CN101820592A - Method and device for mobile search - Google Patents
Method and device for mobile search Download PDFInfo
- Publication number
- CN101820592A CN101820592A CN200910140119A CN200910140119A CN101820592A CN 101820592 A CN101820592 A CN 101820592A CN 200910140119 A CN200910140119 A CN 200910140119A CN 200910140119 A CN200910140119 A CN 200910140119A CN 101820592 A CN101820592 A CN 101820592A
- Authority
- CN
- China
- Prior art keywords
- search
- interest
- territory
- user
- type
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Abstract
The invention discloses a method and a device for mobile search. The method comprises the following steps of: recovering a search request containing one or more query key words; calculating the score of each search type domain, wherein the score is the score of any of the following items or the comprehensive score of the items: the similarity between the search request and the search type domain, a public search rate, corresponding to the search type domain, of the search request and an individual user interest score of the search type domain, and the public search rate is the number of public searches or the times of the clicking of a public search result; and according to the scores of the various search type domains, selecting one or more search type domains to be searched for the query key words. By using the method and the device, individual and accurate search results can be provided for a user.
Description
Technical field
The present invention relates to mobile communication technology, be specifically related to a kind of mobile search method and device.
Background technology
At present, the combination-mobile search as two big hot topic fields of search engine and these two current information industries of mobile communication has become new bright spot of mobile value-added service and growth point.The mobile search framework is a platform based on the opening of unit's search, and it integrates the ability of many specialty/vertical search engines, for the user provides a comprehensive search capability.
When the user uses mobile search, usually directly search for behind the inputted search keyword and select the type field (domain) of search.Therefore, how correct understanding user's search intention for the user provides personalized accurate search results, does not also have good solution in the prior art.
Summary of the invention
The embodiment of the invention provides a kind of mobile search method and device, can provide personalized Search Results accurately for the user.
The embodiment of the invention provides a kind of mobile search method, comprising:
Receive searching request, comprise one or more keys word of the inquiry in the described searching request;
Calculate the score value in each search-type territory, described score value is following any one score value or multinomial comprehensive grading value: described searching request and the popular searching rate in the similarity in described search-type territory, the corresponding described search-type of described searching request territory, the personalized user interest scores value in search-type territory;
Select one of them or the described key word of the inquiry of several search-type domain search according to the score value in each search-type territory.
The embodiment of the invention provides a kind of mobile search device, comprising:
Receiving element is used to receive searching request, comprises one or more keys word of the inquiry in the described searching request;
Computing unit, be used to calculate the score value in each search-type territory, described score value is following any one score value or multinomial comprehensive grading value: described searching request and the popular searching rate in the similarity in described search-type territory, the corresponding described search-type of described searching request territory, the personalized user interest scores value in search-type territory;
Selected cell is selected one of them or several search-type territory according to the score value in each search-type territory;
Search unit is used to the described key word of the inquiry of search-type domain search that utilizes described selected cell to select.
Mobile search method that the embodiment of the invention provides and device by the popular interest of analysis user and user's personalized interest, are determined user's personalized enquire classification, thereby the accurate search results of personalization is provided for the user.
Description of drawings
Fig. 1 is the flow chart of embodiment of the invention mobile search method;
Fig. 2 is a kind of realization flow figure of embodiment of the invention mobile search method;
Fig. 3 is the another kind of realization flow figure of embodiment of the invention mobile search method;
Fig. 4 is the another kind of realization flow figure of embodiment of the invention mobile search method;
Fig. 5 is the another kind of realization flow figure of embodiment of the invention mobile search method;
Fig. 6 is the structural representation of embodiment of the invention mobile search device;
Fig. 7 is a kind of concrete structure schematic diagram of embodiment of the invention mobile search device;
Fig. 8 is the another kind of concrete structure schematic diagram of embodiment of the invention mobile search device;
Fig. 9 is the another kind of concrete structure schematic diagram of embodiment of the invention mobile search device;
Figure 10 is a kind of structural representation that interest model extracts subelement in the device shown in Figure 9;
Figure 11 is the another kind of structural representation that interest model extracts subelement in the device shown in Figure 9;
Figure 12 is the another kind of concrete structure schematic diagram of embodiment of the invention mobile search device.
Embodiment
In order to make those skilled in the art person understand the scheme of the embodiment of the invention better, the embodiment of the invention is described in further detail below in conjunction with drawings and embodiments.
Embodiment of the invention mobile search method and device, searching request at the user, by the popular interest of analysis user correspondence and user's personalized interest, determine user's personalized enquire classification, particularly, calculate the score value in each search-type territory, described score value is following any one score value or multinomial comprehensive grading value: described searching request and the popular searching rate in the similarity in described search-type territory, the corresponding described search-type of described searching request territory, the personalized user interest scores value in search-type territory; Described popular searching rate is: popular searching times, perhaps popular Search Results number of clicks; Then, select one of them or the described key word of the inquiry of several search-type domain search, thereby provide personalized accurate search results for the user according to the score value in each search-type territory.
As shown in Figure 1, be the flow chart of embodiment of the invention mobile search method.
In embodiments of the present invention, divide time-like at the personalized enquire of determining the user, multiple implementation can be arranged, such as, can be similarity, select one or several high search-type territory of similarity to search for according to described searching request and described search-type territory; Also can be according to the popular searching rate in the corresponding described search-type of described searching request territory, select one or several high search-type territory of popular searching rate to search for; Can also select one or several high search-type territory of personalized user interest scores value to search for according to the personalized user interest scores value in search-type territory.Certainly, can also be to take all factors into consideration above-mentioned several, calculate the comprehensive grading value in each search-type territory, select one or several high search-type territory of comprehensive grading value to search for.Below this is described in detail respectively for example.
With reference to Fig. 2, be a kind of realization flow figure of embodiment of the invention mobile search method.
In this embodiment, according to the similarity in described searching request and described search-type territory, select the search-type territory to search for, so that the Search Results accurately of personalization is provided for the user.
Can corresponding weights be set for the key word of the inquiry in the described searching request, by the weight generated query of described key word of the inquiry vector Query (q1, q2 ... qn '); Wherein, q1, q2 ... qn ' is the weight of corresponding each key word of the inquiry; Particularly, all keywords can be provided with identical weight, such as weight=1; Also different weights can be set for different keywords, such as, for coming top keyword weight limit is set, such as weight=1, the weight of middle size is set for the keyword in the middle of coming, such as 0.5＜weight＜1, for coming last keyword minimal weight is set, such as weight=0.5.
Generate territory vector that should the search-type territory by the weight of each speech in described search-type territory, such as all descriptor and the related term of giving each search-type territory certain weight is set, by the set of weights of these descriptor and related term in pairs should the search-type territory territory vector Domain (t1, t2 ... tn), wherein, t1, t2,, tn is the weight of each speech in this search-type territory.By calculating the similarity that described query vector and territory vector obtain described searching request and search-type territory.
As follows compute vector Domian (t1, t2 ..., tn) with vectorial Query (q1, q2 ..., qn ') between similarity:
Wherein, t
I1, t
I2..., t
In' be respectively vectorial Domian (t1, t2 ..., tn) in weight q1, q2 ..., the weight of the speech correspondence that the key word of the inquiry that qn ' is corresponding is identical.
Suppose to have m search-type territory, corresponding territory vector is respectively Domain1 (t1, t2, ..., tn), Domain2 (t1, t2 ..., tn), ..., Domainm (t1, t2, ..., tn), then by formula (1) distinguishes compute vector Query (q1, q2 ..., qn ') with the similarity of above-mentioned each territory vector.
In this embodiment, descriptor, related term in each search-type territory, and the weight of each speech can have multiple mode to be provided with.
1. manual allocation mode
For descriptor maximum weight is set, the weight of size is provided with minimal weight for weak related term in the middle of being provided with for the strong correlation speech.
Such as: it is 1 that descriptor (as " Sichuan cuisine " in the food and drink search-type territory) is provided with weight, and it is 0.8 that strong correlation speech (as " peppery " in the food and drink search-type territory) is provided with weight, and it is 0.5 that weak related term (as " perfume (or spice) " in the food and drink search-type territory) is provided with weight.
2. by the automatic method of salary distribution of study
Detailed process is as follows:
(1), obtains training text language material sample that should the search-type territory for each search-type territory;
(2) described language material sample is cut speech, generate the dictionary in this search-type territory;
(3) weight of each speech in the described dictionary of calculating, weight=the TF*GIDF of each speech, wherein TF is this speech total word frequency in all language material samples of this search-type territory, GIDF is an oppositely document frequency of the overall situation, GIDF=log (1+N/GDF), wherein N is the total quantity of all language material samples in all search-type territories, and GDF is overall language material sample frequency, is the quantity that comprises all language material samples of this speech in all search-type territories;
(4) determine descriptor and related term in the described search-type territory according to the weight of each speech;
Suppose total n speech in the dictionary in certain search-type territory, the weight of correspondence is T1, T2 ..., Tn, wherein, T1＞T2＞...＞Tn, like this, can think the speech of the T1 correspondence speech that is the theme, other speech are related term.
Further, all speech in the described dictionary can also be divided into the set of different class according to weight, for the set of each class is provided with final score value, and with the final score value of each class weight as each speech in this class.Such as, total L shelves are first grade of the highest score value of setting, the score value of size in the middle of middle bay is provided with, and the L shelves are provided with minimum score value.Like this, the territory vector that can form corresponding search-type territory by speech in the part of speech and final score value thereof.
Certainly, the embodiment of the invention is not limited in above-mentioned these set-up modes, and for descriptor, related term in each search-type territory, and the weight of each speech can also adopt other modes to be provided with, and describes in detail no longer one by one at this.
Embodiment of the invention mobile search method, searching request at the user, the similarity of the query vector by calculating searching request and the territory vector in each search-type territory, select one or several high search-type territory of similarity to search for, thereby can determine the personalized enquire classification for the user, for the user provides personalized accurate search results.
With reference to Fig. 3, be the another kind of realization flow figure of embodiment of the invention mobile search method.
In this embodiment, according to the popular searching rate in the corresponding described search-type of described searching request territory, select the search-type territory to search for, so that the Search Results accurately of personalization is provided for the user.
Step 301 receives searching request, comprises one or more keys word of the inquiry in the described searching request.
Step 302 is calculated the popular searching rate in corresponding each the search-type territory of described searching request according to described key word of the inquiry.
Step 303 selects the high one or more search-type territory of popular searching rate to search for.
In embodiments of the present invention, described popular searching rate specifically can be: popular searching times, perhaps popular Search Results number of clicks etc.
Describe the popular searching times in corresponding each the search-type territory of the described searching request of calculating and the process of popular Search Results number of clicks below respectively in detail.
The process of popular searching times in certain search-type territory of calculating described searching request correspondence is as follows:
(1) masses that calculate certain search-type territory of each keyword correspondence in the described searching request search for total degree;
Can be according to historical record, collect all users select the number of times searched for certain search-type territory about the searching request that comprises certain keyword in the described searching request summation, the total degree of this search-type territory being searched for as the masses of this keyword correspondence is promptly searched for total degree to masses that should the search-type territory;
(2) masses in this search-type territory of all keyword correspondences in the described searching request are searched for total degree and, search for total degree as the masses in this search-type territory of described searching request correspondence.
Equally, it is as follows to calculate the process of popular Search Results number of clicks in certain search-type territory of described searching request correspondence:
(1) the popular Search Results that calculates certain search-type territory of each keyword correspondence in the described searching request is clicked total degree;
Can be according to historical record, collect all users select the Search Results number of clicks of searching for certain search-type territory about the searching request that comprises certain keyword in the described searching request summation, the total degree of the Search Results in this search-type territory being clicked as the masses of this keyword correspondence is promptly clicked total degree to popular Search Results that should the search-type territory;
(2) the popular Search Results in this search-type territory of all keyword correspondences in the described searching request is clicked total degree and, click total degree as the popular Search Results in this search-type territory of described searching request correspondence.
Embodiment of the invention mobile search method, searching request at the user, by calculating the popular searching rate in corresponding each the search-type territory of described searching request, select one or several high search-type territory of popular searching rate to search for, thereby can determine the personalized enquire classification for the user, for the user provides personalized accurate search results.
With reference to Fig. 4, be the another kind of realization flow figure of embodiment of the invention mobile search method.
In this embodiment, according to the personalized user interest scores value in search-type territory, select the high search-type territory of score value to search for, so that the Search Results accurately of personalization is provided for the user.
Step 401 receives searching request, comprises one or more keys word of the inquiry in the described searching request.
Step 402 is extracted user's interest model from user data.
Described user's the interest model vector that to be described user data form at the score value of a plurality of interest dimensions, such as IM (I1, I2 ..., In), wherein Ii is the score value of i interest dimension of user.Can from user individual data (click historical data, present business information, local information etc.), extract user interest model such as static archives, search; Also can go out corresponding user interest model and preservation from the user individual extracting data in advance, when needed, directly the user interest model from these preservations extracts required user interest model.
Described user's interest model can be static interest model or dynamic interest model, certainly, also can be the interest model that comprehensive static interest model and dynamic interest model generate.
Can extract user's static interest model from user's static archives, detailed process can have following dual mode:
(1) calculates the word frequency sum that belongs to all speech of each interest dimension in user's the static archives, and with its score value, generate described user interest model as vector by the score value of corresponding each interest dimension as corresponding each interest dimension;
(2) calculate user's the static archives and the similarity score value of each interest dimension, and with its score value, generate described user interest model as vector by the score value of each interest dimension of correspondence as corresponding each interest dimension;
Extract user's dynamic interest model from user data, detailed process can have following dual mode:
(1) the word frequency sum that belongs to all speech of each interest dimension in the historical record is clicked in the search of calculating the user, and, generate described user's dynamic interest model as vector by the score value of corresponding each interest dimension with its score value as corresponding each interest dimension;
(2) calculate the similarity score value that historical record and each interest dimension are clicked in search, and, generate described user's dynamic interest model as vector by the score value of corresponding each interest dimension its score value as corresponding each interest dimension.
The interest model of comprehensive static interest model and dynamically interest model generation can be:
(1) at first respectively described static interest model and described dynamic interest model are carried out normalized, calculate then one or more static interest model after the normalized and one or more dynamic interest models and, and will be somebody's turn to do and as described user's interest model.
(2) at first one or more described static interest models and one or more described dynamic interest model are weighted addition, and then with weighting summation and carry out normalized, and with the interest model of the result after the normalized as described user.
Step 403 is with the score value sum of one or more interest dimensions of the corresponding described user interest model in the described search-type territory personalized user interest scores value as described search-type territory.
Step 404 is selected the high described key word of the inquiry of one or more search-type domain search of score value.
For example, user's interest is represented with n dimension, as: news, physical culture, amusement, finance and economics, science and technology, house property, recreation, women, forum, weather, commodity, household electrical appliances, music, reading, blog, mobile phone, military affairs, education, tourism, multimedia message, CRBT, food and drink, civil aviaton, industry, agricultural, computer, geography etc.Described user interest model be the vectorial W that the user forms the score value of the interest of each dimension (r1, r2, r3 ..., rn).
From user individual extracting data user interest model the time, can from user's static archives, extract, also can from the historical data of user search, extract.
From user's static archives, extract user interest model W1 following several mode can be arranged:
(1) W1=(p1, p2, p3 ..., pn), wherein pi is the word frequency sum that type belongs to all speech of i interest dimension in the static archives.
(2) W1=(p1, p2, p3 ..., pn), wherein pi is the similarity score value of static archives and i interest dimension.
Wherein, it is as follows to calculate the process of similarity pi of static archives and certain interest dimension:
(a) the feature dictionary of extraction grader is specially:
(i) each interest dimension of user is collected corresponding corpus respectively, generate corpus;
(ii) speech cut in described corpus, form a series of entries;
Judge (iii) whether the entry cut behind the speech is the feature speech, specifically can adopt chi algorithm (CHI):
Wherein, the implication of each parameter is as follows: t: a certain entry; C: a certain classification; N: training text sum; A: belong to c and comprise the training text number of t; B: do not belong to the textual data that c still comprises t; C: belong to c but do not comprise the textual data of t; D: do not belong to the textual data that c does not comprise t yet.If C, D are 0, so χ
2(t, c)=0;
Entry t may be defined as the CHI value of whole training set:
Or
The entry that is lower than assign thresholds can not be considered as the feature speech.
Wherein the computational process of P (c) is as follows:
If classification is C
1, C
2..., C
n,
Then
Wherein, N (C
i) be classification C
iThe quantity of the training text that is comprised;
Perhaps,
Wherein, M (C
i) be classification C
iThe entry sum that all training texts comprised, M is the entry sum that all training texts comprise.
The feature entry that finally obtains is designated as t1, t2 ..., tn.
Certainly, judge when whether the entry cut behind the speech is the feature speech, be not limited in above-mentioned CHI algorithm, can also adopt other algorithms, such as, χ
2(t, c)=| AD-BC|.
(b) the feature speech that obtains according to (a) step, generate i interest dimension characteristic vector W i=(wi1, wi2 ..., wii ..., win), wherein wii is the weight of feature speech ti in i interest dimension.
Wii=TFi*log (1+N/GDFi), TFi is the word frequency that feature speech ti occurs in belonging to all language materials of i interest dimension, N is feature speech ti number of documents in all language materials of interesting dimension, comprises the number of documents of feature speech ti in GDFi (global document frequency) all language materials by interesting dimension.
(c) the feature speech that obtains according to (a) step, the characteristic vector S=of the static archives of generation user (s1, s2 ..., sn), wherein si is the weight of feature speech ti in the static archives of user.
The word frequency that Si=feature speech ti occurs in static archives.
(d) calculate similarity between the characteristic vector W i of static archives vector of user and i interest dimension, obtain the score value pi of similarity,
From the historical data of user search, extract user interest model W2 following several mode can be arranged:
W2=d1+d2+d3+......dm, wherein di is the pairing interest model vector of certain click document of user;
Obtain the pairing interest model vector of certain click document two kinds of methods arranged:
(1) di=(t1, t2, t3 ..., tn), clicked this document when the user is up-to-date, tj equals the word frequency sum that type in the document belongs to all speech of j interest dimension.
(2) di=(t1, t2, t3 ..., tn), wherein di is the similarity score value of document and i interest dimension.The process of calculating di is as follows:
(a) the feature dictionary of extraction grader is specially:
(i) each interest dimension of user is collected corresponding corpus respectively, generate corpus;
(ii) described corpus is carried out participle, form a series of entries;
(iii) judge the entry cut behind the speech, feature speech whether, specifically can adopt the CHI algorithm:
Wherein, the implication of each parameter is as follows: t: a certain entry; C: a certain classification; N: training text sum; A: belong to c and comprise the textual data of t; B: do not belong to the textual data that c still comprises t; C: belong to c but do not comprise the textual data of t; D: do not belong to the textual data that c does not comprise t yet; If C, D are 0, so χ
2(t, c)=0.
Entry t may be defined as the CHI value of whole training set:
Or
The entry that is lower than assign thresholds can not be considered as the feature speech.
The setting classification is C
1, C
2..., C
n, the computational process of P (c) is as follows:
Perhaps,
Wherein, M (C
i) be classification C
iThe entry sum that all training texts comprised, M is the entry sum that all training texts comprise.
The feature entry that finally obtains is designated as t1, t2 ..., tn.
Certainly, judge when whether the entry cut behind the speech is the feature speech, be not limited in above-mentioned CHI algorithm, can also adopt other algorithms, such as, χ
2(t, c)=| AD-BC|.
(b) the feature speech that obtains according to (a) step, generate i interest dimension characteristic vector W i=(wi1, wi2 ..., wii ..., win), wherein wii is the weight of feature speech ti in i interest dimension.
Wii=TFi*log (1+N/GDFi), TFi is the word frequency that feature speech ti occurs in belonging to all language materials of i interest dimension, N is feature speech ti number of documents in all language materials of interesting dimension, comprises the number of documents of feature speech ti in GDFi (global document frequency) all language materials by interesting dimension.
(c) the feature speech that obtains according to (a) step, the characteristic vector V=of generation document (v1, v2 ..., vn), wherein vi is the weight of feature speech ti in document, the word frequency that vi=feature speech ti occurs in document.
(d) calculate similarity between the characteristic vector W i of the characteristic vector v of document and i interest dimension, obtain the score value di of similarity:
If the user estimates certain document of clicking, if be evaluated as, the di vector multiply by a positive constant c, and the importance of expression document increases, promptly di=c*di=(c*ti, c*t2, c*t3 ..., c*tn); If be evaluated as badly, the di vector multiply by the inverse of a positive constant c, and the importance of expression document reduces, promptly di=1/c*di=(1/c*ti, 1/c*t2,1/c*t3 ..., 1/c*tn);
After a period of time, the value of tj reduces certain percentage automatically, and expression As time goes on its importance weakens, and the value of long time tj is kept to till zero up to having crossed, at this moment di can be deleted from historical record.
Respectively W1 and W2 are done normalization, obtain user interest model W=r1*W1+r2*W2, wherein r1+r2=1.
Embodiment of the invention mobile search method, searching request at the user, by calculating the personalized user interest scores value in each search-type territory, select one or several high search-type territory of score value to search for, thereby can determine the personalized enquire classification for the user, for the user provides personalized accurate search results.
In the above among each embodiment, when carrying out the selection of search-type territory, the foundation of selecting as the search-type territory with the personalized user interest scores value in the popular searching rate in described searching request and the similarity in described search-type territory, the corresponding described search-type of described searching request territory and search-type territory respectively, determine user's personalized enquire classification, for the user provides personalized accurate search results.
In embodiments of the present invention, can also take all factors into consideration above-mentioned any two or multinomial, calculate the comprehensive grading value in each search-type territory, select one or several high search-type territory of comprehensive grading value to search for.Be example to take all factors into consideration above-mentioned three foundations of selecting as the search-type territory below, the embodiment of the invention is described in detail.
With reference to Fig. 5, be the another kind of realization flow figure of embodiment of the invention mobile search method.
Step 501 receives searching request, comprises one or more keys word of the inquiry in the described searching request.
Step 502 is calculated the personalized user interest scores value in described searching request and the similarity in each search-type territory, the popular searching rate in corresponding each the search-type territory of described searching request, described search-type territory respectively.
Step 503 is carried out normalized with each value that obtains corresponding described search-type territory, obtains the comprehensive grading value in each search-type territory.
Such as, calculate the similarity in described searching request and certain search-type territory, and with its normalization, the value of obtaining Score1;
Calculate described searching request to popular searching rate that should the search-type territory, and with its normalization, the value of obtaining Score2;
Calculate the personalized user interest scores value in this search-type territory, and with its normalization, the value of obtaining Score3;
Calculate the comprehensive grading value=r1*score1+r2*score2+r3*score3 in this search-type territory, wherein, r1, r2, r3 is respectively Score1, Score2, the weights of Score3, r1+r2+r3+r4=1.
The comprehensive grading value also can have other account forms, as:
Comprehensive grading value=score1*score2*score3, perhaps
The comprehensive grading value=(score1+score2+score3)/3, etc.
Step 504 selects the high one or more search-type territory of comprehensive grading value to search for.
As seen, in embodiments of the present invention, take all factors into consideration multinomial factor and determined user's personalized enquire classification, calculate the comprehensive grading value in each search-type territory, select one or several high search-type territory of comprehensive grading value to search for, thereby the accurate search results of personalization is provided for the user.
One of ordinary skill in the art will appreciate that all or part of step that realizes in the foregoing description method is to instruct relevant hardware to finish by program, described program can be stored in the computer read/write memory medium, described storage medium, as: ROM/RAM, magnetic disc, CD etc.
The embodiment of the invention also provides a kind of mobile search device, as shown in Figure 6, is the structural representation of this device:
In this embodiment, described device comprises: receiving element 601, computing unit 602, selected cell 603 and search unit 604.Wherein:
Receiving element 601 is used to receive searching request, comprises one or more keys word of the inquiry in the described searching request;
The comprehensive grading value that computing unit 602 calculates each search-type territory is: according to multinomial calculating product score value, average score value or weighted scoring value in the personalized user interest scores value in the popular searching rate in searching request and the corresponding search-type of similarity, the searching request territory in search-type territory and search-type territory.
In embodiments of the present invention, determine that at computing unit 602 and selected cell 603 user's personalized enquire divides time-like, multiple implementation can be arranged, such as, can be similarity, select one or several high search-type territory of similarity to search for according to described searching request and described search-type territory; Also can be according to the popular searching rate in the corresponding described search-type of described searching request territory, select one or several high search-type territory of popular searching rate to search for; Can also select one or several high search-type territory of personalized user interest scores value to search for according to the personalized user interest scores value in search-type territory.Certainly, can also be to take all factors into consideration above-mentioned several, calculate the comprehensive grading value in each search-type territory, select one or several high search-type territory of comprehensive grading value to search for.Therefore, described computing unit 602 comprises following any one or a plurality of unit:
Similarity calculated is used to calculate the similarity in described searching request and each search-type territory;
Popular searching rate computing unit is used to calculate the popular searching rate in corresponding each the search-type territory of described searching request;
User interest scores value computing unit is used to calculate the personalized user interest scores value in each search-type territory.
Below this is described in detail respectively for example.
As shown in Figure 7, be a kind of concrete structure schematic diagram of embodiment of the invention mobile search device.
In this embodiment, described device comprises: receiving element 701, similarity calculated 702, selected cell 703 and search unit 704.Wherein, described receiving element 701, selected cell 703 and search unit 704 with embodiment illustrated in fig. 6 in each corresponding unit consistent, be not described in detail at this.
Described similarity calculated 702 comprises: weight is provided with subelement 721, query vector generates subelement 722, territory vector generation unit 723 and first computation subunit 724.Wherein: weight is provided with subelement 721, is used to described key word of the inquiry that weight is set; Query vector generates subelement 722, is used for the weight generated query vector by described key word of the inquiry; Vector generation unit 723 in territory is used for weight by each speech in described search-type territory and generates territory vector that should the search-type territory; First computation subunit 724 is used for by calculating the similarity that described query vector and territory vector obtain described searching request and search-type territory.
In this embodiment, described device also can further comprise: unit (not shown) or unit 705 are set.Wherein, the described unit that is provided with is used for determining by manual type the descriptor and the related term in described search-type territory and the weight of each speech; Described unit 705 is used for determining by automatic mode of learning the descriptor and the related term in described search-type territory and the weight of each speech.
Described unit 705 comprises: the language material sample obtains subelement 751, dictionary generates subelement 752, weight calculation subelement 753 and descriptor and determines subelement 754.Wherein: the language material sample obtains subelement 751, is used for for each search-type territory, obtains training text language material sample that should the search-type territory; Dictionary generates subelement 752, is used for described language material sample is cut speech, generates the dictionary in this search-type territory;
In embodiments of the present invention, described unit 705 also can further comprise: class is divided subelement 755 and score value is provided with subelement 756.Wherein, class is divided subelement 755, is used for all speech of described dictionary are divided into according to weight the set of different class; Score value is provided with subelement 756, is used to the set of each class that final score value is set, and with the final score value of each class weight as each speech in this class.
Embodiment of the invention mobile search device, searching request at the user, by calculating the similarity in searching request and each search-type territory, select one or several high search-type territory of similarity to search for, thereby can determine the personalized enquire classification for the user, for the user provides personalized accurate search results.Description during detailed process is can be with reference to the front embodiment illustrated in fig. 2 does not repeat them here.
As shown in Figure 8, be the another kind of concrete structure schematic diagram of embodiment of the invention mobile search device.
In this embodiment, described device comprises: receiving element 801, popular searching rate computing unit 802, selected cell 803 and search unit 804.Wherein, described receiving element 801, selected cell 803 and search unit 804 with embodiment illustrated in fig. 6 in each corresponding unit consistent, be not described in detail at this.
Described popular searching rate computing unit 802 comprises second computation subunit 821 and addition subelement 822, and wherein, second computation subunit 821 is used for calculating the popular searching rate in each search-type territory of each key word of the inquiry correspondence of described searching request; Addition subelement 822, be used for the popular searching rate in the same search-type territory of all key word of the inquiry correspondences of described searching request and as described searching request to popular searching rate that should the search-type territory.
In embodiments of the present invention, described popular searching rate specifically can be popular searching times.When the masses that described second computation subunit 821 is calculated certain search-type territory of each keyword correspondence in the described searching request search for total degree, can be according to historical record, collect all users select the number of times searched for certain search-type territory about the searching request that comprises certain keyword in the described searching request summation, the total degree of this search-type territory being searched for as the masses of this keyword correspondence is promptly searched for total degree to masses that should the search-type territory; Described then addition subelement 822 with the masses in this search-type territory of all keyword correspondences in the described searching request search for total degree and, search for total degree as the masses in this search-type territory of described searching request correspondence.
In embodiments of the present invention, described popular searching rate specifically can also be popular Search Results number of clicks.When the popular Search Results that described second computation subunit 821 is calculated certain search-type territory of each keyword correspondence in the described searching request is clicked total degree, can be according to historical record, collect all users select the Search Results number of clicks of searching for certain search-type territory about the searching request that comprises certain keyword in the described searching request summation, the total degree of the Search Results in this search-type territory being clicked as the masses of this keyword correspondence is promptly clicked total degree to popular Search Results that should the search-type territory; Described then addition subelement 822 with the popular Search Results in this search-type territory of all keyword correspondences in the described searching request click total degree and, click total degree as the popular Search Results in this search-type territory of described searching request correspondence.
Embodiment of the invention mobile search device, searching request at the user, by calculating the popular searching rate in corresponding each the search-type territory of described searching request, select one or several high search-type territory of popular searching rate to search for, thereby can determine the personalized enquire classification for the user, for the user provides personalized accurate search results.Description during detailed process is can be with reference to the front embodiment illustrated in fig. 3 does not repeat them here.
As shown in Figure 9, be the another kind of concrete structure schematic diagram of embodiment of the invention mobile search device.
In this embodiment, described device comprises: receiving element 901, user interest scores value computing unit 902, selected cell 903 and search unit 904.Wherein, described receiving element 901, selected cell 903 and search unit 904 with embodiment illustrated in fig. 6 in each corresponding unit consistent, be not described in detail at this.
Described user interest scores value computing unit 902 comprises that interest model extracts subelement 921 and the 3rd computation subunit 922, wherein, interest model extracts subelement 921, be used for extracting from user data user's interest model, described user's interest model is the vector that described user data is formed at the score value of a plurality of interest dimensions; The 3rd computation subunit 922 is used for the score value sum of one or more interest dimensions of the corresponding described user interest model in the described search-type territory personalized user interest scores value as described search-type territory.
In this embodiment, described user's interest model is: static interest model or dynamic interest model can also be comprehensive described static interest model or dynamically interest model and the interest model that generates.For this reason, described interest model extracts subelement 921 can multiple frame mode.
Described interest model extracts subelement 921 can include only the first extraction subelement (not shown), the static archives that are used for calculating the user belong to the word frequency sum of all speech of each interest dimension, and, generate described user interest model as vector by the score value of corresponding each interest dimension with its score value as corresponding each interest dimension;
Described interest model extracts subelement 921 can also include only the second extraction subelement (not shown), be used for calculating the word frequency sum that belongs to all speech of each interest dimension in the clicked document of the historical record historical record of user search, and, generate described user's dynamic interest model as vector by the score value of corresponding each interest dimension with its score value as corresponding each interest dimension.
As shown in figure 10, described interest model extracts subelement 921 can also comprise that described first extracts subelement 1001 and the described second extraction subelement 1002, and first handles the subelement 1003 and the first weighting subelement 1004.Wherein, first handles subelement 1003, is used for respectively described static interest model and described dynamic interest model being carried out normalized; The first weighting subelement 1004, be used to calculate after the normalized static interest model and dynamically interest model and, and will be somebody's turn to do and as described user's interest model.
As shown in figure 11, described interest model extracts subelement 921 can also comprise that described first extracts subelement 1101 and the described second extraction subelement 1102, and the second weighting subelement 1103 and second is handled subelement 1104.Wherein, the second weighting subelement 1103 is used for described static interest model and described dynamic interest model are weighted addition; Second handles subelement 1104, be used for the result of described second weighting subelement output is carried out normalized, and with the interest model of the result after the normalized as described user.
Embodiment of the invention mobile search device, searching request at the user, by calculating the personalized user interest scores value in each search-type territory, select one or several high search-type territory of score value to search for, thereby can determine the personalized enquire classification for the user, for the user provides personalized accurate search results.Detailed process can be with reference to the description in the embodiment of the invention mobile search method of front.
In the above in the mobile search device of each embodiment, when carrying out the selection of search-type territory, the foundation of selecting as the search-type territory with the personalized user interest scores value in the popular searching rate in described searching request and the similarity in described search-type territory, the corresponding described search-type of described searching request territory and search-type territory respectively, determine user's personalized enquire classification, for the user provides personalized accurate search results.
In embodiments of the present invention, can also take all factors into consideration above-mentioned any two or multinomial, calculate the comprehensive grading value in each search-type territory, select one or several high search-type territory of comprehensive grading value to search for.Be example to take all factors into consideration above-mentioned three foundations of selecting as the search-type territory below, the embodiment of the invention is described in detail.
With reference to Figure 12, be the another kind of structure chart of embodiment of the invention mobile search device.
In this embodiment, described device comprises: receiving element 1201, computing unit 1202, selected cell 1203 and search unit 1204.Wherein, receiving element 1201 is used to receive searching request, comprises one or more keys word of the inquiry in the described searching request; Computing unit 1202, be used to calculate the score value in each search-type territory, described score value is following any one score value or multinomial comprehensive grading value: described searching request and the popular searching rate in the similarity in described search-type territory, the corresponding described search-type of described searching request territory, the personalized user interest scores value in search-type territory; Selected cell 1203 is selected one of them or several search-type territory according to the score value in each search-type territory; Search unit 1204 is used to the described key word of the inquiry of search-type domain search that utilizes described selected cell to select.
In this embodiment, described computing unit 1202 comprises: similarity calculated 1221, popular searching rate computing unit 1222, user interest scores value computing unit 1223, normalized unit 1224 and integrated treatment unit 1225.Wherein, similarity calculated 1221 is used to calculate the similarity in described searching request and each search-type territory; Popular searching rate computing unit 1222 is used to calculate the popular searching rate in corresponding each the search-type territory of described searching request; User interest scores value computing unit 1223 is used to calculate the personalized user interest scores value in each search-type territory; Normalized unit 1224, the value that is used for respectively described similarity calculated, described popular searching rate computing unit and described user interest scores value computing unit being calculated is carried out normalized; Integrated treatment unit 1225, the value after any two or more normalization that are used for normalized unit 1224 is obtained is carried out COMPREHENSIVE CALCULATING, and for example: product, average or weighting summation etc. obtain the score value in each search-type territory.
As seen, the mobile search device of the embodiment of the invention, take all factors into consideration multinomial factor and determined user's personalized enquire classification, calculate the comprehensive grading value in each search-type territory, select one or several high search-type territory of comprehensive grading value to search for, thereby can provide personalized accurate search results for the user.
More than the embodiment of the invention is described in detail, used embodiment herein the present invention set forth, the explanation of above embodiment just is used for help understanding method and apparatus of the present invention; Simultaneously, for one of ordinary skill in the art, according to thought of the present invention, the part that all can change in specific embodiments and applications, in sum, this description should not be construed as limitation of the present invention.
Claims (27)
1. a mobile search method is characterized in that, comprising:
Receive searching request, comprise one or more keys word of the inquiry in the described searching request;
Calculate the score value in each search-type territory, described score value is following any one score value or multinomial comprehensive grading value: described searching request and the popular searching rate in the similarity in described search-type territory, the corresponding described search-type of described searching request territory, the personalized user interest scores value in search-type territory;
Select one of them or the described key word of the inquiry of several search-type domain search according to the score value in each search-type territory.
2. method according to claim 1, it is characterized in that the comprehensive grading value in each search-type territory of described calculating is according to multinomial calculating product score value, average score value or weighted scoring value in the personalized user interest scores value in the popular searching rate in described searching request and the similarity in described search-type territory, the corresponding described search-type of described searching request territory and search-type territory.
3. method according to claim 1 is characterized in that, the similarity in described searching request of described calculating and described search-type territory comprises:
For described key word of the inquiry is provided with weight;
Weight generated query vector by described key word of the inquiry;
Generate territory vector that should the search-type territory by the weight of each speech in described search-type territory;
By calculating the similarity that described query vector and territory vector obtain described searching request and search-type territory.
4. method according to claim 3 is characterized in that, described method also comprises:
Determine descriptor and the related term in the described search-type territory and the weight of each speech by manual type; Perhaps
Determine descriptor and the related term in the described search-type territory and the weight of each speech by automatic mode of learning.
5. method according to claim 4 is characterized in that, describedly determines descriptor and related term in the described search-type territory by automatic mode of learning, and the weight of each speech comprises:
For each search-type territory, obtain training text language material sample that should the search-type territory;
Described language material sample is cut speech, generate the dictionary in this search-type territory;
Calculate the weight of each speech in the described dictionary;
Determine descriptor and related term in the described search-type territory according to the weight of each speech.
6. method according to claim 5 is characterized in that, describedly determines descriptor and related term in the described search-type territory by automatic mode of learning, and the weight of each speech also comprises:
All speech in the described dictionary are divided into the set of different class according to weight;
For the set of each class is provided with final score value, and with the final score value of each class weight as each speech in this class.
7. method according to claim 3 is characterized in that, describedly weight is set comprises for described key word of the inquiry:
For whole keys word of the inquiry are provided with identical weight; Perhaps
For coming the most preceding keyword weight limit is set,, minimal weight is set for coming last keyword for the keyword in the middle of coming is provided with middle big or small weight.
8. method according to claim 1 is characterized in that, the popular searching rate in the corresponding described search-type of the described searching request of described calculating territory comprises:
Calculate the popular searching rate in each search-type territory of each key word of the inquiry correspondence in the described searching request;
With the popular searching rate in the same search-type territory of all key word of the inquiry correspondences in the described searching request and as described searching request to popular searching rate that should the search-type territory.
9. method according to claim 8 is characterized in that, described popular searching rate is: popular searching times, perhaps popular Search Results number of clicks.
10. method according to claim 1 is characterized in that, the personalized user interest scores value in the described search-type of described calculating territory comprises:
Extract user's interest model from user data, described user's interest model is the vector that described user data is formed at the score value of a plurality of interest dimensions;
With the score value sum of one or more interest dimensions of the corresponding described user interest model in described search-type territory personalized user interest scores value as described search-type territory.
11. method according to claim 10 is characterized in that, described user's interest model is: static interest model or dynamic interest model;
The static interest model that extracts the user from user data comprises:
The word frequency sum that belongs to all speech of each interest dimension in calculating user's the static archives, and with its score value as corresponding each interest dimension; Perhaps, calculate user's the static archives and the similarity score value of each interest dimension, and with its score value as corresponding each interest dimension;
Score value by corresponding each interest dimension generates described user interest model as vector;
The dynamic interest model that extracts the user from user data comprises:
Calculate user's search and click the word frequency sum that belongs to all speech of each interest dimension in the historical record, and with its score value as corresponding each interest dimension; Perhaps, calculate the similarity score value that historical record and each interest dimension are clicked in search, and with its score value as corresponding each interest dimension;
Generate described user's dynamic interest model as vector by the score value of corresponding each interest dimension.
12. method according to claim 11 is characterized in that, the described interest model that extracts the user from user data also comprises:
Respectively described static interest model and described dynamic interest model are carried out normalized;
Calculate one or more static interest model after the normalized and one or more dynamic interest models and, and will be somebody's turn to do and as described user's interest model.
13. method according to claim 11 is characterized in that, the described interest model that extracts the user from user data also comprises:
One or more described static interest models and one or more described dynamic interest model are weighted addition;
With weighting summation and carry out normalized, and with the interest model of the result after the normalized as described user.
14. method according to claim 1 is characterized in that, the weighted scoring value in each search-type territory of described calculating comprises:
Calculate the similarity in described searching request and described search-type territory, and with its normalized;
Calculate the popular searching rate in the corresponding described search-type of described searching request territory, and with its normalized;
Calculate the personalized user interest scores value in described search-type territory, and with its normalized;
Value after above-mentioned any two or more normalizeds is weighted addition, obtains the weighted scoring value in described search-type territory.
15. a mobile search device is characterized in that, comprising:
Receiving element is used to receive searching request, comprises one or more keys word of the inquiry in the described searching request;
Computing unit, be used to calculate the score value in each search-type territory, described score value is following any one score value or multinomial comprehensive grading value: described searching request and the popular searching rate in the similarity in described search-type territory, the corresponding described search-type of described searching request territory, the personalized user interest scores value in search-type territory;
Selected cell is selected one of them or several search-type territory according to the score value in each search-type territory;
Search unit is used to the described key word of the inquiry of search-type domain search that utilizes described selected cell to select.
16. device according to claim 15, it is characterized in that the comprehensive grading value that described computing unit calculates each search-type territory is for according to multinomial calculating product score value, average score value or weighted scoring value in the personalized user interest scores value in the popular searching rate in described searching request and the similarity in described search-type territory, the corresponding described search-type of described searching request territory and search-type territory.
17. device according to claim 15 is characterized in that, described computing unit comprises following any one or a plurality of unit:
Similarity calculated is used to calculate the similarity in described searching request and each search-type territory;
Popular searching rate computing unit is used to calculate the popular searching rate in corresponding each the search-type territory of described searching request;
User interest scores value computing unit is used to calculate the personalized user interest scores value in each search-type territory.
18. device according to claim 17 is characterized in that, described similarity calculated comprises:
Weight is provided with subelement, is used to described key word of the inquiry that weight is set;
Query vector generates subelement, is used for the weight generated query vector by described key word of the inquiry;
Vector generation unit in territory is used for weight by each speech in described search-type territory and generates territory vector that should the search-type territory;
First computation subunit is used for by calculating the similarity that described query vector and territory vector obtain described searching request and search-type territory.
19. device according to claim 18 is characterized in that, described device also comprises:
The unit is set, is used for determining the descriptor and the related term in described search-type territory by manual type, and the weight of each speech; Perhaps
Unit is used for determining by automatic mode of learning the descriptor and the related term in described search-type territory and the weight of each speech.
20. device according to claim 19 is characterized in that, described unit comprises:
The language material sample obtains subelement, is used for for each search-type territory, obtains training text language material sample that should the search-type territory;
Dictionary generates subelement, is used for described language material sample is cut speech, generates the dictionary in this search-type territory;
The weight calculation subelement is used for calculating the weight of described each speech of dictionary;
Descriptor is determined subelement, is used for determining according to the weight of each speech the descriptor and the related term in described search-type territory.
21. device according to claim 20 is characterized in that, described unit also comprises:
Class is divided subelement, is used for all speech of described dictionary are divided into according to weight the set of different class;
Score value is provided with subelement, is used to the set of each class that final score value is set, and with the final score value of each class weight as each speech in this class.
22. device according to claim 17 is characterized in that, described popular searching rate computing unit comprises:
Second computation subunit is used for calculating the popular searching rate in each search-type territory of each key word of the inquiry correspondence of described searching request;
The addition subelement, be used for the popular searching rate in the same search-type territory of all key word of the inquiry correspondences of described searching request and as described searching request to popular searching rate that should the search-type territory.
23. device according to claim 17 is characterized in that, described user interest scores value computing unit comprises:
Interest model extracts subelement, is used for extracting from user data user's interest model, and described user's interest model is the vector that described user data is formed at the score value of a plurality of interest dimensions;
The 3rd computation subunit is used for the score value sum of one or more interest dimensions of the corresponding described user interest model in the described search-type territory personalized user interest scores value as described search-type territory.
24. device according to claim 23 is characterized in that, described user's interest model is: static interest model or dynamic interest model;
Described interest model extracts subelement and comprises:
First extracts subelement, the static archives that are used for calculating the user belong to the word frequency sum of all speech of each interest dimension, and with its score value as corresponding each interest dimension, perhaps calculate user's the static archives and the similarity score value of each interest dimension, and, generate described user interest model as vector by the score value of corresponding each interest dimension with its score value as corresponding each interest dimension; Perhaps
Second extracts subelement, the word frequency sum that historical record belongs to all speech of each interest dimension is clicked in the search that is used for calculating the user, and with its score value as corresponding each interest dimension, perhaps calculate the similarity score value that historical record and each interest dimension are clicked in search, and, generate described user's dynamic interest model as vector by the score value of corresponding each interest dimension with its score value as corresponding each interest dimension.
25. device according to claim 23 is characterized in that, described interest model extracts subelement and also comprises:
First handles subelement, is used for respectively described static interest model and described dynamic interest model being carried out normalized;
The first weighting subelement, be used to calculate one or more static interest model after the normalized and one or more dynamic interest models and, and will be somebody's turn to do and as described user's interest model.
26. device according to claim 23 is characterized in that, described interest model extracts subelement and also comprises:
The second weighting subelement is used for one or more described static interest models and one or more described dynamic interest model are weighted addition;
Second handles subelement, be used for the result of described second weighting subelement output is carried out normalized, and with the interest model of the result after the normalized as described user.
27. device according to claim 23 is characterized in that, described computing unit also comprises:
The normalized unit, the value that is used for respectively described similarity calculated, described popular searching rate computing unit and described user interest scores value computing unit being calculated is carried out normalized;
Weighting processing unit, the value after any two or more normalization that are used for described normalized unit is obtained is weighted addition, obtains the score value in each search-type territory.
Priority Applications (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200910140119A CN101820592A (en) | 2009-02-27 | 2009-07-01 | Method and device for mobile search |
| PCT/CN2009/074758 WO2010096986A1 (en) | 2009-02-27 | 2009-11-05 | Mobile search method and device |
| US13/219,058 US20110314059A1 (en) | 2009-02-27 | 2011-08-26 | Mobile search method and apparatus |
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200910118632.7 | 2009-02-27 | ||
| CN200910140119A CN101820592A (en) | 2009-02-27 | 2009-07-01 | Method and device for mobile search |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101820592A true CN101820592A (en) | 2010-09-01 |
Family
ID=42655489
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200910140119A Pending CN101820592A (en) | 2009-02-27 | 2009-07-01 | Method and device for mobile search |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101820592A (en) |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102364467A (en) * | 2011-09-29 | 2012-02-29 | 北京亿赞普网络技术有限公司 | Network search method and system |
| CN102436495A (en) * | 2011-11-14 | 2012-05-02 | 百度在线网络技术（北京）有限公司 | Method and device for providing dynamic searching page |
| CN102436496A (en) * | 2011-11-14 | 2012-05-02 | 百度在线网络技术（北京）有限公司 | Method for providing personated searching labels and device thereof |
| CN102521350A (en) * | 2011-12-12 | 2012-06-27 | 浙江大学 | Selection method of distributed information retrieval sets based on historical click data |
| CN102955813A (en) * | 2011-08-29 | 2013-03-06 | 中国移动通信集团四川有限公司 | Information searching method and information searching system |
| CN102999521A (en) * | 2011-09-15 | 2013-03-27 | 北京百度网讯科技有限公司 | Method and device for identifying search requirement |
| CN102999520A (en) * | 2011-09-15 | 2013-03-27 | 北京百度网讯科技有限公司 | Method and device for identifying search request |
| CN103339623A (en) * | 2010-09-08 | 2013-10-02 | 纽昂斯通讯公司 | Internet search related methods and apparatus |
| CN103455499A (en) * | 2012-05-29 | 2013-12-18 | 北京百度网讯科技有限公司 | Method and system for automatically matching search types according to search terms in mobile terminal |
| CN103530385A (en) * | 2013-10-18 | 2014-01-22 | 北京奇虎科技有限公司 | Method and device for searching for information based on vertical searching channels |
| CN103729359A (en) * | 2012-10-12 | 2014-04-16 | 阿里巴巴集团控股有限公司 | Method and system for recommending search terms |
| CN104699737A (en) * | 2013-12-09 | 2015-06-10 | 国际商业机器公司 | Method and system for managing a search |
| CN104915429A (en) * | 2015-06-15 | 2015-09-16 | 小米科技有限责任公司 | Keyword searching method and device |
| CN104933090A (en) * | 2015-05-18 | 2015-09-23 | 深圳市金立通信设备有限公司 | Information searching method and terminal |
| CN105245589A (en) * | 2015-09-28 | 2016-01-13 | 小米科技有限责任公司 | Information display method and device |
| CN105512298A (en) * | 2015-12-10 | 2016-04-20 | 成都陌云科技有限公司 | Interested content prediction method based on machine learning |
| CN105550282A (en) * | 2015-12-10 | 2016-05-04 | 成都陌云科技有限公司 | User interest forecasting method by utilizing multidimensional data |
| CN108415903A (en) * | 2018-03-12 | 2018-08-17 | 武汉斗鱼网络科技有限公司 | Judge evaluation method, storage medium and the equipment of search intention identification validity |
-
2009
- 2009-07-01 CN CN200910140119A patent/CN101820592A/en active Pending
Cited By (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103339623A (en) * | 2010-09-08 | 2013-10-02 | 纽昂斯通讯公司 | Internet search related methods and apparatus |
| CN102955813B (en) * | 2011-08-29 | 2015-11-25 | 中国移动通信集团四川有限公司 | A kind of information search method and system |
| CN102955813A (en) * | 2011-08-29 | 2013-03-06 | 中国移动通信集团四川有限公司 | Information searching method and information searching system |
| CN102999520B (en) * | 2011-09-15 | 2016-04-27 | 北京百度网讯科技有限公司 | A kind of method and apparatus of search need identification |
| CN102999521B (en) * | 2011-09-15 | 2016-06-15 | 北京百度网讯科技有限公司 | A kind of method and device identifying search need |
| CN102999521A (en) * | 2011-09-15 | 2013-03-27 | 北京百度网讯科技有限公司 | Method and device for identifying search requirement |
| CN102999520A (en) * | 2011-09-15 | 2013-03-27 | 北京百度网讯科技有限公司 | Method and device for identifying search request |
| CN102364467A (en) * | 2011-09-29 | 2012-02-29 | 北京亿赞普网络技术有限公司 | Network search method and system |
| CN102436496A (en) * | 2011-11-14 | 2012-05-02 | 百度在线网络技术（北京）有限公司 | Method for providing personated searching labels and device thereof |
| CN102436495A (en) * | 2011-11-14 | 2012-05-02 | 百度在线网络技术（北京）有限公司 | Method and device for providing dynamic searching page |
| CN102521350A (en) * | 2011-12-12 | 2012-06-27 | 浙江大学 | Selection method of distributed information retrieval sets based on historical click data |
| CN102521350B (en) * | 2011-12-12 | 2014-07-16 | 浙江大学 | Selection method of distributed information retrieval sets based on historical click data |
| CN103455499A (en) * | 2012-05-29 | 2013-12-18 | 北京百度网讯科技有限公司 | Method and system for automatically matching search types according to search terms in mobile terminal |
| US9489688B2 (en) | 2012-10-12 | 2016-11-08 | Alibaba Group Holding Limited | Method and system for recommending search phrases |
| CN103729359A (en) * | 2012-10-12 | 2014-04-16 | 阿里巴巴集团控股有限公司 | Method and system for recommending search terms |
| CN103729359B (en) * | 2012-10-12 | 2017-03-01 | 阿里巴巴集团控股有限公司 | A kind of method and system recommending search word |
| CN103530385A (en) * | 2013-10-18 | 2014-01-22 | 北京奇虎科技有限公司 | Method and device for searching for information based on vertical searching channels |
| US9996588B2 (en) | 2013-12-09 | 2018-06-12 | International Business Machines Corporation | Managing a search |
| CN104699737A (en) * | 2013-12-09 | 2015-06-10 | 国际商业机器公司 | Method and system for managing a search |
| US10176227B2 (en) | 2013-12-09 | 2019-01-08 | International Business Machines Corporation | Managing a search |
| US11176124B2 (en) | 2013-12-09 | 2021-11-16 | International Business Machines Corporation | Managing a search |
| CN104933090A (en) * | 2015-05-18 | 2015-09-23 | 深圳市金立通信设备有限公司 | Information searching method and terminal |
| CN104915429A (en) * | 2015-06-15 | 2015-09-16 | 小米科技有限责任公司 | Keyword searching method and device |
| CN104915429B (en) * | 2015-06-15 | 2018-09-04 | 小米科技有限责任公司 | Keyword search methodology and device |
| CN105245589A (en) * | 2015-09-28 | 2016-01-13 | 小米科技有限责任公司 | Information display method and device |
| CN105245589B (en) * | 2015-09-28 | 2019-06-14 | 小米科技有限责任公司 | Information displaying method and device |
| CN105550282A (en) * | 2015-12-10 | 2016-05-04 | 成都陌云科技有限公司 | User interest forecasting method by utilizing multidimensional data |
| CN105512298A (en) * | 2015-12-10 | 2016-04-20 | 成都陌云科技有限公司 | Interested content prediction method based on machine learning |
| CN108415903A (en) * | 2018-03-12 | 2018-08-17 | 武汉斗鱼网络科技有限公司 | Judge evaluation method, storage medium and the equipment of search intention identification validity |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101820592A (en) | Method and device for mobile search | |
| CN109815308B (en) | Method and device for determining intention recognition model and method and device for searching intention recognition | |
| CN104376406B (en) | A kind of enterprise innovation resource management and analysis method based on big data | |
| CN106547871B (en) | Neural network-based search result recall method and device | |
| US8380697B2 (en) | Search and retrieval methods and systems of short messages utilizing messaging context and keyword frequency | |
| CN103593425B (en) | Preference-based intelligent retrieval method and system | |
| CN105045875B (en) | Personalized search and device | |
| CN103064945B (en) | Based on the Situational searching method of body | |
| US8554854B2 (en) | Systems and methods for identifying terms relevant to web pages using social network messages | |
| CN102663139B (en) | Method and system for constructing emotional dictionary | |
| US9928296B2 (en) | Search lexicon expansion | |
| CN102033955A (en) | Method for expanding user search results and server | |
| US20120317104A1 (en) | Using Aggregate Location Metadata to Provide a Personalized Service | |
| CN106339502A (en) | Modeling recommendation method based on user behavior data fragmentation cluster | |
| CN102999625A (en) | Method for realizing semantic extension on retrieval request | |
| CN109388743B (en) | Language model determining method and device | |
| US20110314059A1 (en) | Mobile search method and apparatus | |
| CN104731954A (en) | Music recommendation method and system based on group perspective | |
| CN105095433A (en) | Recommendation method and device for entities | |
| CN103870505A (en) | Query term recommending method and query term recommending system | |
| CN101532848A (en) | Information retrieval apparatus, informatin retrieval system, and information retrieval method | |
| CN105975596A (en) | Query expansion method and system of search engine | |
| CN106227714A (en) | A kind of method and apparatus obtaining the key word generating poem based on artificial intelligence | |
| CN103886020A (en) | Quick search method of real estate information | |
| Babu et al. | Improving Quality of Content Based Image Retrieval with Graph Based Ranking |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20100901 |