CN107958406A - Inquire about acquisition methods, device and the terminal of data - Google Patents
Inquire about acquisition methods, device and the terminal of data Download PDFInfo
- Publication number
- CN107958406A CN107958406A CN201711243962.XA CN201711243962A CN107958406A CN 107958406 A CN107958406 A CN 107958406A CN 201711243962 A CN201711243962 A CN 201711243962A CN 107958406 A CN107958406 A CN 107958406A
- Authority
- CN
- China
- Prior art keywords
- query
- inquiry
- query characteristics
- maximum entropy
- disaggregated model
- 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
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- 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
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- 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
- G06F16/35—Clustering; Classification
Abstract
The invention discloses a kind of acquisition methods, device and terminal for inquiring about data, is related to Internet technical field, main purpose is to improve the accuracy for obtaining inquiry data, while reduces unnecessary manual resource expense.The main technical schemes of the present invention include:The query characteristics for obtaining target query result are extracted according to searching keyword;The query characteristics are inputted default inquiry disaggregated model to be calculated, obtain the maximum entropy of the query characteristics, wherein, the default inquiry disaggregated model is used to be trained to obtain the corresponding maximum entropy of each query characteristics according to training examples, and the query characteristics of maximum entropy correspond to a query categories;Correspondence between query characteristics and query categories based on the maximum entropy, determines the query categories corresponding to the maximum entropy of the query characteristics;Return and belong to the same query categories the and relevant target query result with the searching keyword.Present invention is mainly applied to during inquiry data are obtained in taking out platform.
Description
Technical field
The present invention relates to Internet technical field, more particularly to a kind of acquisition methods, device and terminal for inquiring about data.
Background technology
With the very fast development of Internet technology and social economy, internet increasingly influences the daily work of people
Make, learn and live.For example, the life of people for convenience, in addition to providing traditional catering services, is going back in taking out platform
The respective services such as supermarket, fresh, medicine are provided.
In the client for taking out platform, when user is in the form of word or voice are converted into word, input inquiry closes
When key word inquires about end article, server can inquire about matching end article according to key word of the inquiry, so that user obtains.
At present, common querying method is by the way of based on dictionary, and common are two kinds of dictionaries:Queries dictionary and query fragment
Dictionary.The inquiry form of two kinds of dictionaries is consistent, and comprising multiple entries, each entry is by a character string and corresponding inquiry class
Do not form.But the priority of two kinds of dictionaries is different, the priority of queries dictionary is higher than query fragment dictionary, i.e., when user is defeated
When entering a key word of the inquiry, all entries in priority match queries dictionary, if key word of the inquiry Full Match Search dictionary
In any one character string, then returned as a result with the corresponding query categories of character string;If it fails to match, continue to match
Query fragment dictionary, if a sub-piece of inquiry have matched any one character string in query fragment dictionary completely, with
The corresponding classification that is intended to of character string returns as a result;If query fragment dictionary pattern matching fails, null value is returned.
During inventor states invention in realization, find to be based in queries dictionary and query fragment dictionary in the prior art
Any dictionary carry out end article inquiry when, a key word of the inquiry may match multiple entries in dictionary, therefore
The conflict between multiple queries classification can occur, so that mistake occurs in the query result for causing to return;In addition, this kind is based on dictionary
The mode of inquiry depends critically upon dictionary, and queries dictionary and query fragment dictionary are required to largely manually go to arrange, cost
It is higher, and error rate is also higher.
The content of the invention
In view of this, a kind of acquisition methods, device and terminal for inquiring about data provided by the invention, main purpose are to carry
Height obtains the accuracy of inquiry data, while reduces unnecessary manual resource expense.
In a first aspect, the present invention provides a kind of acquisition methods for inquiring about data, including:
The query characteristics for obtaining target query result are extracted according to searching keyword;
The query characteristics are inputted default inquiry disaggregated model to be calculated, obtain the maximum entropy of the query characteristics,
Wherein, the default inquiry disaggregated model is used to be trained to obtain the corresponding maximum entropy of each query characteristics according to training examples, most
The query characteristics of big entropy correspond to a query categories;
Correspondence between query characteristics and query categories based on the maximum entropy, determines the query characteristics most
Query categories corresponding to big entropy;
Return and belong to the same query categories the and relevant target query result with the searching keyword.
Optionally, calculated by the default inquiry disaggregated model of query characteristics input, obtain the query characteristics
Maximum entropy before, the method further includes:
The query characteristics for defining the default inquiry disaggregated model are unitary word and/or polynary word;
Build the default inquiry disaggregated model.
Optionally, the structure default inquiry disaggregated model includes:
Training examples are obtained, the training examples are to be determined according to historical query keyword and click inquiry data, described
Click in inquiry data and include trade company's classification;
According to the training examples and the corresponding number of clicks for clicking on inquiry data, determine to click on each trade company's classification
Weight;
According to the weight maximum for clicking on each trade company's classification, determine between the training examples and trade company's classification
Correspondence, completes the structure of the default inquiry disaggregated model.
Optionally, obtaining training examples includes:
The historical query keyword in preset time period is obtained from history log data and corresponding click on inquires about data;
Merchant identification ID is determined according to the inquiry data of clicking on, and trade company's classification is determined according to the merchant identification ID;
Mapping relations between the historical query keyword, the merchant identification ID and trade company's classification are made carry out
Storage, is denoted as a line;
The historical query feature of the historical query keyword is extracted according to the unitary word of definition and/or polynary word, each
Historical query feature is as a training examples.
Optionally, the query characteristics of acquisition target query result are extracted according to searching keyword to be included:
The searching keyword is segmented, each participle can be used as the query characteristics;
According to the definition of the participle and the default inquiry disaggregated model, the unitary word for extracting the searching keyword is looked into
Ask feature and/or polynary word query characteristics.
Optionally, the query characteristics are inputted default inquiry disaggregated model to be calculated, obtains the query characteristics
Maximum entropy includes:
The query characteristics are matched with the historical query feature in the default inquiry disaggregated model;
If successful match, the weight limit corresponding to matching historical query feature is obtained;
Calculate the maximum entropy of the query characteristics according to the weight limit and default characteristic function, the weight limit with
Maximum entropy is directly proportional.
Optionally, the method is applied to take out in platform, described to take out the inquiry data for including plurality of classes in platform.
Second aspect, the present invention provide a kind of acquisition device for inquiring about data, including:
Extraction unit, the query characteristics of target query result are obtained for being extracted according to searching keyword;
Processing unit, the default inquiry disaggregated model of query characteristics input for the acquiring unit to be obtained carry out
Calculate, obtain the maximum entropy of the query characteristics, wherein, the default inquiry disaggregated model is used for trained according to training examples
To the corresponding maximum entropy of each query characteristics, the query characteristics of maximum entropy correspond to a query categories;
Determination unit, between the query characteristics and query categories for the maximum entropy obtained based on the processing unit
Correspondence, determine the query categories corresponding to the maximum entropy of the query characteristics;
Returning unit, belongs to the same query categories and relevant described for returning with the searching keyword
Target query result.
Optionally, described device further includes:
Definition unit, based on the default inquiry disaggregated model of query characteristics input is carried out in the processing unit
Calculate, before obtaining the maximum entropy of the query characteristics, define it is described it is default inquiry disaggregated model query characteristics for unitary word and/
Or polynary word;
Construction unit, for building the default inquiry disaggregated model.
Optionally, the construction unit includes:
Acquisition module, for obtaining training examples, the training examples are according to historical query keyword and click on inquiry
Data determine that described click in inquiry data includes trade company's classification;
First determining module, for clicking on inquiry number with corresponding according to the training examples that the acquisition module obtains
According to number of clicks, determine to click on the weight of each trade company's classification;
Second determining module, the weight of each trade company's classification of click for being determined according to first determining module are maximum
Value, determines the correspondence between the training examples and trade company's classification, completes the structure of the default inquiry disaggregated model
Build.
Optionally, the acquisition module includes:
Acquisition submodule, for obtaining historical query keyword and correspondence in preset time period from history log data
Click inquiry data;
Determination sub-module, the click inquiry data for being obtained according to the acquisition submodule determine merchant identification
ID, and trade company's classification is determined according to the merchant identification ID;
Sub-module stored, for the historical query keyword, the merchant identification for determining the determination sub-module
Mapping relations between ID and trade company's classification are made are stored, and are denoted as a line;
Extracting sub-module, going through for the historical query keyword is extracted for the unitary word according to definition and/or polynary word
History query characteristics, each historical query feature is as a training examples.
Optionally, the extraction unit includes:
Word-dividing mode, for the searching keyword to be segmented, each participle can be used as the query characteristics;
Extraction module, for determining for the participle that is obtained according to the word-dividing mode and the default inquiry disaggregated model
Justice, extracts the unitary word query characteristics of the searching keyword and/or polynary word query characteristics.
Optionally, the processing unit includes:
Matching module, for the query characteristics and the historical query feature in the default inquiry disaggregated model to be carried out
Matching;
Acquisition module, for when the matching module successful match, obtaining corresponding to matching historical query feature most
Big weight;
Computing module, for calculating the maximum entropy of the query characteristics according to the weight limit and default characteristic function,
The weight limit is directly proportional to maximum entropy.
Optionally, described device is applied to take out in platform, described to take out the inquiry data for including plurality of classes in platform.
The third aspect, the present invention also provides a kind of acquisition terminal for inquiring about data, the terminal includes processor and storage
Device, wherein, the memory is used to store one or more computer instruction, and one or more computer instruction is described
Processor performs the step of acquisition methods to realize the inquiry data any one of first aspect.
Fourth aspect, the present invention also provides a kind of computer-readable recording medium, is stored thereon with computer instruction, described
The step of acquisition methods of the inquiry data any one of first aspect are realized when computer instruction is executed by processor.
By above-mentioned technical proposal, acquisition methods, device and the terminal of inquiry data provided by the invention, are closed according to inquiry
After keyword extraction obtains the query characteristics of target query result, query characteristics are inputted into default inquiry disaggregated model and are counted
Calculate, obtain the maximum entropy of query characteristics, wherein, default inquiry disaggregated model is used to be trained to obtain each inquiry according to training examples
The corresponding maximum entropy of feature, the query characteristics of maximum entropy correspond to a query categories, query characteristics and inquiry based on maximum entropy
Correspondence between classification, determines the query categories corresponding to the maximum entropy of query characteristics, returns same with searching keyword
Query categories, relevant query result;Compared with prior art, the present invention saves labour turnover without artificial participation, and
And looked into by the target that the maximum entropy that training in default inquiry disaggregated model obtains can be obtained accurately corresponding to searching keyword
Ask as a result, improving the accuracy returned the result.
Described above is only the general introduction of technical solution of the present invention, in order to better understand the technological means of the present invention,
And can be practiced according to the content of specification, and in order to allow above and other objects of the present invention, feature and advantage can
Become apparent, below especially exemplified by the embodiment of the present invention.
Brief description of the drawings
By reading the detailed description of hereafter preferred embodiment, it is various other the advantages of and benefit it is common for this area
Technical staff will be clear understanding.Attached drawing is only used for showing the purpose of preferred embodiment, and is not considered as to the present invention
Limitation.And in whole attached drawing, identical component is denoted by the same reference numerals.In the accompanying drawings:
Fig. 1 shows a kind of flow chart of the querying method of data provided in an embodiment of the present invention;
Fig. 2 shows a kind of schematic diagram of APP display interfaces provided in an embodiment of the present invention;
Fig. 3 shows the flow chart of the querying method of another data provided in an embodiment of the present invention;
Fig. 4 shows a kind of flow chart for building the default inquiry disaggregated model provided in an embodiment of the present invention;
Fig. 5 shows a kind of composition frame chart of the inquiry unit of data provided in an embodiment of the present invention;
Fig. 6 shows the composition frame chart of the inquiry unit of another data provided in an embodiment of the present invention.
Embodiment
The exemplary embodiment of the disclosure is more fully described below with reference to accompanying drawings.Although the disclosure is shown in attached drawing
Exemplary embodiment, it being understood, however, that may be realized in various forms the disclosure without should be by embodiments set forth here
Limited.On the contrary, these embodiments are provided to facilitate a more thoroughly understanding of the present invention, and can be by the scope of the present disclosure
Completely it is communicated to those skilled in the art.
In order to improve the correctness that target query result is returned according to query intention, while cost of labor is reduced, the present invention
Embodiment provides a kind of querying method of data, and this method can be applied in kinds of platform, those platforms provide at two kinds of numbers
According to the service of type, for example, online shopping store, take-away platform, video playing platform, news browsing platform etc. are comprising a variety of
The platform of data type, for the ease of the description of following embodiment, the embodiment of the present invention is illustrated exemplified by taking out platform, should
Take out comprising food and beverage sevice, supermarket service, medicine service, the service of fresh flower cake, the fresh service of fruit etc. in platform, it is different
Service type corresponds to different query categories.It should be noted that in a manner of explanation of the present invention exemplified by take out platform, it is not easy
It is applicable only to take out platform limiting the embodiment of the present invention.As shown in Figure 1, the described method includes:
101st, the query characteristics for obtaining target query result are extracted according to searching keyword.
Take out platform and include front end and backstage, wherein, front end is generally the application program for taking out platform
(Application, APP), user can be browsed by APP, be clicked on, obtaining the operations such as inquiry data, and backstage is generally outer
Sell the server of platform, there is provided the function such as renewal of the page, inquiry data return, after the method described in the embodiment of the present invention is applied to
End.Wherein, acquisition terminal of the server as an inquiry data, it can be independent server, or with clothes
Business device has the Relay Server of data interaction relation, and the embodiment of the present invention does not make specifically the type for inquiring about the acquisition terminal of data
Limit.
As shown in Fig. 2, Fig. 2 shows a kind of schematic diagram of APP display interfaces provided in an embodiment of the present invention, when user is led to
APP input inquiry keyword " dragging in Beijing seabed " in input frame is crossed, and after clicking on search, background server receives inquiry and closes
Keyword.Background server extracts the query characteristics for obtaining target query result according to searching keyword, and the query characteristics are to look into
The full content or partial content of keyword are ask, if for example, query characteristics are the full content of searching keyword, then from looking into
Ask and a two-spot word " dragging in Beijing-seabed " is extracted in keyword, if query characteristics are the partial content of searching keyword, such as " north
Capital ", " dragging in seabed ", then two unitary word features " Beijing " and " dragging in seabed " can be extracted according to searching keyword.
In practical applications, when extracting query characteristics according to searching keyword, the query characteristics of extraction are more, obtain
Target query fruiting quantities are fewer, and content is more accurate, and the query characteristics of extraction are fewer, and obtained target query fruiting quantities are got over
It is more.For example, if query characteristics are two-spot word " dragging in Beijing-seabed ", then as long as including " Beijing-seabed in target query result
Drag for ";If query characteristics are two unitary words " Beijing ", " dragging in seabed ", then corresponding target query result is comprising " north
The summation of the query result in capital " and the query result comprising " dragging in seabed ".
102nd, the query characteristics are inputted default inquiry disaggregated model to be calculated, obtains the maximum of the query characteristics
Entropy.
Wherein, the default inquiry disaggregated model is used to train to obtain each query characteristics according to training examples corresponding most
Big entropy, the query characteristics of maximum entropy correspond to a query categories.Training examples are according to historical query keyword and click on inquiry number
According to determining, maximum entropy model (default inquiry disaggregated model) is obtained by the machine learning to training examples, main purpose is
Realize the category classification of inquiry data, it is not required that the independence between query characteristics.
Entropy is the parameter for describing randomness, and entropy is bigger, illustrates that randomness is stronger, and entropy is smaller, is illustrated more orderly.
In embodiments of the present invention, by the study to training examples, a query characteristics (being determined by historical query keyword) can
The category classification (classification according to inquiry data are clicked on determines) of a variety of inquiry data can be corresponded to, still, those classifications are looked into
The size of entropy has differences between asking data, and entropy is more big more can represent such inquiry data and belongs to the probability of user's query intention more
Greatly;Entropy it is smaller more can represent such inquiry data belong to user's query intention probability it is smaller, therefore, in order to improve user inquiry
The accuracy of intention, can be by the corresponding query categories of maximum entropy, as the classification returned the result.
When being trained study to default inquiry disaggregated model, the corresponding different query categories of query characteristics are determined
Entropy size, when using default inquiry disaggregated model, inquiry disaggregated model is preset in the query characteristics input that step 101 is obtained
The entropy that the query characteristics correspond to different query categories can be obtained.Exemplary, when query characteristics are " dragging in Beijing-seabed "
When, the entropy of chafing dish restaurant query categories is 0.2, and the entropy of supermarket's query categories is 0.5, and the entropy that query categories are taken out in restaurant is
0.8 etc., it may be determined that maximum entropy of the query characteristics " dragging in Beijing-seabed " in default inquiry disaggregated model is 0.2.It is to show above
The explanation of example property, the embodiment of the present invention do not limit query characteristics and entropy size specifically.
103rd, the correspondence between query characteristics and query categories based on the maximum entropy, determines the query characteristics
Maximum entropy corresponding to query categories.
Except recording the corresponding pass having between the query characteristics of maximum entropy and query categories in default inquiry disaggregated model
System is outer, and also record has the query categories corresponding to any entropy of query characteristics, as shown in table 1.In order to lift inquiry velocity, this
In inventive embodiments after the maximum entropy of query characteristics is determined by the method for step 102, query categories can be directly determined, and simultaneously
The query categories corresponding to each entropy are not determined one by one.It should be noted that the showing of only providing for convenience of description of table 1
Example, in practical applications, the entropy size of query characteristics, and its it is not limited to table with the storage mode of the correspondence of query categories
Lattice, particular content are not limited to the content shown in table 1.
Table 1
104th, return and the same query categories of the searching keyword, relevant target query results.
The a variety of merchandise newss, more included under same query categories can be recorded in background server, in database
A shops, branch information etc..Exemplary, when the target query classification of identification searching keyword " dragging in Beijing-seabed " is chafing dish
During shop, can combine according to the current geographic position of user or, the positive rating of shops is dragged in seabed, to user recommend it is multiple on
Target query that seabed is dragged for as a result, such as nearest apart from user's current geographic position merchant information, with the specific business under the trade company
Product information etc., accurately identifies the query intention of user.
It is not single to be only returned only to maximum it should be noted that the present invention is implemented when returning to target query result
The corresponding query categories of entropy query characteristics, corresponding target query is as a result, the corresponding target query result of other query categories
Also can return, simply when returning to target query result, the corresponding target query result of maximum entropy is in the top, other inquiry classes
Not corresponding target query result ranks behind, and is really query demand with meet different user.Exemplary, when inquiry is crucial
When word is " dragging in Beijing-seabed ", in the target query result of return, ranking first five be chafing dish restaurant information, ranking are dragged in seabed
6th, the 7th target query result drags for chafing dish bottom flavorings, the target query result of eight to the tenth of ranking for the seabed that supermarket sells
Commodity etc. are dragged for for the seabed of other businessmans.Specifically the quantity, the ranking that return to target query result are not limited.
It can be seen that the acquisition methods of inquiry data provided in an embodiment of the present invention, inquiry with reference to above-mentioned implementation
After the terminal of data extracts the query characteristics for obtaining target query result according to searching keyword, query characteristics are inputted default
Inquiry disaggregated model is calculated, and obtains the maximum entropy of query characteristics, wherein, preset inquiry disaggregated model and be used for according to training sample
Example training obtains the corresponding maximum entropy of each query characteristics, and the query characteristics of maximum entropy correspond to a query categories, based on maximum
Correspondence between the query characteristics and query categories of entropy, determines the query categories corresponding to the maximum entropy of query characteristics, returns
Return and the same query categories of searching keyword, relevant query results;Compared with prior art, the embodiment of the present invention is without people
The participation of work, saves labour turnover, and can accurately be obtained by the maximum entropy that training in default inquiry disaggregated model obtains
Target query corresponding to searching keyword is as a result, improve the accuracy returned the result.
In order to which the acquisition methods of inquiry data provided in an embodiment of the present invention are explained in more detail, particularly this method should
For taking out the use in platform, it is intended to return to accurate target query as a result, to carry to user according to the acquisition data of user
Rise the inquiry experience of user.Specific steps as shown in figure 3, including:
201st, the query characteristics for defining the default inquiry disaggregated model are unitary word and/or polynary word.
It is to extract the query characteristics for obtaining target query result that query characteristics, which are defined, as the purpose of unitary word or polynary word, should
Query characteristics are extracted according to searching keyword (or historical query keyword), it depends on the character string of searching keyword
Length, therefore, in order to adapt to the change of the string length of different searching keywords, when query characteristics are extracted in definition, needs
Define various ways extracting mode.
For the unitary word to be extracted whole searching keyword as an entirety, polynary word is from searching keyword
Middle two or more polynary words of extraction.The number of polynary word is more, it is meant that each polynary when obtaining target query result
Word will be matched with target query result, and the inquiry data of acquisition are more accurate, still, match the number also phase of target query result
To reducing.
In order to improve matched accuracy, when the string length of searching keyword allows to be split as multiple participles, really
Fixed corresponding query characteristics can include a polynary word query characteristics and multiple unitary word query characteristics, to match more business
Family.When the string length of searching keyword do not allow to check the mark for multiple participles when, it is one one to determine corresponding query characteristics
First word.
202nd, the query characteristics for obtaining target query result are extracted according to searching keyword.
When extracting query characteristics, in the following manner is may be employed without limitation of, such as:The searching keyword is carried out
Participle, each participle can be used as the query characteristics, according to the participle and the default definition for inquiring about disaggregated model,
Extract the unitary word query characteristics of the searching keyword and/or polynary word query characteristics.
Exemplary, if searching keyword is " vitamin ", segmented according to any one method of the prior art,
Participle vitamin is obtained, the vitamin is can not be split participle, therefore the query characteristics extracted are " vitamin ".If inquiry is closed
Keyword is " a bright spicy soup ", is obtained after participle:Zhang Liang, spicy soup, according to the definition of default inquiry disaggregated model, it may be determined that look into
It can be a polynary word " Zhang Liang-spicy soup " and two unitary words " Zhang Liang ", " spicy soup " etc. to ask feature.Specifically, this
Inventive embodiments do not limit searching keyword and corresponding query characteristics.
203rd, the query characteristics are inputted default inquiry disaggregated model to be calculated, obtains the maximum of the query characteristics
Entropy.
From step 102, which is used to be trained to obtain each query characteristics according to training examples
Corresponding maximum entropy, the query characteristics of maximum entropy correspond to a query categories.In specific implementation process, query characteristics are determined
Afterwards, the query characteristics are matched with the historical query feature in the default inquiry disaggregated model, the historical query is special
When levying inquiry disaggregated model default for training study, used query characteristics, if successful match, obtain matching historical query
Weight limit corresponding to feature, the maximum entropy of the query characteristics is calculated according to the weight limit and default characteristic function.
If matching is unsuccessful, corresponding target query knot is obtained from database according to the similarity of query characteristics or query characteristics
Fruit.The definite of the historical query feature is that historical query keyword in history log data determines, rather than root
According to the generation of a certain regular random, therefore, historical query feature has certain practicality, representativeness.In practical applications, lead to
When crossing training examples and carrying out machine learning and obtain default inquiry disaggregated model, its historical query feature used is more comprehensively, often
With, therefore, occur query characteristics and it is described it is default inquiry disaggregated model in the unmatched probability of historical query feature it is smaller.
When weight limit corresponding to the historical query feature is inquired about for user by historical query keyword, use
The maximum probability for returning the result (a certain query categories), same historical query feature are clicked in family, and user clicks on same trade company's class
Other number is more, and probability is bigger, and weight is bigger, and corresponding entropy is with regard to smaller (entropy of maximum entropy is minimum).
When calculating the maximum entropy of the query characteristics according to the weight limit and default characteristic function, it can be used
Following manner:argmaxy∑iwifi(x, y), wherein, wi represents the weight of i-th of query characteristics, and fi represents ith feature letter
Number, x training samples, y represent a query categories, and the weight limit is directly proportional to maximum entropy it can be seen from above-mentioned formula,
Weight is bigger, and entropy is smaller, it is represented, and the probability of a kind of trade company of click is bigger, and the description in relation to characteristic function refer to existing skill
Any explanation in art, the embodiment of the present invention are no longer repeated herein.
204th, the correspondence between query characteristics and query categories based on the maximum entropy, determines the query characteristics
Maximum entropy corresponding to query categories.
For determining the specific determination mode of query categories corresponding to the maximum entropy of query characteristics with the above-described embodiments
Step 103 in be illustrated, details are not described herein again.
205th, return and belong to the same query categories the and relevant target query knot with the searching keyword
Fruit.
For returning to target query result specific implementation to be said in step 104 in the above-described embodiments
Bright, details are not described herein again.
The embodiment of the present invention goes to solve to judge that user realizes identification by the machine learning thought of default inquiry disaggregated model
The problem of, without the mode of manual sorting dictionary is needed in the prior art, economize on resources, avoid manual intervention and reduce standard
True property.The embodiment of the present invention also provides a kind of method for building the default inquiry disaggregated model, as shown in figure 4, including:
301st, training examples are obtained, the training examples are to be determined according to historical query keyword and click inquiry data,
Described click in inquiry data includes trade company's classification.
Specific implementation for this step, obtains the mode of training examples:
(1) the historical query keyword in preset time period is obtained from history log data and corresponding click on inquires about number
According to.For building default inquiry disaggregated model by way of machine learning, its first step is to data progress to be learned
Mark, and learnt the click inquiry data of user as labeled data in the embodiment of the present invention, the intention of user is known
It is not converted into and how determines in inquiry classification.Except including historical query keyword and corresponding click in history log data
Inquire about outside data, further include the account information of user, browsing time, click on the time, the information such as connection redirects, therefore, it is necessary to
Historical query keyword is extracted from history log data and corresponding click on inquires about data.
In theory, by inquiring about data to the historical query keyword in history log data and corresponding click on
During habit, the period of history log data is longer, and obtained statistical result is more accurate, meanwhile, calculation amount can also increase.In reality
, can be by the historical query keyword in the continuous history log data in 1 month or 2 months and corresponding in
Click on inquiry data to be learnt, specific it is not limited in the embodiment of the present invention.
In the embodiment of the present invention, specific inquiry data of clicking on can be that user checks trade company or checked in trade company's page
The data that the click behavior of specific commodity is generated, can also be concern trade company, commodity are added the click datas such as shopping cart.
(2) determine merchant identification ID according to the inquiry data of clicking on, and trade company's class is determined according to the merchant identification ID
Not.
Merchant identification ID is obtained from the information database of trade company according to inquiry data are clicked on, merchant identification ID is used for
Take out one trade company of unique mark in platform.Merchant type is obtained according to merchant identification ID, merchant type includes:Food and drink, fresh flower,
Pharmacy, supermarket, fruit are fresh etc..
(3) by the mapping relations between the historical query keyword, the merchant identification ID and trade company's classification work
Stored, be denoted as a line.
The mapping relations between historical query keyword, the merchant identification ID and trade company's classification three are established, and
Mapping relations are stored.Training learning process in, can count multiple historical query keywords, the merchant identification ID and
Trade company's classification, therefore, for the ease of distinguishing, in storage, every kind of mapping relations branch is stored, in order to which follow-up makes
With.
(4) the historical query feature of the historical query keyword is extracted according to the unitary word of definition and/or polynary word, often
A historical query feature is as a training examples.
The method for extracting historical query feature is identical with the method that query characteristics are extracted in step 202, related extraction history
The detailed description of query characteristics, refers to the related explanation of step 202.
302nd, according to the training examples and the corresponding number of clicks for clicking on inquiry data, determine to click on each trade company's class
Other weight.
Same training examples (historical query feature), the corresponding trade company's classification clicked in inquiry data may exist
Difference, for example, being directed to same training examples, clicks in inquiry data and includes 4 businessmans altogether, wherein, the number for clicking on trade company A is
20 times, the number for clicking on trade company B is 200 times, and the number for clicking on trade company C is 56 times, and the number for clicking on trade company D is 24 times, thus
As can be seen that under same query characteristics, the probability that user clicks on trade company B is relatively large.
As a kind of achievable mode of the embodiment of the present invention, when calculating the weight of historical query feature, by some business
The total degree of trade company is clicked under the number of clicks at family divided by same training examples.In by a upper example, under same training examples, point
The total degree for hitting trade company is 300, wherein, the weight for clicking on trade company B is 200/300=0.67, and the weight for clicking on trade company A is 20/
300=0.067 etc., wherein, the weight summation that different trade companies are clicked under same training examples is 1.
303rd, according to the weight maximum for clicking on each trade company's classification, determine the training examples and trade company's classification it
Between correspondence, complete it is described it is default inquiry disaggregated model structure.
After the weight for clicking on each trade company's classification is calculated in step 302, weight maximum is obtained, and establish weight most
The correspondence being worth greatly between corresponding trade company's classification and training examples (historical query feature).In practical applications, it is to be learned
Historical query be characterized as multiple, calculate each click of each training examples (historical query feature) successively by the above method
The weight of trade company's classification, and determine weight maximum, complete the structure of the default inquiry disaggregated model.
Further, as the realization to above method embodiment, another embodiment of the present invention additionally provides a kind of inquiry
The acquisition device of data.The device embodiment is corresponding with preceding method embodiment, and for ease of reading, present apparatus embodiment is no longer right
Detail content in preceding method embodiment is repeated one by one, it should be understood that the device in the present embodiment can correspond to reality
Full content in existing preceding method embodiment.As shown in figure 5, including:
Extraction unit 41, the query characteristics of target query result are obtained for being extracted according to searching keyword;Looked into basis
When asking keyword extraction query characteristics, the query characteristics of extraction are more, and obtained target query fruiting quantities are fewer, and content is more smart
Really, the query characteristics of extraction are fewer, and obtained target query fruiting quantities are more
Processing unit 42, for the default inquiry disaggregated model of query characteristics input for obtaining the acquiring unit 41
Calculated, obtain the maximum entropy of the query characteristics, wherein, the default inquiry disaggregated model is used to be instructed according to training examples
The corresponding maximum entropy of each query characteristics is got, the query characteristics of maximum entropy correspond to a query categories;By to training sample
The machine learning of example obtains maximum entropy model (default inquiry disaggregated model), and main purpose is the classification point for realizing inquiry data
Class, it is not required that the independence between query characteristics;Entropy is the parameter for describing randomness, and entropy is bigger, illustrates randomness
Stronger, entropy is smaller, illustrates more orderly.In embodiments of the present invention, the study to training examples, a query characteristics are passed through
(being determined by historical query keyword) may correspond to the category classification of a variety of inquiry data (according to the class for clicking on inquiry data
Not Que Ding), still, the size of entropy has differences between the inquiry data of those classifications, entropy it is more big can represent such inquiry number
It is bigger according to the probability for belonging to user's query intention;The smaller probability that can more represent such inquiry data and belong to user's query intention of entropy
It is smaller, therefore, can be by the corresponding query categories of maximum entropy, as returning the result in order to improve the accuracy of user's query intention
Classification;
Determination unit 43, for the query characteristics and query categories of the maximum entropy obtained based on the processing unit 42
Between correspondence, determine the query categories corresponding to the maximum entropy of the query characteristics;
Returning unit 44, belongs to the same query categories and relevant institute for returning with the searching keyword
State target query result;The present invention is implemented when returning to target query result, is not that the single maximum entropy that is only returned only to is inquired about
The corresponding query categories of feature, corresponding target query is as a result, the corresponding target query result of other query categories can also be returned
Return, simply when returning to target query result, the corresponding target query result of maximum entropy is in the top, other query categories correspond to
Target query result rank behind, be really query demand with meet different user.
Further, as shown in fig. 6, described device further includes:
Definition unit 45, carries out for the query characteristics to be inputted default inquiry disaggregated model in the processing unit 42
Calculate, before obtaining the maximum entropy of the query characteristics, the query characteristics for defining the default inquiry disaggregated model are unitary word
And/or polynary word;The query characteristics are extracted according to searching keyword (or historical query keyword), it is dependent on inquiry
The string length of keyword, therefore, in order to adapt to the change of the string length of different searching keywords, is extracted in definition
During query characteristics, various ways extracting mode need to be defined;
Construction unit 46, for building the default inquiry disaggregated model.
The embodiment of the present invention goes to solve to judge that user realizes identification by the machine learning thought of default inquiry disaggregated model
The problem of, without the mode of manual sorting dictionary is needed in the prior art, economize on resources, avoid manual intervention and reduce standard
True property.As shown in fig. 6, the construction unit 46 includes:
Acquisition module 461, for obtaining training examples, the training examples are to be looked into according to historical query keyword and click
Ask data to determine, described click in inquiry data includes trade company's classification;
First determining module 462, for the training examples obtained according to the acquisition module 461 and corresponding click
The number of clicks of data is inquired about, determines to click on the weight of each trade company's classification;
Second determining module 463, for the power of each trade company's classification of click determined according to first determining module 462
Weight maximum, determines the correspondence between the training examples and trade company's classification, completes the default inquiry classification mould
The structure of type.
Further, as shown in fig. 6, the acquisition module 461 includes:
Acquisition submodule 4611, for from history log data obtain preset time period in historical query keyword and
It is corresponding to click on inquiry data;
Determination sub-module 4612, the click inquiry data for being obtained according to the acquisition submodule 4611 determine business
Family identifies ID, and determines trade company's classification according to the merchant identification ID;
Sub-module stored 4613, for the historical query keyword, described for determining the determination sub-module 4612
Mapping relations between merchant identification ID and trade company's classification are made are stored, and are denoted as a line;
Extracting sub-module 4614, the historical query keyword is extracted for the unitary word according to definition and/or polynary word
Historical query feature, each historical query feature is as a training examples.
Further, as shown in fig. 6, the extraction unit 41 includes:
Word-dividing mode 411, for the searching keyword to be segmented, each participle can be used as an inquiry special
Sign;
Extraction module 412, for the participle obtained according to the word-dividing mode and the default inquiry disaggregated model
Definition, extract the unitary word query characteristics of the searching keyword and/or polynary word query characteristics.
Further, as shown in fig. 6, the processing unit 42 includes:
Matching module 421, for by the query characteristics with it is described it is default inquiry disaggregated model in historical query feature
Matched;
Acquisition module 422, for when the matching module successful match, obtaining corresponding to matching historical query feature
Weight limit;
Computing module 423, for calculating the maximum of the query characteristics according to the weight limit and default characteristic function
Entropy, the weight limit are directly proportional to maximum entropy.
Further, described device is applied to take out in platform, described to take out the inquiry number for including plurality of classes in platform
According to.
Further, the embodiment of the present invention, which carries, also provides a kind of acquisition terminal for inquiring about data, and the terminal includes processing
Device and memory, wherein, the memory is used to store one or more computer instruction, and one or more computer refers to
Order is performed the step of acquisition methods to realize the inquiry data shown in above-mentioned Fig. 1, Fig. 3 or Fig. 4 by the processor.
Further, the embodiment of the present invention, which carries, also provides a kind of computer-readable recording medium, is stored thereon with computer
Instruction, the computer instruction realize the acquisition methods of the inquiry data shown in above-mentioned Fig. 1, Fig. 3 or Fig. 4 when being executed by processor
The step of.
It can be seen that acquisition device and the end of inquiry data provided in an embodiment of the present invention with reference to above-mentioned implementation
End, is inquired about after query characteristics of the terminal of data according to searching keyword extraction acquisition target query result, by query characteristics
The default inquiry disaggregated model of input is calculated, and obtains the maximum entropy of query characteristics, wherein, default inquiry disaggregated model is used for root
To train to obtain the corresponding maximum entropy of each query characteristics according to training examples, the query characteristics of maximum entropy correspond to a query categories,
Correspondence between query characteristics and query categories based on maximum entropy, determines the inquiry corresponding to the maximum entropy of query characteristics
Classification, returns to the same query categories of searching keyword, relevant query result;Compared with prior art, the present invention is implemented
Example saves labour turnover, and train obtained maximum entropy can by presetting in inquiry disaggregated model without artificial participation
The accurate target query obtained corresponding to searching keyword is as a result, improve the accuracy returned the result.
Following scheme is also disclosed in the embodiment of the present invention:
A1, a kind of acquisition methods for inquiring about data, including:
The query characteristics for obtaining target query result are extracted according to searching keyword;
The query characteristics are inputted default inquiry disaggregated model to be calculated, obtain the maximum entropy of the query characteristics,
Wherein, the default inquiry disaggregated model is used to be trained to obtain the corresponding maximum entropy of each query characteristics according to training examples, most
The query characteristics of big entropy correspond to a query categories;
Correspondence between query characteristics and query categories based on the maximum entropy, determines the query characteristics most
Query categories corresponding to big entropy;
Return and belong to the same query categories the and relevant target query result with the searching keyword.
A2, the method according to A1, are calculated the query characteristics are inputted default inquiry disaggregated model, are obtained
Before the maximum entropy of the query characteristics, the method further includes:
The query characteristics for defining the default inquiry disaggregated model are unitary word and/or polynary word;
Build the default inquiry disaggregated model.
A3, the method according to A2, the structure default inquiry disaggregated model include:
Training examples are obtained, the training examples are to be determined according to historical query keyword and click inquiry data, described
Click in inquiry data and include trade company's classification;
According to the training examples and the corresponding number of clicks for clicking on inquiry data, determine to click on each trade company's classification
Weight;
According to the weight maximum for clicking on each trade company's classification, determine between the training examples and trade company's classification
Correspondence, completes the structure of the default inquiry disaggregated model.
A4, the method according to A3, obtaining training examples includes:
The historical query keyword in preset time period is obtained from history log data and corresponding click on inquires about data;
Merchant identification ID is determined according to the inquiry data of clicking on, and trade company's classification is determined according to the merchant identification ID;
Mapping relations between the historical query keyword, the merchant identification ID and trade company's classification are made carry out
Storage, is denoted as a line;
The historical query feature of the historical query keyword is extracted according to the unitary word of definition and/or polynary word, each
Historical query feature is as a training examples.
A5, the method according to A2, the query characteristics of acquisition target query result are extracted according to searching keyword to be included:
The searching keyword is segmented, each participle can be used as the query characteristics;
According to the definition of the participle and the default inquiry disaggregated model, the unitary word for extracting the searching keyword is looked into
Ask feature and/or polynary word query characteristics.
The query characteristics are inputted default inquiry disaggregated model and are calculated, obtain institute by A6, the method according to A3
Stating the maximum entropy of query characteristics includes:
The query characteristics are matched with the historical query feature in the default inquiry disaggregated model;
If successful match, the weight limit corresponding to matching historical query feature is obtained;
Calculate the maximum entropy of the query characteristics according to the weight limit and default characteristic function, the weight limit with
Maximum entropy is directly proportional.
A7, the method according to any one of A1-A6, the method are applied to take out in platform, the take-away platform
In include the inquiry data of plurality of classes.
B8, a kind of acquisition device for inquiring about data, including:
Extraction unit, the query characteristics of target query result are obtained for being extracted according to searching keyword;
Processing unit, the default inquiry disaggregated model of query characteristics input for the acquiring unit to be obtained carry out
Calculate, obtain the maximum entropy of the query characteristics, wherein, the default inquiry disaggregated model is used for trained according to training examples
To the corresponding maximum entropy of each query characteristics, the query characteristics of maximum entropy correspond to a query categories;
Determination unit, between the query characteristics and query categories for the maximum entropy obtained based on the processing unit
Correspondence, determine the query categories corresponding to the maximum entropy of the query characteristics;
Returning unit, belongs to the same query categories and relevant described for returning with the searching keyword
Target query result.
B9, the device according to B8, described device further include:
Definition unit, based on the default inquiry disaggregated model of query characteristics input is carried out in the processing unit
Calculate, before obtaining the maximum entropy of the query characteristics, define it is described it is default inquiry disaggregated model query characteristics for unitary word and/
Or polynary word;
Construction unit, for building the default inquiry disaggregated model.
B10, the device according to B9, the construction unit include:
Acquisition module, for obtaining training examples, the training examples are according to historical query keyword and click on inquiry
Data determine that described click in inquiry data includes trade company's classification;
First determining module, for clicking on inquiry number with corresponding according to the training examples that the acquisition module obtains
According to number of clicks, determine to click on the weight of each trade company's classification;
Second determining module, the weight of each trade company's classification of click for being determined according to first determining module are maximum
Value, determines the correspondence between the training examples and trade company's classification, completes the structure of the default inquiry disaggregated model
Build.
B11, the device according to B10, the acquisition module include:
Acquisition submodule, for obtaining historical query keyword and correspondence in preset time period from history log data
Click inquiry data;
Determination sub-module, the click inquiry data for being obtained according to the acquisition submodule determine merchant identification
ID, and trade company's classification is determined according to the merchant identification ID;
Sub-module stored, for the historical query keyword, the merchant identification for determining the determination sub-module
Mapping relations between ID and trade company's classification are made are stored, and are denoted as a line;
Extracting sub-module, going through for the historical query keyword is extracted for the unitary word according to definition and/or polynary word
History query characteristics, each historical query feature is as a training examples.
B12, the device according to B9, the extraction unit include:
Word-dividing mode, for the searching keyword to be segmented, each participle can be used as the query characteristics;
Extraction module, for determining for the participle that is obtained according to the word-dividing mode and the default inquiry disaggregated model
Justice, extracts the unitary word query characteristics of the searching keyword and/or polynary word query characteristics.
B13, the device according to B10, the processing unit include:
Matching module, for the query characteristics and the historical query feature in the default inquiry disaggregated model to be carried out
Matching;
Acquisition module, for when the matching module successful match, obtaining corresponding to matching historical query feature most
Big weight;
Computing module, for calculating the maximum entropy of the query characteristics according to the weight limit and default characteristic function,
The weight limit is directly proportional to maximum entropy.
B14, the device according to any one of B8-B13, described device are applied to take out in platform, and described take out is put down
The inquiry data of plurality of classes are included in platform.
C15, a kind of acquisition terminal for inquiring about data, the terminal include processor and memory, wherein, the memory
For storing one or more computer instruction, one or more computer instruction is performed by the processor to realize A1
To the inquiry data any one of A7 acquisition methods the step of.
D16, a kind of computer-readable recording medium, are stored thereon with computer instruction, and the computer instruction is processed
Device realizes the step of acquisition methods of the inquiry data any one of A1 to A7 when performing.
In the above-described embodiments, the description to each embodiment all emphasizes particularly on different fields, and does not have the portion being described in detail in some embodiment
Point, it may refer to the associated description of other embodiment.
It is understood that the correlated characteristic in the above method and device can be referred to mutually.In addition, in above-described embodiment
" first ", " second " etc. be to be used to distinguish each embodiment, and do not represent the quality of each embodiment.
It is apparent to those skilled in the art that for convenience and simplicity of description, the system of foregoing description,
The specific work process of device and unit, may be referred to the corresponding process in preceding method embodiment, details are not described herein.
Algorithm and display be not inherently related to any certain computer, virtual system or miscellaneous equipment provided herein.
Various general-purpose systems can also be used together with teaching based on this.As described above, required by constructing this kind of system
Structure be obvious.In addition, the present invention is not also directed to any certain programmed language.It should be understood that it can utilize various
Programming language realizes the content of invention described herein, and the description done above to language-specific is to disclose this hair
Bright preferred forms.
In the specification that this place provides, numerous specific details are set forth.It is to be appreciated, however, that the implementation of the present invention
Example can be put into practice in the case of these no details.In some instances, known method, structure is not been shown in detail
And technology, so as not to obscure the understanding of this description.
Similarly, it will be appreciated that in order to simplify the disclosure and help to understand one or more of each inventive aspect,
Above in the description to the exemplary embodiment of the present invention, each feature of the invention is grouped together into single implementation sometimes
In example, figure or descriptions thereof.However, the method for the disclosure should be construed to reflect following intention:I.e. required guarantor
The application claims of shield features more more than the feature being expressly recited in each claim.It is more precisely, such as following
Claims reflect as, inventive aspect is all features less than single embodiment disclosed above.Therefore,
Thus the claims for following embodiment are expressly incorporated in the embodiment, wherein each claim is in itself
Separate embodiments all as the present invention.
Those skilled in the art, which are appreciated that, to carry out adaptively the module in the equipment in embodiment
Change and they are arranged in one or more equipment different from the embodiment.Can be the module or list in embodiment
Member or component be combined into a module or unit or component, and can be divided into addition multiple submodule or subelement or
Sub-component.In addition at least some in such feature and/or process or unit exclude each other, it can use any
Combination is disclosed to all features disclosed in this specification (including adjoint claim, summary and attached drawing) and so to appoint
Where all processes or unit of method or equipment are combined.Unless expressly stated otherwise, this specification (including adjoint power
Profit requires, summary and attached drawing) disclosed in each feature can be by providing the alternative features of identical, equivalent or similar purpose come generation
Replace.
In addition, it will be appreciated by those of skill in the art that although some embodiments described herein include other embodiments
In included some features rather than further feature, but the combination of the feature of different embodiments means in of the invention
Within the scope of and form different embodiments.For example, in the following claims, embodiment claimed is appointed
One of meaning mode can use in any combination.
The all parts embodiment of the present invention can be with hardware realization, or to be run on one or more processor
Software module realize, or realized with combinations thereof.It will be understood by those of skill in the art that it can use in practice
Microprocessor or digital signal processor (DSP) realize the acquisition methods of inquiry data according to embodiments of the present invention, dress
Put and some or all functions of some or all components in terminal.The present invention is also implemented as being used to perform here
The some or all equipment or program of device of described method are (for example, computer program and computer program production
Product).Such program for realizing the present invention can store on a computer-readable medium, or can have one or more
The form of signal.Such signal can be downloaded from internet website and obtained, and either be provided or on carrier signal to appoint
What other forms provides.
It should be noted that the present invention will be described rather than limits the invention for above-described embodiment, and ability
Field technique personnel can design alternative embodiment without departing from the scope of the appended claims.In the claims,
Any reference symbol between bracket should not be configured to limitations on claims.Word "comprising" does not exclude the presence of not
Element or step listed in the claims.Word "a" or "an" before element does not exclude the presence of multiple such
Element.The present invention can be by means of including the hardware of some different elements and being come by means of properly programmed computer real
It is existing.In if the unit claim of equipment for drying is listed, several in these devices can be by same hardware branch
To embody.The use of word first, second, and third does not indicate that any order.These words can be explained and run after fame
Claim.
Claims (10)
- A kind of 1. acquisition methods for inquiring about data, it is characterised in that including:The query characteristics for obtaining target query result are extracted according to searching keyword;The query characteristics are inputted default inquiry disaggregated model to be calculated, obtain the maximum entropy of the query characteristics, wherein, The default inquiry disaggregated model is used to be trained to obtain the corresponding maximum entropy of each query characteristics according to training examples, maximum entropy Query characteristics correspond to a query categories;Correspondence between query characteristics and query categories based on the maximum entropy, determines the maximum entropy of the query characteristics Corresponding query categories;Return and belong to the same query categories the and relevant target query result with the searching keyword.
- 2. according to the method described in claim 1, it is characterized in that, the query characteristics are being inputted into default inquiry disaggregated model Calculated, before obtaining the maximum entropy of the query characteristics, the method further includes:The query characteristics for defining the default inquiry disaggregated model are unitary word and/or polynary word;Build the default inquiry disaggregated model.
- 3. according to the method described in claim 2, it is characterized in that, the structure default inquiry disaggregated model includes:Training examples are obtained, the training examples are to be determined according to historical query keyword and click inquiry data, the click Include trade company's classification in inquiry data;According to the training examples and the corresponding number of clicks for clicking on inquiry data, determine to click on the power of each trade company's classification Weight;According to the weight maximum for clicking on each trade company's classification, determine corresponding between the training examples and trade company's classification Relation, completes the structure of the default inquiry disaggregated model.
- 4. according to the method described in claim 3, it is characterized in that, obtaining training examples includes:The historical query keyword in preset time period is obtained from history log data and corresponding click on inquires about data;Merchant identification ID is determined according to the inquiry data of clicking on, and trade company's classification is determined according to the merchant identification ID;Mapping relations between the historical query keyword, the merchant identification ID and trade company's classification are made are deposited Storage, is denoted as a line;The historical query feature of the historical query keyword, each history are extracted according to the unitary word of definition and/or polynary word Query characteristics are as a training examples.
- 5. according to the method described in claim 2, obtain target query result it is characterized in that, being extracted according to searching keyword Query characteristics include:The searching keyword is segmented, each participle can be used as the query characteristics;According to the definition of the participle and the default inquiry disaggregated model, the unitary word inquiry for extracting the searching keyword is special Sign and/or polynary word query characteristics.
- 6. according to the method described in claim 3, it is characterized in that, by the query characteristics input default inquiry disaggregated model into Row calculates, and obtaining the maximum entropy of the query characteristics includes:The query characteristics are matched with the historical query feature in the default inquiry disaggregated model;If successful match, the weight limit corresponding to matching historical query feature is obtained;The maximum entropy of the query characteristics, the weight limit and maximum are calculated according to the weight limit and default characteristic function Entropy is directly proportional.
- 7. according to the method any one of claim 1-6, it is characterised in that the method is applied to take out in platform, The inquiry data of plurality of classes are included in the take-away platform.
- A kind of 8. acquisition device for inquiring about data, it is characterised in that including:Extraction unit, the query characteristics of target query result are obtained for being extracted according to searching keyword;Processing unit, carry out by the default inquiry disaggregated model of query characteristics input for obtaining the acquiring unit based on Calculate, obtain the maximum entropy of the query characteristics, wherein, the default inquiry disaggregated model is used to train to obtain according to training examples The corresponding maximum entropy of each query characteristics, the query characteristics of maximum entropy correspond to a query categories;Determination unit, for pair between the query characteristics and query categories of the maximum entropy obtained based on the processing unit It should be related to, determine the query categories corresponding to the maximum entropy of the query characteristics;Returning unit, belongs to the same query categories the and relevant target for returning with the searching keyword Query result.
- A kind of 9. acquisition terminal for inquiring about data, it is characterised in that the terminal includes processor and memory, wherein, it is described Memory be used for store one or more computer instruction, one or more computer instruction by the processor perform with The step of realizing the acquisition methods of the inquiry data any one of claim 1 to 7.
- 10. a kind of computer-readable recording medium, is stored thereon with computer instruction, it is characterised in that the computer instruction The step of acquisition methods of the inquiry data any one of claim 1 to 7 are realized when being executed by processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711243962.XA CN107958406A (en) | 2017-11-30 | 2017-11-30 | Inquire about acquisition methods, device and the terminal of data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711243962.XA CN107958406A (en) | 2017-11-30 | 2017-11-30 | Inquire about acquisition methods, device and the terminal of data |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107958406A true CN107958406A (en) | 2018-04-24 |
Family
ID=61962766
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711243962.XA Pending CN107958406A (en) | 2017-11-30 | 2017-11-30 | Inquire about acquisition methods, device and the terminal of data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107958406A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108763200A (en) * | 2018-05-15 | 2018-11-06 | 达而观信息科技(上海)有限公司 | Chinese word cutting method and device |
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 |
CN110909231A (en) * | 2019-11-15 | 2020-03-24 | 上海易点时空网络有限公司 | Method and system for inquiring violation data and storage medium |
CN111061835A (en) * | 2019-12-17 | 2020-04-24 | 医渡云(北京)技术有限公司 | Query method and device, electronic equipment and computer readable storage medium |
CN111831867A (en) * | 2020-04-02 | 2020-10-27 | 北京嘀嘀无限科技发展有限公司 | Address query method and device, electronic equipment and computer readable storage medium |
CN111858922A (en) * | 2019-09-24 | 2020-10-30 | 北京嘀嘀无限科技发展有限公司 | Service side information query method and device, electronic equipment and storage medium |
CN112035599A (en) * | 2020-11-06 | 2020-12-04 | 苏宁金融科技(南京)有限公司 | Query method and device based on vertical search, computer equipment and storage medium |
CN113342983A (en) * | 2021-06-30 | 2021-09-03 | 中国平安人寿保险股份有限公司 | Resume distribution method, device and equipment based on machine learning and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101887436A (en) * | 2009-05-12 | 2010-11-17 | 阿里巴巴集团控股有限公司 | Retrieval method, device and system |
CN102033877A (en) * | 2009-09-27 | 2011-04-27 | 阿里巴巴集团控股有限公司 | Search method and device |
EP2583190A1 (en) * | 2010-06-18 | 2013-04-24 | Alibaba Group Holding Limited | Determining and using search term weightings |
CN103207881A (en) * | 2012-01-17 | 2013-07-17 | 阿里巴巴集团控股有限公司 | Query method and unit |
-
2017
- 2017-11-30 CN CN201711243962.XA patent/CN107958406A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101887436A (en) * | 2009-05-12 | 2010-11-17 | 阿里巴巴集团控股有限公司 | Retrieval method, device and system |
CN102033877A (en) * | 2009-09-27 | 2011-04-27 | 阿里巴巴集团控股有限公司 | Search method and device |
EP2583190A1 (en) * | 2010-06-18 | 2013-04-24 | Alibaba Group Holding Limited | Determining and using search term weightings |
CN103207881A (en) * | 2012-01-17 | 2013-07-17 | 阿里巴巴集团控股有限公司 | Query method and unit |
Non-Patent Citations (3)
Title |
---|
中国索引学会主办等: "《中国索引 第2辑》", 30 September 2017, 复旦大学出版社 * |
王爱功等: "《数字图书档案技术》", 31 May 2006, 时代文艺出版社 * |
赵欣: ""基于最大熵的中文术语抽取系统的设计与实现"", 《中国优秀硕士学位论文全文数据库(信息科技辑)》 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108763200A (en) * | 2018-05-15 | 2018-11-06 | 达而观信息科技(上海)有限公司 | Chinese word cutting method and device |
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 |
CN111858922A (en) * | 2019-09-24 | 2020-10-30 | 北京嘀嘀无限科技发展有限公司 | Service side information query method and device, electronic equipment and storage medium |
CN110909231A (en) * | 2019-11-15 | 2020-03-24 | 上海易点时空网络有限公司 | Method and system for inquiring violation data and storage medium |
CN111061835A (en) * | 2019-12-17 | 2020-04-24 | 医渡云(北京)技术有限公司 | Query method and device, electronic equipment and computer readable storage medium |
CN111061835B (en) * | 2019-12-17 | 2023-09-22 | 医渡云(北京)技术有限公司 | Query method and device, electronic equipment and computer readable storage medium |
CN111831867A (en) * | 2020-04-02 | 2020-10-27 | 北京嘀嘀无限科技发展有限公司 | Address query method and device, electronic equipment and computer readable storage medium |
CN112035599A (en) * | 2020-11-06 | 2020-12-04 | 苏宁金融科技(南京)有限公司 | Query method and device based on vertical search, computer equipment and storage medium |
CN113342983A (en) * | 2021-06-30 | 2021-09-03 | 中国平安人寿保险股份有限公司 | Resume distribution method, device and equipment based on machine learning and storage medium |
CN113342983B (en) * | 2021-06-30 | 2023-02-07 | 中国平安人寿保险股份有限公司 | Resume distribution method, device and equipment based on machine learning and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107958406A (en) | Inquire about acquisition methods, device and the terminal of data | |
CN111784455B (en) | Article recommendation method and recommendation equipment | |
US10977710B2 (en) | Systems and methods for tracking consumer tasting preferences | |
CN109299994B (en) | Recommendation method, device, equipment and readable storage medium | |
CN103246980B (en) | Information output method and server | |
US8856129B2 (en) | Flexible and scalable structured web data extraction | |
CN104199833B (en) | The clustering method and clustering apparatus of a kind of network search words | |
US20220405607A1 (en) | Method for obtaining user portrait and related apparatus | |
CN110383263B (en) | Creating cognitive intelligent queries from multiple data corpora | |
WO2020187168A1 (en) | Resume pushing method and apparatus, and task pushing method and apparatus | |
CN112434151A (en) | Patent recommendation method and device, computer equipment and storage medium | |
CN108038161A (en) | Information recommendation method, device and computing device based on photograph album | |
CN109460519B (en) | Browsing object recommendation method and device, storage medium and server | |
CN110647696B (en) | Business object sorting method and device | |
CN107247728B (en) | Text processing method and device and computer storage medium | |
Jaeger et al. | Analysing conjoint data with OLS and PLS regression: a case study with wine | |
CN110175895A (en) | A kind of item recommendation method and device | |
US9256805B2 (en) | Method and system of identifying an entity from a digital image of a physical text | |
US20220107980A1 (en) | Providing an object-based response to a natural language query | |
CN107958270A (en) | Classification recognition methods, device, electronic equipment and computer-readable recording medium | |
CN115982463A (en) | Resource recommendation method, device, equipment and storage medium | |
CN109472025B (en) | Dish name extraction method and device | |
KR101498944B1 (en) | Method and apparatus for deciding product seller related document | |
KR102082900B1 (en) | System for providing optimal keyword of sale items | |
US10217080B1 (en) | Item classification using visible attributes |
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 | ||
CB02 | Change of applicant information |
Address after: Building N3, building 12, No. 27, Chengzhong Road, Jiancai, Haidian District, Beijing 100096 Applicant after: Beijing Xingxuan Technology Co.,Ltd. Address before: Room 202, 2 floors, 1-3 floors, No. 11 Shangdi Information Road, Haidian District, Beijing 100085 Applicant before: Beijing Xiaodu Information Technology Co.,Ltd. |
|
CB02 | Change of applicant information | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180424 |
|
RJ01 | Rejection of invention patent application after publication |