CN104750781A - Advertisement retrieval method - Google Patents
Advertisement retrieval method Download PDFInfo
- Publication number
- CN104750781A CN104750781A CN201510099290.4A CN201510099290A CN104750781A CN 104750781 A CN104750781 A CN 104750781A CN 201510099290 A CN201510099290 A CN 201510099290A CN 104750781 A CN104750781 A CN 104750781A
- Authority
- CN
- China
- Prior art keywords
- conjunction
- assignment
- value
- sizeof
- advertisement
- 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
Abstract
The invention belongs to the technical field of internet application and relates to an advertisement retrieval method. The method aims to solve the problem that in the prior art, if a certain attribute of a user is multiple values, advertisements which can be matched with the attribute can not be searched. The advertisement retrieval method comprises the following steps that 1, a DNF expression is set up; 2, a first layer retrieval system is set up; 3, a second layer retrieval system is set up; 4, retrieval is achieved. According to the advertisement retrieval method, the network advertisements can meet the requirement of the user, and particularly the user advertisements with multiple characteristics can be retrieved.
Description
Technical field
The present invention is specifically related to a kind of advertisement retrieval method, belongs to technical field of internet application.
Background technology
From the demand angle of advertising business, in fact advertisement beam search is appreciated that the matching problem of the stereotactic conditions being user tag and advertisement, and these couplings all belong to content matching.For advertisement condition: an attribute can be monodrome and can be many-valued; Many-valued, not only can be "or" relation but also can be "AND" relation; Condition as advertisement A can be " throwing in the user liking electronic product ", " throwing in liking electronic product or liking the user of fashion " also can be " throwing in not only liking electronic product but also liking the user of fashion ".For user tag, an attribute can be monodrome and can be many-valued, and many-valued can only be "AND" relation; Also user tag can be " not only liking electronic product but also like fashion " as can be " liking electronic product ".Therefore, if use user tag to do directed coupling, six kinds of situations are in fact had.
From the angle of DNF, because it has the feature of inverted index, so in search procedure: user tag attribute can only be monodrome; In index process of establishing, advertisement condition attribute can be monodrome and can be many-valued, and many-valued can only be "or" relation.Therefore DNF algorithm solve only two kinds of situations, brings great restriction like this, if some attributes of user are multiple value, so just can not find the advertisement that all can match with it to the demand that advertiser customizes.
DNF (disjunctive normal form) is a kind of standard normal form in Boolean logic, DNF by several Clause's or form, and each Clause has some words (argument or argument non-) composition.Inverted index comes from practical application needs the value according to attribute to search record.Each in this concordance list all comprises a property value and has the address of each record of this property value.Owing to not being determine property value by recording, but being determined the position of recording by property value, being thus called inverted index (inverted index).
In advertisement delivery system, the data needing pin to have more user decide this and throw which type of advertisement to user, and this process is referred to as " orientation ", and inverted index is then the data structure that must generate in orientation.The performance quality of an ad system and the directed implementation of this ad system have inseparable relation.The quality of inverted index also determines that this searches the speed of advertisement, affects Consumer's Experience equally.
Both above, the stereotactic conditions of an advertiser can use DNF express by we, as ad=(age ∈ { 3} ∩ state ∈ { NY}) ∪ (state ∈ { CA} ∩ gender ∈ { M}) represents that the input object of this advertisement is: " user of 3 years old resident NY " or " the men's family of resident CA ".
In DNF expression formula, to each unit in stereotactic conditions DNF just like defining below:
Conjunction
Each disjunctive normal form (DNF) can be broken down into the union of one or more Conjunction; such as above that example, ad can be decomposed into j1=(age ∈ { 3} ∩ state ∈ { NY}) and j2=(state ∈ { CA} ∩ gender ∈ { M}) two Conjunction; For same attribute, only can occur once in a Conjunction.
Assignment
Each Conjunction can be decomposed into again the common factor of one or more Assignment further, as j1 can be decomposed into age ∈ { 3} and state ∈ { NY} two Assignment.
Sizeof( Conjunction)
Represent the Assignment number of " ∈ " contained in a Conjunction, namely equal the Assignment number that it comprises non-" " (empty string).
Summary of the invention
Therefore, if the present invention is directed to some attributes of user in prior art is multiple value, so just can not find the problem of the advertisement that all can match with it, a kind of advertisement retrieval method is provided, it is characterized in that, said method comprising the steps of:
Step one: set up DNF expression formula, converts DNF expression formula form to by the stereotactic conditions of each advertisement document;
Step 2: set up ground floor searching system, participle is carried out to the DNF expression formula of each advertisement document, extract and its Conjunction comprised of iteration: if new Conjunction, the Conjunction-ID that then distribution one is new, otherwise with the Conjunction-ID distributed before, by containing adding special marking " ^ " after the attribute of "AND", the Sizeof value of Conjunction is constant simultaneously, be still attribute number, for each advertisement document distributes Ad-ID, described Conjunction-ID and Ad-ID is from 0 by the value that round values increases progressively, and set up the relation of falling row of Conjunction-ID to Ad-ID, form the searching system of ground floor,
Step 3: set up second layer searching system, each Conjunction step one extracted is decomposed into Assignment, for each Assignment, keyword is attribute and property value, extract the value of the Sizeof of the Conjunction of its correspondence, the new keywords of each Assignment is redefined with the value of Sizeof and keyword, each Assignment redefining new keywords is set up with the Conjunction corresponding with it relation of falling row, wherein the property value of Conjunction is divided into and belongs to, do not belong to two kinds of values, the Conjunction-ID being zero by all Sizeof values is placed in inverted list, and compose with " ∈ " operational character (Z, ∈), relation is arranged in write, by the relation of falling row of foundation according to the ascending sequence of the value of Sizeof, set up the second layer searching system that Assignment and Conjunction is corresponding, Assignment is made up of new keywords, Conjunction part is made up of Conjunction-ID and property value thereof,
Step 4: realize retrieval, according to the DNF expression formula of the client demand received, first the total quantity of its Conjunction is determined, each value of these total quantitys such as Sizeof is less than in second layer searching system is retrieved, for each Conjunction of the DNF expression formula of client demand, when it is single Assignment composition, this Assignment of extracting directly is at the property value of Conjunction-ID and Conjunction corresponding to second layer searching system, when it is multiple Assignment composition, extract the result composition union of each Assignment at the property value of Conjunction-ID and Conjunction corresponding to second layer searching system, judge all results extracted under each Sizeof value, when the situation occurrence number that the property value of certain Conjunction-ID and this Conjunction is all identical equals current Sizeof value, this Conjunction-ID exports as a result, judge that the rear Conjunction-ID according to all Output rusults retrieves the Ad-ID corresponding to Conjunction-ID in ground floor searching system, the advertisement document needing to provide is provided.
Beneficial effect of the present invention is: the DNF expression formula of the client demand adopting advertisement retrieval method of the present invention to receive, first retrieve in second layer searching system, eliminate Sizeof value to be greater than maximum possible and to be worth situation, whole recall precision is improved, owing to establishing double-deck searching system, achieve the retrieval of the client of multi-valued attribute, accurately for the client of multi-valued attribute provides all satisfactory advertisement document.The Conjunction-ID being zero by all Sizeof values is placed in inverted list, solves the problem realizing when Sizeof value is zero retrieving.
Accompanying drawing explanation
Fig. 1 is the step schematic diagram of a kind of advertisement retrieval method of the present invention;
Embodiment
Below in conjunction with accompanying drawing, the specific embodiment of the present invention is described:
Following targeted ads is had in advertisement base:
cast1 = ( age∈{ 3} ∩area∈{ NY，CA，^} ) ∪(area∈{ CA} ∩sex∈{ M} ) 。
cast2 = ( age ∈{ 3 } ∩sex∈{ F} ) ∪(area ?{ CA; NY} )
cast3 = ( age ∈ { 3 } ∩sex∈ { M } ∩area ?{ CA} ) ∪(area∈{ CA} ∩sex∈{ F} )
cast4 = ( age∈{ 3，4} ) ∪(area∈{ CA} ∩sex∈{ M} )
cast5 = (area ?{ CA，NY} ) ∪( age∈{ 3，4} )
cast6 = (area ?{ CA，NY} ) ∪( age ∈{ 3} ∩area∈{ NY} ) ∪(area∈{ CA} ∩sex∈{ M} )
cast7 = ( age ∈{ 3 } ∩area∈{ NY} ) ∪(area∈{ CA} ∩sex∈{ F} )
Ground floor index is set up
Conjunction is as follows:
c1 age ∈ { 3} ∩area∈ { NY，CA，^} 2
c2 age ∈ { 3} ∩sex∈ { F} 2
c3 age ∈ { 3} ∩sex∈ { M} ∩state?{ CA} 2
c4 area∈ { CA} ∩sex∈ { M} 2
c5 age ∈ { 3; 4} 1
c6 area ?{ CA; NY} 0
c7 age ∈ { 3} ∩area∈ { NY} 2
c8 area∈ { CA} ∩sex∈ { F} 2
Ground floor inverted index:
c1 →cast1，cast6，cast7
c2 →cast2
c3 →cast3
c4 →cast1，cast4，cast6
c5 →cast4，cast5
c6 →cast2，cast5，cast6
c7 →cast6，cast7
c8 →cast3，cast7
Second layer index is set up
Carry out participle to Conjunction, be cut into Assignment stream, wherein keyword (term) is " attribute+value ".As for age ∈ { 3; 4} cuts out two keywords: " age ∈ 3} ", " age ∈ 4} "; For area { CA; NY} is also cut into two keywords: " area CA} ", " area NY} ".Do not belong to " " and belong to " ∈ " and be embodied in down on row's slide fastener.Calculate the size of Conjunction and be embodied on keyword, final keyword consists of Sizeof ［ Conjunction ］, attribute and value.Be the Conjunction of 0 for size, algorithm introduces a special keyword, is that the Conjunction of zero is placed in its inverted list by all size, and composes with " ∈ " operational character (Z, ∈).Write arranges relation, term → (Conjunction-ID, ∈ |).
Second layer inverted index:
0(area，NY) ［( c6，?) ］
(area，CA) ［( c6，?) ］
Z ?［( c6，∈) ］
1( age，3) ［( c5，∈) ］
( age，4) ［( c5，∈) ］
2( age，3) ［( c1，∈) ，( c2，∈) ，( c3，∈) ］
(area，NY) ［( c1，∈) ］
(area，CA) ［( c1，∈) ，( c3，? ) ，( c4，∈) ， ( c8，∈) ］
(sex，F) ［( c2，∈) ，( c8，∈) ］
(sex，M) ［( c3，∈) ，( c4，∈) ］
(3) retrieve
Use have label " three years old ", " resident CA or NY ", " man " audient carry out advertisement retrieval.Set up query statement: age=3, area=CA/ NY, sex=M, it can thus be appreciated that Sizeof (Query)=3, namely user is with three attribute, then the Conjunction of size > 3 does not then pay attention to, because the advertisement request user containing such Conjunction has 4 attributes at least.Inquiry size=2, size=1, the Conjunction of size=0 layer, during the layer of inquiry size=2, to there being area attribute, " NY ", " CA " two the ConjunctionId-ID that hits of value get union, result is as follows:
( age，3)［( c1，∈)，( c2，∈)，( c3，∈)］
(area，NY / CA)［( c1，∈) ，( c3，?) ，( c4，∈)， ( c8，∈)］
(sex，M)［( c3，∈) ，( c4，∈) ］
Shown in the row's of falling slide fastener, first chained list is consistent with the Conjunction-ID of second chained list head, and i.e. (c1, ∈), c1 occurs twice altogether, and without " ", then c1 meets the requirements.
In like manner, after completing all inquiries, find that c1, c4, c5 meet the requirements.Inquiry ground floor inverted index known cast1, cast4, cast5, cast6 and cast7 are the advertisement meeting audient.
The above is the preferred embodiment of the present invention; it should be pointed out that for those skilled in the art, under the prerequisite not departing from principle of the present invention; can also make some improvements and modifications, these improvements and modifications also should be considered as protection scope of the present invention.
Claims (2)
1. an advertisement retrieval method, is characterized in that, said method comprising the steps of:
Step one: set up DNF expression formula, converts DNF expression formula form to by the stereotactic conditions of each advertisement document;
Step 2: set up ground floor searching system, participle is carried out to the DNF expression formula of each advertisement document, extract and its Conjunction comprised of iteration, if new Conjunction, the Conjunction-ID that then distribution one is new, otherwise with the Conjunction-ID distributed before, the Sizeof value of Conjunction is constant simultaneously, be still attribute number, for each advertisement document distributes Ad-ID, described Conjunction-ID and Ad-ID is from 0 by the value that round values increases progressively, and set up the relation of falling row of Conjunction-ID to Ad-ID, form the searching system of ground floor,
Step 3: set up second layer searching system, each Conjunction step one extracted is decomposed into Assignment, for each Assignment, keyword is attribute and property value, extract the value of the Sizeof of the Conjunction of its correspondence, the new keywords of each Assignment is redefined with the value of Sizeof and keyword, each Assignment redefining new keywords is set up with the Conjunction corresponding with it relation of falling row, wherein the property value of Conjunction is divided into and belongs to, do not belong to two kinds of values, the Conjunction-ID being zero by all Sizeof values is placed in inverted list, and compose with " ∈ " operational character (Z, ∈), relation is arranged in write, by the relation of falling row of foundation according to the ascending sequence of the value of Sizeof, set up the second layer searching system that Assignment and Conjunction is corresponding, Assignment is made up of new keywords, Conjunction part is made up of the Conjunction property value of Conjunction-ID and correspondence,
Step 4: realize retrieval, according to the DNF expression formula of the client demand received, first its Conjunction total quantity is determined, each value Sizeof in second layer searching system being less than or equal to this total quantity is retrieved, for each Conjunction of the DNF expression formula of client demand, when it is single Assignment composition, this Assignment of extracting directly is at the property value of Conjunction-ID and Conjunction corresponding to second layer searching system, when it is multiple Assignment composition, extract the result composition union of each Assignment at the property value of Conjunction-ID and Conjunction corresponding to second layer searching system, judge all results extracted under each Sizeof value, when the situation occurrence number that the property value of certain Conjunction-ID and this Conjunction is all identical equals current Sizeof value, this Conjunction-ID exports as a result, judge that the rear Conjunction-ID according to all Output rusults retrieves the Ad-ID corresponding to Conjunction-ID in ground floor searching system, the advertisement document needing to provide is provided.
2. advertisement retrieval method as claimed in claim 1, is characterized in that, extracts and in its Conjunction process comprised of iteration in step 2, by containing adding special marking " ^ " after the attribute of "AND".
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510099290.4A CN104750781A (en) | 2015-03-06 | 2015-03-06 | Advertisement retrieval method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510099290.4A CN104750781A (en) | 2015-03-06 | 2015-03-06 | Advertisement retrieval method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104750781A true CN104750781A (en) | 2015-07-01 |
Family
ID=53590465
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510099290.4A Pending CN104750781A (en) | 2015-03-06 | 2015-03-06 | Advertisement retrieval method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104750781A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105426349A (en) * | 2015-11-03 | 2016-03-23 | 北京锐安科技有限公司 | Transmission exchange method and device of logic expression |
| CN105786950A (en) * | 2015-12-30 | 2016-07-20 | 北京锐安科技有限公司 | Method and device for processing data |
| CN111177501A (en) * | 2019-12-13 | 2020-05-19 | 杭州首展科技有限公司 | Label processing method, device and system |
| CN113112168A (en) * | 2021-04-22 | 2021-07-13 | 上海刀奇智能科技有限公司 | Pollution source supervision and tracing platform based on Internet of things |
| WO2023165557A1 (en) * | 2022-03-03 | 2023-09-07 | 掌阅科技股份有限公司 | Inverted index table construction method, resource retrieval method, device, and storage medium |
-
2015
- 2015-03-06 CN CN201510099290.4A patent/CN104750781A/en active Pending
Non-Patent Citations (1)
| Title |
|---|
| 王茜: "基于倒排索引的DNF算法改进", 《信息技术》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105426349A (en) * | 2015-11-03 | 2016-03-23 | 北京锐安科技有限公司 | Transmission exchange method and device of logic expression |
| CN105786950A (en) * | 2015-12-30 | 2016-07-20 | 北京锐安科技有限公司 | Method and device for processing data |
| CN111177501A (en) * | 2019-12-13 | 2020-05-19 | 杭州首展科技有限公司 | Label processing method, device and system |
| CN111177501B (en) * | 2019-12-13 | 2023-11-17 | 杭州首展科技有限公司 | Label processing method, device and system |
| CN113112168A (en) * | 2021-04-22 | 2021-07-13 | 上海刀奇智能科技有限公司 | Pollution source supervision and tracing platform based on Internet of things |
| WO2023165557A1 (en) * | 2022-03-03 | 2023-09-07 | 掌阅科技股份有限公司 | Inverted index table construction method, resource retrieval method, device, and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20110196737A1 (en) | Semantic advertising selection from lateral concepts and topics | |
| US8380563B2 (en) | Using previous user search query to target advertisements | |
| US8171021B2 (en) | Query identification and association | |
| CN104750781A (en) | Advertisement retrieval method | |
| JP5311378B2 (en) | Feature word automatic learning system, content-linked advertisement distribution computer system, search-linked advertisement distribution computer system, text classification computer system, and computer programs and methods thereof | |
| JP5679993B2 (en) | Method and query system for executing a query | |
| CN101401062A (en) | Method and system for determining relevant sources, querying and merging results from multiple content sources | |
| US20100293170A1 (en) | Social network message categorization systems and methods | |
| JP2016505993A (en) | Method and apparatus for configuring search terms, delivering advertisements, and retrieving product information | |
| CN105930539A (en) | Topic subscription method and device | |
| US20100036726A1 (en) | Method of reducing cost per action of an internet advertisement campaign, and optimizing to the maximum the number of actions performed by web surfers | |
| Li et al. | Detection and characterization of human trafficking networks using unsupervised scalable text template matching | |
| JP7023865B2 (en) | Improved landing page generation | |
| CN105447186A (en) | Big data platform based user behavior analysis system | |
| CN101398926A (en) | Advertisement bidding and playing method and system based on directional delivery | |
| US11609943B2 (en) | Contextual content distribution | |
| CN111292167A (en) | Big data-based competitive product analysis method and system, storage medium and electronic equipment | |
| US20150213481A1 (en) | Optimization of advertisements | |
| Farseev et al. | Somin. ai: Social multimedia influencer discovery marketplace | |
| JP2019519840A5 (en) | ||
| US20120005021A1 (en) | Selecting advertisements using user search history segmentation | |
| US20130332276A1 (en) | Advertisement management device, advertisement selection device, advertisement management method, advertisement management program and storage medium storing advertisement management program | |
| JP5514703B2 (en) | Search delivery server, program and method for delivering related information according to search log | |
| Becker et al. | What happens after an ad click? quantifying the impact of landing pages in web advertising | |
| US20090043753A1 (en) | Method for generating structured query results using lexical clustering |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20150701 |
|
| RJ01 | Rejection of invention patent application after publication |