CN105095433B - Entity recommended method and device - Google Patents
Entity recommended method and device Download PDFInfo
- Publication number
- CN105095433B CN105095433B CN201510434989.1A CN201510434989A CN105095433B CN 105095433 B CN105095433 B CN 105095433B CN 201510434989 A CN201510434989 A CN 201510434989A CN 105095433 B CN105095433 B CN 105095433B
- Authority
- CN
- China
- Prior art keywords
- related entities
- entity
- rank
- rank related
- feature
- 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.)
- Active
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
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the invention discloses a kind of entity recommended method and devices.Wherein, which comprises receive the search statement that user is inputted by character input modes or voice input mode, identify the searching entities for including in described search sentence；According to the entity associated network of personal connections being pre-created, obtain the N rank related entities of described search entity, wherein the first rank related entities in the N rank related entities are the entity for having direct correlation relationship with described search entity, i-th rank related entities are the entity for having direct correlation relationship with the (i-1)-th rank related entities, the N is the natural number greater than 1, and the i is greater than 2 and is less than or equal to N；Each rank related entities that will acquire are showed.The degree of correlation between the entity for including in recommended entity and search statement can be improved in technical solution provided in an embodiment of the present invention, enhances the interpretation of recommended entity.
Description
Technical field
The present embodiments relate to Internet technical field more particularly to entity recommended method and devices.
Background technique
In current search technique, to excite the more search needs of user, in the search phrase for receiving user's input
After sentence, search engine uses open air in addition to the web page interlinkage relevant to the search statement that will be searched is presented to, and can also pass through
The proposed algorithm of setting calculates the related entities for the entity for including in the search statement, and using the related entities as recommended entity
It is presented to user, while corresponding rationale for the recommendation also being showed.
Wherein, proposed algorithm mainly uses collaborative filtering.Collaborative filtering is divided into two major classes: the association based on user
With filter algorithm and based on the collaborative filtering of recommendation.Currently, more mainstream is using the collaboration based on user
Algorithm is filtered, i.e., using the co-occurrence feature between the entity for including in search statement come the correlation between computational entity, thus to do
Recommend to calculate.
However, the related entities suggested design of the collaborative filtering based on user have following defects that recommended entity with
The degree of correlation for the entity for including in search statement is lower, causes the interpretation of recommended entity poor.
Summary of the invention
The embodiment of the present invention provides entity recommended method and device, to improve the reality for including in recommended entity and search statement
The degree of correlation between body enhances the interpretation of recommended entity.
On the one hand, the embodiment of the invention provides a kind of entity recommended methods, this method comprises:
The search statement that user is inputted by character input modes or voice input mode is received, identifies described search sentence
In include searching entities；
According to the entity associated network of personal connections being pre-created, the N rank related entities of described search entity are obtained, wherein the N
The first rank related entities in rank related entities are the entity for having direct correlation relationship with described search entity, and the i-th rank is related
Entity is the entity for having direct correlation relationship with the (i-1)-th rank related entities, and the N is the natural number greater than 1, and the i is greater than 2
Less than or equal to N；
Each rank related entities that will acquire are showed.
On the other hand, the embodiment of the invention also provides a kind of entity recommendation apparatus, which includes:
Searching entities recognition unit, the search inputted for receiving user by character input modes or voice input mode
Sentence identifies the searching entities for including in described search sentence；
Related entities acquiring unit, for obtaining the N of described search entity according to the entity associated network of personal connections being pre-created
Rank related entities, wherein the first rank related entities in the N rank related entities are to have to be directly linked with described search entity
The entity of relationship, the i-th rank related entities be with the (i-1)-th rank related entities have direct correlation relationship entity, the N be greater than
1 natural number, the i are greater than 2 and are less than or equal to N；
Show unit, each rank related entities for obtaining the related entities acquiring unit show.
Technical solution provided in an embodiment of the present invention, based on the entity associated network of personal connections being pre-created, inquiry is real with search
Body has the multistage related entities of direct correlation relationship and indirect association relationship, is presented to user, Ke Yiti as recommended entity
The degree of correlation between entity for including in high recommended entity and search statement, enhances the interpretation of recommended entity.
Detailed description of the invention
Fig. 1 is a kind of flow diagram for entity recommended method that the embodiment of the present invention one provides；
Fig. 2 is a kind of flow diagram of entity recommended method provided by Embodiment 2 of the present invention；
Fig. 3 is a kind of flow diagram for entity recommended method that the embodiment of the present invention three provides；
Fig. 4 is a kind of flow diagram for entity recommended method that the embodiment of the present invention four provides；
Fig. 5 is a kind of structural schematic diagram for entity recommendation apparatus that the embodiment of the present invention five provides.
Specific embodiment
The present invention is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining the present invention rather than limiting the invention.It also should be noted that in order to just
Only the parts related to the present invention are shown in description, attached drawing rather than entire infrastructure.
It should be mentioned that some exemplary embodiments are described as before exemplary embodiment is discussed in greater detail
The processing or method described as flow chart.Although operations (or step) are described as the processing of sequence by flow chart,
Many of these operations can be implemented concurrently, concomitantly or simultaneously.In addition, the sequence of operations can be pacified again
Row.The processing can be terminated when its operations are completed, it is also possible to have the additional step being not included in attached drawing.Institute
Stating processing can correspond to method, function, regulation, subroutine, subprogram etc..
Embodiment one
Fig. 1 is a kind of flow diagram for entity recommended method that the embodiment of the present invention one provides.The present embodiment can be used for
Guidance user's discovery is really intended to or the related interests of user is caused to search again for.The method of the present embodiment can be by entity
Recommendation apparatus executes, which can be integrated in by software realization and provide search service for miscellaneous terminal device
In search engine.Referring to Fig. 1, entity recommended method provided in this embodiment specifically includes following operation:
Operation 110 receives the search statement that user is inputted by character input modes or voice input mode, identification search
The searching entities for including in sentence.
In the present embodiment, search statement can be the text that user is inputted in search box by keyboard or hand-written operation
The sentence of format is also possible to user and passes through the voice messaging that voice acquisition device (such as microphone) inputs, which is believed
Breath searches for corresponding search statement as this.
After receiving search statement, which is identified, to extract entity included in it, as searching
Suo Shiti.For example, the search statement of input is " Gulang Island is in Xiamen ", then searching entities are " Gulang Island " and " Xiamen ".
Wherein, for the search statement of phonetic matrix, it need to be first converted into the search statement of text formatting, then scanned for
Entity recognition.Specific conversion operation can be completed locally by entity recommendation apparatus, or report speech recognition server, by this
Server is completed, and the present embodiment is not especially limited this.
The entity associated network of personal connections that operation 120, basis are pre-created, obtains the N rank related entities of searching entities.
In the present embodiment, corpus can be excavated in advance, to create entity associated network of personal connections.Entity associated relationship
Net is the semantic network for describing incidence relation between entity.Corpus can by based on set algorithm crawled from internet to
A large amount of websites provided by webpage composition, may include having structural data and/or unstructured data in these webpages.Example
Property, entity associated network of personal connections is knowledge mapping.
Wherein, the first rank related entities in N rank related entities are the entity for having direct correlation relationship with searching entities,
I-th rank related entities are the entity for having direct correlation relationship with the (i-1)-th rank related entities, and N is the natural number greater than 1, and i is greater than
2 and be less than or equal to N.Preferably, if same order related entities identical entity do not occur (for example, in the first rank related entities
Including " Jiu Zhaigou ", there is also " Jiu Zhaigous " in third rank related entities), then it can only retain an entity in identical entity i.e.
It can.
There is so-called two entities direct correlation relationship to refer to: without just having by third entity between the two entities
Standby incidence relation.For example, user has input the searching entities of " red sorghum " this TV play classification, since its protagonist is " Zhou Xun "
The other entity of this figure kind, and the works that " Zhou Xun " was drilled have the reality of the TV plays classification such as " Daming Palace word ", " orange is red "
Body, so: " Zhou Xun " and " red sorghum " has direct correlation relationship, is the first rank related entities of " red sorghum "；And " Daming Palace
Word " or " orange is red " only have direct correlation relationship with " Zhou Xun ", do not have to be directly linked with " red sorghum " and close
System, could will generate indirectly incidence relation with " red sorghum " by means of " Zhou Xun " this intermediate entities, be the second of " red sorghum "
Rank related entities.
For another example searching entities are " Liu Dehua ", it is " Zhu with its related entities with man and wife this direct correlation relationship
It is beautiful pretty ", and having the related entities of this direct correlation relationship of father and daughter with entity " Zhu Liqian " is " Zhu Jiancheng ".Then " Zhu Liqian "
For the first rank related entities of searching entities " Liu Dehua ", and " Zhu Jiancheng " will be by means of " Zhu Liqian " this intermediate entities ability
Incidence relation is generated indirectly with " Liu Dehua ", is the second-order related entities of " Liu Dehua ".
Operation 130, each rank related entities that will acquire are showed.
Technical solution provided in this embodiment, based on the entity associated network of personal connections being pre-created, inquiry has with searching entities
There are the multistage related entities of direct correlation relationship and indirect association relationship, is presented to user as recommended entity, can be improved and push away
The degree of correlation between the entity for including in entity and search statement is recommended, the interpretation of recommended entity is enhanced.
Embodiment two
Fig. 2 is a kind of flow diagram of entity recommended method provided by Embodiment 2 of the present invention.The present embodiment is above-mentioned
On the basis of embodiment one, the operation of " creation entity associated network of personal connections " and " obtain and show rationale for the recommendation " is increased.Ginseng
See Fig. 2, entity recommended method provided in this embodiment specifically includes following operation:
Operation 210 obtains knowledge mapping, and wherein knowledge mapping includes the feature of at least one entity and entity.
Include multiple entities in knowledge mapping, each entity has a feature, this feature include: attributive character and/or
Relationship characteristic, and Characteristic Number can for one, it is two or more.Correspondingly, each entity should also have and each feature pair
The value answered, the value can be a new entity, be also possible to a concept.Classification belonging to entity is different, excavates
Feature is also just different.For entity other for figure kind, attributive character can be age, weight, occupation etc., and relationship is special
Sign can be wife, father, daughter etc..For example, " Liu Dehua " is exactly an entity, the attributive character of Liu Dehua include: height,
Weight, constellation, age, wife and daughter etc.；" Yao Ming " is also an entity, and the attributive character of Yao Ming includes, height, weight,
Occupation, wife, daughter, blood group and constellation etc..
For the entity of variety show classification, attributive character can be play time, play satellite TV etc., relationship
Feature can be host, welcome guest etc..
In the present embodiment, knowledge mapping can be considered a huge network diagramming, and each node in network diagramming indicates
Entity or concept, and the side in network diagramming is then made of attributive character or relationship characteristic.Knowledge mapping is obtained, is mainly exactly to be used for
Incidence relation between building and maintenance entity, the map can be adjusted with the update of corpus.As the present embodiment
A kind of preferred embodiment can in advance excavate corpus, to generate the knowledge mapping of different field.For wherein every
The knowledge mapping in one field only includes the feature of each entity in the art.
It operates 220, according to knowledge mapping, creates entity associated network of personal connections.
It illustratively, can be directly by knowledge mapping, as entity associated network of personal connections.
But, it is contemplated that the content that the knowledge mapping got is included is often more, is not easy to search, and some of
The feature of entity is not extremely to pay close attention to for vast search user, can first be carried out thus to each entity in knowledge mapping
Feature Selection pretreatment, with filter out some users in knowledge mapping dare not interest content.According to knowledge mapping, creation is real
Body incidence relation net, comprising: according to the historical search sentence of search log recording, the heat of the feature of entity in Extracting Knowledge map
Men Du, to carry out Feature Selection；By the knowledge mapping after Feature Selection, as entity associated network of personal connections.
Illustratively, each entity in knowledge mapping can be traversed, is performed the following operations: the entity that statistics currently traverses
With the co-occurrence feature of the feature of the entity in the historical search sentence of search log recording, the frequency for example, occurred jointly
(abbreviation co-occurrence frequency) or the number occurred jointly (abbreviation co-occurrence number), the popular degree of the feature as the entity；It will be popular
The feature that degree is greater than given threshold is retained in the knowledge mapping as the popular feature of the entity, and is filtered out and described known
Know the non-popular feature of the entity in map.For example, sharing 5 features (respectively for this entity of someone
One feature is to fifth feature), and each feature in this 5 features and the entity are in the historical search language of search log recording
Co-occurrence frequency in sentence is respectively as follows: 0.9,0.7,0.65,0.5 and 0.3.Assuming that given threshold is 0.6, then in 5 features
Fisrt feature, second feature and third feature be popular feature.Such as: the popular feature of entity " Liu Dehua " has: constellation,
Wife and daughter；And the popular feature of entity " Yao Ming " has: height, occupation, wife and daughter.
Operation 230 receives the search statement that user is inputted by character input modes or voice input mode, identification search
The searching entities for including in sentence.
It operates 240, according to entity associated network of personal connections, obtains the N rank related entities and each rank related entities of searching entities
Rationale for the recommendation.
Wherein, the first rank related entities in N rank related entities are the entity for having direct correlation relationship with searching entities,
I-th rank related entities are the entity for having direct correlation relationship with the (i-1)-th rank related entities, and N is the natural number greater than 1, and i is greater than
2 and be less than or equal to N.The rationale for the recommendation of any rank related entities is obtained according to the generation path computing of the rank related entities
's.Specifically, the corresponding rationale for the recommendation of the i-th rank related entities can be obtained according to following content determination: order in N rank related entities
There are in related entities less than i with the i-th rank related entities the related entities of direct correlation relationship and/or indirect association relationship,
And corresponding incidence relation.
With the increase of the order of related entities, the feature of high-order related entities may be more and more sparse.As this reality
A kind of preferred embodiment of example is applied, the order N value of related entities is 2.Specifically, according to the entity associated being pre-created
Network of personal connections obtains the N rank related entities of searching entities, comprising:
The feature of query search entity from entity associated network of personal connections, to obtain single order feature；
If the value of single order feature belongs to entity type, using the value of single order feature as searching entities, real with search
Body has the first rank related entities of direct correlation relationship, and is managed according to the recommendation that single order feature generates the first rank related entities
By；
The feature of the first rank related entities is inquired, from entity associated network of personal connections to obtain second order feature；
If the value of second order feature belongs to entity type, using the value of second order feature as searching entities, same first rank
Related entities have the second-order related entities of direct correlation relationship, and according to single order feature, the value of single order feature and second order
Feature generates the rationale for the recommendation of second-order related entities.
For the clearer acquisition scheme for illustrating related entities provided in this embodiment and its rationale for the recommendation, now illustrate
Explanation.For example, searching entities are " blame sincere not faze ",
(1) the popular feature of " blame sincere not faze " is inquired from entity associated network of personal connections are as follows: " host " and " welcome guest " makees
For single order feature；
Correspondingly, the value of single order feature " host " is " Meng Fei ", the value of single order feature " welcome guest " is " Huang Han ", " peaceful wealth
Mind " etc.；
Therefore, can the first rank by " Meng Fei ", " Huang Han " to " the peaceful mammon " as searching entities " blame sincere not faze " it is related
Entity；
(2) single order characteristic value " Meng Fei ", " Huang Han ", " the peaceful mammon " these three entities are inquired from entity associated network of personal connections
Popular feature be respectively " daughter ", " husband " and " wife ", as second order feature；
Correspondingly, the value of the popular feature " daughter " of " Meng Fei " is " Meng Xingya ", the popular feature " husband " of " Huang Han "
Value is " Yuan Jian "；The value of the popular feature " wife " of " the peaceful mammon " is " Cheng Jiaoe "；
It therefore, can be related as the second-order of searching entities " blame sincere not faze " by " Meng Xingya ", " Yuan Jian ", " Cheng Jiaoe "
Entity.
Wherein, the rationale for the recommendation of the first rank related entities " Meng Fei " is " host "；First rank related entities " Huang Han " and
The rationale for the recommendation of " the peaceful mammon " is " welcome guest "；
The rationale for the recommendation of second-order related entities " Meng Xingya " is " daughter of host Meng Fei "；Second-order related entities
The rationale for the recommendation of " Yuan Jian " is " husband of welcome guest Huang Han "；The rationale for the recommendation of second-order related entities " Cheng Jiaoe " is that " welcome guest is peaceful
The wife of the mammon ".
The rationale for the recommendation for operating 250, each rank related entities and each rank related entities that will acquire is showed.
Technical solution provided in this embodiment, on the one hand using by legitimate verification knowledge mapping as lookup library, from
The related entities to be recommended are obtained in the lookup library, it can be ensured that the controllability of related entities in terms of content；Another party
Face can also go out rationale for the recommendation automatically according to the generation path computing of related entities, show user, recommend with human-edited
The scheme of reason is compared, and the formation efficiency of rationale for the recommendation is substantially increased.
Embodiment three
Fig. 3 is a kind of flow diagram for entity recommended method that the embodiment of the present invention three provides.The present embodiment is above-mentioned
On the basis of embodiment one and embodiment two, before each rank related entities that will acquire are showed, increase to acquisition
The operation that each rank related entities are ranked up.Referring to Fig. 3, entity recommended method provided in this embodiment specifically includes following behaviour
Make:
Operation 310 receives the search statement that user is inputted by character input modes or voice input mode, identification search
The searching entities for including in sentence.
The entity associated network of personal connections that operation 320, basis are pre-created, obtains the N rank related entities of searching entities.
It operates 330, according to the historical search sentence of search log recording, each rank related entities of acquisition is ranked up.
In the present embodiment, since each rank related entities number got might not only one, for example, the first rank
Related entities number be 10, second-order related entities number be 8, so can according to search log recording historical search sentence,
The each related entities for belonging to phase same order in N rank related entities are ranked up.
It illustratively, can be according to the first rank related entities and searching entities in search log recording in the case where N is 2
Historical search sentence in co-occurrence feature, each first rank related entities are ranked up；According to same second-order related entities
The first rank related entities and searching entities being total in the historical search sentence of search log recording with the relationship of direct correlation
Existing feature, is ranked up each second-order related entities.Preferably, the ranking of all second-order related entities wants forward in
The ranking of single order related entities.
Under normal conditions, often search user is relatively more well known and searching entities have relatively strong pass for the first rank related entities
The entity of connection degree, but known to entity, be less susceptible to cause search user's note that therefore following sequence rule can be used
Then, be ranked up to each first rank related entities: the value of co-occurrence feature is higher, and ranking is more rearward.But due to second-order phase
Closing entity is the entity for having indirect association relationship with searching entities, generally can be seldom searched known to user, and consider
It is not answered to the degree of association between searching entities too small yet, therefore following ordering rule can be used, to each second-order related entities
Be ranked up: the first rank related entities and searching entities for having direct correlation relationship with second-order related entities are in search log
The value of co-occurrence feature in the historical search sentence of record is higher, and ranking is more forward.
For example, searching entities are " blame sincere not faze ", the first rank related entities have " Meng Fei ", " Huang Han " and " the peaceful mammon ", right
The second-order related entities answered have " Meng Xingya ", " Yuan Jian " and " Cheng Jiaoe "." Meng Fei ", " Huang Han " and " the peaceful mammon " with " non-really
Do not disturb " search log recording historical search sentence in co-occurrence frequency be followed successively by 0.6,0.4 and 0.7, then these three first
The ranking results of rank related entities are as follows: " Huang Han ", Meng Fei " and " the peaceful mammon ".Also, " Meng Xingya ", " Yuan Jian " and " Cheng Jiaoe "
Ranking results, be according to " Meng Fei ", " Huang Han " and " the peaceful mammon " and " blame sincere not faze " in the historical search for searching for log recording
What the sequence determination of co-occurrence frequency from high to low in sentence obtained, are as follows: " Cheng Jiaoe ", " Meng Xingya " and " Yuan Jian ".
Certainly, it can also realize by other means according to each rank of the historical search sentence to acquisition for searching for log recording
The operation that related entities are ranked up.For example, not being ranked up respectively for each rank related entities, but unifies basis and search
Co-occurrence feature of the Suo Shiti in the historical search sentence of search log recording, to each correlation in the N rank related entities
Entity is ranked up.For example, " Meng Fei ", " Huang Han ", " the peaceful mammon ", " Meng Xingya ", " Yuan Jian " and " Cheng Jiaoe ", with searching entities
Co-occurrence frequency in the historical search sentence of search log recording is successively are as follows: and 0.5,0.45,0.8,0.2,0.6,0.72, then it arranges
Sequence result are as follows: " the peaceful mammon ", " Cheng Jiaoe ", " Yuan Jian ", " Meng Fei ", " Huang Han " and " Meng Xingya ".
Alternatively, estimating search based on ctr (Click-Through-Rate, the clicking rate) predictive algorithm being pre-created and using
Clicking rate of the family to each related entities in the N rank related entities；And then according to the estimation results and search log recording
Historical search sentence, each related entities are ranked up.
Operate 340, according to ranking results, each rank related entities that will acquire are showed.
On the basis of above-described embodiment one and embodiment two, it is preferred that entity recommended method provided in this embodiment,
After the N rank related entities for obtaining searching entities, further includes:
It is more than the thresholding set in advance for the rank related entities for each rank related entities got, such as sporocarp number
Value is then based on preset screening rule, carries out Screening Treatment to the rank related entities, so that the rank filtered out is related real
The number of body is less than or equal to the threshold value.
Preferably, after each rank related entities that will acquire are showed, further includes: according to real to the correlation showed
The clicking rate of body corrects the screening rule.For example, the screening rule is wheel exhibition rule, namely every rank phase is first selected at random
The related entities that number in entity meets the threshold value are closed, are showed；If related to some in current rank related entities
After the clicking rate of entity is lower than preset superseded threshold value, then when showing next time, it is related real to be replaced with current rank
The related entities that other in body were not demonstrated.
Example IV
Fig. 4 is a kind of flow diagram for entity recommended method that the embodiment of the present invention four provides.The present embodiment is with above-mentioned
Based on all embodiments, a preferred embodiment is provided.Referring to fig. 4, entity recommended method provided in this embodiment specifically includes
Following operation:
Operation 410 receives the search statement that user is inputted by character input modes or voice input mode, identification search
The searching entities for including in sentence.
Operation 420, the input for recommending computation model using the searching entities as entity recommend computation model based on the entity
Carry out the recommendation of related entities.Specifically recommendation process includes:
The feature of query search entity from knowledge mapping, to obtain single order feature；If the value of single order feature belongs to reality
Body type, then using the value of single order feature as searching entities, have the first rank of direct correlation relationship related with searching entities
Entity, and according to the rationale for the recommendation of single order feature the first rank related entities of generation；
The feature of the first rank related entities is inquired, from knowledge mapping to obtain second order feature；If the value of second order feature
Belong to entity type, then there is direct correlation relationship using the value of second order feature as searching entities, same first rank related entities
Second-order related entities generate second-order related entities and according to single order feature, the value of single order feature and second order feature
Rationale for the recommendation.
Wherein, the knowledge mapping is original to know previously according to the historical search sentence of search log recording to what is got
Know the new map obtained after the popular Feature Selection operation of map progress entity.Specific screening operation can be found in above-described embodiment
To the explanation of operation 220 in two, details are not described herein.
Operation 430, the entity recommended models obtain recommendation results and are exported.
Wherein, recommendation results include: the rationale for the recommendation of the first rank related entities and the first rank related entities, second-order phase
Close the rationale for the recommendation of entity and second-order related entities.
After obtaining recommendation results, based on the ctr predictive algorithm being pre-created, search user is estimated to two rank related entities
In each related entities clicking rate；And then according to the estimation results and the historical search sentence of search log recording, to each
A related entities are ranked up.
It is the controllable related entities of content that technical solution provided in this embodiment, which enables to recommended entity, and can be mentioned
Relevance between high recommended entity and search statement, automatically calculates out rationale for the recommendation.
Embodiment five
Fig. 5 is a kind of structural schematic diagram for entity recommendation apparatus that the embodiment of the present invention five provides.The device can be by soft
Part is realized, is integrated in and is provided in the search engine of search service for miscellaneous terminal device, for guiding user's discovery true
Just it is intended to or the related interests of user is caused to search again for.Specific structure referring to Fig. 5, the entity recommendation apparatus is as follows:
Searching entities recognition unit 510 is inputted by character input modes or voice input mode for receiving user
Search statement identifies the searching entities for including in described search sentence；
Related entities acquiring unit 520, for obtaining described search entity according to the entity associated network of personal connections being pre-created
N rank related entities, wherein the first rank related entities in the N rank related entities are to have directly pass with described search entity
The entity of connection relationship, the i-th rank related entities are the entity for having direct correlation relationship with the (i-1)-th rank related entities, and the N is big
In 1 natural number, the i is greater than 2 and is less than or equal to N；
Show unit 530, each rank related entities for obtaining the related entities acquiring unit show.
Illustratively, entity recommendation apparatus provided in this embodiment further include:
Rationale for the recommendation acquiring unit 525, for obtaining the rationale for the recommendation of each rank related entities in the N rank related entities,
The rationale for the recommendation of any rank related entities is obtained according to the generation path computing of the rank related entities；
It is described to show unit 530, specifically for the rationale for the recommendation of each rank related entities and each rank related entities that will acquire
Showed.
Illustratively, entity recommendation apparatus provided in this embodiment further include:
Knowledge mapping acquiring unit 500, for obtaining the N of described search entity in the related entities acquiring unit 520
Before rank related entities, knowledge mapping is obtained, wherein the knowledge mapping includes: the feature of at least one entity and entity；
Network of personal connections creating unit 505, for creating entity associated network of personal connections according to the knowledge mapping.
Illustratively, the network of personal connections creating unit 505, is specifically used for:
By the knowledge mapping, as the entity associated network of personal connections；Or
According to the historical search sentence of search log recording, the popular degree of the feature of entity in the knowledge mapping is excavated,
To carry out Feature Selection；By the knowledge mapping after Feature Selection, as entity associated network of personal connections.
Illustratively, the related entities acquiring unit 520, is specifically used for:
The feature of described search entity is inquired, from the entity associated network of personal connections to obtain single order feature；
If the value of the single order feature belongs to entity type, using the value of the single order feature as described search entity
, with described search entity have direct correlation relationship the first rank related entities, and according to the single order feature generate described in
The rationale for the recommendation of first rank related entities；
The feature of the first rank related entities is inquired, from the entity associated network of personal connections to obtain second order feature；
If the value of the second order feature belongs to entity type, using the value of the second order feature as described search entity
, the second-order related entities with the first rank related entities with direct correlation relationship, and according to the single order feature, institute
The value and the second order feature for stating single order feature, generate the rationale for the recommendation of the second-order related entities.
Based on the above technical solution, entity recommendation apparatus provided in this embodiment further include:
Sequencing unit 528, for the historical search sentence according to search log recording, to the related entities acquiring unit
Each rank related entities obtained are ranked up.
Based on the above technical solution, entity recommendation apparatus provided in this embodiment further include:
Related entities screening unit 526, each rank for getting for the related entities acquiring unit 520 are related real
Body is then based on preset screening rule if sporocarp number is more than the threshold value set in advance for the rank related entities, right
The rank related entities carry out Screening Treatment, so that the number of the rank related entities filtered out is less than or equal to the threshold value；
Screening rule amending unit 540, for according to the click for showing the related entities that unit 530 is showed
Rate corrects the screening rule.
Method provided by any embodiment of the invention can be performed in the said goods, has the corresponding functional module of execution method
And beneficial effect.
Note that the above is only a better embodiment of the present invention and the applied technical principle.It will be appreciated by those skilled in the art that
The invention is not limited to the specific embodiments described herein, be able to carry out for a person skilled in the art it is various it is apparent variation,
It readjusts and substitutes without departing from protection scope of the present invention.Therefore, although being carried out by above embodiments to the present invention
It is described in further detail, but the present invention is not limited to the above embodiments only, without departing from the inventive concept, also
It may include more other equivalent embodiments, and the scope of the invention is determined by the scope of the appended claims.
Claims (12)
1. a kind of entity recommended method characterized by comprising
The search statement that user is inputted by character input modes or voice input mode is received, identifies and is wrapped in described search sentence
The searching entities contained；
According to the entity associated network of personal connections being pre-created, the N rank related entities of described search entity are obtained, wherein the N rank phase
Closing the first rank related entities in entity is the entity for having direct correlation relationship with described search entity, the i-th rank related entities
To have the entity of direct correlation relationship with the (i-1)-th rank related entities, the N is natural number greater than 1, and the i is equal to or greatly
In 2 and be less than or equal to N；Each rank related entities that will acquire are showed；
Wherein, before the N rank related entities for obtaining described search entity, further includes:
Knowledge mapping is obtained, wherein the knowledge mapping includes: the feature of at least one entity and entity；
According to search log recording historical search sentence, excavate the popular degree of the feature of entity in the knowledge mapping, with into
Row Feature Selection；By the knowledge mapping after Feature Selection, as entity associated network of personal connections.
2. the method according to claim 1, wherein further include: it is related to obtain each rank in the N rank related entities
The rationale for the recommendation of entity, the rationale for the recommendation of any rank related entities are obtained according to the generation path computing of the rank related entities
It arrives；
Each rank related entities that will acquire are showed, comprising: each rank related entities that will acquire and each rank related entities push away
Reason is recommended to be showed.
3. the method according to claim 1, wherein obtaining institute according to the entity associated network of personal connections being pre-created
State the N rank related entities of searching entities, comprising:
The feature of described search entity is inquired, from the entity associated network of personal connections to obtain single order feature；
If the value of the single order feature belongs to entity type, using the value of the single order feature as described search entity,
There are the first rank related entities of direct correlation relationship with described search entity, and generate described first according to the single order feature
The rationale for the recommendation of rank related entities；
The feature of the first rank related entities is inquired, from the entity associated network of personal connections to obtain second order feature；
If the value of the second order feature belongs to entity type, using the value of the second order feature as described search entity,
There are the second-order related entities of direct correlation relationship with the first rank related entities, and according to the single order feature, described
The value of single order feature and the second order feature, generate the rationale for the recommendation of the second-order related entities.
4. method according to any one of claim 1-3, which is characterized in that each rank related entities that will acquire are opened up
Before now, further includes:
According to the historical search sentence of search log recording, each rank related entities of acquisition are ranked up.
5. method according to any one of claim 1-3, which is characterized in that in the N rank phase for obtaining described search entity
After the entity of pass, further includes:
It is more than the threshold value set in advance for the rank related entities for each rank related entities got, such as sporocarp number,
It is then based on preset screening rule, Screening Treatment is carried out to the rank related entities, so that the rank related entities filtered out
Number be less than or equal to the threshold value.
6. according to the method described in claim 5, it is characterized in that, after each rank related entities that will acquire are showed,
Further include: according to the clicking rate to the related entities showed, correct the screening rule.
7. a kind of entity recommendation apparatus characterized by comprising
Searching entities recognition unit, the search phrase inputted for receiving user by character input modes or voice input mode
Sentence identifies the searching entities for including in described search sentence；
Related entities acquiring unit, for obtaining the N rank phase of described search entity according to the entity associated network of personal connections being pre-created
Entity is closed, wherein the first rank related entities in the N rank related entities are to have direct correlation relationship with described search entity
Entity, the i-th rank related entities be with the (i-1)-th rank related entities have direct correlation relationship entity, the N be greater than 1
Natural number, the i are equal to or more than 2 and are less than or equal to N；Show unit, for obtaining the related entities acquiring unit
Each rank related entities showed；
Described device further include:
Knowledge mapping acquiring unit, for obtaining the N rank related entities of described search entity in the related entities acquiring unit
Before, knowledge mapping is obtained, wherein the knowledge mapping includes: the feature of at least one entity and entity；
Network of personal connections creating unit excavates entity in the knowledge mapping for the historical search sentence according to search log recording
Feature popular degree, to carry out Feature Selection；By the knowledge mapping after Feature Selection, as entity associated network of personal connections.
8. device according to claim 7, which is characterized in that further include:
Rationale for the recommendation acquiring unit, it is any for obtaining the rationale for the recommendation of each rank related entities in the N rank related entities
The rationale for the recommendation of rank related entities is obtained according to the generation path computing of the rank related entities；
It is described to show unit, it is opened up specifically for the rationale for the recommendation of each rank related entities and each rank related entities that will acquire
It is existing.
9. device according to claim 7, which is characterized in that the related entities acquiring unit is specifically used for:
The feature of described search entity is inquired, from the entity associated network of personal connections to obtain single order feature；
If the value of the single order feature belongs to entity type, using the value of the single order feature as described search entity,
There are the first rank related entities of direct correlation relationship with described search entity, and generate described first according to the single order feature
The rationale for the recommendation of rank related entities；
The feature of the first rank related entities is inquired, from the entity associated network of personal connections to obtain second order feature；
If the value of the second order feature belongs to entity type, using the value of the second order feature as described search entity,
There are the second-order related entities of direct correlation relationship with the first rank related entities, and according to the single order feature, described
The value of single order feature and the second order feature, generate the rationale for the recommendation of the second-order related entities.
10. the device according to any one of claim 7-9, which is characterized in that further include:
Sequencing unit obtains the related entities acquiring unit for the historical search sentence according to search log recording
Each rank related entities are ranked up.
11. the device according to any one of claim 7-9, which is characterized in that further include:
Related entities screening unit, each rank related entities for being got for the related entities acquiring unit, such as fruit
Body number is more than then to be based on preset screening rule in advance for the threshold value of rank related entities setting, to rank correlation
Entity carries out Screening Treatment, so that the number of the rank related entities filtered out is less than or equal to the threshold value.
12. device according to claim 11, which is characterized in that further include:
Screening rule amending unit, for according to the clicking rate for showing the related entities that unit is showed, described in amendment
Screening rule.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510434989.1A CN105095433B (en) | 2015-07-22 | 2015-07-22 | Entity recommended method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510434989.1A CN105095433B (en) | 2015-07-22 | 2015-07-22 | Entity recommended method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105095433A CN105095433A (en) | 2015-11-25 |
| CN105095433B true CN105095433B (en) | 2019-07-05 |
Family
ID=54575870
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510434989.1A Active CN105095433B (en) | 2015-07-22 | 2015-07-22 | Entity recommended method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105095433B (en) |
Families Citing this family (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106095858A (en) * | 2016-06-02 | 2016-11-09 | 海信集团有限公司 | A kind of audio video searching method, device and terminal |
| CN106528610A (en) * | 2016-09-28 | 2017-03-22 | 厦门理工学院 | Knowledge graph representation learning method based on path tensor decomposition |
| CN106547887B (en) * | 2016-10-27 | 2020-04-07 | 北京百度网讯科技有限公司 | Search recommendation method and device based on artificial intelligence |
| CN108536702B (en) * | 2017-03-02 | 2022-12-02 | 腾讯科技（深圳）有限公司 | Method and device for determining related entities and computing equipment |
| CN107169010A (en) * | 2017-03-31 | 2017-09-15 | 北京奇艺世纪科技有限公司 | A kind of determination method and device of recommendation search keyword |
| CN107885875B (en) * | 2017-11-28 | 2022-07-08 | 北京百度网讯科技有限公司 | Synonymy transformation method and device for search words and server |
| CN108509479B (en) * | 2017-12-13 | 2022-02-11 | 深圳市腾讯计算机***有限公司 | Entity recommendation method and device, terminal and readable storage medium |
| CN108334632B (en) * | 2018-02-26 | 2021-03-23 | 深圳市腾讯计算机***有限公司 | Entity recommendation method and device, computer equipment and computer-readable storage medium |
| CN108416051A (en) * | 2018-03-20 | 2018-08-17 | 杭州声讯网络科技有限公司 | A kind of Intelligent dialogue logical algorithm for sector application |
| CN108829854B (en) * | 2018-06-21 | 2021-08-31 | 北京百度网讯科技有限公司 | Method, apparatus, device and computer-readable storage medium for generating article |
| CN110717011B (en) * | 2018-06-27 | 2022-08-16 | 北京百度网讯科技有限公司 | Session message processing method and device |
| CN109582878A (en) * | 2018-11-05 | 2019-04-05 | 咪咕文化科技有限公司 | Method and device for realizing search prompt and computer readable storage medium |
| CN110196947A (en) * | 2019-05-31 | 2019-09-03 | 三角兽（北京）科技有限公司 | Method, apparatus, electronic equipment and the storage medium that recommendation information determines |
| CN110175242B (en) * | 2019-05-31 | 2024-05-17 | 嗷呜世界(深圳)网络科技有限公司 | Human-computer interaction association method, device and medium based on knowledge graph |
| CN110245357B (en) * | 2019-06-26 | 2023-05-02 | 北京百度网讯科技有限公司 | Main entity identification method and device |
| CN110457452B (en) * | 2019-07-08 | 2022-06-14 | 汉海信息技术(上海)有限公司 | Recommendation reason generation method and device, electronic equipment and readable storage medium |
| CN112732926A (en) * | 2020-12-31 | 2021-04-30 | 平安资产管理有限责任公司 | Text retrieval method and device, computer equipment and storage medium |
| CN116888590A (en) * | 2021-03-29 | 2023-10-13 | 深圳市欢太科技有限公司 | Entity heat generation method and device, storage medium and electronic equipment |
| CN118051672A (en) * | 2024-03-15 | 2024-05-17 | 广州凡拓数字创意科技股份有限公司 | Multi-target dynamic recommendation method and device based on knowledge graph |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101441667A (en) * | 2008-12-29 | 2009-05-27 | 北京搜狗科技发展有限公司 | Music recommend method and apparatus |
| CN103294778A (en) * | 2013-05-13 | 2013-09-11 | 百度在线网络技术（北京）有限公司 | Method and system for pushing messages |
| CN104484339A (en) * | 2014-11-21 | 2015-04-01 | 百度在线网络技术（北京）有限公司 | Method and system for recommending relevant entities |
| CN104503978A (en) * | 2014-11-26 | 2015-04-08 | 百度在线网络技术（北京）有限公司 | Related entity recommending method and system |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10223411B2 (en) * | 2013-03-06 | 2019-03-05 | Nuance Communications, Inc. | Task assistant utilizing context for improved interaction |
-
2015
- 2015-07-22 CN CN201510434989.1A patent/CN105095433B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101441667A (en) * | 2008-12-29 | 2009-05-27 | 北京搜狗科技发展有限公司 | Music recommend method and apparatus |
| CN103294778A (en) * | 2013-05-13 | 2013-09-11 | 百度在线网络技术（北京）有限公司 | Method and system for pushing messages |
| CN104484339A (en) * | 2014-11-21 | 2015-04-01 | 百度在线网络技术（北京）有限公司 | Method and system for recommending relevant entities |
| CN104503978A (en) * | 2014-11-26 | 2015-04-08 | 百度在线网络技术（北京）有限公司 | Related entity recommending method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105095433A (en) | 2015-11-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105095433B (en) | Entity recommended method and device | |
| CN109241538B (en) | Chinese entity relation extraction method based on dependency of keywords and verbs | |
| Ben-David et al. | PADA: Example-based Prompt Learning for on-the-fly Adaptation to Unseen Domains | |
| CN102708100B (en) | Method and device for digging relation keyword of relevant entity word and application thereof | |
| CN104102723B (en) | Search for content providing and search engine | |
| US8312022B2 (en) | Search engine optimization | |
| CN104484339B (en) | A kind of related entities recommend method and system | |
| CN106776881A (en) | A kind of realm information commending system and method based on microblog | |
| CN107133345A (en) | Exchange method and device based on artificial intelligence | |
| Liu et al. | Relational memory-augmented language models | |
| WO2012000013A1 (en) | Method and system for determining word senses by latent semantic distance | |
| CN109255012B (en) | Method and device for machine reading understanding and candidate data set size reduction | |
| CN104978332A (en) | UGC label data generating method, UGC label data generating device, relevant method and relevant device | |
| CN104281565A (en) | Semantic dictionary constructing method and device | |
| CN110309271A (en) | Intelligent knowledge study and question and answer technology | |
| Marujo et al. | Hourly traffic prediction of news stories | |
| Dubuisson Duplessis et al. | Utterance retrieval based on recurrent surface text patterns | |
| CN107784019A (en) | Word treatment method and system are searched in a kind of searching service | |
| CN109472032A (en) | A kind of determination method, apparatus, server and the storage medium of entity relationship diagram | |
| Kumar et al. | Meeting summarization: A survey of the state of the art | |
| Mahalakshmi et al. | An effective multilingual retrieval with query optimization using deep learning technique | |
| Kwak et al. | Interactive Story Maker: Tagged Video Retrieval System for Video Re-creation Service | |
| Chen | A cost-effective solution for blog search | |
| Wanderley et al. | Knowledge discovery in task-oriented dialogue | |
| CN115795057B (en) | Audit knowledge processing method and system based on AI technology |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |