and method for publishing information according to mutual exclusive indication | |
CN109376287B (en) | House property map construction method, device, computer equipment and storage medium | |
CN101266615B (en) | Method for browsing a data communications 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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |