CN108052554B - The method and apparatus of various dimensions expansion keyword - Google Patents
The method and apparatus of various dimensions expansion keyword Download PDFInfo
- Publication number
- CN108052554B CN108052554B CN201711229068.7A CN201711229068A CN108052554B CN 108052554 B CN108052554 B CN 108052554B CN 201711229068 A CN201711229068 A CN 201711229068A CN 108052554 B CN108052554 B CN 108052554B
- Authority
- CN
- China
- Prior art keywords
- keyword
- app
- similarity
- words
- expanded
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention relates to the method and apparatus that various dimensions expand keyword.The described method includes: on the one hand determining associated 2nd APP of APP to be expanded after receiving wait expand APP, obtaining the first expansion keyword set based on the 2nd APP;On the other hand the emphasis keyword for determining APP to be expanded, obtains the second expansion keyword set based on emphasis keyword;Then keyword screening is carried out on the basis of first expands keyword set and second expands keyword set, obtains the final expansion keyword of APP to be expanded.The present invention can carry out the expansion of keyword based on two dimensions of competing product APP and emphasis keyword, improve the quality of keyword expansion and comprehensive.
Description
Technical field
The present invention relates to data analysis technique field, the method and apparatus for expanding keyword more particularly to various dimensions.
Background technique
Since Most users are the various APP of application library platform (i.e. application shop) downloading in an intelligent terminal
(application is also referred to as applied), therefore APP developer is to improve itself APP in the search quality of application shop, is needed
The key word analysis of APP is carried out to optimize itself APP.
Specific industry knowledge background based on intelligent terminal application shop, the keyword of traditional application shop APP expand compared with
Mostly by judgement expansion is manually carried out, expands quality and be affected by human subjective's human-subject test, keyword expands result
Unstable quality defect.Also, the existing feature progress keyword expansion expanded thinking and be normally based on APP itself, because
This is difficult to expand keyword comprehensively.
Summary of the invention
Based on this, the present invention provides the method and apparatus that various dimensions expand keyword, can overcome existing application
Keyword expands unstable quality and expands incomplete defect.
Scheme provided in an embodiment of the present invention includes:
A kind of method that various dimensions expand keyword, comprising:
The first keyword that APP to be expanded is covered in application library platform is obtained, is being applied according to each first keyword
The APP that library platform searches obtains associated 2nd APP of APP to be expanded;Obtain what each 2nd APP was covered in application library platform
Second keyword obtains associated 3rd APP of APP to be expanded in the APP that application library platform searches according to each second keyword;
The keyword that each 3rd APP is covered in application library platform is obtained, the first candidate is obtained according to the keyword that each 3rd APP is covered
Keyword set;It determines the similarity that each 3rd APP gathers relative to the 2nd APP, determines each in the first candidate key set of words
Specific gravity shared by keyword calculates each keyword in the first candidate key set of words according to the similarity and the specific gravity
First similarity score;The first setting quantity is filtered out from the first candidate key set of words according to first similarity score
Keyword, obtain the first expansion keyword set;
Emphasis keyword is filtered out from the first keyword, is searched according to each emphasis keyword in application library platform
APP obtains associated 4th APP of APP to be expanded;The second time is obtained according to the keyword that the 4th APP is covered in application library platform
Select keyword set;Determine that each keyword is similar relative to the synthesis of emphasis keyword set in the second candidate key set of words
Degree, determines the specific gravity of each keyword in the second candidate key set of words, according to the specific gravity and the comprehensive similarity meter
Calculate the second similarity score of each keyword in the second candidate key set of words;According to second similarity score from described
The keyword that the second setting quantity is filtered out in two candidate key set of words obtains the second expansion keyword set;
Keyword set and second is expanded from first and expands in keyword set the keyword chosen third and set quantity, is obtained
To the expansion keyword of APP to be expanded;
Wherein, the keyword of APP covering need to meet condition: include in the search result of application library platform in the keyword
There is the APP.
A kind of various dimensions expand the device of keyword, comprising:
First opens up word module, the first keyword covered in application library platform for obtaining APP to be expanded, according to each
First keyword obtains associated 2nd APP of APP to be expanded in the APP that application library platform searches;Each 2nd APP is obtained to answer
With the second keyword covered in the platform of library, obtained in the APP that application library platform searches wait expand according to each second keyword
Associated 3rd APP of APP;The keyword that each 3rd APP is covered in application library platform is obtained, according to each 3rd APP covering
Keyword obtains the first candidate key set of words;It determines the similarity that each 3rd APP gathers relative to the 2nd APP, determines first
Specific gravity shared by each keyword in candidate key set of words calculates the first candidate key according to the similarity and the specific gravity
First similarity score of each keyword in set of words;According to first similarity score from the first candidate key set of words
The keyword for filtering out the first setting quantity obtains the first expansion keyword set;
Second opens up word module, for filtering out emphasis keyword from the first keyword, is existed according to each emphasis keyword
The APP that application library platform searches obtains associated 4th APP of APP to be expanded;It is covered in application library platform according to the 4th APP
Keyword obtain the second candidate key set of words;Determine that each keyword is relative to emphasis key in the second candidate key set of words
The comprehensive similarity of set of words determines the specific gravity of each keyword in the second candidate key set of words, according to the specific gravity and
The comprehensive similarity calculates the second similarity score of each keyword in the second candidate key set of words;According to second phase
The keyword for filtering out the second setting quantity from the second candidate key set of words like degree score obtains the second expansion key
Set of words;And
Screening module expands selection third setting number in keyword set for expanding keyword set and second from first
The keyword of amount obtains the expansion keyword of APP to be expanded;
Wherein, the keyword of APP covering need to meet condition: include in the search result of application library platform in the keyword
There is the APP.
A kind of computer readable storage medium, is stored thereon with computer program, realization when which is executed by processor
The step of method described above.
A kind of computer equipment can be run on a memory and on a processor including memory, processor and storage
The step of computer program, the processor realizes method described above when executing described program.
Implement above-described embodiment, after receiving wait expand APP, on the one hand determines associated 2nd APP of APP to be expanded
(competing product APP) obtains the first expansion keyword set based on the 2nd APP;On the other hand determine that the emphasis of APP to be expanded closes
Keyword obtains the second expansion keyword set based on emphasis keyword;Then keyword set and second is expanded first to expand
Keyword screening is carried out on the basis of keyword set, obtains the final expansion keyword of APP to be expanded.Above-mentioned technical proposal energy
It is enough to can be improved according to the APP for treating expansion based on the expansion of two dimensions progress keywords of competing product APP and emphasis keyword
The quality and comprehensive that keyword is expanded.In addition, keyword expanding method through the foregoing embodiment, be also convenient for batch export to
The corresponding keyword of the APP of expansion opens up word scheme, realizes that efficiency is also highly improved;Both it realizes volume production, ensures that simultaneously
Expand quality.
Detailed description of the invention
Fig. 1 is the schematic flow chart that the various dimensions of an embodiment expand the method for keyword;
Fig. 2 is the schematic flow chart that the various dimensions of another embodiment expand the method for keyword;
Fig. 3 is the schematic diagram that the various dimensions of an embodiment expand the device of keyword.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
Referenced herein " multiple " refer to two or more."and/or", the association for describing affiliated partner are closed
System, indicates may exist three kinds of relationships, and character "/" typicallys represent the relationship that forward-backward correlation object is a kind of "or".
Although the step in various embodiments of the present invention is arranged with label, it is not used to successive time that limits step
Sequence, based on the order of step or the execution of certain step need other steps unless expressly stated, the otherwise phase of step
Order is adjustable.
Fig. 1 is the schematic flow chart that the various dimensions of an embodiment expand the method for keyword;As shown in Figure 1, this implementation
Example in various dimensions expand keyword method comprising steps of
S11 obtains the first keyword that APP to be expanded is covered in application library platform, is existed according to each first keyword
The APP that application library platform searches obtains associated 2nd APP of APP to be expanded;Each 2nd APP is obtained to cover in application library platform
Second keyword of lid, obtains the associated third of APP to be expanded in the APP that application library platform searches according to each second keyword
APP;The keyword that each 3rd APP is covered in application library platform is obtained, obtains first according to the keyword that each 3rd APP is covered
Candidate key set of words;It determines the similarity that each 3rd APP gathers relative to the 2nd APP, determines the first candidate key set of words
In specific gravity shared by each keyword, each key in the first candidate key set of words is calculated according to the similarity and the specific gravity
First similarity score of word;The first setting is filtered out from the first candidate key set of words according to first similarity score
The keyword of quantity obtains the first expansion keyword set.
Keyword in the embodiment of the present invention can be used for searching for the character of APP, such as the Chinese in application library platform including all
Word, English word or letter, number or other letter symbols, can also be the combining form of several characters.Described first
Keyword can be to be obtained by the historical search information for analyzing application library platform, and key is included in the historical search information
The mapping relations of word and APP can also be preassigned based on experience value;First keyword of APP to be expanded covering be it is multiple,
Associated 2nd APP of APP to be expanded also is multiple.
Wherein, the keyword of APP covering need to meet condition: include described in the corresponding search result of the keyword
APP.I.e. each first keyword includes the APP to be expanded in the search result of application library platform.
Wherein, second keyword can be is obtained by the historical search information for analyzing application library platform, can also
It is preassigned based on experience value.Second keyword of one the 2nd APP covering, need to meet condition: the second keyword is being answered
It include the 2nd APP in search result with library platform.The second keyword, the 3rd APP of each 2nd APP covering are covered
Keyword be it is multiple, associated 3rd APP of APP to be expanded also is multiple.
Wherein, the keyword of the 3rd APP covering can be is obtained by the historical search information for analyzing application library platform
It arrives, can also be preassigned based on experience value.The keyword of one the 3rd APP covering, need to meet condition: the keyword
It include the 3rd APP in the search result of application library platform.
Wherein, the similarity of the 3rd APP and corresponding 2nd APP indicates that the 3rd APP is closed with the comprehensive of corresponding 2nd APP
Connection degree.In one embodiment, if corresponding 2nd APP of the 3rd APP is one, the 3rd APP is obtained with corresponding 2nd APP's
Similarity, the similarity gathered as the 3rd APP relative to the 2nd APP;If corresponding 2nd APP of the 3rd APP is two
More than, then the similarity of the 3rd APP and each corresponding 2nd APP are obtained respectively, similarity mean value are calculated with this, with the phase
The similarity gathered as the 3rd APP relative to the 2nd APP like degree mean value.Wherein, the 3rd APP and single second
The similarity of APP can be predetermined, and be also possible to what the search record based on application platform calculated in real time.The calculating
Similarity mean value both includes calculating absolute average, also includes calculating weighted average.
Wherein, each keyword specific gravity shared in the first candidate key set of words is based on keyword for third
What the different degree of APP determined.That is keyword characterizes the APP in the search result of the keyword for the different degree of an APP
Ranking information.Keyword can be the pre- historical search record data for first passing through application library platform for the different degree of APP
The different degree that data are analyzed is also possible to preset different degree.It further include root in one embodiment if the former
Information is recorded according to the historical search of application library platform, predefines the step of each keyword searches it different degree of APP.
S12 filters out emphasis keyword from the first keyword, is searched for according to each emphasis keyword in application library platform
To APP obtain associated 4th APP of APP to be expanded;Is obtained according to the keyword that the 4th APP is covered in application library platform
Two candidate key set of words;Determine synthesis phase of each keyword relative to emphasis keyword set in the second candidate key set of words
Like degree, the specific gravity of each keyword in the second candidate key set of words is determined, according to the specific gravity and the comprehensive similarity
Calculate the second similarity score of each keyword in the second candidate key set of words;According to second similarity score from described
The keyword that the second setting quantity is filtered out in second candidate key set of words obtains the second expansion keyword set.
Wherein, the emphasis keyword need to meet condition: in the search result of emphasis keyword, the row of APP to be expanded
Name is forward, such as APP ranking to be expanded is at first 10.
It optionally, further include being screened to each emphasis keyword in the APP that application library platform searches in this step,
It only chooses wherein ranking and thus obtains APP to be expanded the associated 4th in the APP of first 100 (quantity can specifically be set)
APP.Such as 500 APP that an emphasis keyword can search in application library platform, only choose wherein ranking at first 100
APP, thus can reduce the computation complexity of subsequent key word expansion, meanwhile, ranking more rearward, shows keyword and APP
Relevance is lower, therefore will come subsequent APP and reject, and also can guarantee the accuracy that keyword is expanded.Wherein, one the 4th
The second level keyword of APP covering need to meet condition: second level keyword includes this in the search result of application library platform
4th APP.
Wherein, the similarity of each keyword and single emphasis keyword indicates that keyword exists in the second candidate key set of words
The degree of association in same application platform can reflect and respectively search the registration of APP, and the similarity of keyword and keyword can
To be predetermined, it is also possible to the search record based on application platform and is calculated in real time.Optionally, described according to second
The similarity of the corresponding emphasis keyword of each keyword, each keyword and corresponding emphasis keyword, meter in candidate key set of words
It includes: to obtain second candidate to close that each keyword, which is calculated in the second candidate key set of words, relative to the comprehensive similarity of emphasis keyword
The similarity of the corresponding emphasis keyword of each keyword, each keyword and corresponding emphasis keyword, calculates second in keyword set
The average value of each keyword and the similarity of corresponding emphasis keyword in candidate key set of words, as the second candidate keywords
Comprehensive similarity of each keyword relative to emphasis keyword set in set.The average value includes absolute average, is also wrapped
Include weighted average.
Such as: assuming that emphasis keyword: " shopping ", " Taobao ";Keyword " shopping " expands out the key that APP is covered
Word: [Jingdone district, Suning easily purchase];Keyword " Taobao " expands out the keyword that APP is covered: [Jingdone district, day cat];So second waits
Keyword set is selected to be combined into [Jingdone district, Suning easily purchase, day cat].
Wherein, the comprehensive similarity of " Jingdone district " relative to emphasis keyword set in the second candidate key set of words are as follows:
Sim (Jingdone district)=[sim (shopping, Jingdone district)+sim (Taobao, Jingdone district)]/2.
The comprehensive similarity of " Suning easily purchases " relative to emphasis keyword set in second candidate key set of words are as follows:
Sim (Suning easily purchases)=sim (shopping, Suning easily purchases).
The comprehensive similarity of " day cat " relative to emphasis keyword set in second candidate key set of words are as follows:
Sim (day cat)=sim (Taobao, day cat).
Wherein, each keyword specific gravity shared in the second candidate key set of words is based on keyword for corresponding APP
Different degree determine, keyword characterizes ranking of the APP in the keyword search results for the different degree of an APP
Information.Keyword can be the data point of the pre- historical search record data for first passing through application library platform for the different degree of APP
Obtained different degree is analysed, preset different degree is also possible to.It further include according to application in one embodiment if the former
The historical search of library platform records information, predefines the step of each keyword searches it different degree of APP.
S13 expands keyword set and second from first and expands in keyword set the key chosen third and set quantity
Word obtains the expansion keyword of APP to be expanded.
Keyword expanding method through the foregoing embodiment is receiving after receiving wait expand APP wait expand
After APP, on the one hand determines associated 2nd APP (competing product APP) of APP to be expanded, the first expansion is obtained based on the 2nd APP
Keyword set;On the other hand the emphasis keyword for determining APP to be expanded, obtains the second expansion keyword based on emphasis keyword
Set;Then keyword screening is carried out on the basis of first expands keyword set and second expands keyword set, obtained
The final expansion keyword of APP to be expanded.Above-mentioned technical proposal can be based on competing product APP and emphasis according to the APP for treating expansion
Two dimensions of keyword carry out the expansion of keyword, can be improved the quality of keyword expansion and comprehensive.
In one embodiment, in above-mentioned steps S11, first similarity can be chosen from the first candidate key set of words
Score from high to low ranking it is preceding setting quantity keyword, obtain the first expansion keyword set.
In one embodiment, second similarity score can be chosen from the second candidate key set of words to arrange from high to low
The keyword of the preceding setting quantity of name, obtains the second expansion keyword set.
Further, in one embodiment, in above-mentioned steps S13, keyword set and second is expanded from first and expands pass
The keyword that third setting quantity is chosen in keyword set, obtains the expansion keyword of APP to be expanded, comprising: remember first respectively
It expands keyword set and is combined into W(1), second expansion keyword set be combined into W(2), keyword set, the second expansion key are expanded by first
Set of words obtains third and expands keyword set, is denoted as W(3), keyword set W is expanded to third(3)In each keyword first
Similarity score or the second similarity score are normalized;It obtains third and expands keyword set W(3)In each keyword
Searchable index W is calculated according to the similarity score after the searchable index of each keyword and normalized(3)In each key
The final similarity score of word;According to final similarity score from W(3)The middle keyword for choosing setting quantity, obtains wait expand
The expansion keyword of APP.
Optionally, in the following way to W(3)In i-th1First similarity score of a keyword or the second similarity obtain
Divide and be normalized:
Wherein,For W(3)In i-th1The first similarity score or the second similarity score of a keyword, sminWith smaxPoint
It Biao Shi not W(3)The minimum value and maximum value of middle similarity score,For W(3)In i-th1Similarity after a keyword normalization obtains
Point.
Optionally, the method to the final similarity score for calculating each keyword are as follows:
Inquire W(3)In each keyword searchable index, by the normalized principle of above-mentioned similarity score to W(3)In each key
The searchable index of word is normalized, and obtains the searchable index correction value p ' of each keyword;It calculates in the following way each
The final similarity score of keyword:
Wherein, weight system α ∈ [0,1] is preset.
The final similarity score of basis is from W(3)The middle keyword for choosing setting quantity, obtains the expansion of APP to be expanded
Keyword includes: from W(3)The middle keyword for choosing the final similarity score preceding setting quantity of ranking from high to low, obtain to
Expand the expansion keyword of APP;Alternatively, can also be according to the sequence of final similarity score from high to low, from W(3)Middle selection setting
The keyword phrase of number includes multiple keywords in each keyword phrase, and the multiple groups for obtaining APP to be expanded expand keyword.
As it can be seen that keyword expanding method through the foregoing embodiment, it is also convenient for batch and exports the corresponding keyword of APP to be expanded opening up word
Scheme realizes that efficiency is also highly improved;Both it realizes volume production, while ensuring that expansion quality.
In one embodiment, in above-mentioned steps S11, it is crucial to obtain APP to be expanded is covered in application library platform first
The process of word can include: whole keywords of APP covering to be expanded are obtained according to the historical search of application library platform record;It treats
The whole keywords for expanding APP covering carry out screening anomaly, to delete abnormal keyword therein, obtain APP covering to be expanded
The first keyword.Wherein, the abnormal keyword includes: searchable index exception, keyword search results data exception, APP
The keyword of at least one of ranking exception, number of characters exception feature in search result.Wherein, searchable index is that basis is set
Determine the cumulative number (volumes of searches) for carrying out APP search in statistical time in application library platform using the keyword, while considering to search
What the factors such as rope magnitude were calculated, both searchable index and volumes of searches are that positive relationship is presented;Searchable index and volumes of searches two
Person's relationship can be simple with right and wrong linear increase relationship.Searchable index refers to that searchable index is less than setting numerical value extremely;Search
Results abnormity refers to that the APP quantity that keyword search is arrived is less than setting quantity;Different degree is referred to extremely in keyword
APP ranking is more rearward in search result;Number of words refers to that number of words is too short or too long extremely.
Correspondingly, the process of second keyword for obtaining each 2nd APP covering can include: according to application library platform
Historical search record obtains whole keywords of each 2nd APP covering;Whole keywords of each 2nd APP covering are carried out abnormal
Screening obtains the second keyword of the 2nd APP covering to delete abnormal keyword therein.And described obtain each the
The process for the keyword that three APP are covered in application library platform can include: recorded and obtained according to the historical search of application library platform
Whole keywords of each 3rd APP covering;Screening anomaly is carried out to whole keywords of each 3rd APP covering, to delete wherein
Abnormal keyword, obtain the keyword of the 3rd APP covering.
The purpose of above-mentioned keyword filtration treatment is to carry out screening anomaly to keyword, such as keyword search results are too
Less, searchable index is too low, search rank rearward, number of words it is too short or it is too long etc. belong to keyword abnormal conditions, rejected, with
Interference of the abnormal data to subsequent expansion is prevented, the accuracy that keyword is expanded is improved.
It in one embodiment, further include predefining keyword for the different degree of corresponding APP before above-mentioned steps S11
The step of, which specifically includes: according to the ranking information of APP in keyword search results, to keyword for the important of APP
Spend assignment:
V_2 (w)=(15,14,13,12,11,10,9,8,7,6,5,4,3,2,1,0.5)
V_3 (r)=(0,1,3,6,10,16,22,30,40,50,65,80,100,120,150,200, ∞)
wi=V_2 (w)t;V_3(r)t< rank≤V_3 (r)t+1
Wherein, [1,16] i ∈;V_2 (w) is different degree weight vectors;V_3 (r) is ranking interval vector;∞ indicates ranking
Positive infinity;Rank indicates the ranking of APP in search result;wiIndicate keyword kiTo the different degree of APP.For example, APP is being closed
Keyword kiSearch result in ranking be the 2nd, then keyword kiDifferent degree to the APP is wi=V_2 (w)2=14;V_3
(r)2< rank≤V_3 (r)3.Wherein, V_2 (w), V_3 (r) can be preset according to different application library platforms.
In one embodiment, in above-mentioned steps S12, emphasis is filtered out from the first keyword that APP to be expanded is covered and is closed
The detailed process of keyword includes: to obtain each first keyword for the different degree of APP to be expanded, chooses different degree and is greater than or waits
Emphasis keyword in the first keyword of the first setting different degree threshold value, as APP to be expanded covering;Wherein, keyword pair
In the different degree of APP to be expanded, ranking information of the APP to be expanded in the search result of the keyword is characterized.
In an alternative embodiment, acquisition of information APP covering to be expanded is recorded according to the historical search of application library platform
It further include that the record information of the historical search to application library platform carries out pretreated step before keyword.Such as based on nearest
Information is recorded in the search that the application library platform occurs within one week, historical search record information includes the keyword letter for search
Breath and the corresponding search result information of each keyword.Such as nearest one week keyword search results, APP information (may include
The dimensions such as APPID, APP title, affiliated list), key word information (including keyword ID, keyword, searchable index, search knot
The dimensions such as fruit).
In an alternative embodiment, carrying out pretreated step to the historical search record information of application library platform can be wrapped
Include: the historical search for obtaining application library platform in set period of time records information, records information according to the historical search and determines
Corresponding first mapping relations of each keyword;It include the corresponding APP information of keyword and APP in first mapping relations
Ranking information in the multiple search result of the keyword.Then, multiple passes in information are recorded according to the historical search
First mapping relations of keyword determine corresponding second mapping relations of each APP;It include APP in second mapping relations
Corresponding keyword further includes different degree of each keyword for the APP, and the different degree is for indicating APP described
Ranking information in the search result of keyword, APP in the search result of keyword ranking more before, the keyword is for this
The different degree of APP is bigger.Further, the application library platform is established according to first mapping relations and the second mapping relations
Corresponding data mapping library.
Mapping library based on the data, it is described that acquisition of information APP to be expanded is recorded according to the historical search of application library platform
Corresponding first keyword can include: inquire the data mapping library, obtain corresponding second mapping relations of APP to be expanded, root
The different degree of corresponding first keyword of APP to be expanded and first keyword is obtained according to second mapping relations.Institute
It states and obtains the APP information that each first order emphasis keyword is covered in application library platform can include: inquire the data mapping
Library obtains corresponding first mapping relations of each first order emphasis keyword, obtains each according to first mapping relations
The APP information of level-one emphasis keyword covering.The each emphasis keyword of acquisition can be wrapped in the APP that application library platform searches
It includes: inquiring the data mapping library, obtain corresponding first mapping relations of each emphasis keyword, closed according to first mapping
System obtains the APP information of each emphasis keyword covering.
In one embodiment, in above-mentioned steps S11, the APP that is searched according to each first keyword in application library platform
Obtain associated 2nd APP of APP to be expanded include: according to historical search record in each first keyword setting historical period in
Multiple search result, obtain the frequency sequencing information of APP in the corresponding multiple search result of first keyword;Obtain the frequency
The APP of the preceding setting quantity of sequence row, the APP information arrived as each first keyword search.According to all the first keywords,
Each first keyword search to APP information obtain an APP matrix;The frequency of occurrence of each APP in the APP matrix is counted, is selected
Frequency of occurrence in the APP matrix is taken to be greater than or equal to the first APP for setting the frequency, as associated 2nd APP of APP to be expanded.
In one embodiment, in above-mentioned steps S11, it is described according to each second keyword search to APP obtain wait expand
Associated 3rd APP of APP include: according to historical search record in each second keyword setting historical period in multiple search
As a result, obtaining the frequency sequencing information of APP in the corresponding multiple search result of second keyword;Before the acquisition frequency comes
Setting quantity APP, the APP information arrived as each second keyword search;It is closed according to all the second keywords, each second
The APP information that keyword searches obtains an APP matrix;The frequency of occurrence for counting each APP in the APP matrix chooses the APP matrix
Middle frequency of occurrence is greater than or equal to the APP of the second setting frequency, as associated 3rd APP of the 2nd APP.
In one embodiment, in above-mentioned steps S12, it is described according to emphasis keyword search to APP obtain the 4th APP packet
Include: according to historical search record in each emphasis keyword setting historical period in multiple search result, obtain the emphasis
The frequency sequencing information of APP in the corresponding multiple search result of keyword;Obtain the preceding setting quantity of frequency sequence row
APP, the APP arrived as the emphasis keyword search;It is arrived according to whole emphasis keywords, each emphasis keyword search
APP obtains an APP matrix;The frequency of occurrence for counting each APP in the APP matrix, choose frequency of occurrence in the APP matrix be greater than or
Equal to the APP of the setting frequency, the 4th APP is obtained.
Since the same keyword may be searched for repeatedly in (such as in one week) setting historical period, and search plain knot
Fruit changes with the variation of search time.Statistics is carried out to search result to summarize, and finally obtains keyword k0Corresponding APP collection
Close A (k0) and frequency ordering vector V (k0),
A(k0)=(appid1,appid2,…,appidn)
V(k0)=(count1,count2,…,countn)
Wherein k0Indicate keyword, countnIt indicates to use keyword k in setting historical period0There is appid in searchnIt is right
Cope with the frequency of app.Wherein, the frequency sequencing information of APP refers to that the frequency sorts in the corresponding multiple search result of keyword
Vector V (k0) described in the corresponding frequency of APP.
In one embodiment, it after obtaining the 2nd APP, before the second keyword for obtaining each 2nd APP covering, also wraps
It includes step: obtaining APP to be expanded application list affiliated in application library platform, delete and belong to different application with APP to be expanded
2nd APP of list.It optionally, further include step before the keyword for obtaining each 3rd APP covering after obtaining the 3rd APP
It is rapid: to obtain APP to be expanded application list affiliated in application library platform, delete and belong to different application list with APP to be expanded
The 3rd APP.Optionally, it after obtaining the 4th APP, before the keyword for obtaining each 4th APP covering, further comprises the steps of:
Obtain APP to be expanded in application library platform belonging to apply list, delete and belong to the of different application list with APP to be expanded
Four APP.Thus the accuracy of subsequent key word expansion can be improved.
In one embodiment, the similarity of each 3rd APP and single 2nd APP is the similarity calculated in real time, specific to count
Calculation process includes: to obtain the feature vector of the 2nd APP according to the second keyword of the 2nd APP covering, is covered according to each 3rd APP
Keyword obtain the feature vector of each 3rd APP;By One-Hot coding to the feature vector and the 3rd APP of the 2nd APP
Feature vector handled, obtain the sparse features vector of the 2nd APP and the sparse features vector of the 3rd APP;According to
The sparse features vector of two APP and the sparse features vector of the 3rd APP calculate the phase of each 3rd APP and corresponding 2nd APP
Like degree.Wherein, the sparse features vector of the 2nd APP is equal with the dimension of sparse features vector of the 3rd APP, and meets condition:
dV≤m+n;M indicates the dimension of the feature vector of the 2nd APP, and n indicates the dimension of the feature vector of the 3rd APP, dVIndicate described dilute
Dredge the dimension of feature vector.
Such as: APP such as to be expanded is APP(1), it is assumed that its corresponding 2nd APP includes (APP(2) 1、APP(2) 2), wherein
2nd APPAPP(2) 1The keyword of covering is (KW(2) 1, KW(2) 2, KW(2) 3), in this, as the 2nd APPAPP(2) 1Feature vector,
Feature vector dimension is 3;2nd APPAPP(2) 2The keyword of covering is (KW(2) 2, KW(2) 3, KW(2) 4,KW(2) 5), in this, as
Two APPAPP(2) 2Feature vector, feature vector dimension be 4.
Further, the 2nd APPAPP(2) 1Corresponding 3rd APP includes (APP(3) 1, APP(3) 2, APP(3) 3);Second
APPAPP(2) 2Corresponding 3rd APP includes (APP(3) 3,APP(3) 4, APP(3) 5);This makes it possible to obtain the 3rd APP to gather (APP(3) 1,
APP(3) 2, APP(3) 3,APP(3) 4, APP(3) 5).In the 3rd APP set, APP(3) 1Corresponding 2nd APP only has APP(2) 1, because
This, APP(3) 1With similarity, that is, APP of the 2nd APP(3) 1With APP(2) 1Similarity;APP(3) 3Corresponding 2nd APP has APP(2) 1
And APP(2) 2, therefore, APP is obtained respectively(3) 3With APP(2) 1Similarity, APP(3) 3With APP(2) 2Similarity, phase is calculated with this
Like degree mean value, using the similarity mean value as APP(3)3 and the 2nd APP similarity.
Further, the 3rd APPAPP(3) 1The keyword of covering is (KW(3) 1, KW(3) 2, KW(3) 3), in this, as third
APPAPP(3) 1Feature vector, feature vector dimension be 3;3rd APPAPP(3) 2The keyword of covering is (KW(3) 4, KW(3) 2, KW(3) 3, KW(3) 5), in this, as the 3rd APPAPP(3) 2Feature vector, feature vector dimension be 4.Wherein, KW(3) 2With KW(2) 2For
Same keyword.
2nd APPAPP as a result,(2) 1Feature vector (KW(2) 1, KW(2) 2, KW(2) 3), the 3rd APPAPP(3) 1Feature vector
(KW(3) 1, KW(3) 2, KW(3) 3), KW(3) 2With KW(2) 2For same keyword, therefore the two is in the feature vector that real number space is constituted
(KW(2) 1, KW(2) 2, KW(2) 3, KW(3) 1, KW(3) 3), dimension is 5≤3+3, and the sparse features vector for obtaining the two is respectively as follows: second
APPAPP(2) 1Sparse features vector: (1,1,1,0,0), the 3rd APPAPP(3) 1Sparse features vector: (0,1,0,1,1).
Based on the above embodiment, optionally, the similarity of each 3rd APP and single 2nd APP are calculated by the following formula:
In formula, APP(2) tIndicate t-th of the 2nd APP;Indicate i-th2A 3rd APP;Indicate APP(2) tSparse features vector withSparse features vector inner product;Indicate APP(2) tSparse features vector withSparse features vector 2- norm
Product.
It should be understood that between two APP similarity calculation method, it is including but not limited to above-mentioned similar based on cosine
Degree calculates the algorithm of similarity, can also be used to calculate the algorithm of similarity using other.
In one embodiment, the first candidate key set of words is obtained according to the keyword that each 3rd APP is covered, comprising: root
According to the keyword that the 2nd APP associated 3rd APP and each 3rd APP are covered, the 2nd associated keyword matrix of APP is obtained.
To the crucial conflation of words statistics in the keyword matrix, the first candidate key set of words KW is obtained(3)=(kw(3) 1,kw(3) 2,…,
kw(3) n) and corresponding keyword frequency vector be C(3)=(c1,c2,…,cn)。
Further, the first candidate key set of words KW(3)In i-th3Specific gravity shared by a keyword are as follows:
In formula, i3=1,2 ..., n, n indicate the first candidate key set of words KW(3)In include keyword sum.
In one embodiment, described to be calculated in the first candidate key set of words respectively according to the similarity and the specific gravity
First similarity score of keyword includes: corresponding according to specific gravity, the keyword of keyword in the first candidate key set of words
The product of the similarity of 3rd APP and the 2nd APP obtains the first similarity of keyword described in the first candidate key set of words
Score.Specifically for example: the first similarity score of each keyword in the first candidate key set of words is calculated by following formula:
Wherein,Indicate the first candidate key set of words KW(3)In i-th3A keyword,It indicatesIt is right
The similarity of the 3rd APP and the 2nd APP answered,It indicatesShared specific gravity;i3=1,2 ..., n, n indicate first
Candidate key set of words KW(3)In include keyword sum.
It should be understood that above-mentioned corresponding according to the specific gravity of keyword, the keyword in the first candidate key set of words
The product of the similarity of 3rd APP and the 2nd APP obtains the first similarity of keyword described in the first candidate key set of words
Score, it can be direct product, can also be multiplied by the product after proportionality coefficient.
Finally the first candidate key set of words is screened according to first similarity score, obtains the first expansion pass
Keyword set.Above-described embodiment can realize that first expands keyword set based on competing product APP according to the APP for treating expansion
It expands, expands high-efficient.
In one embodiment, in above-mentioned steps S12, further include determine the second candidate key set of words in each keyword with it is right
The step of similarity for the emphasis keyword answered, the step detailed process include:
According to each keyword search in the second candidate key set of words to APP obtain the feature of each keyword to
Amount, according to each emphasis keyword search to APP obtain the feature vector of each emphasis keyword;By the second candidate keywords
The feature vector of each keyword carries out One-Hot coded treatment with the feature vector of each emphasis keyword respectively in set, obtains
To keyword described in the second candidate key set of words sparse features vector and each emphasis keyword sparse features to
Amount;According to the sparse spy of the sparse features vector of keyword described in the second candidate key set of words and each emphasis keyword
Vector is levied, the similarity of keyword and each emphasis keyword described in the second candidate key set of words is calculated.
Further, it is similar with corresponding emphasis keyword that keyword described in the second candidate key set of words can be calculated
The average value of degree, the comprehensive similarity as keyword described in the second candidate key set of words and corresponding emphasis keyword.Institute
Stating average value can be absolute average, can also be weighted average.
Optionally, i-th in the second candidate key set of words4The determination of a keyword and the similarity of corresponding emphasis keyword
Mode is as follows:
In formula, KW(1)′Indicate emphasis keyword set, KW(1)′ kIndicate k-th emphasis keyword;Indicate second
I-th in candidate key set of words4A keyword;Indicate KW(1)′ kSparse features vector withSparse features vector inner product;Indicate KW(1)′ kSparse features vector withSparse features vector 2- norm product.
It should be understood that between two keywords similarity calculation method, it is including but not limited to above-mentioned based on cosine
The algorithm of similarity calculation similarity can also be used to calculate the algorithm of similarity using other.
In one embodiment, in above-mentioned steps S12, the second candidate keywords are obtained according to the keyword that the 4th APP is covered
Set includes: to obtain a keyword matrix according to the keyword of all the 4th APP coverings;To the keyword in the keyword matrix
Merger statistics, obtains the second candidate key set of words KW(2)=(KW(2) 1,KW2 (2),…,KWn (2)) and the second candidate pass
The corresponding keyword frequency vector C of keyword set(2)=(c1′,c2′,…,cn′)。
Second candidate key set of words KW(2)In i-th4Specific gravity shared by a keyword are as follows:
In formula, i4=1,2 ..., n, n indicate the second candidate key set of words KW(2)In include keyword sum.
In one embodiment, in above-mentioned steps S12, it is candidate that second is calculated according to the specific gravity and the comprehensive similarity
Second similarity score of each keyword in keyword set, comprising: according to the ratio of keyword in the second candidate key set of words
The product of weight and the keyword relative to the comprehensive similarity of emphasis keyword, obtains the pass in the second candidate key set of words
Second similarity score of keyword.Specifically for example: each keyword in the second candidate key set of words can be calculated by following formula
The second similarity score:
Wherein,It indicates i-th in the second candidate key set of words4A keyword,Indicate the second candidate pass
Keyword set KW(2)In i-th4Specific gravity shared by a keyword,Indicate the second candidate key set of words KW(2)In i-th4A pass
The comprehensive similarity of keyword.
It should be understood that above-mentioned according to the specific gravity of keyword in the second candidate key set of words and the keyword and corresponding
The product of the comprehensive similarity of emphasis keyword, the second similarity for obtaining the keyword in the second candidate key set of words obtain
Point, it can it is direct product, can also be multiplied by the product after proportionality coefficient.
Finally the second candidate key set of words is screened according to second similarity score, obtains the second expansion pass
Keyword set.Above-mentioned technical proposal can obtain the second expansion keyword set based on the emphasis keyword of APP to be expanded, protect
It has demonstrate,proved the range that the second expansion keyword set is expanded and has guaranteed that keyword expands quality.
Below by taking apple application shop as an example, the keyword expanding course of the embodiment of the present invention is described further,
Principle is identical therewith for other application library platform.
Shown in Figure 2, the expanding course of the first expansion keyword set includes the following steps.
1, key words content grabs
The nearest historical search in one week of apple application shop is obtained using apple developer API and records data, including but not
Be limited to Apply Names, keyword details, keyword search index, keyword search results, using list etc..
2, history keyword word search record data prediction
The Direct mapping relationship of 2.1 keywords and APP, is denoted as A (k), indicates the search result of keyword k, under appid
The practical ranking of APP is searched in mark index expression with keyword k,
A (k)=(appid1,appid2,…,appidn) (2-1)
N is positive integer in formula.
It should be noted that APP can be identified by appid, and appid is unified by application library platform in the embodiment of the present invention
Distribution, for identifying different APP.
The reverse Mapping relationship of 2.2app and keyword are denoted as K (a), indicate all keywords covered using a:
K (a)=(keyword1,...,keywordn) (2-2)
N is positive integer in formula.
3, competing product APP (i.e. associated 2nd APP) is obtained
3.1 remember that the appid of APP to be expanded is APP(1);
3.2 obtain APP by K (a)(1)Keyword set K (the APP of covering(1)), the first of APP covering to be expanded is crucial
Word;
3.3 couples of keyword set K (APP(1)) carry out screening anomaly.Keyword search results are very little, searchable index is too low,
Search rank rearward, number of words it is too short or it is too long belong to data exception situation, rejected;
3.4 obtain keyword set K (APP by A (k)(1)) in each keyword correspond to appid, be denoted as A (K (APP(1)));
3.5 couples of A (K (APP(1))) merger statistics is carried out, the appid of n before wherein frequency collating is taken, APP set APP is denoted as
′(2);
3.6 reject APP set APP '(2)In with APP(1)It is not belonging to the APP of same application list, finally only takes k conduct
Competing product APP is denoted as competing product APP set APP(2), that is, associated 2nd APP of APP to be expanded.
4, APP expands keyword
NoteFor competing product APP set APP(2)In i-th5A APP traverses competing product APP set APP(2), steps are as follows:
4.1 obtain association appid
With preceding 5 steps in step 3, the associated AP P of competing product APP is obtained, is denoted as the 3rd APP, corresponding collection shares APP(3)
It indicates:
Further, the keyword matrix that (4-1) is covered can be obtained:
4.2 characteristic vector pickup.
By competing product APPFeature vector and APP(3)In each APP covered feature vector (i.e. with it is right in (4-2)
Answer a line keyword) One-Hot coding is carried out, thus obtain competing product APPSparse features vectorWith
And APP(3)In the sparse features vector that is covered of each APP
4.3 calculate APP similarity.
Based on 4.2 as a result, calculating APP(3)In each APP withSimilarity, it is as follows:
In formula,Indicate APP(3)In i-th2A 3rd APP;It indicates
Sparse features vector withSparse features vector inner product;It indicates
Sparse features vector withSparse features vector 2- norm product.
To the crucial conflation of words statistics in (4-2), the first candidate key set of words KW is obtained(3)=(KW(3) 1,KW(3) 2,…,
KW(3) n) and corresponding frequency vector be C(3)=(c1,c2,…,cn);
First candidate key set of words KW(3)In i-th3The specific gravity of a keyword are as follows:
In formula, i3=1,2 ..., n, n indicate the first candidate key set of words KW(3)In include keyword sum.
4.4 calculate the first similarity score of each keyword in the first candidate key set of words.
According to the specific gravity of the similarity of (4-3) and (4-4), the first candidate key set of words KW can be calculated(3)In each key
First similarity score of word.
Finally, to the first candidate key set of words KW(3)Middle keyword carries out inverted order (by height according to the first similarity score
To low), take KW(3)In preceding M, obtain the first expansion keyword set W(1)。
Shown in Figure 2, the expanding course of the second expansion keyword set includes the following steps.
5, emphasis keyword is obtained
Remember that the appid of APP to be expanded is APP(1);First closes in the expanding course for expanding keyword set with above-mentioned first
The method of determination of keyword obtains the first keyword of APP covering to be expanded, is denoted as K (APP(1))。
To the first keyword K (APP(1)) screening anomaly is carried out, keyword search results are very little, searchable index is too low, search
Rank behind, number of words is too short or it is too long belong to data exception situation, it is rejected from the corresponding set of the first keyword;So
Afterwards according to A (k), APP in search result is chosen(1)Ranking preceding k keyword as emphasis keyword, be denoted as emphasis keyword
Set KW(1)′;
6, keyword expands keyword
NoteFor emphasis keyword set KW(1)′In i-th6A keyword traverses KW(1)′, steps are as follows:
6.1 obtain keyword according to A (k)Corresponding appid, takes the APP of k before ranking, is denoted as
6.2 obtain according to K (a)In the keyword that is covered of each APP, be denoted asMerger
Keyword is counted, the frequency of keyword is obtained, takes the k keyword that the frequency is forward, obtain the second candidate key set of words are as follows: KW(2)=(KW(2) 1,KW2 (2),…,KWn (2)), frequency vector are as follows: C(2)=(c1′,c2′,…,cn′);
Define the second candidate key set of words KW(2)In i-th4The specific gravity of a keyword are as follows:
In formula, i4=1,2 ..., n,For the second candidate key set of words KW(2)In i-th4The frequency of a keyword.
6.3 obtain emphasis keyword according to A (k)With the second candidate key set of words KW(2)In each keyword pair
The appid answered, and in this, as the feature vector of keyword, based on One-Hot coding obtain corresponding sparse features to
Amount.It, can calculation stress keyword based on the corresponding sparse features vector of keywordWith the second candidate key word set
Close KW(2)In i-th4The cosine similarity of a keyword, is denoted as
In formula,Indicate i-th6A emphasis keyword;It indicates i-th in the second candidate key set of words4A pass
Keyword;It indicatesSparse features vector withSparse features vector it is interior
Product;It indicatesSparse features vector withSparse features vector 2- model
Several products.
The average value for calculating keyword and the similarity of corresponding emphasis keyword in the second candidate key set of words, as
The comprehensive similarity of keyword described in second candidate key set of words and corresponding emphasis keyword, the comprehensive similarity of keyword
It is denoted as
6.4 calculate the second similarity score of keyword in the second candidate key set of words are as follows:
Wherein,It indicates i-th in the second candidate key set of words4A keyword,Indicate the second candidate pass
Keyword set KW(2)In i-th4Specific gravity shared by a keyword,Indicate the second candidate key set of words KW(2)In i-th4A pass
The comprehensive similarity of keyword.
6.5 finally, to the second candidate key set of words KW(2)Middle keyword according to the second similarity score carry out inverted order (by
It is high to Low), take KW(2)In preceding M keyword, thus obtain the second expansion keyword set W(2)。
7. normalization
To eliminate dimension, keyword set W is expanded by first(1)Keyword set W is expanded with second(2)The phase of middle keyword
Section [0,1] is normalized to as follows respectively like degree score, is specifically included:
Keyword set W is expanded by first(1), second expand keyword set W(2)Obtain keyword set W(3), under
Formula is to set W(3)In each keyword the first similarity score or the second similarity score be normalized:
Wherein,For W(3)In i-th1The first similarity score or the second similarity score of a keyword, sminWith smaxPoint
It Biao Shi not W(3)The minimum value and maximum value of the corresponding similarity score of middle keyword,For W(3)In i-th1A keyword normalization
Similarity score afterwards.
8. screening anomaly
W(3)Middle keyword search index is too low, length is too small or the too big, complex form of Chinese characters, there is no itself APP and competing product APP
Covering word etc. belong to abnormal conditions, reject.
9. calculating scoring
Inquire W(3)The searchable index of middle keyword, the principle based on formula (7-1) are normalized to obtain W(3)In
The searchable index correction value p ' of each keyword is arranged weight α ∈ [0,1], calculates W as follows(3)In each keyword it is final
Similarity score:
10. keyword output scheme
10.1 to W(3)Keyword by final similarity score inverted order arrange;
10.2 choose whether to need to reject repetition values to increase scheme information amount;If so, step 10.3 is executed, if it is not, then
Limitation when meet setting number of words (such as 100 words) when just export keyword scheme (calculated by word, not number according to keyword, than
If " shopping " and " social activity " be two keywords, but be four words), and so on until three sets of keyword schemes of output, and
And determine the scoring of every set keyword scheme are as follows:
In formula, i7=1,2,3, m be the number of keyword in every set keyword scheme.
If desired 10.3 reject repetition values to increase scheme information amount, then first seek W(3)In i-th1A keywordWith
i1+ 1 keywordThe public word string str (i of maximum1,i1+ 1), if str (i1,i1+ 1) length is greater than setting length,
Then willStr (i1,i1+ 1) it replaces withIt can be combined to " mobile phone Taobao purchase such as " mobile phone Taobao " and " Taobao's shopping "
Object ";Otherwise it is separated by with comma.Similarly 10.2 three sets of keyword schemes of output, determine the scoring of every set keyword scheme.
In above-mentioned steps, 1~2 can be off-line calculation, regularly update, for example update one time again weekly.Step 3~10
It is corresponding appid to be obtained to each APP name query data mapping library of user's input, and then can be from two in line computation
Dimension expands out the corresponding keyword of the APP comprehensively.
It should be noted that for the various method embodiments described above, describing for simplicity, it is all expressed as a series of
Combination of actions, but those skilled in the art should understand that, the present invention is not limited by the sequence of acts described, because according to
According to the present invention, certain steps can use other sequences or carry out simultaneously.In addition, also any group can be carried out to above-described embodiment
It closes, obtains other embodiments.
Based on thought identical with the various dimensions expansion method of keyword in above-described embodiment, the present invention also provides multidimensional
Degree expands the device of keyword, which can be used for executing the method that above-mentioned various dimensions expand keyword.For ease of description, more
Dimension is expanded in the structural schematic diagram of the Installation practice of keyword, illustrate only part related to the embodiment of the present invention,
It will be understood by those skilled in the art that the restriction of schematic structure not structure twin installation, may include more more or less than illustrating
Component, perhaps combine certain components or different component layouts.
Fig. 3 is the schematic diagram that the various dimensions of one embodiment of the invention expand the device of keyword;As shown in figure 3,
The device that the various dimensions of the present embodiment expand keyword includes:
First opens up word module, the first keyword covered in application library platform for obtaining APP to be expanded, according to each
First keyword obtains associated 2nd APP of APP to be expanded in the APP that application library platform searches;Each 2nd APP is obtained to answer
With the second keyword covered in the platform of library, obtained in the APP that application library platform searches wait expand according to each second keyword
Associated 3rd APP of APP;The keyword that each 3rd APP is covered in application library platform is obtained, according to each 3rd APP covering
Keyword obtains the first candidate key set of words;It determines the similarity that each 3rd APP gathers relative to the 2nd APP, determines first
Specific gravity shared by each keyword in candidate key set of words calculates the first candidate key according to the similarity and the specific gravity
First similarity score of each keyword in set of words;According to first similarity score from the first candidate key set of words
The keyword for filtering out the first setting quantity obtains the first expansion keyword set;
Second opens up word module, for filtering out emphasis keyword from the first keyword, is existed according to each emphasis keyword
The APP that application library platform searches obtains associated 4th APP of APP to be expanded;It is covered in application library platform according to the 4th APP
Keyword obtain the second candidate key set of words;Determine that each keyword is relative to emphasis key in the second candidate key set of words
The comprehensive similarity of set of words determines the specific gravity of each keyword in the second candidate key set of words, according to the specific gravity and
The comprehensive similarity calculates the second similarity score of each keyword in the second candidate key set of words;According to second phase
The keyword for filtering out the second setting quantity from the second candidate key set of words like degree score obtains the second expansion key
Set of words;And
Screening module expands selection third setting number in keyword set for expanding keyword set and second from first
The keyword of amount obtains the expansion keyword of APP to be expanded;
Wherein, the keyword of APP covering need to meet condition: include in the search result of application library platform in the keyword
There is the APP.
It should be noted that the various dimensions of above-mentioned example are expanded in the embodiment of the device of keyword, between each module
The contents such as information exchange, implementation procedure, due to being based on same design, bring technology with preceding method embodiment of the present invention
Effect is identical as preceding method embodiment of the present invention, and for details, please refer to the description in the embodiment of the method for the present invention, herein not
It repeats again.
In addition, the various dimensions of above-mentioned example are expanded in the embodiment of the device of keyword, the logic of each program module is drawn
Divide and be merely illustrative of, can according to need in practical application, such as the configuration requirement of corresponding hardware or the reality of software
Above-mentioned function distribution is completed by different program modules, i.e., the various dimensions is expanded to the dress of keyword by existing convenient consideration
The internal structure set is divided into different program modules, to complete all or part of the functions described above.
It will appreciated by the skilled person that realizing all or part of the process in above-described embodiment method, being can
It is completed with instructing relevant hardware by computer program, the program can be stored in a computer-readable storage and be situated between
In matter, sells or use as independent product.When being executed, the complete of the method such as the various embodiments described above can be performed in described program
Portion or part steps.Wherein, the storage medium can be magnetic disk, CD, read-only memory (Read-Only
Memory, ROM) or random access memory (Random Access Memory, RAM) etc..
Accordingly, a kind of storage medium is also provided in one embodiment, is stored thereon with computer program, wherein the journey
The method for expanding keyword such as any one various dimensions in the various embodiments described above is realized when sequence is executed by processor.
In addition, the storage medium it is also settable with a kind of computer equipment in, further include place in the computer equipment
Manage device, when the processor executes the program in the storage medium, can be realized the method for the various embodiments described above whole or
Part steps.
Accordingly, a kind of computer equipment is also provided in one embodiment, which includes memory, processor
And store the computer program that can be run on a memory and on a processor, wherein processor is realized when executing described program
The method for expanding keyword such as any one various dimensions in the various embodiments described above.
In the above-described embodiments, it all emphasizes particularly on different fields to the description of each embodiment, there is no the portion being described in detail in some embodiment
Point, it may refer to the associated description of other embodiments.It is appreciated that term " first ", " second " used in wherein etc. is at this
For distinguishing object in text, but these objects should not be limited by these terms.It is of the invention several above described embodiment only expresses
Kind embodiment, should not be understood as limitations on the scope of the patent of the present invention.It should be pointed out that for the ordinary skill of this field
For personnel, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to the present invention
Protection scope.Therefore, the scope of protection of the patent of the invention shall be subject to the appended claims.
Claims (15)
1. a kind of method that various dimensions expand keyword characterized by comprising
The first keyword that APP to be expanded is covered in application library platform is obtained, it is flat in application library according to each first keyword
The APP that platform searches obtains associated 2nd APP of APP to be expanded;The second keyword for obtaining each 2nd APP covering, according to each
Second keyword search to APP obtain associated 3rd APP of APP to be expanded;Obtain the keyword of each 3rd APP covering, root
The first candidate key set of words is obtained according to the keyword that each 3rd APP is covered;Determine that each 3rd APP gathers relative to the 2nd APP
Similarity, specific gravity shared by each keyword in the first candidate key set of words is determined, according to the similarity and the ratio
First similarity score of each keyword in re-computation the first candidate key set of words;According to first similarity score from
The keyword that the first setting quantity is filtered out in one candidate key set of words obtains the first expansion keyword set;
Emphasis keyword is filtered out from the first keyword, according to each emphasis keyword search to APP obtain APP to be expanded
Associated 4th APP;The second candidate key set of words is obtained according to the keyword that the 4th APP is covered;Determine the second candidate key
Comprehensive similarity of each keyword relative to emphasis keyword set in set of words determines each in the second candidate key set of words
The specific gravity of keyword calculates each keyword in the second candidate key set of words according to the specific gravity and the comprehensive similarity
Second similarity score;The second setting is filtered out from the second candidate key set of words according to second similarity score
The keyword of quantity obtains the second expansion keyword set;
From first expand keyword set and second expand keyword set in choose third setting quantity keyword, obtain to
Expand the expansion keyword of APP;
Wherein, the keyword of APP covering need to meet condition: in the keyword comprising in the search result of application library platform
State APP;The emphasis keyword need to meet condition: in the search result of emphasis keyword, APP's to be expanded is in the top;
Wherein, the similarity of the 3rd APP and corresponding 2nd APP indicates the Synthesis Relational Grade of the 3rd APP and corresponding 2nd APP;
If corresponding 2nd APP of the 3rd APP is one, the similarity of the 3rd APP and corresponding 2nd APP are obtained, as described the
The similarity that three APP gather relative to the 2nd APP;If corresponding 2nd APP of the 3rd APP is two or more, the is obtained respectively
The similarity of three APP and each corresponding 2nd APP are gathered using similarity mean value as the 3rd APP relative to the 2nd APP
Similarity;Wherein, the similarity of the 3rd APP and single 2nd APP is predetermined, or is based on application platform
Search record in real time calculate;
Wherein, each keyword specific gravity shared in the first candidate key set of words is based on keyword for the 3rd APP
Different degree determine;Different degree of the keyword for APP, the historical search record data including first passing through application library platform in advance
The data different degree or preset different degree analyzed.
2. the method that various dimensions according to claim 1 expand keyword, which is characterized in that
It is described that obtain the first keyword that APP to be expanded is covered in application library platform include: the history according to application library platform
Search record obtains whole keywords of APP covering to be expanded;Treat the whole keywords progress screening anomaly for expanding APP covering
To delete abnormal keyword therein, the first keyword of APP covering to be expanded is obtained;
And/or
Second keyword for obtaining each 2nd APP covering, comprising: obtained according to the historical search of application library platform record each
Whole keywords of 2nd APP covering;It is therein different to delete that screening anomaly is carried out to whole keywords of each 2nd APP covering
Normal keyword obtains the second keyword of the 2nd APP covering;
And/or
The keyword for obtaining each 3rd APP covering, comprising: each third is obtained according to the historical search of application library platform record
Whole keywords of APP covering;Screening anomaly is carried out to whole keywords of each 3rd APP covering to delete abnormal pass therein
Keyword obtains the keyword of the 3rd APP covering;
The exception keyword includes: that searchable index exception, keyword search results data exception, APP are arranged in search result
The keyword of at least one of name is abnormal, number of characters is abnormal feature.
3. the method that various dimensions according to claim 2 expand keyword, which is characterized in that
It is described that associated 2nd APP of APP to be expanded is obtained in the APP that application library platform searches according to each first keyword,
Include: according to historical search record in each first keyword setting historical period in multiple search result, obtain this first
The frequency sequencing information of APP in the corresponding multiple search result of keyword;Obtain the preceding setting quantity of frequency sequence row
APP, the APP arrived as each first keyword search;According to the APP that all the first keyword, each first keyword search are arrived
Information obtains an APP matrix;The frequency of occurrence for counting each APP in the APP matrix is chosen frequency of occurrence in the APP matrix and is greater than
Or associated 2nd APP of APP to be expanded is used as equal to the APP of the first setting frequency;
And/or
The APP arrived according to each second keyword search obtains associated 3rd APP of APP to be expanded, comprising: according to history
Multiple search result of each second keyword in setting historical period, it is corresponding more to obtain second keyword in search record
The frequency sequencing information of APP in secondary search result;The APP for obtaining the preceding setting quantity of frequency sequence row, is closed as each second
The APP that keyword searches;According to the APP that all the second keyword, each second keyword search are arrived, an APP matrix is obtained;System
The frequency of occurrence for counting each APP in the APP matrix chooses frequency of occurrence in the APP matrix and is greater than or equal to the second setting frequency
APP is as associated 3rd APP of the 2nd APP;
And/or
It is described according to emphasis keyword search to APP obtain the 4th APP, comprising: according to historical search record in each emphasis close
Multiple search result of the keyword in setting historical period, obtains APP in the corresponding multiple search result of the emphasis keyword
Frequency sequencing information;The APP for obtaining the preceding setting quantity of frequency sequence row, is arrived as the emphasis keyword search
APP;According to whole emphasis keywords, each emphasis keyword search to APP obtain an APP matrix;It counts in the APP matrix
The frequency of occurrence of each APP chooses the APP that frequency of occurrence in the APP matrix is greater than or equal to the setting frequency, obtains the 4th APP.
4. the method that various dimensions according to claim 3 expand keyword, which is characterized in that
After obtaining the 2nd APP, before the second keyword for obtaining each 2nd APP covering, further includes: obtain APP to be expanded
List is applied belonging in application library platform, deletes the 2nd APP for belonging to different application list with APP to be expanded;
And/or
After obtaining the 3rd APP, before the keyword for obtaining each 3rd APP covering, further includes: obtain APP to be expanded and answering
With application list affiliated in the platform of library, the 3rd APP for belonging to different application list with APP to be expanded is deleted;
And/or
After obtaining the 4th APP, before the keyword for obtaining each 4th APP covering, further includes: obtain APP to be expanded and answering
With application list affiliated in the platform of library, the 4th APP for belonging to different application list with APP to be expanded is deleted.
5. the method that various dimensions according to claim 1 expand keyword, which is characterized in that
The similarity that each 3rd APP of determination gathers relative to the 2nd APP, comprising: if corresponding 2nd APP of the 3rd APP is
One, then the similarity of the 3rd APP and corresponding 2nd APP are obtained, is gathered as the 3rd APP relative to the 2nd APP
Similarity;If corresponding 2nd APP of the 3rd APP is two or more, the 3rd APP and each corresponding 2nd APP are obtained respectively
Similarity, similarity mean value is calculated with this, is gathered using the similarity mean value as the 3rd APP relative to the 2nd APP
Similarity;
And/or
Determine comprehensive similarity of each keyword relative to emphasis keyword set in the second candidate key set of words, comprising: if
The corresponding emphasis keyword of a keyword is one in second candidate key set of words, then obtains the keyword and corresponding emphasis
The similarity of keyword, the comprehensive similarity as the keyword relative to emphasis keyword set;If the second candidate keywords
The corresponding emphasis keyword of a keyword is two or more in set, then obtains the keyword respectively and close with each corresponding emphasis
The similarity of keyword calculates similarity mean value with this, and similar relative to the synthesis of emphasis keyword set as the keyword
Degree.
6. the method that various dimensions according to claim 5 expand keyword, which is characterized in that further include:
Determine that the step of the 3rd APP is with the similarity of corresponding 2nd APP, the step include: second according to the 2nd APP covering
Keyword obtains the feature vector of the 2nd APP, obtains the feature vector of each 3rd APP according to the keyword that each 3rd APP is covered;
The feature vector of the 2nd APP and the feature vector of the 3rd APP are handled by One-Hot coding, obtain the 2nd APP's
The sparse features vector of sparse features vector and the 3rd APP;According to the sparse features vector of the 2nd APP and the 3rd APP
Sparse features vector calculates the similarity of each 3rd APP and corresponding 2nd APP;
And/or
Determine the similarity of each keyword and corresponding emphasis keyword in the second candidate key set of words, comprising: wait according to second
Select each keyword search in keyword set to APP obtain the feature vector of each keyword, it is crucial according to each emphasis
The APP that word searches obtains the feature vector of each emphasis keyword;By the spy of each keyword in the second candidate key set of words
It levies vector and carries out One-Hot coded treatment with the feature vector of corresponding emphasis keyword respectively, obtain the second candidate key word set
The sparse features vector of the sparse features vector of keyword described in conjunction and corresponding emphasis keyword;According to the second candidate key
The sparse features vector of the sparse features vector of keyword described in set of words and corresponding emphasis keyword, it is candidate to calculate second
The similarity of keyword described in keyword set and corresponding emphasis keyword.
7. the method that various dimensions according to claim 6 expand keyword, which is characterized in that be calculated by the following formula each
The similarity of 3rd APP and corresponding 2nd APP:
In formula, APP(2) tIndicate t-th of the 2nd APP;Indicate i-th2A 3rd APP;Table
Show APP(2) tSparse features vector withSparse features vector inner product;It indicates
APP(2) tSparse features vector withSparse features vector 2- norm product;
And/or
It is calculated by the following formula i-th in the second candidate key set of words4A keyword is similar with corresponding emphasis keyword
Degree:
In formula, KW(1)′ kIndicate k-th emphasis keyword;It indicates i-th in the second candidate key set of words4A keyword;Indicate KW(1)′ kSparse features vector withSparse features vector inner product;Indicate KW(1)′ kSparse features vector withSparse features vector 2- norm
Product.
8. the method that various dimensions according to claim 7 expand keyword, which is characterized in that
The first candidate key set of words is obtained according to the keyword that each 3rd APP is covered, comprising: according to the 2nd APP associated the
The keyword of three APP and each 3rd APP covering obtains a keyword matrix;To the crucial conflation of words in the keyword matrix
Statistics, obtains the first candidate key set of words KW(3)=(KW(3) 1,KW(3) 2,…,KW(3) n) and the corresponding keyword frequency to
Amount is C(3)=(c1,c2,…,cn), each element of the keyword frequency vector respectively corresponds the first candidate key set of words
In each keyword frequency of occurrence;
The first candidate key set of words KW(3)In i-th3Specific gravity shared by a keyword are as follows:
In formula, i3=1,2 ..., n, n indicate the first candidate key set of words KW(3)In include keyword sum;
And/or
The second candidate key set of words is obtained according to the keyword that the 4th APP is covered, comprising: according to all the 4th APP coverings
Keyword obtains a keyword matrix;To the crucial conflation of words statistics in the keyword matrix, the second candidate key word set is obtained
Close KW(2)=(KW(2) 1,KW2 (2),…,KWn (2)) and the corresponding keyword frequency vector C of the second candidate key set of words(2)=(c '1,c′2,…,c′n);Each element of the keyword frequency vector respectively corresponds in the second candidate key set of words
The frequency of occurrence of each keyword;
The second candidate key set of words KW(2)In i-th4Specific gravity shared by a keyword are as follows:
In formula, i4=1,2 ..., n, n indicate the second candidate key set of words KW(2)In include keyword sum.
9. the method that various dimensions according to any one of claims 1 to 8 expand keyword, which is characterized in that
It is obtained according to the first similarity that the similarity and the specific gravity calculate each keyword in the first candidate key set of words
Point, comprising: according to corresponding 3rd APP of the specific gravity of keyword, the keyword and corresponding second in the first candidate key set of words
The product of the similarity of APP obtains the first similarity score of keyword described in the first candidate key set of words;
And/or
It is similar according to second of each keyword in the specific gravity and the comprehensive similarity the second candidate key set of words of calculating
Spend score, comprising: according to the specific gravity of keyword in the second candidate key set of words and the keyword and corresponding emphasis keyword
Comprehensive similarity product, obtain the second similarity score of the keyword in the second candidate key set of words.
10. the method that various dimensions according to any one of claims 1 to 8 expand keyword, which is characterized in that from the first key
Emphasis keyword is filtered out in word, comprising:
According to each first keyword for the different degree of APP to be expanded, it is important more than or equal to the first setting to choose different degree
The first keyword for spending threshold value, as emphasis keyword;
Wherein, keyword characterizes row of the APP to be expanded in the search result of the keyword for the different degree of APP to be expanded
Name information.
11. the method that various dimensions according to claim 1 expand keyword, which is characterized in that expand keyword from first
Set and second expands in keyword set the keyword for choosing third setting quantity, obtains the expansion keyword of APP to be expanded,
Include:
It is obtained third by the first expansion keyword set, the second expansion keyword set and is expanded keyword set, third is expanded
The first similarity score of each keyword or the second similarity score are normalized in keyword set;Third is obtained to open up
The searchable index for opening up each keyword in keyword set, according to similar after the searchable index of each keyword and normalized
Score is spent, the final similarity score that third expands each keyword in keyword set is calculated;According to final similarity score from
Third expands in keyword set the keyword for choosing third setting quantity, obtains the expansion keyword of APP to be expanded.
12. the method that various dimensions according to claim 11 expand keyword, which is characterized in that
It is obtained third by the first expansion keyword set, the second expansion keyword set and is expanded keyword set, comprising: first opens up
Exhibition keyword set, the second expansion keyword set obtain a keyword set, are picked to the keyword set and are handled again, obtained
Keyword set is expanded to third;
And/or
The first similarity score of each keyword or the second similarity in keyword set is expanded to third by following formula to obtain
Divide and be normalized:
Wherein,It is expanded i-th in keyword set for third1The first similarity score or the second similarity score of a keyword,
sminWith smaxMinimum value and maximum value that third expands the corresponding similarity score of keyword in keyword set are respectively indicated,
It is expanded i-th in keyword set for third1Similarity score after a keyword normalization;
And/or
Third is calculated as follows to expand i-th in keyword set1The final similarity score of a keyword:
Wherein,It is expanded i-th in keyword set for third1Similarity score after a keyword normalization, p ' are keyword
Searchable index correction value, weight α ∈ [0,1];The searchable index correction value of keyword by keyword searchable index normalization
Reason obtains;
The keyword for choosing third setting quantity is expanded in keyword set from third according to final similarity score, is obtained wait open up
Open up the expansion keyword of APP, comprising:
The keyword expanded in keyword set to third is arranged by final similarity score inverted order, is successively selected according to rank results
The keyword for taking setting number of words, obtains a set of keyword scheme, and so on obtain covering keyword schemes more;And using following public
Formula determines the scoring of a set of keyword scheme are as follows:
In formula, i7=1,2 ... q, q are total tricks of keyword scheme, and m is the number of keyword in a set of keyword scheme,
score(1) jThird expands the final similarity score of j-th of keyword in keyword set.
13. the device that a kind of various dimensions expand keyword characterized by comprising
First opens up word module, the first keyword covered in application library platform for obtaining APP to be expanded, according to each first
Keyword obtains associated 2nd APP of APP to be expanded in the APP that application library platform searches;Obtain the of each 2nd APP covering
Two keywords, according to each second keyword search to APP obtain associated 3rd APP of APP to be expanded;Obtain each 3rd APP
The keyword of covering obtains the first candidate key set of words according to the keyword that each 3rd APP is covered;Determine each 3rd APP phase
For the similarity of the 2nd APP set, specific gravity shared by each keyword in the first candidate key set of words is determined, according to the phase
The first similarity score of each keyword in the first candidate key set of words is calculated like degree and the specific gravity;According to described first
Similarity score filters out the keyword of the first setting quantity from the first candidate key set of words, obtains the first expansion keyword
Set;
Second opens up word module, for filtering out emphasis keyword from the first keyword, is arrived according to each emphasis keyword search
APP obtain associated 4th APP of APP to be expanded;The second candidate key word set is obtained according to the keyword that the 4th APP is covered
It closes;It determines comprehensive similarity of each keyword relative to emphasis keyword set in the second candidate key set of words, determines second
The specific gravity of each keyword in candidate key set of words calculates the second candidate pass according to the specific gravity and the comprehensive similarity
Second similarity score of each keyword in keyword set;According to second similarity score from second candidate keywords
The keyword that the second setting quantity is filtered out in set obtains the second expansion keyword set;And
Screening module expands selection third setting quantity in keyword set for expanding keyword set and second from first
Keyword obtains the expansion keyword of APP to be expanded;
Wherein, the keyword of APP covering need to meet condition: in the keyword comprising in the search result of application library platform
State APP;The emphasis keyword need to meet condition: in the search result of emphasis keyword, APP's to be expanded is in the top;
Wherein, the similarity of the 3rd APP and corresponding 2nd APP indicates the Synthesis Relational Grade of the 3rd APP and corresponding 2nd APP;
If corresponding 2nd APP of the 3rd APP is one, the similarity of the 3rd APP and corresponding 2nd APP are obtained, as described the
The similarity that three APP gather relative to the 2nd APP;If corresponding 2nd APP of the 3rd APP is two or more, the is obtained respectively
The similarity of three APP and each corresponding 2nd APP are gathered using similarity mean value as the 3rd APP relative to the 2nd APP
Similarity;Wherein, the similarity of the 3rd APP and single 2nd APP is predetermined, or is based on application platform
Search record in real time calculate;
Wherein, each keyword specific gravity shared in the first candidate key set of words is based on keyword for the 3rd APP
Different degree determine;Different degree of the keyword for APP, the historical search record data including first passing through application library platform in advance
The data different degree or preset different degree analyzed.
14. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
The step of claim 1 to 12 any the method is realized when execution.
15. a kind of computer equipment including memory, processor and stores the meter that can be run on a memory and on a processor
Calculation machine program, which is characterized in that the processor realizes the step of claim 1 to 12 any the method when executing described program
Suddenly.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711229068.7A CN108052554B (en) | 2017-11-29 | 2017-11-29 | The method and apparatus of various dimensions expansion keyword |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711229068.7A CN108052554B (en) | 2017-11-29 | 2017-11-29 | The method and apparatus of various dimensions expansion keyword |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108052554A CN108052554A (en) | 2018-05-18 |
CN108052554B true CN108052554B (en) | 2019-04-30 |
Family
ID=62121443
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711229068.7A Active CN108052554B (en) | 2017-11-29 | 2017-11-29 | The method and apparatus of various dimensions expansion keyword |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108052554B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111435374B (en) * | 2019-01-11 | 2023-04-25 | 百度在线网络技术(北京)有限公司 | Display device and method for searching statistical data |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103049470A (en) * | 2012-09-12 | 2013-04-17 | 北京航空航天大学 | Opinion retrieval method based on emotional relevancy |
CN103455613A (en) * | 2013-09-06 | 2013-12-18 | 南京大学 | Interest aware service recommendation method based on MapReduce model |
CN103870505A (en) * | 2012-12-17 | 2014-06-18 | 阿里巴巴集团控股有限公司 | Query term recommending method and query term recommending system |
CN104915405A (en) * | 2015-06-02 | 2015-09-16 | 华东师范大学 | Microblog query expansion method based on multiple layers |
JP2016126567A (en) * | 2015-01-05 | 2016-07-11 | 日本放送協会 | Content recommendation device and program |
-
2017
- 2017-11-29 CN CN201711229068.7A patent/CN108052554B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103049470A (en) * | 2012-09-12 | 2013-04-17 | 北京航空航天大学 | Opinion retrieval method based on emotional relevancy |
CN103870505A (en) * | 2012-12-17 | 2014-06-18 | 阿里巴巴集团控股有限公司 | Query term recommending method and query term recommending system |
CN103455613A (en) * | 2013-09-06 | 2013-12-18 | 南京大学 | Interest aware service recommendation method based on MapReduce model |
JP2016126567A (en) * | 2015-01-05 | 2016-07-11 | 日本放送協会 | Content recommendation device and program |
CN104915405A (en) * | 2015-06-02 | 2015-09-16 | 华东师范大学 | Microblog query expansion method based on multiple layers |
Also Published As
Publication number | Publication date |
---|---|
CN108052554A (en) | 2018-05-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105701216B (en) | A kind of information-pushing method and device | |
CN105893609B (en) | A kind of mobile APP recommended method based on weighted blend | |
Masucci et al. | Network properties of written human language | |
CN110532451A (en) | Search method and device for policy text, storage medium, electronic device | |
CN103914478B (en) | Webpage training method and system, webpage Forecasting Methodology and system | |
CN107862022B (en) | Culture resource recommendation system | |
CN106547871A (en) | Method and apparatus is recalled based on the Search Results of neutral net | |
CN106874292A (en) | Topic processing method and processing device | |
CN103916436B (en) | Information-pushing method, device, terminal and server | |
CN109582969A (en) | Methodology for Entities Matching, device and electronic equipment | |
CN107729468A (en) | Answer extracting method and system based on deep learning | |
CN103309869B (en) | Method and system for recommending display keyword of data object | |
CN110334178A (en) | Data retrieval method, device, equipment and readable storage medium storing program for executing | |
CN108647322B (en) | Method for identifying similarity of mass Web text information based on word network | |
CN107180093A (en) | Information search method and device and ageing inquiry word recognition method and device | |
CN103617213B (en) | Method and system for identifying newspage attributive characters | |
CN110033342A (en) | A kind of training method and device, a kind of recommended method and device of recommended models | |
CN109189990A (en) | A kind of generation method of search term, device and electronic equipment | |
CN109582849A (en) | A kind of Internet resources intelligent search method of knowledge based map | |
CN109933660A (en) | The API information search method based on handout and Stack Overflow towards natural language form | |
CN113239268B (en) | Commodity recommendation method, device and system | |
US20190324979A1 (en) | System, method, and non-transitory computer-readable storage media for generating synonyms of a search query | |
CN110134792A (en) | Text recognition method, device, electronic equipment and storage medium | |
CN108280124A (en) | Product classification method and device, ranking list generation method and device, electronic equipment | |
CN108052554B (en) | The method and apparatus of various dimensions expansion keyword |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |