CN101984437B - Music resource individual recommendation method and system thereof - Google Patents
Music resource individual recommendation method and system thereof Download PDFInfo
- Publication number
- CN101984437B CN101984437B CN2010105556951A CN201010555695A CN101984437B CN 101984437 B CN101984437 B CN 101984437B CN 2010105556951 A CN2010105556951 A CN 2010105556951A CN 201010555695 A CN201010555695 A CN 201010555695A CN 101984437 B CN101984437 B CN 101984437B
- Authority
- CN
- China
- Prior art keywords
- label
- music sources
- user
- weight
- seed
- 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
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a music resource individual recommendation method and a system thereof. The method includes that: text information related to music resource is captured from network; statistics on label seed in related text information of each music resource and frequency of each label seed and statistics on label seed in related text information of all the music resource and frequency of each label seed are carried out, thus obtaining global weight of each label seed; for each single music resource, monomer weight of each label seed corresponding to single music resource is obtained according to the frequency of each label seed and global weight thereof; label seeds monomer weight of which are consistent with preset conditions are determined to be a label of music resource; and individual recommendation is carried out according to the label of music resource. By adopting the invention, automatic and relatively accurate individual recommendation can be realized.
Description
Technical field
The present invention relates to the recommendation of personalized information technical field, particularly relate to music sources personalized recommendation method and system.
Background technology
Along with Internet development, music sources has been called very large ingredient in the internet, applications.Each big music site, music software etc. provide millions of music sources for the user.But the user but can't understand all at short notice, therefore also is difficult to find easily the own music of being liked.This just need be through coming for the user recommends the interested music of its possibility intelligently also promptly personalized music recommend someway.
The method of current personalized recommendation mainly is collaborative filtering recommending algorithm (CollaborativeFiltering Recommendation).It is through collection user's music behavioural habits, and the user calculates the music list of each user's preferences to the evaluation of music; And then find out the similar customer group of interest, and the music that this customer group is paid close attention to is carried out analysis-by-synthesis, the unique user in this group, the music of other user's preferences in this group is exactly the music that this user possibly like.As shown in Figure 1; The vertical rectangle song of listening of representative of consumer A, B, C, D is respectively tabulated; Collaborative filtering is at first found out user A, B, C, the D of part same music hobby, and with user B, C, D total and music recommend that user A does not have is given user A.
But this collaborative filtering recommending algorithm has following shortcoming at least, and the eternal recommendation of never being paid close attention to by Any user of concert is not come out.For this reason; Prior art also provides the information filtering proposed algorithm; This method need add label (tag) for each music sources, because the label of music sources has been represented information such as the described classification of music sources, for example jazz, rock and roll or the like; Therefore, can carry out personalized recommendation according to the label of music sources and user's hobby.
Yet prior art needs a large amount of professionals to come the label of music is extracted, marks, and workload is very big, and efficient is low; And annotation results receives the influence of factor and individual subjective factor to a certain extent, maybe be inaccurate.
Summary of the invention
The present invention provides music sources personalized recommendation method and system, can realize robotization and personalized recommendation relatively accurately.
The invention provides following scheme:
A kind of music sources personalized recommendation method comprises:
Grasp the text message relevant from network with music sources;
Said text message is cut speech; Add up the number of times that the label seed that occurs in the relevant textual information of each music sources and each label seed occur, and the number of times that occurs of the label seed that occurs in the relevant textual information of all music sources and each label seed;
The number of times that the label seed that occurs in the relevant textual information according to said all music sources and each label seed occur obtains the overall weight of each label seed;
For single music sources, number of times and overall weight thereof according to each label seed in its relevant textual information occurs obtain the monomer weight of each label seed with respect to single music sources;
The label seed that the monomer weight is met prerequisite is confirmed as the label of music sources;
Label according to said music sources carries out personalized recommendation.
Preferably, said label according to said music sources carries out personalized recommendation and comprises:
For music sources set, merge the label of each music sources in this set;
According to each label with respect to the monomer weight in each music sources, the weight of each label after obtaining merging;
The label that weight is met prerequisite is confirmed as the label that this music sources is gathered;
Label according to said music sources set carries out personalized recommendation.
Preferably, said music sources set comprises each singer's music sources, and the label of said music sources set comprises each singer's label, and said label according to said music sources carries out personalized recommendation and comprises:
The weight of each singer's label is mapped to vector space, obtains each singer's label vector;
Cosine angle between the computation tag vector obtains each singer similarity between any two in twos;
Carry out personalized recommendation according to the similarity between the singer.
Preferably, said music sources set comprises all music sources that each user listened, and the label of said music sources set comprises each user's label, and said label according to said music sources carries out personalized recommendation and comprises:
Label based on each music sources is set up inverted list, obtains the corresponding music sources tabulation of each label;
Personalized recommendation is carried out in label and the corresponding music sources tabulation of label according to the user.
Preferably, also comprise:
According to each user's the behavior of listening to historical record, adjust each label weight of each user relatively.
Preferably, also comprise:
According to weight rank order from big to small, and will come the label of the preset number of front to user's label, confirm as this user's popular label;
Said label and the corresponding music sources tabulation of label according to the user carried out personalized recommendation and comprised: personalized recommendation is carried out in popular label and the corresponding music sources tabulation of label according to said user.
Preferably, also comprise:
According to the weight of each user's label, calculate the average weight of each label in all users;
This user's the label size according to the merchant of this user's relatively weight and said average weight is sorted, and will come the label of the preset number of front, confirm as this user's long-tail label;
Said label and the corresponding music sources tabulation of label according to the user carried out personalized recommendation and comprised: personalized recommendation is carried out in long-tail label and the corresponding music sources tabulation of label according to said user.
A kind of music sources personalized recommendation system comprises:
The information placement unit is used for grasping the text message relevant with music sources from network;
Statistic unit; Be used for said text message is cut speech; Add up the number of times that the label seed that occurs in the relevant textual information of each music sources and each label seed occur, and the number of times that occurs of the label seed that occurs in the relevant textual information of all music sources and each label seed;
Overall situation weight acquiring unit is used for the number of times that label seed that the relevant textual information according to said all music sources occurs and each label seed occur, and obtains the overall weight of each label seed;
Monomer weight acquiring unit is used for for single music sources, and number of times and overall weight thereof according to each label seed in its relevant textual information occurs obtain the monomer weight of each label seed with respect to single music sources;
Music label is confirmed the unit, is used for the label seed that the monomer weight meets prerequisite is confirmed as the label of music sources;
Recommendation unit is used for carrying out personalized recommendation according to the label of said music sources.
Preferably, said recommendation unit comprises:
Label merges subelement, is used for merging the label of each music sources in this set for music sources set;
Weight merges subelement, is used for according to the monomer weight of each label with respect to each music sources, the weight of each label after obtaining merging;
The set label is confirmed subelement, is used for the label that weight meets prerequisite is confirmed as the label of this music sources set;
First recommends subelement, is used for carrying out personalized recommendation according to the label of said music sources set.
Preferably, said music sources set comprises each singer's music sources, and the label of said music sources set comprises each singer's label, and said first recommends subelement to comprise:
The label vector obtains subelement, is used for the weight of each singer's label is mapped to vector space, obtains each singer's label vector;
The similarity computation subunit is used for the cosine angle between the computation tag vector in twos, obtains each singer similarity between any two;
The singer recommends subelement, is used for carrying out personalized recommendation according to the similarity between the singer.
Preferably, said music sources set comprises all music sources that each user listened, and the label of said music sources set comprises each user's label, and said first recommends subelement to comprise:
Arrange subelement, be used for setting up inverted list, obtain the corresponding music sources tabulation of each label based on the label of each music sources;
Label is recommended subelement, is used for carrying out personalized recommendation according to user's label and the corresponding music sources tabulation of label.
Preferably, also comprise:
Weight adjustment unit is used for the behavior of the listening to historical record according to each user, adjusts each label weight of each user relatively.
Preferably, also comprise:
Popular label acquiring unit is used for label to the user according to weight rank order from big to small, and will come the label of the preset number of front, confirms as this user's popular label;
Said label recommends subelement specifically to be used for: personalized recommendation is carried out in popular label and the corresponding music sources tabulation of label according to said user.
Preferably, also comprise:
The average weight computing unit is used for the weight according to each user's label, calculates the average weight of each label in all users;
Long-tail label acquiring unit is used for this user's the label size according to the merchant of this user's relatively weight and said average weight is sorted, and will comes the label of the preset number of front, confirms as this user's long-tail label;
Said label recommends subelement specifically to be used for: personalized recommendation is carried out in long-tail label and the corresponding music sources tabulation of label according to said user.
According to specific embodiment provided by the invention, the invention discloses following technique effect:
The present invention is through grasp the relevant textual information of music sources automatically from network; Automatically from relevant textual information, extract the label of music sources then; So that the label according to music sources carries out personalized music sources recommendation to the user; Therefore, can realize robotization and personalized recommendation relatively accurately.
Description of drawings
In order to be illustrated more clearly in the embodiment of the invention or technical scheme of the prior art; To do to introduce simply to the accompanying drawing of required use among the embodiment below; Obviously, the accompanying drawing in describing below only is some embodiments of the present invention, for those of ordinary skills; Under the prerequisite of not paying creative work, can also obtain other accompanying drawing according to these accompanying drawings.
Fig. 1 is a recommend method synoptic diagram of the prior art;
Fig. 2 is the process flow diagram of the method that provides of the embodiment of the invention;
Fig. 3 is the synoptic diagram of the system that provides of the embodiment of the invention.
Embodiment
To combine the accompanying drawing in the embodiment of the invention below, the technical scheme in the embodiment of the invention is carried out clear, intactly description, obviously, described embodiment only is the present invention's part embodiment, rather than whole embodiment.Based on the embodiment among the present invention, the every other embodiment that those of ordinary skills obtained belongs to the scope that the present invention protects.
Referring to Fig. 2, the embodiment of the invention provides a kind of music sources personalized recommendation method, and this method may further comprise the steps:
S201: grasp the text message relevant with music sources from network;
Wherein, The operation of grasping can be accomplished through the program of robotization; The text message that the music sources that wherein grasps is relevant can comprise website or the online friend label (tag) to the music mark, the described classified information of music, related album--name, list title or the like.
S202: said text message is cut speech; Add up the number of times that the label seed that occurs in the relevant textual information of each music sources and each label seed occur, and the number of times that occurs of the label seed that occurs in the relevant textual information of all music sources and each label seed;
Because the text message that grasps from network possibly be a sentence, also possibly be article paragraph or the like, therefore, also need cut speech to the text message that grasps.Also promptly, with sentence or article paragraph according in be syncopated as word one by one according to grammatical and semantic etc.Certainly, the process of cutting also is to accomplish automatically, can adopt the method that provides in the prior art when specifically carrying out automatic segmentation, repeats no more here.
Statistics for label; Because the label of music sources is used to identify the characteristic that music sources has; Therefore, when specifically realizing, some label seeds can be set in advance; For example " popular ", " rock and roll ", " popular " or the like are confirmed the label of each music sources through these label seeds.For this reason; Also need be after the relevant textual information of certain music sources be carried out cutting; At first count the number of times that occurs each label seed in the word that these cuttings obtain, for example possibly occur " popular " 5 times in the relevant textual information of certain music sources, " rock and roll " 8 times; " popular " 0 time, or the like.In addition, also need add up the number of times that occurs each label seed in all music sources relevant textual information, for example; Total music sources 100 head in the relevant textual information of these music sources, " popular " possibly occur 500 times; " rock and roll " 200 times, " popular " 300 times, or the like.
S203: the number of times that the label seed that occurs in the relevant textual information according to said all music sources and each label seed occur obtains the overall weight of each label seed;
During concrete the realization, can each music sources as a file, be adopted the overall weight (for example, can be designated as W_tag_global) of calculating each label seed based on the method for TF-IDF with the label seed as term.Wherein, TF-IDF is a kind of statistical method, in order to assess the significance level of a words for a copy of it file in a file set or the corpus.The number of times that the importance of words occurs along with it the hereof increase that is directly proportional, but the decline that can be inversely proportional to along with the frequency that it occurs in file set or corpus simultaneously.
For a label seed, overall weight is high more, representes that then this label seed is representative more, and discrimination is also just high more.By this method, not only the label seed of some entanglements is filtered owing to occurrence number is very few, and some occurrence numbers are a lot of but do not have the label seed of discrimination (for example " Chinese ", " popular " etc.) also can be filtered.
S204: for single music sources, number of times and overall weight thereof according to each label seed in its relevant textual information occurs obtain the monomer weight of each label seed with respect to single music sources;
The monomer weight can be represented by W_tag_single.
S205: the label seed that the monomer weight is met prerequisite is confirmed as the label of music sources;
For example, for a music sources, the label seed that can the monomer weight be ranked in the top is as the label of this music sources.
S206: the label according to said music sources carries out personalized recommendation.
Realized obtaining automatically of music sources label through aforementioned several steps, like this, just can carry out the personalized recommendation of music sources according to the label of each music sources.
Specifically when the label that utilizes music sources carries out personalized recommendation, multiple implementation can be arranged, at length introduce respectively below.
In the practical application, can have some music sources set usually, for example, music list, singer's special edition or the like are for these music sources set.Can calculate the label of music sources set, and carry out the personalized recommendation of music sources according to the label of collection of music.During the label of concrete computational music resource collection, can at first merge the label of each music sources in this set, according to each label with respect to the monomer weight in each music sources and the significance level of each music sources, the weight of each label after obtaining merging; The label that weight is met prerequisite is confirmed as the label that this music sources is gathered.
For example, comprise music sources a, b, c in certain music sources set A, have three labels 1,2,3 respectively, wherein for music sources a, the weight of each label is respectively (t
A1, t
A2, t
A3), for music sources b, the weight of each label is respectively (t
B1, t
B2, t
B3), for music sources c, the weight of each label is respectively (t
C1, t
C2, t
C3), then after the label merging with each music sources, the weight of each label that obtains is (t
A1+ t
B1+ t
C1, t
A2+ t
B2+ t
C2, t
A3+ t
B3+ t
C3).Then, the weight after just can each being merged is arranged according to order from big to small, and the several weights as this music sources set that come the front are got final product.
Need to prove; Because each music sources during a music sources is gathered possibly have different labels; Therefore, when the label to each music sources merges, be the equal of the union of getting the list of labels of each music sources; Simultaneously, the monomer weight addition of same label in each music sources.For example, suppose to comprise music sources a, b, c in certain music sources set A that wherein, music sources a has three labels 1,2,3, and for this music sources a, the weight of each label is respectively (t
A1, t
A2, t
A3); Music sources b has two labels 1,2, and for this music sources b, the weight of each label is respectively (t
B1, t
B2); Music sources c has three labels 2,3,4, and for this music sources c, the weight of each label is respectively (t
C2, t
C3, t
C4), then after the label merging with each music sources, the weight of each label that obtains is (t
A1+ t
B1, t
A2+ t
B2+ t
C2, t
A3+ t
C3, t
C4).
Certainly, when extracting the label of music sources set, can also combine the significance level of each music sources to calculate.Also promptly, when the label of computational music resource collection, important music sources, the weight of its label has higher weighted value.Wherein, the significance level of music sources can adopt its popular degree to calculate.For example, suppose to comprise music sources a, b in certain music sources set A that wherein, music sources a has two labels 1,2, and for this music sources a, the weight of each label is respectively (t
A1, t
A2); Music sources b has two labels 1,2, and for this music sources b, the weight of each label is respectively (t
B1, t
B2); If the importance degree of music sources a is higher than music sources b, then after the label merging with each music sources, the weight of each label that obtains can be (x* (t
A1+ t
B1), y* (t
A2+ t
B2)), wherein, 0＜y＜x＜1, x+y=1.
In practical application; The user may submit a plurality of music sources to modes such as music lists; At this moment, can calculate the label of this music list, then according to the label of this list according to preceding method; And the label of the music sources that calculates set, recommend other music sources set to the user.
Perhaps, if each music sources of a singer is formed a music sources set, the label of then aforementioned music sources set just can be represented each singer's label.Also promptly, the label of a plurality of music sources of a singer is merged, finally can obtain this singer's label, in order to the characteristics of representing this singer to have.On this basis, can also calculate the similarity between each singer,, carry out the personalized recommendation of music sources to the user according to singer's similarity.Concrete, can the weight of each singer's label be mapped to vector space, obtain each singer's label vector, through the mode of the cosine angle between the computation tag vector in twos, just can obtain each singer similarity between any two then.When carrying out personalized recommendation according to the similarity between the singer, can be after the user submit certain singer to, this singer and other singers' similarity is sorted, the singer that rank is forward recommends this user as this singer's similar singer.
Certainly, when calculating the similarity between the singer according to the method described above, the label vector that only has identical vector space just has comparability; Also be; Suppose that certain singer a, b have three labels 1,2,3, wherein for music sources a, the weight of each label is respectively (t
A1, t
A2, t
A3), for music sources b, the weight of each label is respectively (t
B1, t
B2, t
B3), then can calculate this two singers' similarity through following formula:
Otherwise,, then obviously can't utilize the mode of cosine angle to calculate both similarities if the vector space of two singers' label vector is different.But, in this case, can calculate two singers' similarity according to the mode of cosine angle, directly two singers are confirmed as uncorrelated getting final product.For example, singer a has label 1,2,3, and singer b has label 1,4,5, shows obviously that then these two singers' similarity is lower.
Except a singer's music sources is gathered as a music sources; Can also the music sources that a user listened be formed a music sources set, like this, through obtaining the label of music sources set; Can obtain this user's label; In order to the interest of listening to of representative of consumer, like this, can also carry out the personalized recommendation of music sources to the user according to user's label.During concrete the realization; Can obtain user's the daily record of listening to, history is sung in listening of recording user, so just can the music sources that a user listened be formed a music sources set; And extracting the label of music sources set according to preceding method, this label just can be used as this user's label.Certainly, other users also can do similar processing.
When specifically carrying out personalized recommendation; Except obtaining user's label, can also set up inverted list based on the label of all music sources, obtain the corresponding music sources tabulation of each label; Like this; According to user's label, and the corresponding music sources of label, just can carry out personalized recommendation to the user.During concrete the realization; After the label that has obtained each user; Can in system, preserve the corresponding relation between ID and its label; When the user uses the ID of oneself to sign in in the system, just can obtain the corresponding label of this ID, then that this label is corresponding music sources is recommended this user and is got final product.
Need to prove, can also adjust each label weight of each user relatively according to user's the historical record of listening to.Wherein, listen to that historical record can comprise time of listening to, listen to the source (comprise active searching, click list, local disk etc.), user's behavior (comprising online playing or download or the like), whether repeat to listen to, to evaluation of music sources favorable rating or the like.
During concrete the realization,, but might not each label can both well embody user's the interest of listening to because after the label of all music sources that a user is listened to merged, the user tag that obtains had a lot of.Therefore; Can with to user's label according to weight rank order from big to small; And will come the label of the preset number of front, and confirm as this user's popular label, carry out personalized recommendation according to user's popular label and the corresponding music sources tabulation of label then.Wherein, because the ordering of aforementioned user tag is to be decided by the music sources that user self listens to fully, it doesn't matter with music sources that other users listen to, therefore, several labels that come the front in this ordering is called this user's popular label.
In practical application, also possibly there is this situation: some label of certain user; Possibly in this user's list of labels, sort comparatively lean on after; But its weight wants high a lot of with respect to the weight of other users' same label; In fact this label can embody this user's special preferences, therefore, can be referred to as the long-tail label.During concrete the realization; Can add up each user's the weight of label is calculated the average weight (be designated as w_avg) of each label in all users, and the weight of supposing each label of each user is w; Then can this user's the label value according to w/w_avg be sorted; Come the label of front through this sort method, be this user's long-tail label, the long-tail label has been described user's the popular special preferences that is different from.Just can carry out personalized recommendation to this user then according to user's long-tail label and the corresponding music sources tabulation of label.
In a word, the embodiment of the invention can be obtained the label of music sources automatically, carries out the personalized recommendation of music sources then on this basis to the user, therefore, can realize robotization and personalized recommendation relatively accurately.
Corresponding with the music sources personalized recommendation method that the embodiment of the invention provides, the embodiment of the invention also provides a kind of music sources personalized recommendation system, and referring to Fig. 3, this system comprises:
Overall situation weight acquiring unit 303 is used for the number of times that label seed that the relevant textual information according to said all music sources occurs and each label seed occur, and obtains the overall weight of each label seed;
Monomer weight acquiring unit 304 is used for for single music sources, and number of times and overall weight thereof according to each label seed in its relevant textual information occurs obtain the monomer weight of each label seed with respect to single music sources;
Music label is confirmed unit 305, is used for the label seed that the monomer weight meets prerequisite is confirmed as the label of music sources;
Wherein, recommendation unit 306 can comprise:
Label merges subelement, is used for merging the label of each music sources in this set for music sources set;
Weight merges subelement, is used for according to the monomer weight of each label with respect to each music sources, the weight of each label after obtaining merging;
The set label is confirmed subelement, is used for the label that weight meets prerequisite is confirmed as the label of this music sources set;
First recommends subelement, is used for carrying out personalized recommendation according to the label of said music sources set.
Wherein, said music sources set can be each singer's a music sources, and the label of the music sources set that gets access to just can be represented each singer's label, and is corresponding, can carry out similar singer's recommendation, and at this moment, first recommends subelement to comprise:
The label vector obtains subelement, is used for the weight of each singer's label is mapped to vector space, obtains each singer's label vector;
The similarity computation subunit is used for the cosine angle between the computation tag vector in twos, obtains each singer similarity between any two;
The singer recommends subelement, is used for carrying out personalized recommendation according to the similarity between the singer.
Perhaps, all music sources that the music sources set also can make each user listen, at this moment, the label of the music sources set that gets access to just can be represented each user's label, and corresponding, first recommends subelement to comprise:
Arrange subelement, be used for setting up inverted list, obtain the corresponding music sources tabulation of each label based on the label of each music sources;
Label is recommended subelement, is used for carrying out personalized recommendation according to user's label and the corresponding music sources tabulation of label.
In practical application, when recommending, can also sing history according to listening of user according to user tag, user tag is adjusted, at this moment, this system can also comprise:
Weight adjustment unit is used for the behavior of the listening to historical record according to each user, adjusts each label weight of each user relatively.
In addition, this system can also comprise:
Popular label acquiring unit is used for label to the user according to weight rank order from big to small, and will come the label of the preset number of front, confirms as this user's popular label;
Said label recommends subelement specifically to be used for: personalized recommendation is carried out in popular label and the corresponding music sources tabulation of label according to said user.
Perhaps, this system also can also comprise:
The average weight computing unit is used for the weight according to each user's label, calculates the average weight of each label in all users;
Long-tail label acquiring unit is used for this user's the label size according to the merchant of this user's relatively weight and said average weight is sorted, and will comes the label of the preset number of front, confirms as this user's long-tail label;
Said label recommends subelement specifically to be used for: personalized recommendation is carried out in long-tail label and the corresponding music sources tabulation of label according to said user.
More than to a kind of music sources personalized recommendation method provided by the present invention and system; Carried out detailed introduction; Used concrete example among this paper principle of the present invention and embodiment are set forth, the explanation of above embodiment just is used for helping to understand method of the present invention and core concept thereof; Simultaneously, for one of ordinary skill in the art, according to thought of the present invention, part all can change on embodiment and range of application.In sum, this description should not be construed as limitation of the present invention.
Claims (14)
1. a music sources personalized recommendation method is characterized in that, comprising:
Grasp the text message relevant from network with music sources;
Said text message is cut speech; Add up the number of times that the label seed that occurs in the relevant textual information of each music sources and each label seed occur, and the number of times that occurs of the label seed that occurs in the relevant textual information of all music sources and each label seed;
The number of times that the label seed that occurs in the relevant textual information according to said all music sources and each label seed occur obtains the overall weight of each label seed;
For single music sources, number of times and overall weight thereof according to each label seed in its relevant textual information occurs obtain the monomer weight of each label seed with respect to single music sources;
The label seed that the monomer weight is met prerequisite is confirmed as the label of music sources;
Label according to said music sources carries out personalized recommendation.
2. method according to claim 1 is characterized in that, said label according to said music sources carries out personalized recommendation and comprises:
For music sources set, merge the label of each music sources in this set;
According to each label with respect to the monomer weight in each music sources, the weight of each label after obtaining merging;
The label that weight is met prerequisite is confirmed as the label that this music sources is gathered;
Label according to said music sources set carries out personalized recommendation.
3. method according to claim 2; It is characterized in that; Said music sources set comprises each singer's music sources, and the label of said music sources set comprises each singer's label, and said label according to said music sources carries out personalized recommendation and comprises:
The weight of each singer's label is mapped to vector space, obtains each singer's label vector;
Cosine angle between the computation tag vector obtains each singer similarity between any two in twos;
Carry out personalized recommendation according to the similarity between the singer.
4. method according to claim 2; It is characterized in that; Said music sources set comprises all music sources that each user listened, and the label of said music sources set comprises each user's label, and said label according to said music sources carries out personalized recommendation and comprises:
Label based on each music sources is set up inverted list, obtains the corresponding music sources tabulation of each label;
Personalized recommendation is carried out in label and the corresponding music sources tabulation of label according to the user.
5. method according to claim 4 is characterized in that, also comprises:
According to each user's the behavior of listening to historical record, adjust each label weight of each user relatively.
6. according to claim 4 or 5 described methods, it is characterized in that, also comprise:
According to weight rank order from big to small, and will come the label of the preset number of front to user's label, confirm as this user's popular label;
Said label and the corresponding music sources tabulation of label according to the user carried out personalized recommendation and comprised: personalized recommendation is carried out in popular label and the corresponding music sources tabulation of label according to said user.
7. according to claim 4 or 5 described methods, it is characterized in that, also comprise:
According to the weight of each user's label, calculate the average weight of each label in all users;
This user's the label size according to the merchant of this user's relatively weight and said average weight is sorted, and will come the label of the preset number of front, confirm as this user's long-tail label;
Said label and the corresponding music sources tabulation of label according to the user carried out personalized recommendation and comprised: personalized recommendation is carried out in long-tail label and the corresponding music sources tabulation of label according to said user.
8. a music sources personalized recommendation system is characterized in that, comprising:
The information placement unit is used for grasping the text message relevant with music sources from network;
Statistic unit; Be used for said text message is cut speech; Add up the number of times that the label seed that occurs in the relevant textual information of each music sources and each label seed occur, and the number of times that occurs of the label seed that occurs in the relevant textual information of all music sources and each label seed;
Overall situation weight acquiring unit is used for the number of times that label seed that the relevant textual information according to said all music sources occurs and each label seed occur, and obtains the overall weight of each label seed;
Monomer weight acquiring unit is used for for single music sources, and number of times and overall weight thereof according to each label seed in its relevant textual information occurs obtain the monomer weight of each label seed with respect to single music sources;
Music label is confirmed the unit, is used for the label seed that the monomer weight meets prerequisite is confirmed as the label of music sources;
Recommendation unit is used for carrying out personalized recommendation according to the label of said music sources.
9. system according to claim 8 is characterized in that, said recommendation unit comprises:
Label merges subelement, is used for merging the label of each music sources in this set for music sources set;
Weight merges subelement, is used for according to the monomer weight of each label with respect to each music sources, the weight of each label after obtaining merging;
The set label is confirmed subelement, is used for the label that weight meets prerequisite is confirmed as the label of this music sources set;
First recommends subelement, is used for carrying out personalized recommendation according to the label of said music sources set.
10. system according to claim 9 is characterized in that, said music sources set comprises each singer's music sources, and the label of said music sources set comprises each singer's label, and said first recommends subelement to comprise:
The label vector obtains subelement, is used for the weight of each singer's label is mapped to vector space, obtains each singer's label vector;
The similarity computation subunit is used for the cosine angle between the computation tag vector in twos, obtains each singer similarity between any two;
The singer recommends subelement, is used for carrying out personalized recommendation according to the similarity between the singer.
11. system according to claim 9 is characterized in that, said music sources set comprises all music sources that each user listened, and the label of said music sources set comprises each user's label, and said first recommends subelement to comprise:
Arrange subelement, be used for setting up inverted list, obtain the corresponding music sources tabulation of each label based on the label of each music sources;
Label is recommended subelement, is used for carrying out personalized recommendation according to user's label and the corresponding music sources tabulation of label.
12. system according to claim 11 is characterized in that, also comprises:
Weight adjustment unit is used for the behavior of the listening to historical record according to each user, adjusts each label weight of each user relatively.
13. according to claim 11 or 12 described systems, it is characterized in that, also comprise:
Popular label acquiring unit is used for label to the user according to weight rank order from big to small, and will come the label of the preset number of front, confirms as this user's popular label;
Said label recommends subelement specifically to be used for: personalized recommendation is carried out in popular label and the corresponding music sources tabulation of label according to said user.
14. according to claim 11 or 12 described systems, it is characterized in that, also comprise:
The average weight computing unit is used for the weight according to each user's label, calculates the average weight of each label in all users;
Long-tail label acquiring unit is used for this user's the label size according to the merchant of this user's relatively weight and said average weight is sorted, and will comes the label of the preset number of front, confirms as this user's long-tail label;
Said label recommends subelement specifically to be used for: personalized recommendation is carried out in long-tail label and the corresponding music sources tabulation of label according to said user.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2010105556951A CN101984437B (en) | 2010-11-23 | 2010-11-23 | Music resource individual recommendation method and system thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2010105556951A CN101984437B (en) | 2010-11-23 | 2010-11-23 | Music resource individual recommendation method and system thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101984437A CN101984437A (en) | 2011-03-09 |
| CN101984437B true CN101984437B (en) | 2012-08-08 |
Family
ID=43641606
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2010105556951A Active CN101984437B (en) | 2010-11-23 | 2010-11-23 | Music resource individual recommendation method and system thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101984437B (en) |
Families Citing this family (34)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102750289B (en) * | 2011-04-19 | 2015-08-05 | 富士通株式会社 | Based on the method and apparatus that set of tags mixes data |
| CN103064856B (en) * | 2011-10-21 | 2016-03-30 | 中国移动通信集团重庆有限公司 | A kind of resource recommendation method based on Belief Network and device |
| US20130124584A1 (en) * | 2011-11-16 | 2013-05-16 | Google Inc. | Start page for a user's personal music collection |
| CN102567467A (en) * | 2011-12-02 | 2012-07-11 | 华中科技大学 | Method for acquiring hotspot video information based on video tags |
| CN102682132B (en) * | 2012-05-18 | 2013-12-04 | 合一网络技术(北京)有限公司 | Method and system for searching information based on word frequency, play amount and creation time |
| CN103631823B (en) * | 2012-08-28 | 2017-01-18 | 腾讯科技（深圳）有限公司 | Method and device for recommending media content |
| CN103678388B (en) * | 2012-09-19 | 2016-09-14 | 腾讯科技（深圳）有限公司 | Online Music recommends method and device |
| CN102890713B (en) * | 2012-09-20 | 2015-08-12 | 浙江大学 | A kind of music recommend method based on user's current geographic position and physical environment |
| CN103218438B (en) * | 2013-04-18 | 2016-04-20 | 广东欧珀移动通信有限公司 | Method and the mobile terminal of Online Music recommended in broadcasting record based on mobile terminal |
| CN103294778B (en) * | 2013-05-13 | 2019-07-23 | 百度在线网络技术（北京）有限公司 | A kind of method and system pushing information |
| CN104166648A (en) * | 2013-05-16 | 2014-11-26 | 百度在线网络技术（北京）有限公司 | Recommendation data excavation method and device based on labels |
| CN104090912B (en) * | 2014-06-10 | 2017-03-15 | 腾讯科技（深圳）有限公司 | Information-pushing method and device |
| CN104239571B (en) * | 2014-09-30 | 2018-04-24 | 北京奇虎科技有限公司 | It is a kind of to carry out using the method and apparatus recommended |
| CN105989018B (en) * | 2015-01-29 | 2020-04-21 | 深圳市腾讯计算机系统有限公司 | Label generation method and label generation device |
| CN104573105B (en) * | 2015-01-30 | 2017-10-20 | 福州大学 | Hit song, the recommendation method of singer in a kind of Music on Demand network |
| CN104899302B (en) * | 2015-06-10 | 2018-07-17 | 百度在线网络技术（北京）有限公司 | Recommend the method and apparatus of music to user |
| CN104965897A (en) * | 2015-06-26 | 2015-10-07 | 百度在线网络技术（北京）有限公司 | Information recommendation method and device |
| CN105045864B (en) * | 2015-07-10 | 2019-11-05 | 浙江工商大学 | A kind of digitalization resource personalized recommendation method |
| CN105335465B (en) * | 2015-09-23 | 2019-03-05 | 广州酷狗计算机科技有限公司 | A kind of method and apparatus showing main broadcaster's account |
| CN105608105B (en) * | 2015-10-30 | 2018-11-13 | 浙江大学 | It is a kind of that method is recommended based on the music for listening to context |
| CN105868254A (en) * | 2015-12-25 | 2016-08-17 | 乐视网信息技术（北京）股份有限公司 | Information recommendation method and apparatus |
| CN106227816B (en) * | 2016-07-22 | 2019-08-06 | 北京小米移动软件有限公司 | Method and device for pushing song list |
| CN106250557A (en) * | 2016-08-16 | 2016-12-21 | 青岛海信传媒网络技术有限公司 | The recommendation method and device of application |
| CN106528588A (en) * | 2016-09-14 | 2017-03-22 | 厦门幻世网络科技有限公司 | Method and apparatus for matching resources for text information |
| CN107977370B (en) * | 2016-10-21 | 2021-03-12 | 北京酷我科技有限公司 | Singer recommendation method and system |
| CN107423352A (en) * | 2017-05-25 | 2017-12-01 | 杭州回车电子科技有限公司 | Music recommends method and system |
| CN108133011B (en) * | 2017-12-22 | 2022-05-24 | 新奥（中国）燃气投资有限公司 | Information pushing method and device |
| CN108363769A (en) * | 2018-02-07 | 2018-08-03 | 大连大学 | The method for building up of semantic-based music retrieval data set |
| CN108595599A (en) * | 2018-04-19 | 2018-09-28 | 广州优视网络科技有限公司 | Using label generating method, device, storage medium and computer equipment |
| CN110096614B (en) * | 2019-04-12 | 2022-09-20 | 腾讯科技（深圳）有限公司 | Information recommendation method and device and electronic equipment |
| CN110688575A (en) * | 2019-09-25 | 2020-01-14 | 南阳理工学院 | A computer-based art design platform |
| CN110598011B (en) * | 2019-09-27 | 2024-05-28 | 腾讯科技（深圳）有限公司 | Data processing method, device, computer equipment and readable storage medium |
| CN111611432B (en) * | 2020-05-29 | 2023-09-15 | 北京酷我科技有限公司 | Singer classification method based on Labeled LDA model |
| CN112800270B (en) * | 2021-01-27 | 2022-10-14 | 南京邮电大学 | Music recommendation method and system based on music tags and time information |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101276375A (en) * | 2007-03-31 | 2008-10-01 | 索尼德国有限责任公司 | Method for recommending content |
| CN101364222A (en) * | 2008-09-02 | 2009-02-11 | 浙江大学 | A Two-Stage Audio Retrieval Method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080222120A1 (en) * | 2007-03-08 | 2008-09-11 | Nikolaos Georgis | System and method for video recommendation based on video frame features |
-
2010
- 2010-11-23 CN CN2010105556951A patent/CN101984437B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101276375A (en) * | 2007-03-31 | 2008-10-01 | 索尼德国有限责任公司 | Method for recommending content |
| CN101364222A (en) * | 2008-09-02 | 2009-02-11 | 浙江大学 | A Two-Stage Audio Retrieval Method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101984437A (en) | 2011-03-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101984437B (en) | Music resource individual recommendation method and system thereof | |
| US10242120B1 (en) | Selecting a template for a content item | |
| US9020933B2 (en) | Identifying inadequate search content | |
| US8719283B2 (en) | Summarizing reviews | |
| CN104899302A (en) | Method and device for recommending music to user | |
| CN102043781B (en) | Web page resource recommendation method and device | |
| CN103885987B (en) | A kind of music recommends method and system | |
| US20080103886A1 (en) | Determining relevance of a term to content using a combined model | |
| CA2610038A1 (en) | Providing community-based media item ratings to users | |
| CN102163211A (en) | Information processing device, importance calculation method, and program | |
| CN102637178A (en) | Music recommending method, music recommending device and music recommending system | |
| CN106776860A (en) | One kind search abstraction generating method and device | |
| CN104063476A (en) | Social network-based content recommending method and system | |
| CN100527132C (en) | Classified sample set optimizing method and content-related advertising server | |
| CN105868254A (en) | Information recommendation method and apparatus | |
| CN109271550A (en) | A kind of music personalization classification recommended method based on deep learning | |
| CN103455538A (en) | Information processing apparatus, information processing method, and program | |
| CN105069080A (en) | Document retrieval method and system | |
| CN102650991A (en) | Commodity recommending method and system both based on user preference | |
| US20130346385A1 (en) | System and method for a purposeful sharing environment | |
| CN105468649A (en) | Method and apparatus for determining matching of to-be-displayed object | |
| KR101542417B1 (en) | Method and apparatus for learning user preference | |
| CN112800270B (en) | Music recommendation method and system based on music tags and time information | |
| US8868565B1 (en) | Calibrating click duration according to context | |
| CN105468654A (en) | Method and system for selecting reading range of digital resource |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |