CN108319603A - Object recommendation method and apparatus - Google Patents
Object recommendation method and apparatus Download PDFInfo
- Publication number
- CN108319603A CN108319603A CN201710031775.9A CN201710031775A CN108319603A CN 108319603 A CN108319603 A CN 108319603A CN 201710031775 A CN201710031775 A CN 201710031775A CN 108319603 A CN108319603 A CN 108319603A
- Authority
- CN
- China
- Prior art keywords
- candidate target
- word
- query
- historical
- score value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/3332—Query translation
- G06F16/3338—Query expansion
-
- 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)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention relates to a kind of object recommendation method and apparatus, the method includes:Obtain the query word inputted in search box;Determine candidate target set corresponding with the query word;Each candidate target is collected from historical query selected object in the candidate target set;The matching degree of number and each candidate target and the query word is chosen according to each corresponding history of candidate target in the candidate target set, obtains the corresponding recommendation score value of each candidate target;The candidate target for meeting recommendation condition is picked out from the candidate target set according to the recommendation score value;Recommend the candidate target picked out.Object recommendation method provided by the present application improves the accuracy of recommendation.
Description
Technical field
The present invention relates to field of computer technology, more particularly to a kind of object recommendation method and apparatus.
Background technology
With the development of computer technology, daily life is combined more and more closer with internet, more and more
User know various information using network.Internet product nowhere loses in life, and internet product is applied to network
After in equipment, recommendation function is essential application in internet product.
Traditional recommendation function is usually the object for inputting or selecting according to user, and associated popular object recommendation is given
User.However, it is not high using the object of traditional recommendation method recommendation and the degree of association of user demand, it can not be accurately located
The demand of user causes recommendation results accuracy low.
Invention content
Based on this, it is necessary to for the low problem of traditional recommendation method recommendation results accuracy, provide a kind of recommendation
Content processing method and device.
A kind of recommendation processing method, the method includes:
Obtain the query word inputted in search box;
Determine candidate target set corresponding with the query word;In the candidate target set each candidate target collected from
Historical query selected object;
Number and each candidate target are chosen according to each corresponding history of candidate target in the candidate target set
With the matching degree of the query word, the corresponding recommendation score value of each candidate target is obtained;
The candidate target for meeting recommendation condition is picked out from the candidate target set according to the recommendation score value;
Recommend the candidate target picked out.
A kind of object recommendation device, described device include:
Query word acquisition module, for obtaining the query word inputted in search box;
Candidate target determining module, for determining candidate target set corresponding with the query word;The candidate target
Each candidate target is collected from historical query selected object in set;
Recommend score value computing module, it is secondary for being chosen according to each corresponding history of candidate target in the candidate target set
The matching degree of number and each candidate target and the query word obtains the corresponding recommendation score value of each candidate target;
Choosing module meets recommendation condition for being picked out from the candidate target set according to the recommendation score value
Candidate target;
Recommending module, for recommending the candidate target picked out.
Above-mentioned object recommendation method and apparatus build candidate target set according to historical query selected object, are getting
After the query word inputted in search box, candidate target set corresponding with the query word of input is determined, by candidate target set
In historical query selected object recall, number and each candidate are chosen with each corresponding history of candidate target in candidate target set
The matching degree of object and query word is foundation, calculates the corresponding recommendation score value of each candidate target, will meet the candidate of recommendation condition
Object is recommended.This to be fed back according to user's history behavioral data, the recommendation method of analog subscriber real behavior can be compared with
The demand for adequately positioning user, to improve the accuracy of recommendation.
Description of the drawings
Fig. 1 is the applied environment figure of object recommendation method in one embodiment;
Fig. 2 is the internal structure schematic diagram of the server for realizing object recommendation method in one embodiment;
Fig. 3 is the flow diagram of object recommendation method in one embodiment;
Fig. 4 is flow diagram the step of determining candidate target set corresponding with query word in one embodiment;
Fig. 5 is flow diagram the step of updating candidate target set corresponding with query word in one embodiment;
Fig. 6 be number chosen according to the corresponding history of each candidate target in candidate target set in one embodiment, and
The matching degree of each candidate target and query word obtains the flow diagram of each candidate target corresponding the step of recommending score value;
Fig. 7 is the step of obtaining the matching degree of each candidate target and query word in candidate target set in one embodiment
Flow diagram;
Fig. 8 is to choose number according to the corresponding history of each candidate target in candidate target set in another embodiment, with
And the matching degree of each candidate target and query word, obtain the flow diagram of each candidate target corresponding the step of recommending score value;
Fig. 9 is the schematic diagram that terminal recommends interface in one embodiment;
Figure 10 is the flow diagram of object recommendation method in another embodiment;
Figure 11 is the logic diagram based on user's history behavioral data feedback in one embodiment;
Figure 12 is the statistical Butut that server carries out that user when object recommendation selects recommended;
Figure 13 is that user input query word is averaged the statistical Butut of number of characters when server carries out object recommendation;
Figure 14 is the structure diagram of object recommendation device in one embodiment;
Figure 15 is the structure diagram of object recommendation device in another embodiment;
Figure 16 is the structure diagram of object recommendation device in another embodiment.
Specific implementation mode
In order to make the purpose , technical scheme and advantage of the present invention be clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
Fig. 1 is the applied environment figure of object recommendation method in one embodiment.Referring to Fig.1, the object recommendation method application
In object recommendation system.Object recommendation system includes terminal 110 and server 120, and terminal 110 passes through network and server 120
Connection.Terminal 110 can be specifically terminal console or mobile terminal, and mobile terminal specifically can be with mobile phone, tablet computer, notebook
At least one of computer etc..Server 120 can be specifically independent physical server, can also be physical server collection
Group.
Fig. 2 is the internal structure schematic diagram of server in one embodiment.As shown in Fig. 2, the server can be used as in Fig. 1
Server 120 to implement a kind of object recommendation method.The server includes the processor connected by system bus, non-volatile
Property storage medium, built-in storage and network interface.Wherein, the non-volatile memory medium of the server be stored with operating system,
Database and object recommendation device are stored with historical query selected object etc. in database, the object recommendation device for realizing
A kind of object recommendation method suitable for server.For the processor of the server for providing calculating and control ability, support is whole
The operation of a server.The built-in storage of the server provides for the operation of the object recommendation device in non-volatile memory medium
Environment can store computer-readable instruction in the built-in storage, can when which is executed by the processor
So that the processor executes a kind of object recommendation method.The network interface of the server with external terminal for passing through according to this
Network connection communicates, for example obtains the query word etc. inputted in terminal.Server can be either multiple with independent server
The server cluster of server composition is realized.It will be understood by those skilled in the art that structure shown in Figure 2, only with
The block diagram of the relevant part-structure of application scheme, does not constitute the limit for the server being applied thereon to application scheme
Fixed, specific server may include either combining certain components or with not than more or fewer components as shown in the figure
Same component arrangement.
As shown in figure 3, in one embodiment, providing a kind of object recommendation method, the present embodiment is applied in this way
Server 120 in above-mentioned Fig. 1 illustrates.This method specifically comprises the following steps:
S302 obtains the query word inputted in search box.
Wherein, query word refers to the character string for being inquired.Specifically, it can be run with function of search in terminal
Client, the client are configurable to the search box of input inquiry word.Client may be connected to and the client at runtime
Corresponding server, and with server interaction data.Client such as map client or webpage client etc..Terminal can be examined
The input operation for acting on search box is surveyed, after detecting input operation, the query word inputted in acquisition search box will obtain
The query word be uploaded to server.
S304 determines candidate target set corresponding with query word;Each candidate target is collected from going through in candidate target set
History inquires selected object.
Wherein, candidate target set is the set that several recommendable objects are constituted.Object tool in candidate target set
Body can be point of interest, commodity or application program etc..Point of interest (point of interest POI) refers to can be on map
The object positioned, a point of interest include at least title, classification and geographical location.Commodity such as virtual goods or entity
Commodity etc..The object that historical query selected object refers to user to be selected in historical query operation can be server according to
The object that the query word of family input is recommended can also be that the direct search entrance precise search that is provided by terminal of user obtains
Object.Server can after each user is inquired by terminal, the query word of respective record this inquiry operation input and
The object chosen after inquiry is stored as user's history behavioral data into database or caching.
Specifically, server can build candidate target set according to the user's history behavioral data of storage.In the present embodiment
In, server can by history input query word be segmented to obtain prefix set of words, then by history input query word with
It segments obtained prefix word and corresponds to storage, structure and each one-to-one prefix source objects collection of prefix word in prefix set of words
It closes.Again by the storage corresponding with the prefix word that participle obtains of the corresponding historical query selected object of query word of history input, structure
With the one-to-one historical query selected object set of each prefix word in prefix set of words.
In one embodiment, server can also be looked into history using historical query selected object as prefix source objects
It askes selected object to be segmented to obtain prefix word, the prefix word segmented according to historical query selected object is added to prefix
In set of words, and historical query selected object is added in prefix source objects set corresponding with corresponding prefix word.
Server can traverse each prefix word in prefix set of words after the query word for obtaining input, in traversal, will traverse
Prefix word and the query word of acquisition compare, if the prefix word of traversal is consistent with the query word of acquisition, by the prefix word of traversal
Corresponding historical query selected object set is as candidate target set.
In one embodiment, the candidate target in the recommended set involved in object recommendation method includes that prompt is looked into
Ask word and point of interest.Prompting query word refers to the character string for prompting input, such as " Peking University " or " Peking University north
Door " etc..
It illustrates, it is assumed that the client for carrying out query word input is map client, and user is defeated in search box
" Peking University north gate " point of interest is had selected after entering " north gate ".So wherein historical query word is " north gate ", and historical query is chosen
Object is point of interest " Peking University north gate ".The prefix word that server is segmented " north gate " can be " north " and " north
Word segmentation result " north " and " north gate ", are then added to as set element in prefix set of words by door ", then by query word " north gate "
It is added to prefix source objects set corresponding with prefix word " north " and " north gate ", and correspondingly by point of interest " Peking University north
Door " is added to history selected object set corresponding with prefix word " north " and " north gate ".
Server is when it is query word to get user with " north " or " north gate ", based on user's history behavioral data and generally
Rate principle, it is reason to believe that user is in input " north " or " north gate ", it is intended that the object of search is point of interest " Peking University
Point of interest " Peking University north gate " is also recommended user by north gate " as recommended.
Further, server also can be by prefix word " north " that point of interest " Peking University north gate " is segmented, " north
Capital ", " Beijing is big ", " Peking University ", " Peking University north " and " Peking University north gate ".Search based on key sequence input
Custom, it is reason to believe that user is in input " north ", " Beijing ", " Beijing is big ", " Peking University ", " Peking University north " or " north
When capital university north gate ", it is intended that the object of search is point of interest " Peking University north gate ", also i.e. by point of interest " Peking University north gate "
User is recommended as recommended.
S306, according to the corresponding history of each candidate target in candidate target set choose number and each candidate target with
The matching degree of query word obtains the corresponding recommendation score value of each candidate target.
Wherein, it refers to the number that each candidate target was once selected that the corresponding history of each candidate target, which chooses number,.It is each to wait
It refers to the similarity degree of each candidate target and query word to select the matching degree of object and query word.In the present embodiment, server can
The matching degree for whether judging using query word each candidate target and query word as prefix word according to each candidate target, in candidate target
It is to be judged to exactly matching using query word as prefix word.
Such as, it is assumed that query word is " Beijing University ", and candidate target is " Beijing University north gate ", wherein " Beijing University north gate " is with " Beijing University "
Prefix word then judges that candidate target " Beijing University north gate " is exactly matched with query word " Beijing University ".Assuming that query word is " north gate ", it is candidate
Object be " Beijing University north gate ", wherein " Beijing University north gate " not with " north gate " be prefix word, then judge candidate target " Beijing University north gate " with
Query word " Beijing University " Incomplete matching.
In one embodiment, server also can include with query word according to each candidate target identical characters number is sentenced
The more matching degrees of identical characters number of the matching degree of fixed each candidate target and query word, judgement candidate target and query word are more
It is high.For example, query word is " north ", when candidate target is " Beijing University north gate ", " north " is 1 with " Beijing University north gate " identical characters number,
Query word is " Beijing University ", and recommended is " Beijing University north gate ", and " Beijing University " is 2 with " Beijing University north gate " identical characters number, then judges
" Beijing University " is with " Beijing University north gate " matching degree higher than " north " and " Beijing University north gate ".
Specifically, server can be obtained using query word as the corresponding prefix source objects set of prefix word.Statistics was before
When sewing that each prefix source objects are as query word in source objects set, each corresponding history of candidate target chooses number.According to
It is right that the corresponding history of each candidate target of statistics chooses the matching degree of number and candidate target and query word that each candidate is calculated
As the recommendation factor when each prefix source objects is query words in using prefix source objects set, obtained further according to each recommendation factor
To recommendation score value, obtained recommendation score value and each recommendation factor positive correlation.
Wherein, it can be calculated according to preset function to recommend the factor.Preset function can be quadrature function, will
Candidate target corresponding history when prefix source objects is query words in using prefix source objects set chooses number and the time
The direct quadrature of the matching degree of object and query word is selected, obtains recommending the factor accordingly.Such as:Recommend the factor can be according to T=Khis*
Kmat is calculated, and wherein Khis indicates that candidate target is corresponding when prefix source objects is query words in using prefix source objects set
History choose number, Kmat indicates the matching degree of corresponding candidate object and query word.Preset function can also be will be candidate
Object corresponding history when prefix source objects is query words in using prefix source objects set chooses number and the candidate right
As the matching degree with query word is as exponential function of changed factor or logarithmic function etc..Such as:Recommend the factor can be according to T=
eKhis+eKmatIt calculates.
Server can will be when each prefix source objects is query words in using prefix source objects set, each candidate target pair
The corresponding recommendation score value of each candidate target is calculated according to preset positive correlation function in the recommendation factor answered.Preset positive correlation
Function such as direct proportion function, coefficient are positive exponential function or logarithmic function etc..In the present embodiment, each candidate target
Recommend score value can be each candidate target using each prefix source objects in prefix source objects set as query word when corresponding push away
The factor is recommended to sum to obtain according to corresponding Weight.
S308 picks out the candidate target for meeting recommendation condition according to recommendation score value from candidate target set.
Specifically, suitable screening conditions can be set as needed in server, according to the screening conditions from candidate target collection
Qualified object is selected in conjunction, which is constrained according to recommendation score value.Screening conditions such as select to have big
Recommend the corresponding object of recommendation score value of the maximum preset quantity of score value in the object of the recommendation score value of preset value, or selection
Deng.
In one embodiment, a variety of different types of candidate targets are may include in candidate target set.Server can be first
Each candidate target in candidate target set is classified according to the affiliated type of candidate target, obtains belonging to different types of candidate
Object subset, then pick out the object for meeting screening conditions respectively in each candidate target subset.
In one embodiment, the candidate target in the recommended set involved in object recommendation method includes that prompt is looked into
Ask word and point of interest.Step S308 includes:The prompting query word and point of interest for including by candidate target set are respectively according to right
The recommendation score value descending sort answered;Pick out the prompting query word of preset quantity from the prompting query word that descending arranges, and from
The point of interest of preset quantity is picked out in the point of interest of descending arrangement.In the present embodiment, server can be recommended different types of
Candidate target, more accurately to recommend point of interest, is improved and is pushed away by recommending prompting query word further to predict user view
Recommend the accuracy of content.
S310 recommends the candidate target picked out.
Specifically, server produces the recommendation information for the object for recommending to pick out, and recommendation information is pushed to terminal.
Recommendation information may include recommended, can also include object factory text and/or description picture, can also include each selecting
The corresponding recommendation score value of object gone out.
The candidate target in the recommended set involved in object recommendation method includes that prompt is looked into one embodiment
Ask word and point of interest.Step S310 includes the prompting query word and point of interest for recommending to pick out.
Above-mentioned object recommendation method builds candidate target set according to historical query selected object, is being searched for getting
After the query word inputted in frame, candidate target set corresponding with the query word of input is determined, by going through in candidate target set
History inquiry selected object is recalled, with each corresponding history of candidate target in candidate target set choose number and each candidate target with
The matching degree of query word is foundation, calculates the corresponding recommendation score value of each candidate target, will meet the candidate target of recommendation condition into
Row is recommended.This to be fed back according to user's history behavioral data, the recommendation method of analog subscriber real behavior can be more accurate
Ground positions the demand of user, to improve the accuracy of recommendation.
As shown in figure 4, in one embodiment, step S304 specifically comprises the following steps:
S402 is searched using query word as the historical record word of prefix word.
Wherein, historical record word refers to the character string recorded during historical query, including history input inquiry word and is gone through
History inquires selected object.Server can be after each user be inquired by terminal, respective record this inquiry operation input
Query word and inquiry after the object chosen, and stored into database or caching as historical record word.
Specifically, server traverses the historical record word of storage after getting query word, the going through traversal in traversal
The Records of the Historian records word and the query word of acquisition compares, if the historical record word of traversal is consistent with the query word of acquisition, it is determined that the traversal
Query word be the historical record word required to look up.
S406 obtains historical query selected object corresponding with the historical record word found.
Specifically, it when the object that server is chosen after the query word and inquiry for recording each inquiry operation input, can incite somebody to action
Query word is stored as historical record word, will correspondingly input the object chosen after the query word as the query word pair with input
The historical query selected object answered is stored.Server can also be stored the object chosen after inquiry as historical record word,
And using the object as corresponding historical query selected object.
For example, user has selected " Peking University north gate " point of interest after inputting " north gate " in search box.Server
Recordable " north gate " is used as historical record word, and " Peking University north gate " is corresponded to storage as historical query selected object.Service
" Peking University north gate " can be also used as historical record word by device, and " Peking University north gate " is corresponded to as historical query selected object
Storage.
S408, determination include the candidate target set of historical query selected object.
Specifically, server can be using the historical query selected object of acquisition as the candidate target recommended, and structure is waited
Select object set.
In the present embodiment, it using the query word of acquisition as foundation, searches the query word to obtain and remembers as the history of prefix word
Word is recorded, it is the historical record word found is expansible as query word as a result, corresponding according to the historical record word found
Historical query selected object positions the object of the intent query of user, can relatively accurately position the demand of user, to carry
The accuracy of high recommendation.
In one embodiment, further include the step for updating candidate target set corresponding with query word before step S402
Suddenly, which specifically includes as shown in Figure 5:
S502 is periodically collected and is increased historical record word and accordingly newly-increased historical query selected object newly.
Specifically, when the query word that server is inputted when recording user and carrying out each inquiry operation and the object chosen,
Corresponding record query time.Server can be pre-configured with the renewal time of update candidate target set, detect current time
When reaching preconfigured renewal time, the historical record word increased newly in renewal time last time to the period of current time is obtained
And increase historical query selected object newly accordingly.Wherein, it is periodically to periodically carry out certain operation, for example, each natural
Day, or per hour etc..
S504 segments to obtain prefix word to the newly-increased historical record word of collection.
Specifically, server can divide the newly-increased historical record word of collection since first character, then by increasing one by one
Padding continues, and obtains several prefix words.Such as:Newly-increased historical record word is " Peking University ", then from first character
Start, be split successively, " north ", " Beijing ", " Beijing is big " and " Peking University " four prefix words can be obtained.
In one embodiment, server can divide the newly-increased historical record word of collection since first character, and from
Character is picked out in subsequent character is combined to obtain the portmanteau word for meeting natural language rule as prefix word with initial character.
For example " Peking University " can select to obtain " Beijing University " as prefix word.
S506 is searched and to segment obtained prefix word as the historical record word of prefix word.
Specifically, server is after the newly-increased historical record word to collection segments to obtain newly-increased prefix word, by what is increased newly
Prefix word is compared with each prefix word in history prefix set of words.If newly-increased prefix word is present in history prefix set of words
In, then it searches with using newly-increased prefix word as the historical record word of prefix word, the historical query selected object increased newly is added
Into the candidate target set of the corresponding historical query selected object of the historical record word that includes with find.
If newly-increased prefix word is not present in history prefix set of words, newly-increased prefix word is added to history prefix
In set of words, the updated prefix set of words being currently adapted to is obtained, and build the newly-increased corresponding prefix source pair of prefix word
As set and history selected object set.
S508 obtains candidate target set, and the candidate target in the candidate target set of acquisition is and the history that finds
Record the corresponding historical query selected object of word.
Newly-increased historical query selected object is added in the candidate target set of acquisition by S510.
In the present embodiment, candidate target set is built according to historical query selected object by regularly updating, come continuous
It improves according to user's history behavioral data analog subscriber true intention, more accurately to position the demand of user, improves and recommend
The accuracy of content.
As shown in fig. 6, in one embodiment, step S306 specifically comprises the following steps:
S602 obtains the corresponding historical record word of each candidate target in candidate target set.
S604, extraction include the historical query record of historical record word.
Wherein, historical query record refers to the historical query content of record, including historical record word and historical query choose
Object.Server can record user in the form of inquiring and record and carry out the query word of inquiry operation input and pair chosen every time
As.For example, historical query record can be recorded as:" input:Beijing University, selection:Beijing University's north gate point of interest ", wherein " Beijing University " is
Historical record word, " Beijing University's north gate point of interest " are historical query selected object.
Specifically, server can traversal history inquiry record, include by the historical query of traversal record in traversal
Historical record word filters out historical query note compared with obtaining the corresponding historical record word of each candidate target in candidate target set
The identical history of the historical record word that record includes historical record word corresponding with each candidate target in acquisition candidate target set
Inquiry record.
S606 filters out the history for including candidate target corresponding with historical record word from the historical query of extraction record
Inquiry record.
S608 counts the quantity of the historical query record filtered out.
Specifically, the historical query record characterization filtered out is to select the history when history inputs historical record word
The number for the historical query selected object that inquiry record includes.For example, the historical query record " input of statistics:Beijing University, choosing
It selects:The quantity of Beijing University's north gate point of interest " is 100, in the historical operating data for indicating user, when inputting " Beijing University ", chooses " Beijing University
The number of north gate point of interest " is 100 times.
S610, according to the matching degree of the quantity of the statistics corresponding to each candidate target and corresponding candidate object and query word,
Obtain recommendation score value corresponding with each candidate target.
Specifically, server is configurable to calculate the function for recommending score value.In the present embodiment, server can be by correspondence
In the matching degree quadrature of the quantity and corresponding candidate object and query word of the statistics of each candidate target, and add according to corresponding weight
Power summation, obtains recommendation score value corresponding with each candidate target.
In one embodiment, the function for calculating recommendation score value can be:P(Obj|word)≈∑K*count
(Obj|word)*P(CX|Obj).Wherein P (Obj | word) indicates the recommendation score value of candidate target " Obj ", count (Obj |
When word) indicating to inquire as historical query word using the historical record word " word " that query word " CX " is prefix word, selection is candidate
The number of object " Obj ", P (CX | Obj) are the matching degree of query word " CX " and candidate target " Obj ", and K is weights, and ∑ indicates to ask
And calculating.
In the present embodiment, user's history behavioral data is recorded in the form of historical query record, historical query is same in recording
When include historical query word and candidate target, it is different when history inputs different query words counting, input the candidate target
Number when, only need to count not only include candidate target, but also including the corresponding historical record word of the candidate target historical query note
Record improves response efficiency.
Obtained as shown in fig. 7, in one embodiment, in object recommendation method in candidate target set each candidate target with
The step of matching degree of query word includes:
S702 is determined when the candidate target in candidate target set is using query word as prefix word using query word as prefix
The candidate target of word and the matching degree of query word are preset matching degree.
Specifically, server can configured in advance be used to judge the matching condition of candidate target and query word matching degree, and set
Set matching degree when exactly matching that preset matching degree is candidate target and query word.Server can set matching condition to root
The matching degree for whether judging using query word each candidate target and query word as prefix word according to each candidate target, candidate target with
Query word is that prefix word is to be judged to exactly matching.Wherein, preset matching degree can be 1 or other numerical value.
S704, when the candidate target in candidate target set is not using query word as prefix word, it is determined that not with query word
For the corresponding historical record word of candidate target of prefix word.
S706, statistics include the first quantity of the historical query record of historical record word.
Specifically, including the historical query of historical record word record the first quantity characterization be with historical record word be it is defeated
Enter inquiry times when word.
S708, statistics includes historical record word, and includes not using query word as the historical query of the candidate target of prefix word
Second quantity of record.
Specifically, including historical record word, and include not remembering by the historical query of the candidate target of prefix word of query word
It is chosen not using query word as the candidate target of prefix word when what the second quantity of record characterized is using historical record word as input word
Inquiry times.
S710, determination is not the second quantity and first by the matching degree of the candidate target of prefix word and query word of query word
The ratio of quantity;Ratio is less than preset value.
Specifically, including historical record word, and include not remembering by the historical query of the candidate target of prefix word of query word
What the ratio for the first quantity that the second quantity of record is recorded with the historical query including historical record word indicated is with historical data
It under conditions of input is using query word as the historical record word of prefix word, will be selected in history according to Probability principle for foundation
Not using query word as the frequency of the candidate target of prefix word, the probability of the candidate target is chosen when as current input inquiry word.
In the present embodiment, using candidate target and whether using query word as of prefix word judgment candidate target and query word
With degree, meets the search custom inputted based on key sequence, can more accurately predict the query word of user view input.
It illustrates, it is assumed that query word input by user is " north ", and server is right according to the candidate that query word " north " determines
Include " Peking University " and " mansions Chuan Jing " as gathering, using query word " north " as prefix word, corresponding prefix source objects collection
It closes including " north ", " Beijing ", " Beijing is big ", " Peking University ", " Beijing wound ", " Beijing wound scape ", " it is big that scape is created in Beijing " and " Beijing
The mansions Chuan Jing ".Wherein, including the historical query in " north " record quantity is 20, wherein 15 historical queries record going through of including
It is " Peking University " that history, which inquires selected object, and the historical query selected object that 5 historical query records include is that " wound scape is big
Tall building ".Due to " Peking University " with " north " for prefix word, then by 1 as query word " north " and candidate target " Peking University "
Matching degree;Due to " mansions Chuan Jing " not with " north " for prefix word, then by ratio 5/20 as query word " north " and candidate target
The matching degree of " mansions Chuan Jing ".
As shown in figure 8, in one embodiment, the historical query selected object involved in object recommendation method includes history
Selected object and historical search selected object, step S306 is recommended to specifically include:
S802 chooses number and each candidate right according to the corresponding history recommendation of each candidate target in candidate target set
As the matching degree with query word, obtains each candidate target corresponding first and recommend score value.
Specifically, server can be recommended to meet to push away after getting the query word that user is inputted by terminal by terminal
The candidate target for recommending condition, after user chooses the candidate target that server is recommended by terminal, the candidate target chosen is
History recommends selected object.Server can calculate function according to pre-set recommendation score value, and it is history to calculate each candidate target
First when selected object is recommended to recommend score value.Wherein, pre-set recommendation score value calculates function and can be quadrature function, refer to
Number function or repeatedly multinomial function etc..
S804 chooses number and each candidate right according to the corresponding historical search of each candidate target in candidate target set
As the matching degree with query word, obtains each candidate target corresponding second and recommend score value.
Specifically, server can be recommended to meet to push away after getting the query word that user is inputted by terminal by terminal
The candidate target of condition is recommended, if user does not choose the candidate target that server is recommended by terminal, but is referred to by triggering search
After order directly scans for, search result shown in choosing, then the candidate target chosen is historical search selected object.Clothes
Business device can calculate function according to pre-set recommendation score value, calculate first when each candidate target is historical search selected object
Recommend score value.
S806, the recommendation score value for recommending score value each candidate target corresponding with the second recommendation score value acquisition according to first;It pushes away
It recommends score value and first and recommends score value positive correlation, and score value and second is recommended to recommend score value positive correlation.
Specifically, server can recommend score value and second to recommend score value, meter according to preset positive correlation function and first
Calculate the recommendation score value of corresponding each candidate target.Wherein, positive correlation function can be direct proportion function, and coefficient is positive index letter
Number or logarithmic function etc..
In one embodiment, it is P (Obj | word) ≈ ∑s K* that pre-set recommendation score value, which calculates function,
countsug/search(Obj | word) * P (CX | Obj), wherein countsug(Obj | word) indicate that candidate target is recommended for history
Selected object, and using query word " CX " be prefix word historical record word " word " inquire as historical query word when, select time
Select the number of object " Obj ";countsearch(Obj | word) expression candidate target is historical search selected object, and with inquiry
When word " CX " is that the historical record word " word " of prefix word is inquired as historical query word, the number of candidate target " Obj " is selected.
Preset positive correlation function is P (Obj | word) ≈ ∑s K1*countsug(Obj|word)*P(CX|Obj)+K∑K2*
countsearch(Obj|word)*P(CX|Obj)。
In the present embodiment, recommend selected object to carry out comprehensive assessment with historical search selected object history, avoid
The situation for solely leading to recommendation results inaccuracy using the user's history behavioral data under recommendation pattern, will be under recommendation pattern
User's history behavioral data is combined with the user's history behavioral data under precise search pattern, can more accurately be fitted user's row
To optimize the accuracy of recommended.
It is the schematic diagram that terminal recommends interface in one embodiment with reference to figure 9.The interface includes search box 910, search box
The candidate target 930 recommended after the query word 920 and input inquiry word 920 that are inputted in 910.Specifically, in the visitor of terminal operating
Family end provides the search box for scanning for input, and user is in search box after input inquiry word " Beijing University west gate ", terminal
Query word input by user " Beijing University west gate " is obtained, and query word " Beijing University west gate " is uploaded to server.
Whois lookup is with the historical record word that the query word " Beijing University west gate " currently obtained is prefix word, by what is found
The corresponding history selected object of historical record word is collected as candidate target set.According to each candidate target phase in candidate target set
The history answered chooses the matching degree of number and each candidate target and query word " Beijing University west gate ", obtains each candidate target and corresponds to
Recommendation score value;The candidate target for meeting recommendation condition is picked out from candidate target set according to recommendation score value;It will be singled out
Candidate target be issued to terminal, show the candidate target picked out on the recommendation interface of terminal.
As shown in Figure 10, in one embodiment, a kind of method of object recommendation is provided, following steps are specifically included:
S1002 obtains the query word inputted in search box.
S1004 determines the geographical location residing for the terminal for inputting the query word obtained.
Specifically, terminal is when detecting that the query word is uploaded to server by query word input by user, by terminal institute
Place geographical location is uploaded to server.Wherein, geographical location can be can on map status longitude and latitude.
S1006 is periodically collected and is increased historical record word and accordingly newly-increased historical query selected object newly;To the new of collection
Increase historical record word to segment to obtain prefix word;It searches and to segment obtained prefix word as the historical record word of prefix word;It obtains
Candidate target set, the candidate target in the candidate target set of acquisition are that history corresponding with the historical record word found is looked into
Ask selected object;Newly-increased historical query selected object is added in the candidate target set of acquisition.
S1008 is searched using query word as the historical record word of prefix word;It obtains corresponding with the historical record word found
Historical query selected object;Determination includes the candidate target set of historical query selected object.Wherein, each in candidate target set
For candidate target collected from historical query selected object, historical query selected object includes that history recommends selected object and historical search
Selected object, historical query selected object include prompting query word and point of interest.
S1010 is filtered out from candidate target set and is recommended the candidate target that selected object obtains by history, obtains screening
The corresponding historical record word of each candidate target gone out;Extraction includes the historical query record of the historical record word;From going through for extraction
The historical query record including candidate target corresponding with historical record word is filtered out in history inquiry record;What statistics filtered out goes through
The quantity of history inquiry record;It will be corresponding to the matching degree of the quantity of the statistics of each candidate target and corresponding candidate object and query word
Quadrature, and sum according to corresponding Weight, it obtains corresponding with each candidate target first and recommends score value.
S1012 filters out the candidate target obtained by historical search selected object from candidate target set, obtains screening
The corresponding historical record word of each candidate target gone out;Extraction includes the historical query record of the historical record word;From going through for extraction
The historical query record including candidate target corresponding with historical record word is filtered out in history inquiry record;What statistics filtered out goes through
The quantity of history inquiry record;It will be corresponding to the matching degree of the quantity of the statistics of each candidate target and corresponding candidate object and query word
Quadrature, and sum according to corresponding Weight, it obtains corresponding with each candidate target second and recommends score value.
S1014, the recommendation score value for recommending score value each candidate target corresponding with the second recommendation score value acquisition according to first;It pushes away
It recommends score value and first and recommends score value positive correlation, and score value and second is recommended to recommend score value positive correlation.
Each candidate is assessed in S1016, the geographical location corresponding to each candidate target at a distance from the geographical location determined
The additive score of object;Wherein, the corresponding geographical location of the higher candidate target of additive score with determine geographical location away from
From closer.
Specifically, the geographical location corresponding to each candidate target refer to candidate target be point of interest when, the point of interest correspond to
The position coordinates that can be positioned on map.Server can be pressed each candidate target behind the geographical location where determining terminal
Descending sort is carried out at a distance from the geographical location where the terminal determined according to respectively corresponding geographical location.Server can be pressed
The additive score of each candidate target is assessed according to the pre-set additive score valuation functions with apart from negative correlation.Wherein, negative
The additive score valuation functions of pass can be that either coefficient is the exponential function born or logarithmic function etc. to inverse proportion function.
S1018, according to the comprehensive scores for recommending score value each candidate target corresponding with additive score acquisition;Comprehensive scores with
Recommend score value positive correlation, and comprehensive scores and additive score positive correlation.
S1020, the prompting query word and point of interest for including by candidate target set are respectively according to corresponding comprehensive scores
Descending sort.
S1022 picks out the prompting query word of preset quantity from the prompting query word that descending arranges, and is arranged from descending
Point of interest in pick out the point of interest of preset quantity.
Specifically, the prompting query word of preset quantity refers to the quantity of the pre-set prompting query word for needing to recommend.
The point of interest of preset quantity refers to the quantity of the pre-set point of interest for needing to recommend.
S1024 recommends the prompting query word picked out and point of interest.
In the present embodiment, according to for inputting geographical location and user's history row residing for the terminal of the query word obtained
The recommendation score value that candidate target is assessed for aggregation of data, is more in line with user demand, avoids due to recommended object distance user
Current location is too far, and user does not select recommended, and the problem of cause recommendation not meet user demand, it improves
The accuracy of recommendation.
In one embodiment, server can be in advance by the point of interest in historical query selected object according to historical query
Period classifies.For example, each consecutive days 11:00 to 13:Historical query selected object in 00 period.Specifically,
Server obtains current time of the user by terminal input inquiry word after the query word for obtaining input, is chosen pair from inquiry
The historical query selected object for the inquiry for meeting current time is filtered out as in as candidate target.
In one embodiment, server will can also be divided in historical query selected object according to user identifier in advance
Class.Server obtains user identifier after the query word for obtaining input, is filtered out from historical query selected object and acquisition
The matched historical query selected object of user identifier is as candidate target.
It is the logic diagram based on user's history behavioral data feedback in one embodiment with reference to figure 11.User is in terminal
In the search box of offer after input inquiry word, server is recommended according to query word input by user.If user is to server
The object of recommendation carries out clicking operation, chooses recommended, then it represents that recommendation meets user demand;If the non-selected clothes of user
The object that business device is recommended, then it represents that recommendation does not meet user demand, and user may be selected to continue input inquiry word, until recommendation
Content meets user demand;User also can be done directly on the search entrance triggering precise search of search box offer, if user selects
Select the object of server precise search offer, then it represents that recommendation meets user demand, if the precise search knot of current presentation
Fruit does not meet user demand, and user can carry out page turn over operation, update the precise search result of current presentation;User is also replaceable to be looked into
Word is ask, until selector closes the object of user view.It is fed back based on user's history behavioral data and carries out object recommendation, utilize user
Input inquiry selected ci poem selects recommended or the behavior flow of the result of precise search is selected to be looked into determine that user's input is imperfect
The intention for asking word, improves the accuracy of recommendation.
Figure 12 is the statistical Butut that server carries out that user when object recommendation selects recommended.With reference to figure 12, the system
The abscissa of score Butut is timing statistics, and ordinate is the percentage that user selects recommended, wherein user selects recommendation
The percentage of object can select the total degree of recommended divided by user's inquiry total degree to obtain according to user.It can from Figure 12
It is user's history behavior to go out object recommendation method to use the growth rapid growth of initial stage curve 1201 at any time, this propagation process
Data Preparation Process, after user's history behavioral data set-up procedure, user selects the percentage of recommended gradually to become
In stabilization, and more than 8%, than the percentage that user when object recommendation method provided by the present application is not used selects recommended
2% has and is greatly promoted.
Figure 13 is that user input query word is averaged the statistical Butut of number of characters when server carries out object recommendation.Reference chart
13, the abscissa of the statistical Butut is timing statistics, and ordinate is that user input query word is averaged number of characters.Wherein, user
The input inquiry word number of characters that is averaged inputs the average number of characters of all query words with user.User selects the object that server is recommended
When the query word number of characters that inputs it is fewer, indicate that the content that server is recommended is more accurate.Using this Shen as can be seen from Figure 13
After the object recommendation method that please provide carries out object recommendation, downward trend is integrally presented in the user input query word number of characters that is averaged,
Improve user's input efficiency.
As shown in figure 14, in one embodiment, a kind of object recommendation device 1400 is provided, including:Query word obtains
Module 1401, recommends score value computing module 1403, Choosing module 1404 and recommending module at candidate target determining module 1402
1405。
Query word acquisition module 1401, for obtaining the query word inputted in search box.
Candidate target determining module 1402, for determining candidate target set corresponding with query word;Candidate target set
In each candidate target collected from historical query selected object.
Recommend score value computing module 1403, it is secondary for being chosen according to the corresponding history of each candidate target in candidate target set
The matching degree of number and each candidate target and query word, obtains the corresponding recommendation score value of each candidate target.
Choosing module 1404, for picking out the candidate for meeting recommendation condition from candidate target set according to recommendation score value
Object.
Recommending module 1405, the candidate target for recommending to pick out.
Above-mentioned object recommendation device builds candidate target set according to historical query selected object, is being searched for getting
After the query word inputted in frame, candidate target set corresponding with the query word of input is determined, by going through in candidate target set
History inquiry selected object is recalled, with each corresponding history of candidate target in candidate target set choose number and each candidate target with
The matching degree of query word is foundation, calculates the corresponding recommendation score value of each candidate target, will meet the candidate target of recommendation condition into
Row is recommended.This to be fed back according to user's history behavioral data, the recommendation method of analog subscriber real behavior can be more accurate
Ground positions the demand of user, to improve the accuracy of recommendation.
In one embodiment, candidate target determining module 1402 is additionally operable to search and remember by the history of prefix word of query word
Record word;Obtain historical query selected object corresponding with the historical record word found;Determination includes historical query selected object
Candidate target set.
In the present embodiment, it using the query word of acquisition as foundation, searches the query word to obtain and remembers as the history of prefix word
Word is recorded, it is the historical record word found is expansible as query word as a result, corresponding according to the historical record word found
Historical query selected object positions the object of the intent query of user, can relatively accurately position the demand of user, to carry
The accuracy of high recommendation.
In one embodiment, score value computing module 1403 is recommended to be additionally operable to obtain each candidate target in candidate target set
Corresponding historical record word;Extraction includes the historical query record of historical record word;It is screened from the historical query of extraction record
Go out the historical query record including candidate target corresponding with historical record word;Count the number of the historical query record filtered out
Amount;According to the matching degree of the quantity of the statistics corresponding to each candidate target and corresponding candidate object and query word, obtain and each time
Select the corresponding recommendation score value of object.
In the present embodiment, user's history behavioral data is recorded in the form of historical query record, historical query is same in recording
When include historical query word and candidate target, it is different when history inputs different query words counting, input the candidate target
Number when, only need to count not only include candidate target, but also including the corresponding historical record word of the candidate target historical query note
Record improves response efficiency.
In one embodiment, recommend score value computing module 1403 be additionally operable to when the candidate target in candidate target set with
When query word is prefix word, determine by the matching degree of the candidate target of prefix word and query word of query word to be preset matching degree;
When the candidate target in candidate target set is not using query word as prefix word, it is determined that not using query word as the candidate of prefix word
The corresponding historical record word of object;Statistics includes the first quantity of the historical query record of historical record word;Statistics includes history
Word is recorded, and includes the second quantity not recorded as the historical query of the candidate target of prefix word using query word;It determines not to look into
It is the candidate target of prefix word and the ratio that the matching degree of query word is the second quantity and the first quantity to ask word;Ratio is less than default
Matching degree.
In the present embodiment, using candidate target and whether using query word as of prefix word judgment candidate target and query word
With degree, meets the search custom inputted based on key sequence, can more accurately predict the query word of user view input.
In one embodiment, candidate target includes prompting query word and point of interest in candidate target set;Choosing module
1404 are additionally operable to the prompting query word for including by candidate target set and point of interest respectively according to corresponding recommendation score value descending
Sequence;The prompting query word of preset quantity, and the point of interest arranged from descending are picked out from the prompting query word that descending arranges
In pick out the point of interest of preset quantity.Recommending module 1405 is additionally operable to the prompting query word and point of interest recommending to pick out.
In the present embodiment, server can recommend different types of candidate target, by recommending prompting query word further
Prediction user view improves the accuracy of recommendation more accurately to recommend point of interest.
In one embodiment, historical query selected object includes that history recommends selected object and historical search to choose pair
As;Recommendation score value computing module 1403 is additionally operable to be chosen according to the corresponding history recommendation of each candidate target in candidate target set secondary
The matching degree of number and each candidate target and query word obtains each candidate target corresponding first and recommends score value;According to candidate right
As the matching degree of number and each candidate target and query word is chosen in the corresponding historical search of each candidate target in set, obtain
Each candidate target corresponding second recommends score value;Recommend score value each candidate corresponding with the second recommendation score value acquisition right according to first
The recommendation score value of elephant;Recommend score value and first to recommend score value positive correlation, and score value and second is recommended to recommend score value positive correlation.
In the present embodiment, recommend selected object to carry out comprehensive assessment with historical search selected object history, avoid
The situation for solely leading to recommendation results inaccuracy using the user's history behavioral data under recommendation pattern, will be under recommendation pattern
User's history behavioral data is combined with the user's history behavioral data under precise search pattern, can more accurately be fitted user's row
To optimize the accuracy of recommended.
As shown in figure 15, in one embodiment, a kind of object recommendation device 1500 is provided, including:Query word obtains
Module 1501, recommends score value computing module 1503, Choosing module 1504, recommending module 1505 at candidate target determining module 1502
With candidate target update module 1506.
Query word acquisition module 1501, for obtaining the query word inputted in search box.
Candidate target determining module 1502, for determining candidate target set corresponding with query word;Candidate target set
In each candidate target collected from historical query selected object.
Recommend score value computing module 1503, it is secondary for being chosen according to the corresponding history of each candidate target in candidate target set
The matching degree of number and each candidate target and query word, obtains the corresponding recommendation score value of each candidate target.
Choosing module 1504, for picking out the candidate for meeting recommendation condition from candidate target set according to recommendation score value
Object.
Recommending module 1505, the candidate target for recommending to pick out.
Candidate target update module 1506 increases historical record word and accordingly newly-increased historical query newly for periodically collecting
Selected object;The newly-increased historical record word of collection is segmented to obtain prefix word;It searches and to segment obtained prefix word as prefix
The historical record word of word;Acquisition includes the candidate target collection of historical query selected object corresponding with the historical record word found
It closes;Newly-increased historical query selected object is added in the candidate target set of acquisition.
In the present embodiment, candidate target set is built according to historical query selected object by regularly updating, come continuous
It improves according to user's history behavioral data analog subscriber true intention, more accurately to position the demand of user, improves and recommend
The accuracy of content.
As shown in figure 16, in one embodiment, a kind of object recommendation device 1600 is provided, including:Query word obtains
Module 1601, locating module 1602, candidate target determining module 1603, recommend score value computing module 1604, Choosing module 1605,
Recommending module 1606.
Query word acquisition module 1601, for obtaining the query word inputted in search box.
Locating module 1602, for determining geographical location residing for the terminal for inputting the query word obtained.
Candidate target determining module 1603, for determining candidate target set corresponding with query word;Candidate target set
In each candidate target collected from historical query selected object.
Recommend score value computing module 1604, it is secondary for being chosen according to the corresponding history of each candidate target in candidate target set
The matching degree of number and each candidate target and query word, obtains the corresponding recommendation score value of each candidate target.
Choosing module 1605, for the geographical location corresponding to each candidate target at a distance from the geographical location determined
Assess the additive score of each candidate target;Wherein, the corresponding geographical location of the higher candidate target of additive score and the ground determined
The distance for managing position is closer;According to the comprehensive scores for recommending score value each candidate target corresponding with additive score acquisition;Comprehensive point
Value and recommendation score value positive correlation, and comprehensive scores and additive score positive correlation;It is chosen from candidate target set according to comprehensive scores
Select the candidate target for meeting recommendation condition.
Recommending module 1606, the candidate target for recommending to pick out.
In the present embodiment, according to for inputting geographical location and user's history row residing for the terminal of the query word obtained
The recommendation score value that candidate target is assessed for aggregation of data, is more in line with user demand, avoids due to recommended object distance user
Current location is too far, and user does not select recommended, and the problem of cause recommendation not meet user demand, it improves
The accuracy of recommendation.
One of ordinary skill in the art will appreciate that realizing all or part of flow in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the program can be stored in a non-volatile computer and can be read
In storage medium, the program is when being executed, it may include such as the flow of the embodiment of above-mentioned each method.Wherein, the storage is situated between
Matter can be magnetic disc, CD, read-only memory (Read-Only Memory, ROM) etc..
Each technical characteristic of above example can be combined arbitrarily, to keep description succinct, not to above-described embodiment
In each technical characteristic it is all possible combination be all described, as long as however, the combination of these technical characteristics be not present lance
Shield is all considered to be the range of this specification record.
Several embodiments of the invention above described embodiment only expresses, the description thereof is more specific and detailed, but simultaneously
Cannot the limitation to the scope of the claims of the present invention therefore be interpreted as.It should be pointed out that for those of ordinary skill in the art
For, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to the guarantor of the present invention
Protect range.Therefore, the protection domain of patent of the present invention should be determined by the appended claims.
Claims (16)
1. a kind of object recommendation method, the method includes:
Obtain the query word inputted in search box;
Determine candidate target set corresponding with the query word;Each candidate target is collected from history in the candidate target set
Inquire selected object;
Number and each candidate target and institute are chosen according to each corresponding history of candidate target in the candidate target set
The matching degree of query word is stated, the corresponding recommendation score value of each candidate target is obtained;
The candidate target for meeting recommendation condition is picked out from the candidate target set according to the recommendation score value;
Recommend the candidate target picked out.
2. according to the method described in claim 1, it is characterized in that, determination candidate target collection corresponding with the query word
The step of conjunction includes:
It searches using the query word as the historical record word of prefix word;
Obtain historical query selected object corresponding with the historical record word found;
Determination includes the candidate target set of the historical query selected object.
3. according to the method described in claim 2, it is characterized in that, the lookup is remembered by the history of prefix word of the query word
Before recording word, the method further includes:
It periodically collects and increases historical record word and accordingly newly-increased historical query selected object newly;
The newly-increased historical record word of collection is segmented to obtain prefix word;
It searches and to segment the obtained prefix word as the historical record word of prefix word;
Candidate target set is obtained, the candidate target in the candidate target set of acquisition is to remember with the history found
Record the corresponding historical query selected object of word;
The newly-increased historical query selected object is added in the candidate target set of acquisition.
4. according to the method described in claim 1, it is characterized in that, described according to each candidate target in the candidate target set
Corresponding history chooses the matching degree of number and each candidate target and the query word, obtains each candidate target
The step of corresponding recommendation score value includes:
Obtain the corresponding historical record word of each candidate target in the candidate target set;
Extraction includes the historical query record of the historical record word;
The history for including candidate target corresponding with the historical record word is filtered out from the historical query of extraction record
Inquiry record;
Count the quantity of the historical query record filtered out;
According to the matching of the quantity and corresponding candidate object and the query word of the statistics corresponding to each candidate target
Degree obtains recommendation score value corresponding with each candidate target.
5. according to the method described in claim 1, it is characterized in that, the matching of each candidate target and the query word
Degree is obtained by following steps:
When the candidate target in the candidate target set is using the query word as prefix word, before determination is with the query word
The matching degree of the candidate target and the query word of sewing word is preset matching degree;
When the candidate target in the candidate target set is not using the query word as prefix word, then
It determines not using the query word as the corresponding historical record word of the candidate target of prefix word;
Statistics includes the first quantity of the historical query record of the historical record word;
Statistics includes the historical record word, and includes not gone through using the query word as described in the candidate target of prefix word
Second quantity of history inquiry record;
Determine that not using the query word be second number as the matching degree of the candidate target of prefix word and the query word
The ratio of amount and first quantity;The ratio is less than the preset matching degree.
6. according to the method described in claim 1, it is characterized in that, candidate target includes that prompt is looked into the candidate target set
Ask word and point of interest;
Described picked out from the candidate target set according to the recommendation score value meets the candidate target of recommendation condition and includes:
The prompting query word and point of interest for including by the candidate target set are arranged according to corresponding recommendation score value descending respectively
Sequence;
The prompting query word of preset quantity is picked out from the prompting query word that descending arranges, and from described in descending arrangement
The point of interest of preset quantity is picked out in point of interest;
It is described to recommend the candidate target picked out to include:
Recommend the prompting query word picked out and the point of interest.
7. according to the method described in claim 1, it is characterized in that, the historical query selected object include history recommendation choose
Object and historical search selected object;
It is described that number and each candidate target are chosen according to each corresponding history of candidate target in the candidate target set
With the matching degree of the query word, the step of obtaining each candidate target corresponding recommendation score value, includes:
Number and each candidate target are chosen according to the corresponding history recommendation of each candidate target in the candidate target set
With the matching degree of the query word, obtains each candidate target corresponding first and recommend score value;
Number and each candidate target are chosen according to each corresponding historical search of candidate target in the candidate target set
With the matching degree of the query word, obtains each candidate target corresponding second and recommend score value;
The recommendation score value for recommending score value each candidate target corresponding with the second recommendation score value acquisition according to described first;
The recommendation score value recommends score value positive correlation with described first, and the recommendation score value recommends score value positive correlation with described second.
8. according to the method described in claim 1, it is characterized in that, it is described obtain the query word that is inputted in search box after,
The method further includes:
Determine geographical location residing for the terminal for inputting the query word obtained;
The step for picking out the candidate target for meeting recommendation condition from the candidate target set according to the recommendation score value
Suddenly include:
Each time is assessed at a distance from the geographical location determined in geographical location corresponding to each candidate target
Select the additive score of object;Wherein, the higher corresponding geographical location of the candidate target of additive score with determine describedly
The distance for managing position is closer;
According to the comprehensive scores for recommending score value each candidate target corresponding with additive score acquisition;The synthesis
Score value and the recommendation score value positive correlation, and the comprehensive scores and additive score positive correlation;
The candidate target for meeting recommendation condition is picked out from the candidate target set according to the comprehensive scores.
9. a kind of object recommendation device, which is characterized in that described device includes:
Query word acquisition module, for obtaining the query word inputted in search box;
Candidate target determining module, for determining candidate target set corresponding with the query word;The candidate target set
In each candidate target collected from historical query selected object;
Recommend score value computing module, for choosing number according to each corresponding history of candidate target in the candidate target set,
And the matching degree of each candidate target and the query word, obtain the corresponding recommendation score value of each candidate target;
Choosing module, for picking out the candidate for meeting recommendation condition from the candidate target set according to the recommendation score value
Object;
Recommending module, for recommending the candidate target picked out.
10. device according to claim 9, which is characterized in that the candidate target determining module is additionally operable to search with institute
State the historical record word that query word is prefix word;Historical query corresponding with the historical record word found is obtained to choose pair
As;Determination includes the candidate target set of the historical query selected object.
11. device according to claim 10, which is characterized in that described device further includes:
Candidate target update module, for periodically collecting newly-increased historical record word and newly-increased historical query accordingly is chosen pair
As;The newly-increased historical record word of collection is segmented to obtain prefix word;Before lookup is with the prefix word obtained with participle
Sew the historical record word of word;Candidate target set is obtained, the candidate target in the candidate target set of acquisition is and lookup
The corresponding historical query selected object of the historical record word arrived;The newly-increased historical query selected object is added to acquisition
The candidate target set in.
12. device according to claim 9, which is characterized in that the recommendation score value computing module is additionally operable to described in acquisition
The corresponding historical record word of each candidate target in candidate target set;Extraction includes the historical query note of the historical record word
Record;The history including candidate target corresponding with the historical record word is filtered out from the historical query of extraction record to look into
Consultation record;Count the quantity of the historical query record filtered out;According to the institute of the statistics corresponding to each candidate target
It states the matching degree of quantity and corresponding candidate object and the query word, obtains recommendation score value corresponding with each candidate target.
13. device according to claim 9, which is characterized in that the recommendation score value computing module is additionally operable to work as the time
When selecting the candidate target in object set using the query word as prefix word, determine using the query word as the time of prefix word
It is preset matching degree to select the matching degree of object and the query word;When the candidate target in the candidate target set is not with described
When query word is prefix word, it is determined that not using the query word as the corresponding historical record word of the candidate target of prefix word;
Statistics includes the first quantity of the historical query record of the historical record word;Statistics includes the historical record word, and
Including the second quantity not recorded as the historical query of the candidate target of prefix word using the query word;Determine not with
The query word is the candidate target of prefix word and the matching degree of the query word is second quantity and described first
The ratio of quantity;The ratio is less than the preset matching degree.
14. device according to claim 9, which is characterized in that candidate target includes prompt in the candidate target set
Query word and point of interest;
The Choosing module is additionally operable to the prompting query word for including by the candidate target set and point of interest respectively according to right
The recommendation score value descending sort answered;The prompting query word of preset quantity is picked out from the prompting query word that descending arranges,
And the point of interest of preset quantity is picked out from the point of interest that descending arranges;
The recommending module is additionally operable to the prompting query word for recommending to pick out and the point of interest.
15. device according to claim 9, which is characterized in that the historical query selected object includes that history recommends choosing
Middle object and historical search selected object;
The recommendation score value computing module is additionally operable to be recommended according to the corresponding history of each candidate target in the candidate target set
The matching degree of number and each candidate target and the query word is chosen, each candidate target corresponding first is obtained
Recommend score value;Number and each time are chosen according to each corresponding historical search of candidate target in the candidate target set
The matching degree of object and the query word is selected, each candidate target corresponding second is obtained and recommends score value;According to described first
Recommend the recommendation score value of score value each candidate target corresponding with the second recommendation score value acquisition;The recommendation score value and institute
The first recommendation score value positive correlation is stated, and the recommendation score value recommends score value positive correlation with described second.
16. device according to claim 9, which is characterized in that described device further includes:
Locating module, for determining geographical location residing for the terminal for inputting the query word obtained;
The Choosing module is additionally operable to the geographical location corresponding to each candidate target and the determining geographical location
Distance assess the additive score of each candidate target;Wherein, the higher corresponding geography of the candidate target of additive score
Position is closer at a distance from the geographical location determined;It is corresponding each according to the recommendation score value and additive score acquisition
The comprehensive scores of the candidate target;The comprehensive scores and the recommendations score value positive correlation, and the comprehensive scores with add
Score value positive correlation;The candidate target for meeting recommendation condition is picked out from the candidate target set according to the comprehensive scores.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710031775.9A CN108319603A (en) | 2017-01-17 | 2017-01-17 | Object recommendation method and apparatus |
PCT/CN2017/118053 WO2018133624A1 (en) | 2017-01-17 | 2017-12-22 | Object recommendation method and apparatus, server, and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710031775.9A CN108319603A (en) | 2017-01-17 | 2017-01-17 | Object recommendation method and apparatus |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108319603A true CN108319603A (en) | 2018-07-24 |
Family
ID=62891041
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710031775.9A Pending CN108319603A (en) | 2017-01-17 | 2017-01-17 | Object recommendation method and apparatus |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN108319603A (en) |
WO (1) | WO2018133624A1 (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110162695A (en) * | 2019-04-09 | 2019-08-23 | 中国科学院深圳先进技术研究院 | A kind of method and apparatus of information push |
CN110598127A (en) * | 2019-09-05 | 2019-12-20 | 腾讯科技(深圳)有限公司 | Group recommendation method and device |
CN110765370A (en) * | 2019-09-24 | 2020-02-07 | 北京嘀嘀无限科技发展有限公司 | Location information storage method, recommended location query device and server |
CN111125310A (en) * | 2019-12-24 | 2020-05-08 | 北京百度网讯科技有限公司 | Voice query method and device |
CN111460283A (en) * | 2020-03-06 | 2020-07-28 | 拉扎斯网络科技(上海)有限公司 | Information processing method, information processing device, electronic equipment and computer readable storage medium |
CN111859184A (en) * | 2020-06-28 | 2020-10-30 | 北京嘀嘀无限科技发展有限公司 | Boarding point recommendation method and system |
CN112100480A (en) * | 2020-09-15 | 2020-12-18 | 北京百度网讯科技有限公司 | Search method, device, equipment and storage medium |
CN112650931A (en) * | 2021-01-04 | 2021-04-13 | 杭州情咖网络技术有限公司 | Content recommendation method |
WO2021139154A1 (en) * | 2020-01-10 | 2021-07-15 | 百度在线网络技术(北京)有限公司 | Data prefetching method and apparatus, electronic device, and computer-readable storage medium |
CN114936861A (en) * | 2022-07-21 | 2022-08-23 | 威海海洋职业学院 | Financial data sharing method and system based on big data |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2105846A1 (en) * | 2008-03-28 | 2009-09-30 | Sony Corporation | Method of recommending content items |
US20100228710A1 (en) * | 2009-02-24 | 2010-09-09 | Microsoft Corporation | Contextual Query Suggestion in Result Pages |
CN103729351A (en) * | 2012-10-10 | 2014-04-16 | 阿里巴巴集团控股有限公司 | Search term recommendation method and device |
CN104143001A (en) * | 2014-08-01 | 2014-11-12 | 百度在线网络技术(北京)有限公司 | Search term recommending method and device |
CN105224554A (en) * | 2014-06-11 | 2016-01-06 | 阿里巴巴集团控股有限公司 | Search word is recommended to carry out method, system, server and the intelligent terminal searched for |
CN105956149A (en) * | 2016-05-12 | 2016-09-21 | 北京奇艺世纪科技有限公司 | Default search word recommendation method and apparatus |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101593720B1 (en) * | 2011-12-27 | 2016-02-17 | 한국전자통신연구원 | Contents search recommendation apparatus and method based on semantic network |
CN103870505B (en) * | 2012-12-17 | 2017-10-27 | 阿里巴巴集团控股有限公司 | One kind inquiry words recommending method and query word commending system |
CN103631929B (en) * | 2013-12-09 | 2016-08-31 | 江苏金智教育信息股份有限公司 | A kind of method of intelligent prompt, module and system for search |
-
2017
- 2017-01-17 CN CN201710031775.9A patent/CN108319603A/en active Pending
- 2017-12-22 WO PCT/CN2017/118053 patent/WO2018133624A1/en active Application Filing
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP2105846A1 (en) * | 2008-03-28 | 2009-09-30 | Sony Corporation | Method of recommending content items |
US20100228710A1 (en) * | 2009-02-24 | 2010-09-09 | Microsoft Corporation | Contextual Query Suggestion in Result Pages |
CN103729351A (en) * | 2012-10-10 | 2014-04-16 | 阿里巴巴集团控股有限公司 | Search term recommendation method and device |
CN105224554A (en) * | 2014-06-11 | 2016-01-06 | 阿里巴巴集团控股有限公司 | Search word is recommended to carry out method, system, server and the intelligent terminal searched for |
CN104143001A (en) * | 2014-08-01 | 2014-11-12 | 百度在线网络技术(北京)有限公司 | Search term recommending method and device |
CN105956149A (en) * | 2016-05-12 | 2016-09-21 | 北京奇艺世纪科技有限公司 | Default search word recommendation method and apparatus |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110162695A (en) * | 2019-04-09 | 2019-08-23 | 中国科学院深圳先进技术研究院 | A kind of method and apparatus of information push |
CN110162695B (en) * | 2019-04-09 | 2022-04-26 | 中国科学院深圳先进技术研究院 | Information pushing method and equipment |
WO2020207074A1 (en) * | 2019-04-09 | 2020-10-15 | 中国科学院深圳先进技术研究院 | Information pushing method and device |
CN110598127A (en) * | 2019-09-05 | 2019-12-20 | 腾讯科技(深圳)有限公司 | Group recommendation method and device |
CN110765370B (en) * | 2019-09-24 | 2020-10-27 | 北京嘀嘀无限科技发展有限公司 | Location information storage method, recommended location query device and server |
CN110765370A (en) * | 2019-09-24 | 2020-02-07 | 北京嘀嘀无限科技发展有限公司 | Location information storage method, recommended location query device and server |
CN111125310A (en) * | 2019-12-24 | 2020-05-08 | 北京百度网讯科技有限公司 | Voice query method and device |
WO2021139154A1 (en) * | 2020-01-10 | 2021-07-15 | 百度在线网络技术(北京)有限公司 | Data prefetching method and apparatus, electronic device, and computer-readable storage medium |
CN111460283A (en) * | 2020-03-06 | 2020-07-28 | 拉扎斯网络科技(上海)有限公司 | Information processing method, information processing device, electronic equipment and computer readable storage medium |
CN111859184A (en) * | 2020-06-28 | 2020-10-30 | 北京嘀嘀无限科技发展有限公司 | Boarding point recommendation method and system |
CN112100480A (en) * | 2020-09-15 | 2020-12-18 | 北京百度网讯科技有限公司 | Search method, device, equipment and storage medium |
CN112650931A (en) * | 2021-01-04 | 2021-04-13 | 杭州情咖网络技术有限公司 | Content recommendation method |
CN112650931B (en) * | 2021-01-04 | 2023-05-30 | 杭州情咖网络技术有限公司 | Content recommendation method |
CN114936861A (en) * | 2022-07-21 | 2022-08-23 | 威海海洋职业学院 | Financial data sharing method and system based on big data |
CN114936861B (en) * | 2022-07-21 | 2022-10-04 | 威海海洋职业学院 | Financial data sharing method and system based on big data |
Also Published As
Publication number | Publication date |
---|---|
WO2018133624A1 (en) | 2018-07-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108319603A (en) | Object recommendation method and apparatus | |
CN108921221B (en) | User feature generation method, device, equipment and storage medium | |
CN110162695B (en) | Information pushing method and equipment | |
CN105701216B (en) | A kind of information-pushing method and device | |
CN104123332B (en) | The display methods and device of search result | |
CN112612957B (en) | Recommendation method of interest points and training method and device of recommendation model of interest points | |
CN107729578B (en) | Music recommendation method and device | |
CN104143005A (en) | Related searching system and method | |
CN113268641B (en) | User data processing method based on big data and big data server | |
CN111666450A (en) | Video recall method and device, electronic equipment and computer-readable storage medium | |
CN111159563A (en) | Method, device and equipment for determining user interest point information and storage medium | |
CN111191133B (en) | Service search processing method, device and equipment | |
CN104123321B (en) | A kind of determining method and device for recommending picture | |
CN113761084B (en) | POI search ranking model training method, ranking device, method and medium | |
Tan et al. | Preference-oriented mining techniques for location-based store search | |
KR100682552B1 (en) | System, apparatus and method for providing a weight to search engines according to situation of user and computer readable medium processing the method | |
CN115408618B (en) | Point-of-interest recommendation method based on social relation fusion position dynamic popularity and geographic features | |
CN115952426A (en) | Distributed noise data clustering method based on random sampling and user classification method | |
CN110188277A (en) | A kind of recommended method and device of resource | |
CN113890833B (en) | Network coverage prediction method, device, equipment and storage medium | |
CN114722290A (en) | Trust-relationship-fused ranking learning POI recommendation algorithm | |
CN110659406A (en) | Searching method and device | |
CN111741526B (en) | Positioning method, positioning device, electronic equipment and computer storage medium | |
CN113868481A (en) | Component acquisition method and device, electronic equipment and storage medium | |
CN112100670A (en) | Big data based privacy data grading protection method |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180724 |
|
RJ01 | Rejection of invention patent application after publication |