CN106557480A - Implementation method and device that inquiry is rewritten - Google Patents
Implementation method and device that inquiry is rewritten Download PDFInfo
- Publication number
- CN106557480A CN106557480A CN201510621932.2A CN201510621932A CN106557480A CN 106557480 A CN106557480 A CN 106557480A CN 201510621932 A CN201510621932 A CN 201510621932A CN 106557480 A CN106557480 A CN 106557480A
- Authority
- CN
- China
- Prior art keywords
- expansion word
- word
- history
- expansion
- feature
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2453—Query optimisation
- G06F16/24534—Query rewriting; Transformation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/242—Query formulation
- G06F16/2433—Query languages
- G06F16/2448—Query languages for particular applications; for extensibility, e.g. user defined types
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The disclosure provides implementation method and the device that a kind of inquiry is rewritten, and the method can include：Obtain the expansion word set corresponding with the searching keyword of user input；Determine satisfaction degree of each expansion word in the expansion word set to pre-set business target, and the expansion word in the expansion word set is ranked up according to the satisfaction degree；The expansion word of the first predetermined number of arrangement is chosen, using as the expansion word result corresponding with the searching keyword.By the technical scheme of the disclosure, expansion word result can be made to meet pre-set business target as much as possible.
Description
Technical field
It relates to search technique field, more particularly to the implementation method and device of inquiry rewriting.
Background technology
User under many scenes is required for using function of search.When search operation is performed, user can be with
Any search keyword is input into, and corresponding Search Results is provided by search engine.
However, the search keyword of user input is often relatively more random, user can not be directly embodied
Actual intention, cause Search Results meet the actual demand of user.
The content of the invention
In view of this, the disclosure provides implementation method and the device that a kind of inquiry is rewritten, and can make at expansion word
Reason result meets pre-set business target as much as possible.
For achieving the above object, disclosure offer technical scheme is as follows：
According to the first aspect of the disclosure, it is proposed that the implementation method that a kind of inquiry is rewritten, including：
Obtain the expansion word set corresponding with the searching keyword of user input；
Determine satisfaction degree of each expansion word in the expansion word set to pre-set business target, and root
The expansion word in the expansion word set is ranked up according to the satisfaction degree；
The expansion word of the first predetermined number of arrangement is chosen, using as corresponding with the searching keyword
Expand word result.
According to the second aspect of the disclosure, it is proposed that what a kind of inquiry was rewritten realizes device, including：
Acquiring unit, obtains the expansion word set corresponding with the searching keyword of user input；
Determining unit, determines satisfaction of each expansion word in the expansion word set to pre-set business target
Degree；
Sequencing unit, is ranked up to the expansion word in the expansion word set according to the satisfaction degree；
Unit is chosen, the expansion word of the first predetermined number of arrangement is chosen, to inquire about crucial as with described
The corresponding expansion word result of word.
From above technical scheme, the disclosure is expired to pre-set business target by determining each expansion word
Sufficient degree, and realize the sequence to expansion word accordingly and choose so that the expansion word for finally giving can be true
Protect with the maximization of the pre-set business target.
Description of the drawings
Fig. 1 is the flow chart of the implementation method that a kind of inquiry that one exemplary embodiment of the disclosure is provided is rewritten；
Fig. 2 is a kind of flow chart of training pattern that one exemplary embodiment of the disclosure is provided；
Fig. 3 is the flow process of the implementation method that another kind of inquiry that one exemplary embodiment of the disclosure is provided is rewritten
Figure；
Fig. 4 is the structural representation of a kind of electronic equipment that one exemplary embodiment of the disclosure is provided；
Fig. 5 is the block diagram for realizing device that a kind of inquiry that one exemplary embodiment of the disclosure is provided is rewritten.
Specific embodiment
As described by background section, due to the search keyword of user input it is more random, it is past
Toward its true intention can not be embodied, so as to cause Search Results not meet the actual demand of user.For
The technical problem is solved, QR (query rewrite, inquiry rewrite) is proposed in correlation technique and is processed
Means, can be analyzed by the search keyword to user input, and replace with automatically and can embody
The expansion word of the actual intention of user.
In the related, it is proposed that many kinds realize the technological means of QR.For example, such as：
(1) method of text similarity is directly based upon, by bag of words (BOW, bag of word) mould
Search keyword and expansion word are expressed as vector space by type, and similarity is calculated in vector space.
For mode (1), this mode can embody the similarity between text well；But,
This mode cannot be calculated does not have the similarity between the search keyword of co-occurrence word and expansion word (such as cannot
It is determined that the similarity of " Fructus Mali pumilae " and iphone between), and when same word has various explanations,
Be easy to bad (not meeting the actual demand of user) occur expansion word (as " Fructus Mali pumilae fruit basket " with
" i Phone ").
(2) directly QR process is carried out using the search data in same Session (session).
For mode (2), this mode is error-prone with respect to not allowing, and is better than mode in terms of accuracy rate
(1)；But, as user is unwilling to be input into repeatedly search keyword, cause the overlay capacity of this mode
And to push away word amount often little.
(3) using feedback data such as historical search daily records building the bigraph (bipartite graph) of search keyword-expansion word,
Similarity between each search keyword is calculated using modes such as SimRank.
For mode (3), the co-occurrence of this mode usage history search Web log mining clicks on feedback coefficient
According to method, can the information architecture that provides of a large amount of domestic consumers of effectively utilizes click on bigraph (bipartite graph), pass through
The methods such as SimRank can obtain the similarity between Query；But, this mode cannot be captured not
The similarity interosculated between the search keyword hit, the particularly especially many positions of new search key word
It is limited in terms of overlay capacity, such as " Fructus Mali pumilae Mobile phone " and " apple mobile phones ", such as the two words are in point
Hit and click on to transmit jointly on the whole link of bigraph (bipartite graph), then " Fructus Mali pumilae Mobile phone " and " apple
Mutually extension can not be realized between mobile phone ", therefore this mode also seems not enough on word coverage rate is pushed away.
(4) using a large number of services object data (such as title, description details etc.), excavate key word with
Frequent relation between key word；And set up similar between each search keyword using this frequent relation
Degree.
For mode (4), this mode is often searched for because not being that direct search key word sets out
Key word pushes away deficiency in terms of word coverage rate, and the search keyword of many long-tails is not covered substantially.
As can be seen here, there is the advantage of itself in above-mentioned every kind of QR modes, can obtain at corresponding aspect
To extraordinary expansion word；But meanwhile, every kind of QR modes there is also the inferior position of itself, cause pushing away word
Shortcomings in terms of coverage rate, accuracy, and can not often meet actual business objective.
Therefore, the disclosure is processed to expansion word based on pre-set business target, and by correlation technique
In one or more QR technological means combination, make full use of every kind of QR modes expand word in terms of
Advantage, to solve technical problem present in correlation technique.One of skill in the art will understand that, this
Disclosed technical scheme can be scaled up to the various of the internet, applications environment related to search and occupation mode
Under.
It is that the disclosure is further described, there is provided the following example：
Fig. 1 is the flow chart of the implementation method that a kind of inquiry that one exemplary embodiment of the disclosure is provided is rewritten,
As shown in figure 1, the method may comprise steps of：
Step 102, obtains the expansion word set corresponding with the searching keyword of user input.
In the present embodiment, expansion word set there may be various sources, as long as corresponding to user input
Searching keyword, the disclosure are not limited to this.
For example, in one embodiment, these expansion word set can be by third-party application to
The searching keyword of family input is processed and is derived, or directly by the expansion word set of outside input.
In one embodiment, the third-party application can be the remote application or remote entity based on the Internet,
It can also be the third-party application for locally being called.In one embodiment, the outside input can be base
The input that remote entity in the Internet is transmitted.
And one can with reference to or alternative embodiment in, these expansion word set can be according to predefined
One or more inquiry rewriting algorithms, carries out expansion word respectively and processes and obtain to the searching keyword of user input
Arrive,
In the present embodiment, the QR processing modes in correlation technique can be applied to " looking into herein
Ask rewriting algorithms ", the disclosure is not limited to this；In fact, by looking into using a greater variety of
Rewriting algorithms are ask, the technical scheme of the disclosure can be made to obtain the expansion word that these inquiry rewriting algorithms are obtained,
So that more preferably accuracy and coverage rate are realized in expansion word set.
Step 104, determines satisfaction of each expansion word in the expansion word set to pre-set business target
Degree, and the expansion word in the expansion word set is ranked up according to the satisfaction degree.
In the present embodiment, various types of business objectives can be pre-configured with, to meet practical business need
Ask.For example, it is assumed that business demand is to find the expansion word for meeting user's request as far as possible, then business
Target could be arranged to the displaying amount of object search, click volume etc.；It is assumed that business demand is pushed to user
More meet the desired product of user, then business objective could be arranged to product by quantity ordered, the quilt of product
Volumes of searches, clicked amount, sales volume etc.；Or, when needing to perform the popularization of advertising business, business is needed
It can be to find to ask, then business objective could be arranged to search
RPM (the Revenue Per Thousand Ad Impressions, per thousand advertising unit displayings of object
Income) etc..One of skill in the art will understand that, the business demand and business objective of the disclosure can quilts
Extend under the various of the internet, applications environment and occupation mode related to search.
Step 106, chooses the expansion word of the first predetermined number of arrangement, to inquire about crucial as with described
The corresponding expansion word result of word.
In the present embodiment, by determining satisfaction degree of each expansion word to pre-set business target, and according to
This realizes the sequence to expansion word and chooses so that the expansion word for finally giving is able to ensure that with the default industry
The maximization of business target.
It is to be noted that：In the related, QR algorithms itself are only focused on for the generation of expansion word
Process, search keyword of user input etc. such as so that expansion word is more fitted, and be not directed to it is right
The further screening of expansion word, is more not directed to be ranked up expansion word and sieved based on pre-set business target
The processing procedure of choosing.
So, in the technical scheme of the disclosure, in order to ensure satisfaction of the expansion word to pre-set business target
Situation, can be accomplished by：According to pre-configured expansion word analysis model, expansion is respectively obtained
Satisfaction degree of each expansion word in exhibition set of words to pre-set business target；Wherein, the expansion word is analyzed
Model is to be trained to obtain according to historical search daily record.In the technical scheme, searched by obtaining history
Suo Zhi, it can be realized that carry out expanding after word process to historical search key word, the history expansion word for obtaining
Situation is met to the actual of pre-set business target, so as to hereby it is possible to being more accurately predicted out for current
The search keyword of user input, each expansion word in corresponding expansion word set is to pre-set business target
Meet situation.
Therefore, according to the order of occurrence of whole process, the disclosure can include two stages：First, train
Model；2nd, inquiry is rewritten.Below for the above-mentioned two stage, it is described in detail respectively.
Stage one：Training pattern
Fig. 2 is a kind of flow chart of training pattern that one exemplary embodiment of the disclosure is provided, such as Fig. 2 institutes
Show, the training process of model may comprise steps of：
1) data Layer
Step 202, obtains historical search daily record.
In the present embodiment, the historical search daily record in random time section can be obtained；When historical search day
When will is more, trains the model for obtaining be better able to meet actual demand, but need data to be processed
Amount is likely to bigger.In one exemplary embodiment, it is assumed that choose the historical search daily record in 58 days.
In other alternative embodiments, the historical search daily record of 30 days, 90 days, 120 days or other natural law is chosen,
Can apply in the technical scheme of the disclosure.
Step 204, from historical search daily record, extracts historical query key word, history expansion word and each
The historic effect of dimension.
In the present embodiment, there are many historical query key words in historical search daily record, each history is looked into
Ask key word and there are corresponding one or more history expansion words, be all under historic state, to pass through QR algorithms
The expansion word that process is obtained.Wherein, if each historical query key word and a corresponding history are expanded
As a word to (pair), then to there is corresponding historic effect, and history effect in each word to exhibition word
Fruit there may be one or more dimensions, such as the amount of representing dimension, click volume dimension, clicking rate dimension,
Geographical position latitude, time latitude, RPM dimensions etc..
2) characteristic layer
Step 206, according to the data extracted in step 204 and pre-configured feature platform table, it is determined that
Corresponding foundation characteristic and advanced features, and form feature group.
In the present embodiment, " feature group " is by historical query keyword feature and history expansion word feature structure
Into history satisfaction degree of the disclosure exactly according to each feature group to pre-set business target, training are obtained
Final expansion word analysis model.
Wherein, the historical query key word phase in " historical query keyword feature " and historical search daily record
Close, and " history expansion word feature " is related to the history expansion word in historical search daily record.Therefore, walk
Rapid 206, exactly to data such as the historical query key word that obtains in step 204 and history expansion words, generate
Above-mentioned historical query keyword feature and history expansion word feature.
And specifically determining and filtering out " historical query keyword feature " and " history expansion word feature "
When, matching can be carried out according to pre-configured feature platform table from above-mentioned historical search daily record and be obtained.
Wherein, in feature platform table, record has pre-configured special to historical query keyword feature and history expansion word
The characterization information levied, the i.e. condition of execution Feature Selection, such as " historical query key word itself ",
" keyword that historical query key word is included ", " preset attribute of historical query key word ", " go through
History expansion word ", " keyword that history expansion word is included ", " preset attribute of history expansion word ",
" represent amount of the history expansion word in Preset Time window ", " click volume " etc., this also corresponds to down
The characteristic type that text will be included in each feature set mentioned.
Can be with reference to/alternative embodiment according to one, feature platform table is to match somebody with somebody according to different business demand in advance
Put.For example, feature platform table can be the data in the historical search daily record that basis ought be for the previous period
Collected and analyzed derived, a large amount of search and/or inquire about institute's general character that representative ought be interior for the previous period
Characterization information.Feature platform table can be updated regularly, and augment or delete.
(1) historical query keyword feature
The corresponding historical query keyword feature of each historical query key word can include by it is following at least it
The one first foundation feature set for constituting：The pass that the historical query key word, the historical query key word are included
The preset attribute of key word, the historical query key word.
For example, it is assumed that historical query key word is " Fructus Mali pumilae Mobile phone ", then historical query is crucial
The keyword that word is included can be the core word in " Fructus Mali pumilae Mobile phone ", such as " Fructus Mali pumilae ", " new
Money ", " mobile phone " etc.；And the preset attribute of historical query key word, can be crucial for the historical query
Prediction classification of word etc., such as " digital product " etc..
It can be seen that, in first foundation feature set, historical query keyword feature is all that the historical query is crucial
The content directly included in the word of word, or the attribute information of historical query key word, can be direct
Obtain processing without carrying out data statisticss etc., thus " foundation characteristic " referred in belonging to step 206.
(2) history expansion word feature
The corresponding history expansion word feature of each history expansion word includes second be made up of at least one of
Foundation characteristic collection：Keyword that the history expansion word, the history expansion word are included, the history expansion word
Preset attribute.
Second foundation characteristic collection is similar with above-mentioned first foundation feature set, refers in belonging to step 206
" foundation characteristic ".For example, it is assumed that when historical query key word is " Fructus Mali pumilae Mobile phone ",
One history expansion word is " iphone6P ", then the keyword that history expansion word is included can be " iphone6P "
Core word, such as " iphone ", " 6P " etc.；And the preset attribute of history expansion word, Ke Yiwei
Any attribute of the history expansion word, such as in an exemplary application scene, a such as O2O (Online
To Offline, on line under line) businessman commercial articles searching, then the preset attribute of history expansion word can be
(i.e. corresponding commodity of history expansion word, the purchase frequency of such as commodity are higher, then temperature is got over for commodity temperature
It is high), the search frequency (frequency/number of times for operating such as is scanned for using the history expansion word), place an order
The frequency (the lower unifrequency/number of times after operating is scanned for using the history expansion word such as), geographic preferences
(user's more preference commodity in such as a certain area), (such as user more concentrates on time preference
Place an order within certain time period), user preference (such as meet the user of the features such as certain class sex, age
More preference commodity), these attributes can by O2O businessmans order volume over a period to come,
The identity information (comprising customer position information, sex, age etc.) of register user, user price preference
Obtain etc. being analyzed.
Certainly, the technical scheme of the disclosure can also be applied to more scenes.Such as user is using arbitrarily searching
Index is held up under the scene of execution information retrieval, and the preset attribute of history expansion word can be searching for active user
(such as when historical query key word is " one-piece dress ", history expansion word is " red to rope historical data
One-piece dress "), (such as user has recently input searching keyword for " purple connects clothing to current search data
Skirt "), the global search hot word of search engine (such as many users tend to " yellow one-piece dress ")
Deng, so as to be input in user again when " one-piece dress ", can be to " red one-piece dress ", " purple company
Clothing skirt " and " yellow one-piece dress " are extended and show.For another example, if history expansion word is business
The word of bidding (Bidword) of family's purchase, then the preset attribute of the history expansion word can be corresponding including which
Buy depth, recall depth, average bid, maximum bid etc..
One of skill in the art will understand that, the preset attribute of the disclosure can be according to each of internet, applications
Kind with the environment and occupation mode of search correlation and carry out corresponding customizing and/or changing.
Optionally, in addition to the second foundation characteristic collection, the corresponding history expansion word feature of history expansion word
The advanced features collection being made up of at least one of can also be included：The history expansion word is in Preset Time window
Mouthful in represent the click situation statistical value of situation statistical value, the history expansion word in Preset Time window.
Specifically, such as history expansion word " iphone6P " is in continuous 8 days, in 5 days, in 3 days, 1 day
The statistical nature of the interior amount of representing, click volume, clicking rate and other effects feedback data.It can be seen that, advanced features
Collection is the feature that the statistical disposition based on feedback data is furthermore achieved that on the basis of foundation characteristic collection
Set.
In sum, the feature that step 206 is obtained can include：
Foundation characteristic：Historical search key word, the core word of historical search key word, historical search are crucial
Preset attribute, history expansion word, the core word of history expansion word, the history expansion words such as the prediction classification of word
The preset attribute such as purchase depth.Wherein, between historical search keyword feature and history expansion word feature
Simple characteristic crossover is there is likely to be, such as by the core word and history expansion word of historical search key word
Core word carry out intersecting and obtain a new foundation characteristic.
Advanced features：Represent amount of the history expansion word in Preset Time window, click volume and other effects are fed back
The statistical nature of data.
Accordingly, each historical search keyword feature and corresponding each history expansion word feature can be with
A feature group is combined as, and pre-set business target is expired comprising each feature group in historical search daily record
" historic effect " referred in sufficient situation, i.e. step 204.
3) model layer
Step 208, carries out numerical discretization to the corresponding business objective of each feature group.
In the present embodiment, whole feature architecture is constituted by above-mentioned foundation characteristic and advanced features,
Such that it is able to this feature system to be expressed as the input example of model, then each input example include one it is special
Levy group and this feature group and situation is met to pre-set business target.For example, it is assumed that pre-set business mesh
RPM value is designated as, then each corresponding RPM value (" meeting situation ") of input example can be entered
Row discretization, then the value after discretization is exactly the target of model sequence.
Certainly, pre-set business target can also include other types, this industry with the technical scheme of the disclosure
Business scene is related.Such as the business scenario can include at least one of：Information retrieval scene, sale
Scene, advertisement bidding scene etc., the disclosure is not limited to this.So, in information retrieval scene
Under, pre-set business target can be " clicking rate ", " rate is adopted in search " or " click volume " etc.；
Under sale scene, pre-set business target can be " place an order rate ", " amount of placing an order ", or " sales volume "
Deng；Under advertisement bidding scene, pre-set business target can be above-mentioned RPM value etc..
Step 210, the RankSVM algorithms based on pairwise-learning (learning in pairs) are to above-mentioned
Input example be trained, obtain corresponding expansion word analysis model.
In the present embodiment, can by it is all of input example carry out IDization, be then divided into training set and
Two parts of test set, training set part are used for model training herein, and test set part is used for subsequently
Model checking.
In the present embodiment, for inquiry q (query), multiple input examples and extension can be obtained
Input example.It is when using being trained based on the RankSVM algorithms of pairwise-learning, main
If optimization is with minor function：
Where it is assumed that xi、xjFor i-th and j-th input example (instance), which may include m
Individual feature (m is positive integer), then the target of model training is exactly to be madeDuring w when minimum, w are all input examples
All features corresponding to the characteristic vector that constituted of feature weight.C is constant value, can be according to mould
The hypothesis of type carrys out value.
Additionally, P defines the scope of i and j, it is defined as follows：
P ≡ (i, j) | qi=qj, yi＞ yj} (2)
Wherein, qi=qjExpression requires xiExample and xjExample belongs to same inquiry query.And according to industry
The difference of business scene, yiThere can be different expressions.For example, if business scenario is search environment, and
The business objective of concern is that rate is adopted in the search of user, then yiRepresent that i-th instance is corresponding discrete
Rate is adopted in search after change.If business scenario is shopping environment, and the business objective paid close attention to is the amount of placing an order,
Then yi represents the amount of placing an order after the corresponding discretization of i-th instance.If business scenario is advertisement pushed away
Ringoid border, and the business objective paid close attention to is RPM value, then yiRepresent that i-th instance is corresponding discrete
RPM value after change.Therefore, yi＞ yjMean that and required xiBusiness objective after the discretization of example
It is greater than xjExample.
In the present embodiment, by adopting RankSVM algorithms based on pairwise-learning, can be with
Which is made full use of to be directed to the training characteristic of " to (pair) " feature, for other algorithms,
When being trained to above-mentioned feature group, contribute to the accuracy and effectiveness of lift scheme.
Step 212, verifies to the expansion word analysis model that training is obtained.
In the present embodiment, can be according to the input example of above-mentioned " another part ", to the expansion word
Analysis model verified, the NDCG of the expansion word analysis model is such as calculated by test set
(Normalized Discounted Cumulative Gain) value, to assess the ordering scenario of the model.
Through actual test, the NDCG values of the expansion word analysis model obtained based on disclosure training can reach
More than 0.8.
Stage two：Inquiry is rewritten
Fig. 3 is the flow process of the implementation method that another kind of inquiry that one exemplary embodiment of the disclosure is provided is rewritten
Figure, as shown in figure 3, the process for performing inquiry rewriting may comprise steps of：
Step 302, based on one or more QR algorithm, expands to the search keyword of user input
Word process.
In the present embodiment, the arbitrary QR algorithms in correlation technique can apply to the enforcement of the disclosure
In example, the disclosure is not limited to this.
Step 304, merges to the expansion word result that all QR algorithms are obtained, and be expanded set of words.
In the present embodiment, in addition to the set of words that is expanded by QR algorithms, it is also possible to by which
His any-mode obtains the expansion word set；For example, such as by third-party application to user input
Searching keyword is processed and is derived and obtained or directly by outside input etc..
Step 306, carries out the prescreening based on dependency to the expansion word in expansion word set.
In the present embodiment, it is less than to the degree of association of searching keyword in screening out expansion word set default related
The expansion word of degree, to guarantee that expansion word has certain dependency with the search keyword of user input.Lift
For example, when the search keyword of user input is " Fructus Mali pumilae Mobile phone ", if expansion word is " bar
Western Semen Pini ", then as difference between the two is huge, should be screened out.As an exemplary enforcement
Mode, when expansion word belongs to same classification with search keyword, it is believed that the expansion word and searching keyword
Degree of association reach default degree of association, and inhomogeneity purpose expansion word is screened out.
Certainly, in addition to affiliated classification, it is clear that by other means expansion word can also be realized being based on
The prescreening of dependency is processed, and the disclosure is not limited to this.For example, can such as screen out
The expansion word not " paid close attention to by user " in expansion word set；Wherein, under information retrieval scene, " quilt
User pays close attention to " can be understood as：Concentrated the expansion word of search in a short time always by a large number of users；In O2O
Under sale scene, " being paid close attention to by user " can be understood as：In a short time always by a large number of users (example of interest
Such as, put into collection list), bought, or the expansion word commented on；When search platform is needed to user
When carrying out advertisement promotion, " being paid close attention to by user " herein can be understood as：It is word of bidding by businessman's purchase,
Especially when pre-set business target is to lift the advertisement business revenue of search platform, by screening out non-word of bidding,
The word that may insure to bid has more preferable RPM value.
Step 308, according to pre-configured expansion word analysis model, extends to each in expansion word set
Word is ranked up process.
In the present embodiment, by embodiment as shown in Figure 2, training obtains expansion word analysis herein
Model.Embodiment as shown in Figure 2 understands that the weighted value that the model goes out according to training in advance, prediction are worked as
Meet situation of each expansion word of front input to pre-set business target, and situation is met according to this arranged
Sequence process.Wherein, shown in the computational methods of the weighted value such as formula (3).
Score=WTX=ΣJ=1wj*xj (3)
It can be seen that, by the model training of the disclosure and inquiry rewriting process, so as to get expansion word be all with
For the purpose of pre-set business target, as long as thus the business objective needed for being input into during model training, i.e.,
Can ensure that the expansion word for finally giving meets actual demand.
Step 310, according to expansion word quantity set in advance, chooses the first multiple expansion words of arrangement.
In the present embodiment, the consideration based on aspects such as system operations abilities, only needs to certain amount every time
Expansion word, thus the first multiple expansion words of arrangement can be chosen, i.e. the satisfaction to pre-set business target
The optimum multiple expansion words of situation；For example when expansion word quantity is 3, selection is arranged in the extension of first three
Word.
Fig. 4 shows the schematic configuration diagram of the electronic equipment of the exemplary embodiment according to the disclosure.Please
With reference to Fig. 4, in hardware view, the electronic equipment includes processor, internal bus, network interface, interior
Deposit and nonvolatile memory, the hardware being also possible that required for other business certainly.Processor from
Corresponding computer program is read in nonvolatile memory in internal memory and then is run, on logic level
Form inquiry rewriting realizes device.Certainly, in addition to software realization mode, the disclosure is not precluded from
Mode of other implementations, such as logical device or software and hardware combining etc., that is to say, that locate below
The executive agent of reason flow process is not limited to each logical block, or hardware or logical device.
Fig. 5 is refer to, in Software Implementation, what the inquiry was rewritten realizes that device can include expanding word
Unit, determining unit, sequencing unit and selection unit.Wherein：
Acquiring unit, obtains the expansion word set corresponding with the searching keyword of user input；
Determining unit, determines satisfaction of each expansion word in the expansion word set to pre-set business target
Degree；
Sequencing unit, is ranked up to the expansion word in the expansion word set according to the satisfaction degree；
Unit is chosen, the expansion word of the first predetermined number of arrangement is chosen, to inquire about crucial as with described
The corresponding expansion word result of word.
Optionally, the acquiring unit obtains the expansion word set by least one of following manner：
According to predefined one or more inquiry rewriting algorithms, the searching keyword of user input is carried out
Expand word to process, obtain the expansion word set corresponding with the searching keyword of user input；
Reception is processed and derived expansion-word set to the searching keyword of user input by third-party application
Close；And
Receive the expansion word set corresponding with the searching keyword of user input directly by outside input.
Optionally, the determining unit specifically for：
According to pre-configured expansion word analysis model, each extension in the expansion word set is respectively obtained
Satisfaction degree of the word to the pre-set business target；Wherein, the expansion word analysis model is according to history
Search daily record is trained and obtains.
Optionally, the expansion word analysis model is according to extracting from the historical search daily record by going through
The feature group that history searching keyword feature is constituted with history expansion word feature, and each feature group is to described
The history satisfaction degree of pre-set business target is trained and obtains；
Wherein, the historical query keyword feature is crucial with the historical query in the historical search daily record
Word is related, and the history expansion word feature is related to the history expansion word in the historical search daily record.
Optionally, the historical query keyword feature and the history expansion word are characterized in that according to pre-configured
Feature platform table match from the historical search daily record and obtain；
Wherein, in the feature platform table record have it is pre-configured to the historical query keyword feature and
The characterization information of the history expansion word feature.
Optionally, the corresponding historical query keyword feature of each historical query key word include by with down to
One of few first foundation feature set for constituting：
Keyword that the historical query key word, the historical query key word are included, the history are looked into
Ask the preset attribute of key word.
Optionally, the corresponding history expansion word feature of each history expansion word is included by least one of structure
Into the second foundation characteristic collection：
Keyword that the history expansion word, the history expansion word are included, the history expansion word it is pre-
If attribute.
Optionally, the corresponding history expansion word feature of each history expansion word is also included by least one of
The advanced features collection of composition：
The history expansion word represents situation statistical value, the history expansion word in Preset Time window
Click situation statistical value in Preset Time window.
Optionally, the expansion word analysis model is by based on study pairwise-learning in pairs
RankSVM algorithms are trained to the feature group and obtain.
Optionally, also include：
First screens out unit, the expansion word do not paid close attention to by user in screening out the expansion word set.
Optionally, also include：
Second screens out unit, screens out in the expansion word set and is less than with the degree of association of the searching keyword
The expansion word of default degree of association.
Optionally, the pre-set business target is related to business scenario, the business scenario include with down to
It is one of few：
Information retrieval scene, sale scene, advertisement bidding scene.
Optionally, when the business scenario is information retrieval scene, the pre-set business target is following
At least one：Rate, click volume are adopted in clicking rate, search；When the business scenario is to sell scene,
The pre-set business target is at least one of：Place an order rate, the amount of placing an order, sales volume；When the business
Scene be advertisement bidding scene when, the pre-set business target for advertising unit displaying number of times reach it is default
Revenue during number of times.
One it is typical configure, computing device include one or more processors (CPU), input/
Output interface, network interface and internal memory.
Internal memory potentially includes the volatile memory in computer-readable medium, random access memory
And/or the form, such as read only memory (ROM) or flash memory (flash such as Nonvolatile memory (RAM)
RAM).Internal memory is the example of computer-readable medium.
Computer-readable medium includes that permanent and non-permanent, removable and non-removable media can be by
Any method or technique is realizing information Store.Information can be computer-readable instruction, data structure,
The module of program or other data.The example of the storage medium of computer includes, but are not limited to phase transition internal memory
(PRAM), static RAM (SRAM), dynamic random access memory (DRAM),
Other kinds of random access memory (RAM), read only memory (ROM), electrically erasable
Read only memory (EEPROM), fast flash memory bank or other memory techniques, read-only optical disc read only memory
(CD-ROM), digital versatile disc (DVD) or other optical storages, magnetic cassette tape, tape magnetic
Disk storage or other magnetic storage apparatus or any other non-transmission medium, can be used for storage can be counted
The information that calculation equipment is accessed.Define according to herein, computer-readable medium does not include that temporary computer can
Read media (transitory media), the such as data signal and carrier wave of modulation.
Also, it should be noted that term " including ", "comprising" or its any other variant be intended to it is non-
Exclusiveness is included, so that a series of process, method, commodity or equipment including key elements is not only
Including those key elements, but also including other key elements being not expressly set out, or also include for this
The intrinsic key element of process, method, commodity or equipment.In the absence of more restrictions, by language
The key element that sentence "including a ..." is limited, it is not excluded that in the process including the key element, method, business
Also there is other identical element in product or equipment.
The preferred embodiment of the disclosure is the foregoing is only, it is not to limit the disclosure, all at this
Within disclosed spirit and principle, any modification, equivalent substitution and improvements done etc. should be included in
Within the scope of disclosure protection.
Claims (26)
1. the implementation method that a kind of inquiry is rewritten, it is characterised in that include：
Obtain the expansion word set corresponding with the searching keyword of user input；
Determine satisfaction degree of each expansion word in the expansion word set to pre-set business target, and root
The expansion word in the expansion word set is ranked up according to the satisfaction degree；
The expansion word of the first predetermined number of arrangement is chosen, using as corresponding with the searching keyword
Expand word result.
2. method according to claim 1, it is characterised in that the acquisition is looked into user input
Ask key word and mutually obtain corresponding expansion word set, including at least one of：
According to predefined one or more inquiry rewriting algorithms, the searching keyword of user input is carried out
Expand word to process, obtain the expansion word set corresponding with the searching keyword of user input；
Reception is processed and derived expansion-word set to the searching keyword of user input by third-party application
Close；And
Receive the expansion word set corresponding with the searching keyword of user input directly by outside input.
3. method according to claim 1, it is characterised in that the determination expansion word set
In satisfaction degree of each expansion word to pre-set business target, including：
According to pre-configured expansion word analysis model, each extension in the expansion word set is respectively obtained
Satisfaction degree of the word to the pre-set business target；Wherein, the expansion word analysis model is according to history
Search daily record is trained and obtains.
4. method according to claim 3, it is characterised in that the expansion word analysis model is root
According to extracting from the historical search daily record by historical query keyword feature and history expansion word feature structure
Into feature group, and each feature group is trained to the history satisfaction degree of the pre-set business target
Obtain；
Wherein, the historical query keyword feature is crucial with the historical query in the historical search daily record
Word is related, and the history expansion word feature is related to the history expansion word in the historical search daily record.
5. method according to claim 4, it is characterised in that the historical query keyword feature
It is characterized in that according to pre-configured feature platform table from the historical search daily record with the history expansion word
With obtaining；
Wherein, in the feature platform table record have it is pre-configured to the historical query keyword feature and
The characterization information of the history expansion word feature.
6. method according to claim 4, it is characterised in that each historical query key word correspondence
Historical query keyword feature include the first foundation feature set being made up of at least one of：
Keyword that the historical query key word, the historical query key word are included, the history are looked into
Ask the preset attribute of key word.
7. method according to claim 4, it is characterised in that history expansion word is corresponding goes through for each
History expansion word feature includes the second foundation characteristic collection being made up of at least one of：
Keyword that the history expansion word, the history expansion word are included, the history expansion word it is pre-
If attribute.
8. method according to claim 7, it is characterised in that history expansion word is corresponding goes through for each
History expansion word feature also includes the advanced features collection being made up of at least one of：
The history expansion word represents situation statistical value, the history expansion word in Preset Time window
Click situation statistical value in Preset Time window.
9. method according to claim 3, it is characterised in that the expansion word analysis model is logical
Cross and the feature group is trained based on the RankSVM algorithms of study pairwise-learning in pairs
Arrive.
10. method according to claim 1, it is characterised in that also include：
The expansion word do not paid close attention to by user in screening out the expansion word set.
11. methods according to claim 1, it is characterised in that also include：
The expansion of default degree of association is less than in screening out the expansion word set with the degree of association of the searching keyword
Exhibition word.
12. methods according to claim 1, it is characterised in that the pre-set business target and industry
Business scene is related, and the business scenario includes at least one of：
Information retrieval scene, sale scene, advertisement bidding scene.
13. methods according to claim 12, it is characterised in that
When the business scenario is information retrieval scene, the pre-set business target is at least one of：
Rate, click volume are adopted in clicking rate, search；
When the business scenario is to sell scene, the pre-set business target is at least one of：Under
Single rate, the amount of placing an order, sales volume；
When the business scenario is advertisement bidding scene, exhibition of the pre-set business target for advertising unit
Revenue when showing that number of times reaches preset times.
What a kind of 14. inquiries were rewritten realizes device, it is characterised in that include：
Acquiring unit, obtains the expansion word set corresponding with the searching keyword of user input；
Determining unit, determines satisfaction of each expansion word in the expansion word set to pre-set business target
Degree；
Sequencing unit, is ranked up to the expansion word in the expansion word set according to the satisfaction degree；
Unit is chosen, the expansion word of the first predetermined number of arrangement is chosen, to inquire about crucial as with described
The corresponding expansion word result of word.
15. devices according to claim 14, it is characterised in that the acquiring unit is by following
At least one of mode, obtains the expansion word set：
According to predefined one or more inquiry rewriting algorithms, the searching keyword of user input is carried out
Expand word to process, obtain the expansion word set corresponding with the searching keyword of user input；
Reception is processed and derived expansion-word set to the searching keyword of user input by third-party application
Close；And
Receive the expansion word set corresponding with the searching keyword of user input directly by outside input.
16. devices according to claim 14, it is characterised in that the determining unit specifically for：
According to pre-configured expansion word analysis model, each extension in the expansion word set is respectively obtained
Satisfaction degree of the word to the pre-set business target；Wherein, the expansion word analysis model is according to history
Search daily record is trained and obtains.
17. devices according to claim 16, it is characterised in that the expansion word analysis model is
According to extracting from the historical search daily record by historical query keyword feature and history expansion word feature
The feature group of composition, and each feature group instructed to the history satisfaction degree of the pre-set business target
Get；
Wherein, the historical query keyword feature is crucial with the historical query in the historical search daily record
Word is related, and the history expansion word feature is related to the history expansion word in the historical search daily record.
18. devices according to claim 17, it is characterised in that the historical query key word is special
The history expansion word of seeking peace is characterized in that according to pre-configured feature platform table from the historical search daily record
Matching is obtained；
Wherein, in the feature platform table record have it is pre-configured to the historical query keyword feature and
The characterization information of the history expansion word feature.
19. devices according to claim 17, it is characterised in that each historical query key word pair
The historical query keyword feature answered includes the first foundation feature set being made up of at least one of：
Keyword that the historical query key word, the historical query key word are included, the history are looked into
Ask the preset attribute of key word.
20. devices according to claim 17, it is characterised in that each history expansion word is corresponding
History expansion word feature includes the second foundation characteristic collection being made up of at least one of：
Keyword that the history expansion word, the history expansion word are included, the history expansion word it is pre-
If attribute.
21. devices according to claim 20, it is characterised in that each history expansion word is corresponding
History expansion word feature also includes the advanced features collection being made up of at least one of：
The history expansion word represents situation statistical value, the history expansion word in Preset Time window
Click situation statistical value in Preset Time window.
22. devices according to claim 16, it is characterised in that the expansion word analysis model is
By being trained to the feature group based on the RankSVM algorithms of study pairwise-learning in pairs
Obtain.
23. devices according to claim 14, it is characterised in that also include：
First screens out unit, the expansion word do not paid close attention to by user in screening out the expansion word set.
24. devices according to claim 14, it is characterised in that also include：
Second screens out unit, screens out in the expansion word set and is less than with the degree of association of the searching keyword
The expansion word of default degree of association.
25. devices according to claim 14, it is characterised in that the pre-set business target and industry
Business scene is related, and the business scenario includes at least one of：
Information retrieval scene, sale scene, advertisement bidding scene.
26. devices according to claim 25, it is characterised in that
When the business scenario is information retrieval scene, the pre-set business target is at least one of：
Rate, click volume are adopted in clicking rate, search；
When the business scenario is to sell scene, the pre-set business target is at least one of：Under
Single rate, the amount of placing an order, sales volume；
When the business scenario is advertisement bidding scene, exhibition of the pre-set business target for advertising unit
Revenue when showing that number of times reaches preset times.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510621932.2A CN106557480B (en) | 2015-09-25 | 2015-09-25 | Method and device for realizing query rewriting |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510621932.2A CN106557480B (en) | 2015-09-25 | 2015-09-25 | Method and device for realizing query rewriting |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106557480A true CN106557480A (en) | 2017-04-05 |
| CN106557480B CN106557480B (en) | 2020-07-07 |
Family
ID=58414612
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510621932.2A Active CN106557480B (en) | 2015-09-25 | 2015-09-25 | Method and device for realizing query rewriting |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106557480B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108021641A (en) * | 2017-11-29 | 2018-05-11 | 有米科技股份有限公司 | The method and apparatus that the association keyword of application is expanded |
| CN109003146A (en) * | 2018-08-31 | 2018-12-14 | 百度在线网络技术（北京）有限公司 | Business datum promotion method, device, terminal and computer readable storage medium |
| CN109117475A (en) * | 2018-07-02 | 2019-01-01 | 武汉斗鱼网络科技有限公司 | A kind of method and relevant device of text rewriting |
| CN109657145A (en) * | 2018-12-20 | 2019-04-19 | 拉扎斯网络科技（上海）有限公司 | Trade company's searching method and device, electronic equipment and computer readable storage medium |
| CN109857845A (en) * | 2019-01-03 | 2019-06-07 | 北京奇艺世纪科技有限公司 | Model training and data retrieval method, device, terminal and computer readable storage medium |
| CN110110207A (en) * | 2018-01-18 | 2019-08-09 | 北京搜狗科技发展有限公司 | A kind of information recommendation method, device and electronic equipment |
| CN110909217A (en) * | 2018-09-12 | 2020-03-24 | 北京奇虎科技有限公司 | Method and device for realizing search, electronic equipment and storage medium |
| CN110968759A (en) * | 2018-09-30 | 2020-04-07 | 北京奇虎科技有限公司 | Method and device for training rewriting model |
| CN111428119A (en) * | 2020-02-18 | 2020-07-17 | 北京三快在线科技有限公司 | Query rewriting method and device and electronic equipment |
| CN112052303A (en) * | 2019-06-06 | 2020-12-08 | 阿里巴巴集团控股有限公司 | Keyword weight determination method and device and computing equipment |
| CN112256953A (en) * | 2019-07-22 | 2021-01-22 | 腾讯科技（深圳）有限公司 | Query rewriting method and device, computer equipment and storage medium |
| WO2021136009A1 (en) * | 2019-12-31 | 2021-07-08 | 阿里巴巴集团控股有限公司 | Search information processing method and apparatus, and electronic device |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020095405A1 (en) * | 2001-01-18 | 2002-07-18 | Hitachi America, Ltd. | View definition with mask for cell-level data access control |
| CN1954321A (en) * | 2004-03-31 | 2007-04-25 | Google公司 | Query rewriting with entity detection |
| US20080059439A1 (en) * | 2006-08-30 | 2008-03-06 | Lucent Technologies Inc. | Query Translation from XPath to SQL in the Presence of Recursive DTDs |
| CN102682001A (en) * | 2011-03-09 | 2012-09-19 | 阿里巴巴集团控股有限公司 | Method and device for determining suggest word |
| CN103136224A (en) * | 2011-11-24 | 2013-06-05 | 百度时代网络技术（北京）有限公司 | Recommendation method and device for keywords |
| CN103365904A (en) * | 2012-04-05 | 2013-10-23 | 阿里巴巴集团控股有限公司 | Advertising information searching method and system |
| CN104252456A (en) * | 2013-06-25 | 2014-12-31 | 阿里巴巴集团控股有限公司 | Method, device and system for weight estimation |
| CN104636334A (en) * | 2013-11-06 | 2015-05-20 | 阿里巴巴集团控股有限公司 | Keyword recommending method and device |
-
2015
- 2015-09-25 CN CN201510621932.2A patent/CN106557480B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020095405A1 (en) * | 2001-01-18 | 2002-07-18 | Hitachi America, Ltd. | View definition with mask for cell-level data access control |
| CN1954321A (en) * | 2004-03-31 | 2007-04-25 | Google公司 | Query rewriting with entity detection |
| US20080059439A1 (en) * | 2006-08-30 | 2008-03-06 | Lucent Technologies Inc. | Query Translation from XPath to SQL in the Presence of Recursive DTDs |
| CN102682001A (en) * | 2011-03-09 | 2012-09-19 | 阿里巴巴集团控股有限公司 | Method and device for determining suggest word |
| CN103136224A (en) * | 2011-11-24 | 2013-06-05 | 百度时代网络技术（北京）有限公司 | Recommendation method and device for keywords |
| CN103365904A (en) * | 2012-04-05 | 2013-10-23 | 阿里巴巴集团控股有限公司 | Advertising information searching method and system |
| CN104252456A (en) * | 2013-06-25 | 2014-12-31 | 阿里巴巴集团控股有限公司 | Method, device and system for weight estimation |
| CN104636334A (en) * | 2013-11-06 | 2015-05-20 | 阿里巴巴集团控股有限公司 | Keyword recommending method and device |
Non-Patent Citations (1)
| Title |
|---|
| 王秉卿等: ""机器学习的查询扩展在博客检索中的应用"", 《中文信息学报》 * |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108021641A (en) * | 2017-11-29 | 2018-05-11 | 有米科技股份有限公司 | The method and apparatus that the association keyword of application is expanded |
| CN108021641B (en) * | 2017-11-29 | 2019-07-19 | 有米科技股份有限公司 | The method and apparatus that the association keyword of application is expanded |
| CN110110207A (en) * | 2018-01-18 | 2019-08-09 | 北京搜狗科技发展有限公司 | A kind of information recommendation method, device and electronic equipment |
| CN110110207B (en) * | 2018-01-18 | 2023-11-03 | 北京搜狗科技发展有限公司 | Information recommendation method and device and electronic equipment |
| CN109117475B (en) * | 2018-07-02 | 2022-08-16 | 武汉斗鱼网络科技有限公司 | Text rewriting method and related equipment |
| CN109117475A (en) * | 2018-07-02 | 2019-01-01 | 武汉斗鱼网络科技有限公司 | A kind of method and relevant device of text rewriting |
| CN109003146A (en) * | 2018-08-31 | 2018-12-14 | 百度在线网络技术（北京）有限公司 | Business datum promotion method, device, terminal and computer readable storage medium |
| CN110909217A (en) * | 2018-09-12 | 2020-03-24 | 北京奇虎科技有限公司 | Method and device for realizing search, electronic equipment and storage medium |
| CN110968759A (en) * | 2018-09-30 | 2020-04-07 | 北京奇虎科技有限公司 | Method and device for training rewriting model |
| CN109657145A (en) * | 2018-12-20 | 2019-04-19 | 拉扎斯网络科技（上海）有限公司 | Trade company's searching method and device, electronic equipment and computer readable storage medium |
| CN109857845A (en) * | 2019-01-03 | 2019-06-07 | 北京奇艺世纪科技有限公司 | Model training and data retrieval method, device, terminal and computer readable storage medium |
| CN109857845B (en) * | 2019-01-03 | 2021-06-22 | 北京奇艺世纪科技有限公司 | Model training and data retrieval method, device, terminal and computer-readable storage medium |
| CN112052303A (en) * | 2019-06-06 | 2020-12-08 | 阿里巴巴集团控股有限公司 | Keyword weight determination method and device and computing equipment |
| CN112256953A (en) * | 2019-07-22 | 2021-01-22 | 腾讯科技（深圳）有限公司 | Query rewriting method and device, computer equipment and storage medium |
| CN112256953B (en) * | 2019-07-22 | 2023-11-14 | 腾讯科技（深圳）有限公司 | Query rewrite method, query rewrite apparatus, computer device, and storage medium |
| WO2021136009A1 (en) * | 2019-12-31 | 2021-07-08 | 阿里巴巴集团控股有限公司 | Search information processing method and apparatus, and electronic device |
| CN111428119A (en) * | 2020-02-18 | 2020-07-17 | 北京三快在线科技有限公司 | Query rewriting method and device and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106557480B (en) | 2020-07-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106557480A (en) | Implementation method and device that inquiry is rewritten | |
| WO2019233258A1 (en) | Method, apparatus and system for sending information, and computer-readable storage medium | |
| CN104866474B (en) | Individuation data searching method and device | |
| US11243992B2 (en) | System and method for information recommendation | |
| KR101419504B1 (en) | System and method providing a suited shopping information by analyzing the propensity of an user | |
| CN111444395B (en) | Method, system and equipment for obtaining relation expression between entities and advertisement recall system | |
| CN105989004A (en) | Information releasing pretreatment method and device | |
| WO2018053966A1 (en) | Click rate estimation | |
| US8660901B2 (en) | Matching of advertising sources and keyword sets in online commerce platforms | |
| CN109409928A (en) | A kind of material recommended method, device, storage medium, terminal | |
| CN111737418B (en) | Method, apparatus and storage medium for predicting relevance of search term and commodity | |
| US10896458B2 (en) | Method, system, and computer-readable medium for product and vendor selection | |
| CN107644036A (en) | A kind of method, apparatus and system of data object push | |
| CN116308684B (en) | Online shopping platform store information pushing method and system | |
| CN113268656A (en) | User recommendation method and device, electronic equipment and computer storage medium | |
| CN114663197A (en) | Commodity recommendation method and device, equipment, medium and product thereof | |
| Dias et al. | Automating the extraction of static content and dynamic behaviour from e-commerce websites | |
| CN113781139A (en) | Item recommendation method, item recommendation device, equipment and medium | |
| CN115423555A (en) | Commodity recommendation method and device, electronic equipment and storage medium | |
| CN108268519A (en) | A kind of method and apparatus of recommendation network object | |
| CN112036987B (en) | Method and device for determining recommended commodity | |
| CN111931069B (en) | User interest determination method and device and computer equipment | |
| KR102585895B1 (en) | Method and system for increasing keyword marketing efficiency in open market | |
| CN111581492A (en) | Content recommendation method, computing device and readable storage medium | |
| CN105096149B (en) | Business tine method for analyzing product association and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |