CN110362711A - Song recommendations method and device - Google Patents
Song recommendations method and device Download PDFInfo
- Publication number
- CN110362711A CN110362711A CN201910580213.9A CN201910580213A CN110362711A CN 110362711 A CN110362711 A CN 110362711A CN 201910580213 A CN201910580213 A CN 201910580213A CN 110362711 A CN110362711 A CN 110362711A
- Authority
- CN
- China
- Prior art keywords
- song
- songs
- target user
- recommendation
- list
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/60—Information retrieval; Database structures therefor; File system structures therefor of audio data
- G06F16/63—Querying
- G06F16/635—Filtering based on additional data, e.g. user or group profiles
- G06F16/637—Administration of user profiles, e.g. generation, initialization, adaptation or distribution
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/60—Information retrieval; Database structures therefor; File system structures therefor of audio data
- G06F16/63—Querying
- G06F16/638—Presentation of query results
- G06F16/639—Presentation of query results using playlists
Abstract
The disclosure is directed to a kind of song recommendations method and devices.This method comprises: obtaining search instruction, it is that target user determines that recommendation song is single that search instruction, which is used to indicate terminal,;Include at least one keyword in response to search instruction, multiple songs at least one keyword match are obtained from default library as multiple songs to be selected;Do not include keyword in response to search instruction or song at least one keyword match has not been obtained, collaborative filtering is carried out to all songs that default library includes and obtains the multiple songs for meeting target user's preference as multiple songs to be selected;According to multiple songs to be selected, generates and show the recommendation song list for target user.In the technical solution, terminal can provide different song recommendations methods according to whether search instruction includes keyword, meet the individual demand of different user, improve flexibility and the precision of song recommendations, and then improve user experience.
Description
Technical field
This disclosure relates to technical field of information processing more particularly to a kind of song recommendations method and device.
Background technique
With the continuous development of information technology and the continuous expansion of internet scale, the data of magnanimity are stored on network,
The information for being best suitable for user's expectation how is got in these mass datas, becomes the hot issue of information recommendation research.
In the related technology, information recommendation system can combine closely user and information, different users be carried out a
Property screening, help user to filter out the information for meeting user preference from mass data, then by these personalized letters
Breath is presented to the user.
Summary of the invention
To overcome the problems in correlation technique, the embodiment of the present disclosure provides a kind of song recommendations method and device.Institute
It is as follows to state technical solution:
According to the first aspect of the embodiments of the present disclosure, a kind of song recommendations method is provided, comprising:
Search instruction is obtained, it is that target user determines that recommendation song is single that described search instruction, which is used to indicate terminal,;
In response to described search instruction include at least one keyword, from default library obtain with it is described at least one
Multiple songs of keyword match are as multiple songs to be selected;
Do not include keyword or have not been obtained and at least one keyword match in response to described search instruction
Song carries out collaborative filtering acquisition to all songs that the default library includes and meets the multiple of target user's preference
Song is as the multiple song to be selected;
According to the multiple song to be selected, generates and show the recommendation song list for the target user.
The technical scheme provided by this disclosed embodiment can include the following benefits: terminal can be according to search instruction
Whether include keyword to provide different song recommendations methods, meets the individual demand of different user, improve song
The flexibility of recommendation and precision, and then improve user experience.
In one embodiment, described according to the multiple song to be selected, it generates and shows for the target user's
Recommend song single, comprising:
The multiple song to be selected is ranked up based on preset personalized ordering strategy, determines that First ray song is single;
The personalized ordering strategy includes song temperature preference strategy, singer's temperature preference strategy, in song quality preference strategy
At least one;
List is sung according to the First ray, generate and shows the recommendation song list for the target user.
In one embodiment, described that list is sung according to the First ray, it generates and shows for the target user's
Recommend song single, comprising:
Obtain the user behavior data of the target user and/or the context of this search;
The user behavior data include the target user in the multiple song to be selected each song to be selected it is complete
Listen probability and/or the target user to the Information on Collection of each song to be selected in the multiple song to be selected;Described this is searched
The keyword that the search instruction that the context of rope inputs before including the target user includes;
According to the user behavior data and/or this search context, to the First ray song singly include it is more
A song to be selected carries out minor sort again, and it is single to obtain the second sequence song;
List is sung according to second sequence, generate and shows the recommendation song list for the target user.
In one embodiment, described that list is sung according to second sequence, it generates and shows for the target user's
Recommend song single, comprising:
It is complete to obtain the user of the searching times of each song to be selected and/or each song to be selected in the multiple song to be selected
Listen probability;
Listen probability to described according to the user of the searching times of each song to be selected and/or each song to be selected is complete
Multiple songs to be selected that two sequences song singly includes carry out minor sort again, generate and show the recommendation song for the target user
It is single.
In one embodiment, described according to the multiple song to be selected, it generates and shows for the target user's
Recommend song single, comprising:
The multiple song to be selected is screened and/or sorted according to preset feature ordering model, obtains third sequence
Column song is single;
The feature ordering model listens probability for the user according to multiple songs is complete, and multiple users are directed to the receipts of multiple songs
Hide information, at least one characteristic information is trained in the initial Preferences of multiple users and the voiceprint of multiple users
The machine learning model arrived;
List is sung according to the third sequence, generate and shows the recommendation song list for the target user.
In one embodiment, the method also includes the training steps of the feature ordering model:
Determine that each song is for selected in the multiple song according to the historical record that multiple users listen attentively to multiple songs
At least one described characteristic information in each characteristic information characteristic value;
The characteristic value of each characteristic information at least one selected characteristic information is directed to according to each song to institute
It states multiple songs to be screened and/or sorted, it is single to obtain standard song;
The multiple song is inputted to the feature ordering model to be trained established based at least one described characteristic information,
The training song for obtaining the feature ordering model output to be trained is single;
According to the single and described training song of standard song, single comparison result adjusts the feature ordering model, and will be described
Multiple songs input the feature ordering model adjusted, then by the training of the feature ordering model output adjusted
Song is single to be singly compared with standard song again, and then adjusts the feature ordering again according to the comparison result compared again
Model, until the training song list that the feature ordering model exports is matched with standard song list.
In one embodiment, the method also includes the training steps of the feature ordering model:
From corresponding with characteristic information each at least one described characteristic information group of songs is obtained in multiple songs respectively,
Each group of songs includes at least one song;
The multiple song is screened and/or sorted according to the corresponding group of songs of each characteristic information, is obtained
Standard song is single;
The multiple song is inputted to the feature ordering model to be trained established based at least one described characteristic information,
The training song for obtaining the feature ordering model output to be trained is single;
According to the single and described training song of standard song, single comparison result adjusts the feature ordering model, and will be described
Multiple songs input the feature ordering model adjusted, then by the training of the feature ordering model output adjusted
Song is single to be singly compared with standard song again, and then adjusts the feature ordering again according to the comparison result compared again
Model, until the training song list that the feature ordering model exports is matched with standard song list.
In one embodiment, the method also includes:
Obtain the clicking rate of each song and/or the single packet of recommendation song in the multiple songs for recommending song singly to include
The clicking rate of each label in the corresponding multiple labels of the multiple songs included;
According to the clicking rate of each song and/or the clicking rate of each label, it is single to adjust the recommendation song.
In one embodiment, described according to the clicking rate of each song and/or the clicking rate of each label, adjustment
The recommendation is sung
According to the clicking rate of each song and/or the clicking rate of each label, obtains and clicked in the multiple song
Rate is less than or equal to clicking rate in the song to be deleted or multiple labels of default clicking rate threshold value and is less than or equal to default click
The corresponding song to be deleted of the label of rate threshold value;
The song to be deleted is deleted from recommendation song list.
According to the second aspect of an embodiment of the present disclosure, a kind of song recommendations device is provided, comprising:
First obtains module, and for obtaining search instruction, described search instruction is used to indicate terminal and determines for target user
Recommend song single;
Second obtains module, for including at least one keyword in response to described search instruction, from default library
Multiple songs at least one keyword match are obtained as multiple songs to be selected;
Third obtain module, in response to described search instruction do not include keyword or have not been obtained with it is described at least
The song of one keyword match carries out collaborative filtering acquisition to all songs that the default library includes and meets the mesh
Multiple songs of user preference are marked as the multiple song to be selected;
Display module, for according to the multiple song to be selected, generating and showing the recommendation song for the target user
It is single.
In one embodiment, the display module includes:
First acquisition submodule, for being arranged based on preset personalized ordering strategy the multiple song to be selected
Sequence determines that First ray song is single;The personalized ordering strategy include song temperature preference strategy, singer's temperature preference strategy,
At least one of song quality preference strategy;
First shows submodule, single for being sung according to the First ray, generates and shows for the target user's
Recommend song single.
In one embodiment, the first displaying submodule includes:
First acquisition unit, for obtaining the user behavior data of the target user and/or the context of this search;
The user behavior data include the target user in the multiple song to be selected each song to be selected it is complete
Listen probability and/or the target user to the Information on Collection of each song to be selected in the multiple song to be selected;Described this is searched
The keyword that the search instruction that the context of rope inputs before including the target user includes;
Second acquisition unit, for the context according to the user behavior data and/or this search, to described first
Multiple songs to be selected that sequence song singly includes carry out minor sort again, and it is single to obtain the second sequence song;
Display unit, it is single for being sung according to second sequence, it generates and shows the recommendation song for the target user
It is single.
In one embodiment, the display unit is used to obtain searching for each song to be selected in the multiple song to be selected
The user of rope number and/or each song to be selected is complete to listen probability;According to the searching times of each song to be selected and/or each
The user of song to be selected is complete to listen probability to carry out minor sort again to multiple songs to be selected that second sequence song singly includes, and generates simultaneously
It is single to show that the recommendation for the target user is sung.
In one embodiment, the display module includes:
Second acquisition submodule, for being screened according to preset feature ordering model to the multiple song to be selected
And/or sequence, it is single to obtain third sequence song;
The feature ordering model listens probability for the user according to multiple songs is complete, and multiple users are directed to the receipts of multiple songs
Hide information, at least one characteristic information is trained in the initial Preferences of multiple users and the voiceprint of multiple users
The machine learning model arrived;
Second shows submodule, single for being sung according to the third sequence, generates and shows for the target user's
Recommend song single.
In one embodiment, described device further include:
Determining module, the historical record for listening attentively to multiple songs according to multiple users determine each in the multiple song
Characteristic value of the song for each characteristic information at least one selected described characteristic information;
4th obtains module, for being directed to each feature at least one selected characteristic information according to each song
The characteristic value of information is screened and/or is sorted to the multiple song, and it is single to obtain standard song;
5th obtain module, for by the multiple song input based at least one described characteristic information establish wait instruct
Experienced feature ordering model, the training song for obtaining the feature ordering model output to be trained are single;
First training module, for adjusting the feature according to the single comparison result of the single and described training song of standard song
Order models, and the multiple song is inputted into the feature ordering model adjusted, then by the feature adjusted
The training song of order models output is single to be singly compared with standard song again, and then again according to the comparison result compared again
The secondary adjustment feature ordering model sings single match with the standard up to the training song of feature ordering model output is single.
In one embodiment, described device further include:
6th obtain module, for from multiple songs respectively obtain at least one described characteristic information in each feature
The corresponding group of songs of information, each group of songs include at least one song;
7th obtains module, for being sieved according to the corresponding group of songs of each characteristic information to the multiple song
It is single to obtain standard song for choosing and/or sequence;
8th obtain module, for by the multiple song input based at least one described characteristic information establish wait instruct
Experienced feature ordering model, the training song for obtaining the feature ordering model output to be trained are single;
Second training module, for adjusting the feature according to the single comparison result of the single and described training song of standard song
Order models, and the multiple song is inputted into the feature ordering model adjusted, then by the feature adjusted
The training song of order models output is single to be singly compared with standard song again, and then again according to the comparison result compared again
The secondary adjustment feature ordering model sings single match with the standard up to the training song of feature ordering model output is single.
In one embodiment, described device further include:
9th obtain module, for obtain it is described recommend song singly include multiple songs in each song clicking rate, and/
Or the clicking rate for recommending to sing each label in the corresponding multiple labels of multiple songs for singly including;
Module is adjusted, for being pushed away described in adjustment according to the clicking rate of each song and/or the clicking rate of each label
It is single to recommend song.
In one embodiment, the adjustment module includes:
Third acquisition submodule, for obtaining according to the clicking rate of each song and/or the clicking rate of each label
Clicking rate is less than or equal to clicking rate in the song to be deleted or multiple labels of default clicking rate threshold value in the multiple song
Less than or equal to the corresponding song to be deleted of label of default clicking rate threshold value;
Submodule is deleted, for deleting the song to be deleted from recommendation song list.
According to the third aspect of an embodiment of the present disclosure, a kind of song recommendations device is provided, comprising:
Processor;
Memory for storage processor executable instruction;
Wherein, the processor is configured to:
Search instruction is obtained, it is that target user determines that recommendation song is single that described search instruction, which is used to indicate terminal,;
In response to described search instruction include at least one keyword, from default library obtain with it is described at least one
Multiple songs of keyword match are as multiple songs to be selected;
Do not include keyword or have not been obtained and at least one keyword match in response to described search instruction
Song carries out collaborative filtering acquisition to all songs that the default library includes and meets the multiple of target user's preference
Song is as the multiple song to be selected;
According to the multiple song to be selected, generates and show the recommendation song list for the target user.
According to a fourth aspect of embodiments of the present disclosure, a kind of computer readable storage medium is provided, calculating is stored thereon with
Machine instruction, when which is executed by processor the step of realization first aspect any embodiment the method.
It should be understood that above general description and following detailed description be only it is exemplary and explanatory, not
The disclosure can be limited.
Detailed description of the invention
The drawings herein are incorporated into the specification and forms part of this specification, and shows the implementation for meeting the disclosure
Example, and together with specification for explaining the principles of this disclosure.
Fig. 1 a is the flow chart of song recommendations method shown according to an exemplary embodiment.
Fig. 1 b is the flow chart of song recommendations method shown according to an exemplary embodiment.
Fig. 1 c is the flow chart of song recommendations method shown according to an exemplary embodiment.
Fig. 2 is the flow chart of song recommendations method shown according to an exemplary embodiment.
Fig. 3 a is the structural schematic diagram of song recommendations device shown according to an exemplary embodiment.
Fig. 3 b is the structural schematic diagram of song recommendations device shown according to an exemplary embodiment.
Fig. 3 c is the structural schematic diagram of song recommendations device shown according to an exemplary embodiment.
Fig. 3 d is the structural schematic diagram of song recommendations device shown according to an exemplary embodiment.
Fig. 3 e is the structural schematic diagram of song recommendations device shown according to an exemplary embodiment.
Fig. 3 f is the structural schematic diagram of song recommendations device shown according to an exemplary embodiment.
Fig. 3 g is the structural schematic diagram of song recommendations device shown according to an exemplary embodiment.
Fig. 3 h is the structural schematic diagram of song recommendations device shown according to an exemplary embodiment.
Fig. 4 is the structural block diagram of song recommendations device shown according to an exemplary embodiment.
Specific embodiment
Example embodiments are described in detail here, and the example is illustrated in the accompanying drawings.Following description is related to
When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment
Described in embodiment do not represent all implementations consistent with this disclosure.On the contrary, they be only with it is such as appended
The example of the consistent device and method of some aspects be described in detail in claims, the disclosure.
The technical solution that the embodiment of the present disclosure provides is related to terminal, which can be mobile phone, tablet computer, PC
And other equipment that can play audio, the embodiment of the present disclosure are not construed as limiting this.In the related technology, terminal is in order to be based on net
The magnanimity song stored on network recommends to meet its song expected to target user, can be pushed away using the method for collaborative filtering
It recommends, i.e., one group and neighbor user similar in target user's interest is filtered out first, then according to the song of neighbor user preference
The preference for the target user for speculating the song of target user's preference, and then going out by inference recommends song to it.But it uses
This kind of method recommends the precision of song lower, and is not able to satisfy the individual demand of user.What embodiment of the disclosure provided
In technical solution, terminal can provide different song recommendations methods according to whether search instruction includes keyword, meet
The individual demand of different user, improves flexibility and the precision of song recommendations, and then improves user experience.
Fig. 1 a is a kind of flow chart of song recommendations method shown according to an exemplary embodiment, and this method is applied to eventually
End, as shown in Figure 1a, the song recommendations method include the following steps 101 to step 104:
In a step 101, search instruction is obtained, it is that target user determines recommendation song which, which is used to indicate terminal,
It is single.
Exemplary, which can be set search interface, be provided on the search interface search instruction input frame and really
Recognize button.If target user has specific search intention, it can according to need and input song title in the search input frame
The information such as title, artist name or types of songs, then click ACK button.Terminal check gets search instruction at this time, i.e., really
Needing to show to target user calmly recommends song single, includes that the target user is defeated in the search instruction input frame in the search instruction
The text information entered.If target user does not have specific search intention, ACK button can be clicked directly on, terminal can also at this time
Search instruction is got with confirmation, that is, determines and needs to show that recommendation song is single to target user, and the search instruction does not include clear
Information.
Optionally, which is also provided with microphone and search button, and when the search button is pressed, terminal refers to
Show that microphone opens preset time period, i.e. instruction microphone acquires the acoustic information of the preset time period, and by the acoustic information
Search instruction is saved as, terminal, which determines, at this time needs to show that recommendation song is single to target user.Specifically, if target user is with bright
True search intention then can state song using larger volume near terminal as needed after pressing search button
The information such as title, artist name or types of songs.If target user does not have specific search intention, press search button it
Afterwards, it does not need to state any content.
In practical application, terminal also can receive the search instruction by other terminals or server transmission, and the disclosure is real
It applies example and this is not construed as limiting.
In a step 102, include at least one keyword in response to the search instruction, obtain and be somebody's turn to do from default library
Multiple songs of at least one keyword match are as multiple songs to be selected.
Multiple keywords can be stored in advance in terminal, and multiple keyword may include multiple song titles, multiple singers
Name, a variety of types of songs and multiple song labels etc., the embodiment of the present disclosure is it is not limited here.Terminal is determining acquisition
It whether can include that at least one keyword provides different songs for target user according to the search instruction to after search instruction
Bent recommended method.
Exemplary, if target user has specific search intention and inputs search instruction by search interface, this is searched
Suo Zhiling includes the text information that user inputs in the search input frame on search interface, and terminal can identify the text at this time
Whether information includes at least one keyword.If the text information includes at least one keyword, terminal can be according to this extremely
A few keyword carries out information retrieval, obtains from default library and makees with multiple songs of at least one keyword match
For multiple songs to be selected.Optionally, terminal can be obtained and at least one keyword match using Linden search platform
For multiple songs as multiple songs to be selected, which is that the real-time search of distribution realized based on Lucene is put down
Platform, all song resources which can be linked to form the default library.For example, it is assumed that mesh
Mark user search input frame in input " lustily water Liu Dehua ", i.e., search instruction include text information be " lustily water " and
" Liu Dehua ".Terminal can first be compared the text information and multiple keywords, can determine the text by comparing
Information includes a song title " lustily water " and an artist name " Liu Dehua ", i.e., includes two in the search instruction
Keyword, respectively " lustily water " and " Liu Dehua ".It is " lustily that terminal can obtain song title from default library at this time
Water ", and artist name is multiple songs of " Liu Dehua ", i.e. acquisition and the matched multiple songs of the both keyword, and is incited somebody to action
Multiple song is determined as multiple songs to be selected.
Exemplary, if target user has specific search intention and inputs search instruction by voice, which refers to
Enable the voice messaging including target user.Terminal can include to the search instruction first after getting the search instruction
Voice messaging identified, obtain the corresponding feature text information of the voice messaging, then determine this feature text information be
No includes at least one keyword.If this feature text information includes at least one keyword, terminal can be according to this at least
One keyword carries out information retrieval, is obtained from default library using Linden search platform and at least one keyword
Matched multiple songs are as multiple songs to be selected.For example, it is assumed that target user is after pressing search button, near terminal
It is read aloud using larger volume including " lustily water Liu Dehua " text, i.e., includes that the target is used in the voice messaging that search instruction includes
Read aloud the sound of above-mentioned text in family.Terminal can first identify the voice messaging, can determine the voice by identification
Include in the corresponding feature text information of information " lustily water " and " Liu Dehua ".Then terminal can to this feature text information and
Multiple keywords are compared, and can determine that this feature text information includes a song title " lustily water " by comparing, with
And an artist name " Liu Dehua ", i.e., it include both keyword, respectively " lustily water " and " Liu Dehua " in the search instruction.
It is " lustily water " that terminal can obtain song title from default library at this time, and artist name is the multiple of " Liu Dehua "
Song, i.e. acquisition and the matched multiple songs of the both keyword, and multiple song is determined as multiple songs to be selected.
It in step 103, does not include keyword in response to the search instruction or that at least one has not been obtained with this is crucial
The matched song of word presets all songs progress collaborative filtering acquisition that library includes to this and meets target user's preference
Multiple songs are as multiple song to be selected.
Exemplary, if target user does not have specific search intention and inputs search instruction by search interface, this is searched
Suo Zhiling does not include text information, therefore terminal can the be beyond all doubt determination search instruction does not include keyword.Alternatively, if
Target user does not have specific search intention and inputs search instruction, the then sound that terminal includes in the search instruction by voice
Specific feature text information can not be identified in information, and then determines that the search instruction does not include keyword.Terminal can at this time
Multiple songs work that collaborative filtering acquisition meets target user's preference is carried out to preset all songs that library includes to this
For multiple song to be selected, i.e., record is listened to according to the history of the target user and filter out one group and the target from multiple users
Then neighbor user similar in user interest speculates the preference of the target user, Jin Ercong according to the song of neighbor user preference
The multiple songs for meeting target user's preference are obtained as multiple songs to be selected in all songs that the default library includes.
For example, terminal can realize that collaborative filtering obtains the multiple songs for meeting target user's preference, the Cube using Cube framework
The bottom of framework is stored as HBase, cluster management helix, configuration management Zookeeper, can be by the Cube framework energy
All song resources being enough linked to form the default library.Specifically, the Cube framework is a data cube, the number
According to cube by forming according to multiple data cells of multiple default dimension arrangements, multiple default dimension can be the upper of song
Line time, singer, song duration, song languages, song style etc..Each data cell is stored with corresponding metric, this
The metric is song resource in open embodiment.Assuming that multiple data cells that the Cube framework includes are according to singer, song
Qu Shichang, the three dimensions arrangement of song languages, when carrying out collaborative filtering, terminal can be according to three default dimensions successively time
The song resource stored in multiple data cell is gone through, the data sheet for storing the song that the target user once listened to is obtained
Member, the dimension values of the corresponding three default dimensions in position where then determining the data cell, and will be in the Cube framework
The corresponding multiple songs of the dimension values of three default dimensions are determined as multiple songs to be selected of target user.Specifically, assuming
The dimension values of the corresponding three default dimensions in position where the data cell for the song that the target user once listened to are distinguished
For " Liu Dehua ", " 3~4 minutes " and " Guangdong language ", then singer in the Cube framework can be multiple songs of " Liu Dehua " by terminal
Multiple songs and song languages of a length of " 3~4 minutes " are that multiple songs of " Guangdong language " are determined as target user when song, song
Multiple songs to be selected.
Optionally, terminal is at least one keyword for getting search instruction and including, but has not been obtained with this at least
When the song of one keyword match, multiple songs to be selected can also be obtained according to the method for above-mentioned collaborative filtering.
At step 104, it according to multiple song to be selected, generates and shows the recommendation song list for the target user.
Exemplary, multiple song to be selected directly can be sequentially generated pushing away for target user according to random alignment by terminal
It is single to recommend song, and is shown, i.e., show recommendation interface, on the recommendation interface according to the sequence after arrangement successively show it is multiple to
Select song.
Optionally, in order to improve the precision of recommendation, terminal can also be more to this according to preset personalized ordering strategy
A song to be selected is ranked up, and is obtained First ray song list, is then sung list according to the First ray, is obtained and is shown that the target is used
List is sung in the recommendation at family, which may include song temperature preference strategy, singer's temperature preference strategy, song matter
Measure at least one of preference strategy.Specifically, personalized ordering strategy can be arranged in terminal first, the personalized ordering strategy
It may include song temperature preference strategy, singer's temperature preference strategy and song quality preference strategy, i.e. terminal can press first
Multiple song to be selected is ranked up according to one of strategy, sorting identical song to be selected can be according to second therein
Strategy is ranked up again, if there is also the identical song to be selected that sorts, it can be according to the last one strategy therein again
It is ranked up.For example, can be ranked up first, in accordance with song temperature preference strategy to multiple song to be selected, then song is hot
Spend it is identical be ranked up according to singer's temperature preference strategy, singer's temperature it is identical according still further to song quality preference strategy carry out
Sequence.Or song temperature preference strategy, singer's temperature preference strategy and song quality preference strategy can be respectively set in terminal
Weight, it is assumed that the weight of the song temperature preference strategy is a, and the weight of singer's temperature preference strategy is b, and song quality is preferential
The weight of strategy is c, and terminal can obtain the song temperature index M of each song to be selected, Ge Shoure when being ranked up respectively
Index N and song performance figure S is spent, each song composite index F to be selected, composite index F=M*a+N*b+S* are then calculated
C, and then according to the multiple song to be selected of descending arrangement of the composite index of calculated each song to be selected.Specifically
, the song temperature index M can listening to total degree, thumb up total degree or collection total degree etc. for song to be selected, this public affairs
Embodiment is opened not limit this;Singer's temperature index N can thumb up total degree or collection total degree etc. for singer, this
Open embodiment does not limit this;Song performance figure S can be position speed (each second in a data flow of song to be selected
The information content that can pass through) or audio sample rank etc., the embodiment of the present disclosure does not limit this.It is arranged by any one of the above
After program process, it is single that terminal can generate First ray song according to multiple songs to be selected, and is directly by First ray song nonoculture
The recommendation song of the target user is singly shown, i.e., successively showing on recommending interface according to the single sequence of First ray song should
Multiple songs to be selected.
Optionally, which can also be sung in list and be met in advance after getting First ray song list by terminal
If it is required that preceding Q song be determined as that song is recommended list and to be shown, i.e., it is single according to First ray song on recommending interface
The sequence of preceding Q song successively shows multiple song to be selected, and Q is the integer more than or equal to 1.For example, by the First ray
The preceding Q song that composite index is greater than or equal to default index threshold in song list is determined as recommending song list and be shown.
Optionally, which can also obtain the user behavior data of the target user and/or the context of this search,
Then according to the user behavior data and/or the context of this search, the sequence for singly including to First ray song is identical
Multiple songs to be selected carry out minor sort again, obtain the second sequence song list, and then single according to second sequence song, obtain and show this
The recommendation of target user is sung single.The user behavior data includes the target user to each song to be selected in multiple song to be selected
Complete listen probability and/or the target user to the Information on Collection of each song to be selected in multiple song to be selected;This search
Context include keyword that the search instruction inputted before the target user includes.
Specifically, terminal can successively obtain target user to the First ray after getting First ray song list
It sings the complete of each song to be selected in the multiple songs to be selected for singly including and listens probability, i.e., note is listened to according to the history of the target user
Record, obtains the probability that the target user hears out each song to be selected, then listens probability pair according to each the complete of song to be selected
First ray song is single to carry out minor sort again.It, can be with that is, for sorting identical song to be selected in First ray song list
It listens the sequence of probability to be from high to low ranked up according to complete, it is single to obtain the second sequence song.For example, it is assumed that terminal is according to composite index
Multiple songs to be selected are ranked up and get First ray song list, then First ray can be sung composite index phase in list by terminal
Same multiple songs to be selected, that is, the identical multiple songs to be selected that sort listen being arranged from high to low for probability according to complete.
Alternatively, terminal can successively obtain target user and sing to the First ray after getting First ray song list
Whether the Information on Collection of each song to be selected in the multiple songs to be selected for singly including, that is, determine each song to be selected once by the mesh
Mark user collected, and was then arranged again according to the Information on Collection the identical song to be selected that sorts in First ray song list
Sequence, and the song sequence highest of current collecting state, are secondly the song that do not collect collected but current, did not collected finally
Song sequence it is minimum.Such as the song to be selected of collecting state can will be currently in the identical multiple songs to be selected of composite index
Bent sequence is adjusted into the identical multiple songs to be selected of the composite index near preceding position, will once be collected by target user
The sequence for the song to be selected crossed is adjusted into the identical multiple songs to be selected of the composite index by intermediate position, will be not by target
The sequence for the song to be selected that user collected adjusts the rearward position into the composite index identical multiple songs to be selected.
Listen a rule in probability or Information on Collection identical to sorting in First ray song list alternatively, first having used
Minor sort can be continued song to be selected if there is also the identical songs to be selected that sorts after sequence using another rule again
Sequence.
Alternatively, terminal get First ray song list after, it is available this search context, it can obtain
The keyword for including in time search instruction of input is preset before target user, then determines that First ray song singly includes multiple
Whether song to be selected matches with the keyword, and sings single progress minor sort again to the First ray according to matching result, matches journey
Degree is higher, sorts more forward.Such as it can will be to be selected with the keyword match in the identical multiple songs to be selected of composite index
The sequence of song adjusts the position forward into the identical multiple songs to be selected of the composite index, will be unmatched with the keyword
The sequence of song to be selected adjusts the rearward position into the composite index identical multiple songs to be selected.For example, if target user
The keyword for including in the search instruction of last time input is " lyric ", then terminal can sing single sequence according to the First ray
The label for successively obtaining multiple song to be selected, the song to be selected that " lyric " label then will be present are determined as and the keyword
The song to be selected matched, there will be no should " lyric " label song to be selected be determined as with the unmatched song to be selected of the keyword,
And then terminal can be by the sequence tune in the identical multiple songs to be selected of composite index with the song to be selected that there is " lyric " label
The whole position forward into the identical multiple songs to be selected of the composite index, there will be no the suitable of the song to be selected of " lyric " label
Sequence adjusts the rearward position into the composite index identical multiple songs to be selected.
In addition, complete listen probability, Information on Collection, this context searched for can be right with one of them according to preset order
Sort identical song to be selected minor sort again in First ray song list, if there is also the identical song to be selected that sorts after sequence, it can
To continue to sort using second, if being ranked up there is also the song to be selected of identical sequence using the last one.
In practical application, terminal can also determine that second sequence is sung after getting First ray song list first
Sort identical multiple songs to be selected in list, then successively obtains the complete of the identical multiple songs to be selected of the sequence and listens probability, receives
The context of information and/or this search is hidden, then according to the complete context for listening probability, Information on Collection and/or this search
To sorting in First ray song list, identical multiple song to be selected carries out minor sort again, and it is single to obtain the second sequence song.
After above-mentioned sequencer procedure, terminal can sing single second sequence that generates according to First ray and sing list, and can be straight
It connects and is singly shown the recommendation song that second sequence song nonoculture is the target user, i.e., according to second sequence on recommending interface
The single sequence of column song successively shows multiple song to be selected.
Optionally, terminal get second sequence song list after, can also for other users obtain it is multiple to
Select in song that the user of the searching times of each song to be selected and/or each song to be selected is complete to listen probability, it is then each according to this
The user of the searching times of song to be selected and/or each song to be selected is complete listen probability to second sequence song singly include it is multiple to
It selects song to carry out minor sort again, obtain and shows that the recommendation song of the target user is single.Specifically, terminal is getting second sequence
After column song list, the searching times that single sequence successively obtains each song to be selected can be sung according to second sequence, that is, obtained
Multiple and different users search for the total degree of the song to be selected, then according to the searching times of each song to be selected to second sequence
The identical song to be selected that sorts in column song list carries out minor sort again.Assuming that terminal carries out multiple songs to be selected according to composite index
It is single that sequence gets First ray song, and sings single sequence to the First ray according to user behavior data and get the second sequence song
Single, then the second sequence can be sung that composite index is identical in list and meets same user behavior data by terminal, i.e. the second sequence song
It sorts in list identical multiple songs to be selected being arranged from high to low according to searching times.In practical application, terminal is being obtained
After getting second sequence song list, the identical multiple songs to be selected that sort in second sequence song list can also be determined first,
Then the searching times for successively obtaining the identical multiple songs to be selected of the sequence, then according to the searching times to second sequence
The identical multiple song to be selected that sorts in song list carries out minor sort again.
Alternatively, terminal after getting second sequence song list, can sing single sequence successively according to second sequence
Obtain each song to be selected user it is complete listen probability, that is, obtain the probability that multiple and different users hear out the song to be selected, then press
Complete probability is listened to arrange the identical song to be selected that sorts in second sequence song list again according to the user of each song to be selected
Sequence, for example, can multiple songs to be selected that are composite index is identical and meeting same user behavior data listen probability according to user is complete
Arranged from high to low.
Alternatively, first identical to sorting in the second sequence song list using searching times or complete listen in probability a rule
Minor sort can be continued song to be selected if there is also the identical songs to be selected that sorts after sequence using another rule again
Sequence.
After above-mentioned sequencer procedure, terminal can sing the recommendation song for singly generating the target user according to second sequence
It is single, and directly show that recommendation song is single, i.e., it successively shows on recommending interface according to the single sequence of recommendation song multiple to be selected
Song.
In the technical scheme provided by this disclosed embodiment, terminal can be mentioned according to whether search instruction includes keyword
For different song recommendations methods, in target user there are when specific search intention, using information retrieval, personalized ordering plan
It is slightly that user recommends song single in conjunction with the recommended method of user behavior, meets the individual demand of different user, improve
The flexibility of song recommendations and precision, and then improve user experience.
In one embodiment, as shown in Figure 1 b, step 104, i.e., according to multiple song to be selected, generate and show and be directed to
The single step of the recommendation song of the target user, can be realized by step 1041 to step 1042:
In step 1041, multiple song to be selected is ranked up according to preset feature ordering model, or directly into
It is single to obtain third sequence song for row sequence.
In step 1042, list is sung according to the third sequence, generate and shows the recommendation song list for the target user.
Wherein, this feature order models listen probability for the user according to multiple songs is complete, and multiple users are directed to multiple songs
Information on Collection, at least one characteristic information instruction in the initial Preferences of multiple users and the voiceprint of multiple users
The machine learning model got.
Optionally, this feature order models may include GBDT (Gradient Boosting Decison Tree, gradient
Boosted tree), in LR (Logistic Regression, logistic regression) or Wide and Deep (deep learning model) at least
One, the embodiment of the present disclosure is not construed as limiting this.
Exemplary, terminal can first pass through model training in advance and obtain this feature order models, i.e. terminal can be according to multiple
The historical record that user listens attentively to multiple songs determines that each song is for selected at least one feature letter in multiple song
The characteristic value of each characteristic information in breath, then according to each song for each spy at least one selected characteristic information
The characteristic value of reference breath is screened and/or is sorted to multiple song, is obtained standard song list, is then inputted multiple song
Based on the feature ordering model to be trained that at least one characteristic information is established, it is defeated to obtain the feature ordering model to be trained
Training song out is single, then adjusts this feature order models according to the comparison result that standard song is single and training song is single, and will
Multiple song inputs this feature order models adjusted, then sings the training of this feature order models adjusted output
It is single to be singly compared with standard song again, and then this feature order models are adjusted according to the comparison result compared again again,
It is matched until the training song of this feature order models output is single with standard song list, at least one characteristic information can be down at this time
The user including multiple song is complete less listens probability or multiple users to be directed to the Information on Collection of multiple songs.
Specifically, by user is complete listen probability for be illustrated, listen this characteristic information training airplane of probability so that the user is complete
The sequence of multiple songs to input this feature order models may be implemented in the feature ordering model that device learning model obtains.Initially
When change, terminal is available when listening attentively to multiple songs to multiple users to each song listen attentively to time and each song when
To the duration for listening attentively to time and each song of each song when growing, and listening attentively to multiple songs according to multiple user, calculate every
The user of a song is complete to listen probability, i.e. the user of song is complete, and to listen probability be the probability that multiple user hears out the song, then root
It listens probability to be ranked up multiple song according to the user of each song is complete, it is single to obtain standard song.Then terminal can should
Input information input is used as extremely to the duration for listening attentively to time and each song of each song when multiple users listen attentively to multiple songs
Feature ordering model to be trained, then obtains the output data of feature ordering model to be trained, the output data include to
The training song that trained feature ordering model obtains after sorting to multiple song is single.Terminal can compare training song list and it
Before obtained standard song it is single, the parameter of feature ordering model to be trained is adjusted according to the difference of the two, until this feature sorts
The single sequence single with standard song is obtained before of the training song that model training obtains is identical or similarity is greater than or equal to default phase
Threshold speed can illustrate training successfully to this feature order models at this time.
Specifically, being illustrated by taking Information on Collection of multiple users for multiple songs as an example, it is directed to multiple user
The feature ordering model that this characteristic information training machine learning model of the Information on Collection of multiple songs obtains may be implemented to defeated
Enter the sequence of multiple songs of this feature order models.When initialization, terminal is available to be directed to multiple song to multiple users
The Information on Collection of each song in song, i.e., the collection total degree of each song, then according to the collection total degree pair of each song
Multiple song is ranked up, and it is single to obtain standard song.Then terminal extremely can should multiple song as input information input
Feature ordering model to be trained, then obtains the output data of feature ordering model to be trained, the output data include to
The training song that trained feature ordering model obtains after sorting to multiple song is single.Terminal can compare training song list and it
Before obtained standard song it is single, the parameter of feature ordering model to be trained is adjusted according to the difference of the two, until this feature sorts
The single sequence single with standard song is obtained before of the training song that model training obtains is identical or similarity is greater than or equal to default phase
Threshold speed can illustrate training successfully to this feature order models at this time.
It is exemplary, terminal can from multiple songs respectively obtain at least one characteristic information in each characteristic information
Corresponding group of songs, each group of songs include at least one song, and according to the corresponding group of songs of each characteristic information to this
Multiple songs are screened and/or are sorted, and obtain standard song list, the input of multiple song is then based at least one feature
The feature ordering model to be trained that information is established, the training song for obtaining the feature ordering model output to be trained is single, then
This feature order models are adjusted according to the comparison result that standard song is single and training song is single, and multiple song is inputted and is adjusted
This feature order models afterwards, then by the training song of this feature order models adjusted output it is single again with standard song it is single into
Row compares, and then adjusts this feature order models again according to the comparison result compared again, until this feature order models
The training song of output is single to be matched with standard song list.
Specifically, being illustrated by taking the initial Preferences of multiple users as an example, set with the initial preference of multiple user
The feature ordering model that this fixed characteristic information training machine learning model obtains may be implemented to input this feature order models
Multiple songs screening or screening and sequence.When initialization, the available initial Preferences to multiple users of terminal,
Then group of songs corresponding with the initial Preferences of multiple user, each group of songs respectively are obtained from multiple songs includes
At least one song.Assuming that initial Preferences of the terminal by the multiple user of statistics, determine that multiple user's is initial inclined
Good setting can be divided into this four labels of lyric, rock and roll, Guangdong language and Japanese, then be set according to the initial preference of multiple user
It is fixed that multiple songs are screened, i.e., it is filtered out from multiple song and respectively corresponds lyric, rock and roll, Guangdong language and Japanese this four
It is single to obtain standard song then according to the corresponding group of songs of four labels for the group of songs of label.For example, it may be determined that respectively corresponding
The number of songs that each group of songs includes in four group of songs of lyric, rock and roll, the Guangdong language and Japanese, then according to number of songs
The descending song for arranging four group of songs respectively and including, i.e., the group of songs more than number of songs include song arrangement leans on
Before, the song arrangement that the few group of songs of number of songs includes is rearward.In practical application, can also be arranged according to random sequence this four
The song that a group of songs includes, the embodiment of the present disclosure do not limit this.After getting standard song list, terminal can be more by this
A song, to feature ordering model that should be to be trained, then obtains feature ordering model to be trained as input information input
Output data, the output data include the training obtained after the feature ordering model wait train screens and sorts to multiple song
Song is single.Terminal can compare the standard song list that training song is single and obtains before, and spy to be trained is adjusted according to the difference of the two
The parameter of order models is levied, until the single song single with standard song is obtained before of training song that the training of this feature order models obtains
Identical perhaps song is identical and sequence is identical or the single song similarity single with standard song is obtained before of training song is greater than or
Similarity equal to default phase velocity threshold value or song and sequence is all larger than or is equal to default phase velocity threshold value, at this time
Illustrate training successfully to this feature order models.
Alternatively, terminal is after filtering out the group of songs for respectively corresponding this four labels of lyric, rock and roll, Guangdong language and Japanese,
It is single four labels, four standard songs can be corresponded respectively to according to four group of songs generations.Then terminal can will be multiple
One of label in song and four labels is extremely somebody's turn to do feature ordering model to be trained as input information input, then
The output data of feature ordering model to be trained is obtained, which includes feature ordering model to be trained to multiple
The training song obtained after song screening is single.It is single corresponding with the label inputted obtained before that terminal can compare training song
Standard song is single, and the parameter of feature ordering model to be trained is adjusted according to the difference of the two, until the training of this feature order models
The single or similarity identical as the single song of the tag standards inputted song of the training song obtained is greater than or equal to default phase velocity
Threshold value can illustrate training successfully to this feature order models at this time.
Specifically, being illustrated by taking the voiceprint of multiple users as an example, with the voiceprint of multiple user, this is special
Multiple songs to input this feature order models may be implemented in the feature ordering model that reference breath training machine learning model obtains
Bent screening.When initialization, the available voiceprint to multiple users of terminal, it is right respectively then to obtain from multiple songs
The group of songs of different vocal print features is answered, each group of songs includes at least one song.Assuming that terminal is by counting multiple user
Voiceprint, determine that the vocal print feature of multiple user can be divided into male voice feature and female voice feature, then terminal can be with
Multiple songs are screened according to the male voice feature and female voice feature, i.e., is filtered out from multiple song and respectively corresponds male voice
The group of songs of feature and female voice feature, so it is right respectively according to the group of songs acquisition for respectively corresponding the male voice feature and female voice feature
It should two standard song lists of male voice feature and female voice feature.Later, terminal can be by multiple song and the male voice feature or female
Acoustic signature, to feature ordering model that should be to be trained, then obtains feature ordering model to be trained as input information input
Output data, the output data include that the feature ordering model wait train is single to the training song obtained after the screening of multiple song.
Terminal can compare the single standard song list corresponding with the male voice feature or female voice feature label that are inputted of training song, according to the two
Difference adjusts the parameter of feature ordering model to be trained, until the training song that the training of this feature order models obtains is single defeated with institute
The single song of the standard song of the male voice feature or female voice feature that enter is identical or similarity is greater than or equal to default phase velocity threshold
Value, can illustrate training successfully to this feature order models at this time.
In practical application, to the training of this feature order models can be based on multiple rule, can as the case may be into
Row selection, the embodiment of the present disclosure do not limit this.
After the training successfully of feature ordering model, since this feature order models can be based on multiple characteristic informations and more
Kind rule is trained, therefore this feature order models can be realized simultaneously the screening and/or sequence to multiple songs to be selected.Tool
Body, multiple songs to be selected can be input to this feature order models by terminal, by this feature order models screening and/or
Sequence, is screened and/or is sorted to multiple song to be selected, i.e., screened to multiple songs to be selected, or is directly carried out
Sequence, or first screen and be ranked up again, it is single to obtain third sequence song, and the third sequence is directly sung into nonoculture as target use
The recommendation song at family is singly shown, i.e., successively shows on recommending interface according to the single sequence of third sequence song multiple to be selected
Song.
Optionally, which can also obtain the user behavior data of the target user and/or the context of this search,
Then according to the user behavior data and/or the context of this search, the sequence for singly including to third sequence song is identical
Multiple songs to be selected carry out minor sort again, obtain the 4th sequence song list, and then single according to the 4th sequence song, obtain and show this
The recommendation of target user is sung single.The user behavior data includes the target user to each song to be selected in multiple song to be selected
Complete listen probability and/or the target user to the Information on Collection of each song to be selected in multiple song to be selected;This search
Context include keyword that the search instruction inputted before the target user includes.Specifically, terminal is getting third
After sequence sings list, can successively it obtain each to be selected in multiple songs to be selected that target user singly includes to third sequence song
The complete of song listens probability, i.e., according to the historical record of the target user, obtains the target user and hear out each song to be selected
Then probability listens probability to sing single progress minor sort again to the third sequence according to the complete of each song to be selected.Assuming that terminal is pressed
It listens probability training machine learning model to get feature ordering model according to user is complete, and gets according to this feature order models
Three sequences song is single, then third sequence can be sung in list that user is complete to listen the identical multiple songs to be selected of probability according to the target by terminal
The complete of user listens arranging from high to low for probability.
Alternatively, terminal can successively obtain target user and sing to the third sequence after getting third sequence song list
Whether the Information on Collection of each song to be selected in the multiple songs to be selected for singly including, that is, determine each song to be selected once by the mesh
Mark user collected, and then carried out again according to the Information on Collection to the identical multiple songs to be selected that sort in third sequence song list
Minor sort, such as the song to be selected once collected by target user in the identical multiple songs to be selected of probability can be listened by user is complete
Bent sequence, which is adjusted to the user is complete, listens position forward in the identical multiple songs to be selected of probability, will not collected by target user
The sequence for the song to be selected crossed, which is adjusted to the user is complete, listens rearward position in the identical multiple songs to be selected of probability.
Alternatively, terminal get third sequence song list after, it is available this search context, it can obtain
The keyword for including in the search instruction that target user's last time inputs, then determine the third sequence song singly include it is multiple to
Select whether song matches with the keyword, and according to matching result to the identical multiple songs to be selected that sort in third sequence song list
Qu Jinhang minor sort again, for example, can by user it is complete listen it is to be selected with the keyword match in the identical multiple songs to be selected of probability
The sequence of song adjusts the position forward into the identical multiple songs to be selected of the composite index, will be unmatched with the keyword
The sequence of song to be selected, which is adjusted to the user is complete, listens rearward position in the identical multiple songs to be selected of probability.
After above-mentioned sequencer procedure, it is single that terminal can sing single the 4th sequence song that generates according to third sequence, and directly will
The song nonoculture of 4th sequence is that the recommendation song of the target user is singly shown, i.e., sings on recommending interface according to the 4th sequence
Single sequence successively shows multiple song to be selected.
Optionally, terminal can also obtain each in multiple song to be selected after getting the 4th sequence song list
Then the searching times of song to be selected sing the sequence for singly including according to the searching times of each song to be selected to the 4th sequence
Identical multiple songs to be selected carry out minor sort again, obtain and show that the recommendation song of the target user is single.Specifically, terminal is obtaining
After getting the 4th sequence song list, the search that single sequence successively obtains each song to be selected can be sung according to the 4th sequence
Number obtains the total degree that multiple and different users search for the song to be selected, then according to the search of each song to be selected time
It is several that minor sort again is carried out to the identical multiple songs to be selected that sort in the 4th sequence song list.Assuming that terminal is listened generally according to user is complete
Rate training machine learning model gets feature ordering model, and gets third sequence song list according to this feature order models,
And single sequence is sung to the third sequence according to user behavior data and gets the 4th sequence song list, then terminal can be by the 4th sequence
The complete multiple songs to be selected listened probability identical and meet same user behavior data of user are according to searching times by height in song list
It is arranged to low.After above-mentioned sequencer procedure, terminal can be sung according to the 4th sequence and singly generate pushing away for the target user
It is single to recommend song, and directly shows that recommendation song is single, i.e., successively shows on recommending interface according to the single sequence of recommendation song multiple
Song to be selected.
In the technical scheme provided by this disclosed embodiment, terminal can be mentioned according to whether search instruction includes keyword
For different song recommendations methods, meet the individual demand of different user, improve song recommendations flexibility and precisely
Degree, and then improve user experience.
That is, not providing that a variety of ordering strategies using sequence, can be arranged according to the actual situation in the embodiment of the present disclosure
Column combine a variety of ordering strategies.
In one embodiment, as illustrated in figure 1 c, this method further includes step 105 and step 106:
In step 105, obtain the recommendation song singly include multiple songs in each song clicking rate and/or this push away
Recommend the clicking rate of each label in the corresponding multiple labels of multiple songs that song singly includes.
In step 106, according to the clicking rate of each song and/or the clicking rate of each label, recommendation song is adjusted
It is single.
It is exemplary, after terminal successively shows the single multiple songs of recommendation song on recommending interface, it can count respectively
User is to the clicking rate of each song, which can intuitively reflect user to the hope of listening to of each song, and clicking rate is high
Illustrate that user often listens to, clicking rate is low illustrate user listen to the song hope it is lower or user is unwilling to listen to the song
Song, therefore terminal can will click on rate and be determined as song to be deleted less than or equal to the song of default clicking rate threshold value, and again
When secondary display recommendation song list, the song to be deleted is deleted from recommendation song list, so that the entire user for recommending song list is satisfied
Rate is improved.
Alternatively, this can be counted respectively after terminal successively shows the single multiple songs of recommendation song on recommending interface
The clicking rate of each label in the corresponding multiple labels of multiple songs, the label can describe the feature of song, such as song
On-line time, singer, song duration, song languages, song style etc., by every in the corresponding multiple labels of multiple song
The clicking rate of a label can intuitively reflect user to the hope of listening to of the song of a certain label, i.e., user, which listens to, meets certain spy
The song of sign listens to hope, and clicking rate height illustrates that user likes listening to the corresponding song of the label, and clicking rate is low to illustrate user
The hope for listening to the corresponding song of the label is lower or user is unwilling to listen to the corresponding song of the label, therefore terminal can
It is determined as song to be deleted will click on rate less than or equal to the corresponding song of label of default clicking rate threshold value, and is showing again
When showing that recommendation song is single, the song to be deleted is deleted from recommendation song list, so that the entire user's satisfaction rate for recommending song single obtains
To raising.
In practical application, the clicking rate of song and the clicking rate of label can be counted simultaneously, and simultaneously according to this two
A clicking rate is singly adjusted recommendation song, and the embodiment of the present disclosure does not limit this.
In the technical scheme provided by this disclosed embodiment, terminal can determine that recommendation song is single according to the clicking rate of user
It is expected whether to meet user, and adjusts recommendation song list according to the clicking rate, improves the user's satisfaction rate for recommending song single.
Realization process is discussed in detail below by several embodiments.
Fig. 2 is a kind of flow chart of song recommendations method shown according to an exemplary embodiment, and executing subject is terminal,
As shown in Fig. 2, including the following steps 201 to step 216:
In step 201, the search instruction of target user's input is obtained, which can instruct for text class,
It can be phonetic order.
It is that target user determines that recommendation song is single that the search instruction, which is used to indicate terminal,.
In step 202, determine in the search instruction whether include at least one keyword;If the search instruction includes extremely
A few keyword, executes step 203;If the search instruction does not include at least one keyword, step 210 is executed.
The keyword can be song title or singer's title
In step 203, information retrieval is carried out according at least one keyword, obtained from default library with should be to
Multiple songs of a few keyword match are as multiple songs to be selected.
In step 204, multiple song to be selected is ranked up according to preset personalized ordering strategy, obtains first
Sequence song is single.
The personalized ordering strategy includes song temperature preference strategy, singer's temperature preference strategy, the preferential plan of song quality
At least one of slightly.
In step 205, the user behavior data of the target user and/or the context of this search are obtained.
The user behavior data include the target user in multiple songs to be selected each song to be selected it is complete listen probability and/
Or the target user is to the Information on Collection of each song to be selected in multiple song to be selected;The context of this search includes should
The keyword that the search instruction inputted before target user includes.
In step 206, according to the user behavior data and/or the context of this search, single packet is sung to First ray
The identical multiple songs to be selected of the sequence included carry out minor sort again, and it is single to obtain the second sequence song.
In step 207, the searching times and/or each song to be selected of each song to be selected in multiple song to be selected are obtained
Bent user is complete to listen probability.
In a step 208, probability is listened according to the users of the searching times of each song to be selected and/or each song to be selected is complete
Minor sort again is carried out to the identical multiple songs to be selected of sequence that second sequence song singly includes, obtains the recommendation of the target user
Song is single.
In step 209, show that recommendation song is single.
In step 210, collaborative filtering acquisition is carried out to all songs that default library includes and meets the target user
Multiple songs of preference are as multiple songs to be selected.
In step 211, multiple song to be selected is ranked up according to preset feature ordering model, obtains third sequence
Column song is single.
This feature order models listen probability for the user according to multiple songs is complete, and multiple users are directed to the collection of multiple songs
The training of at least one characteristic information obtains in information, the initial Preferences of multiple users and the voiceprint of multiple users
Machine learning model.
In the step 212, the user behavior data of the target user and/or the context of this search are obtained.
The user behavior data include target user in multiple songs to be selected each song to be selected it is complete listen probability and/or
Information on Collection of the target user to each song to be selected in multiple songs to be selected;The context of this search includes target user
The keyword that the search instruction inputted before includes.
In step 213, according to the user behavior data and/or the context of this search, single packet is sung to third sequence
The identical multiple songs to be selected of the sequence included carry out minor sort again, and it is single to obtain the 4th sequence song.
In step 214, the searching times and/or each song to be selected of each song to be selected in multiple song to be selected are obtained
Bent user is complete to listen probability.
In step 215, probability is listened according to the users of the searching times of each song to be selected and/or each song to be selected is complete
Minor sort again is carried out to the identical multiple songs to be selected of sequence that the 4th sequence song singly includes, obtains the recommendation of the target user
Song is single.
In the step 216, show that recommendation song is single.
Embodiment of the disclosure provides a kind of song recommendations method, and whether terminal can include keyword according to search instruction
Different song recommendations methods is provided, the individual demand of different user is met, improve song recommendations flexibility and
Precision, and then improve user experience.
Following is embodiment of the present disclosure, can be used for executing embodiments of the present disclosure.
Fig. 3 a is a kind of structural schematic diagram of song recommendations device 30 shown according to an exemplary embodiment, the device 30
It being implemented in combination with as some or all of of electronic equipment by software, hardware or both.As shown in Figure 3a, the song
Bent recommendation apparatus 30 includes the first acquisition module 301, and second obtains module 302, and third obtains module 303 and display module 304.
Wherein, first module 301 is obtained, for obtaining search instruction, it is target that described search instruction, which is used to indicate terminal,
User, which determines, recommends song single.
Second obtains module 302, for including at least one keyword in response to described search instruction, from default library
Multiple songs of middle acquisition and at least one keyword match are as multiple songs to be selected.
Third obtain module 303, in response to described search instruction do not include keyword or have not been obtained with it is described
The song of at least one keyword match carries out collaborative filtering acquisition to all songs that the default library includes and meets institute
Multiple songs of target user's preference are stated as the multiple song to be selected.
Display module 304, for according to the multiple song to be selected, obtaining and showing the recommendation song of the target user
It is single.
In one embodiment, as shown in Figure 3b, the display module 304 includes the first acquisition submodule 3041 and first
Show submodule 3042.
Wherein, the first acquisition submodule 3041, for being based on preset personalized ordering strategy to the multiple song to be selected
Song is ranked up, and determines that First ray song is single;The personalized ordering strategy includes that song temperature preference strategy, singer's temperature are excellent
First at least one of strategy, song quality preference strategy.
First shows submodule 3042, single for being sung according to the First ray, generates and shows and uses for the target
The recommendation at family is sung single.
In one embodiment, as shown in Figure 3c, described first show that submodule 3042 includes first acquisition unit
3042a, second acquisition unit 3042b and display unit 3042c.
Wherein, first acquisition unit 3042a, for obtain the target user user behavior data and/or this search
The context of rope.
The user behavior data include the target user in the multiple song to be selected each song to be selected it is complete
Listen probability and/or the target user to the Information on Collection of each song to be selected in the multiple song to be selected;Described this is searched
The keyword that the search instruction that the context of rope inputs before including the target user includes.
Second acquisition unit 3042b, for the context according to the user behavior data and/or this search, to institute
It states multiple songs to be selected that First ray song singly includes and carries out minor sort again, it is single to obtain the second sequence song.
Display unit 3042c, it is single for being sung according to second sequence, it generates and shows pushing away for the target user
It is single to recommend song.
In one embodiment, the display unit 3032c is for obtaining each song to be selected in the multiple song to be selected
The user of bent searching times and/or each song to be selected is complete to listen probability;According to the searching times of each song to be selected and/
Or the complete multiple songs to be selected for listening probability singly to include to second sequence song of user of each song to be selected carry out minor sort again,
It generates and shows the recommendation song list for the target user.
In one embodiment, as shown in Figure 3d, the display module 304 includes the second acquisition submodule 3043 and the 4th
Acquisition submodule 3044.
Wherein, the second acquisition submodule 3043 is used for according to preset feature ordering model to the multiple song to be selected
It is screened and/or is sorted, it is single to obtain third sequence song.
The feature ordering model listens probability for the user according to multiple songs is complete, and multiple users are directed to the receipts of multiple songs
Hide information, at least one characteristic information is trained in the initial Preferences of multiple users and the voiceprint of multiple users
The machine learning model arrived.
4th acquisition submodule 3044, it is single for being sung according to the third sequence, it generates and shows and used for the target
The recommendation at family is sung single.
In one embodiment, as shown in Figure 3 e, described device 30 further includes determining module 305, and the 4th obtains module
306, the 5th obtains module 307 and the first training module 308.
Wherein it is determined that module 305, the historical record for listening attentively to multiple songs according to multiple users determines the multiple song
Characteristic value of each song for each characteristic information at least one selected described characteristic information in song.
4th obtains module 306, each at least one selected characteristic information for being directed to according to each song
The characteristic value of characteristic information is screened and/or is sorted to the multiple song, and it is single to obtain standard song.
5th obtains module 307, for establish the input of the multiple song based at least one described characteristic information
Feature ordering model to be trained, the training song for obtaining the feature ordering model output to be trained are single.
First training module 308, for according to the single comparison result adjustment of the single and described training song of standard song
Feature ordering model, and the multiple song is inputted into the feature ordering model adjusted, it then will be adjusted described
The training song of feature ordering model output is single to be singly compared with standard song again, and then according to the comparison knot compared again
Fruit adjusts the feature ordering model again, until the training song of feature ordering model output is single to sing single with the standard
Match.
In one embodiment, as illustrated in figure 3f, described device 30 further includes the 6th acquisition module 309, and the 7th obtains mould
Block 310, the 8th obtains module 311 and the second training module 312.
Wherein, the 6th obtain module 309, for from multiple songs respectively obtain at least one described characteristic information in
The corresponding group of songs of each characteristic information, each group of songs includes at least one song.
7th obtains module 310, for according to the corresponding group of songs of each characteristic information to the multiple song into
It is single to obtain standard song for row screening and/or sequence.
8th obtains module 311, for establish the input of the multiple song based at least one described characteristic information
Feature ordering model to be trained, the training song for obtaining the feature ordering model output to be trained are single.
Second training module 312, for according to the single comparison result adjustment of the single and described training song of standard song
Feature ordering model, and the multiple song is inputted into the feature ordering model adjusted, it then will be adjusted described
The training song of feature ordering model output is single to be singly compared with standard song again, and then according to the comparison knot compared again
Fruit adjusts the feature ordering model again, until the training song of feature ordering model output is single to sing single with the standard
Match.
In one embodiment, as shown in figure 3g, described device further includes the 9th acquisition module 313 and adjustment module 314.
Wherein, the 9th module 313 is obtained, for obtaining the point of each song in the multiple songs for recommending song singly to include
Hit the clicking rate of each label in rate and/or the corresponding multiple labels of multiple songs for recommending song singly to include.
Module 314 is adjusted, for adjusting institute according to the clicking rate of each song and/or the clicking rate of each label
It is single to state recommendation song.
In one embodiment, as illustrated in figure 3h, the adjustment module 314 includes:
Third acquisition submodule 3141, for according to the clicking rate of each song and/or the clicking rate of each label,
It obtains clicking rate in the multiple song and is less than or equal to the song to be deleted for presetting clicking rate threshold value or multiple label midpoints
Hit the corresponding song to be deleted of label that rate is less than or equal to default clicking rate threshold value.
Submodule 3142 is deleted, for deleting the song to be deleted from recommendation song list.
Embodiment of the disclosure provides a kind of song recommendations device, and whether which can include key according to search instruction
Word provides different song recommendations methods, meets the individual demand of different user, improves the flexibility of song recommendations
And precision, and then improve user experience.
The embodiment of the present disclosure provides a kind of song recommendations device, which includes:
Processor;
Memory for storage processor executable instruction;
Wherein, processor is configured as:
Search instruction is obtained, it is that target user determines that recommendation song is single that described search instruction, which is used to indicate terminal,;
In response to described search instruction include at least one keyword, from default library obtain with it is described at least one
Multiple songs of keyword match are as multiple songs to be selected;
Do not include keyword or have not been obtained and at least one keyword match in response to described search instruction
Song carries out collaborative filtering acquisition to all songs that the default library includes and meets the multiple of target user's preference
Song is as the multiple song to be selected;
According to the multiple song to be selected, generates and show the recommendation song list for the target user.
In one embodiment, above-mentioned processor is also configured to: based on preset personalized ordering strategy to described
Multiple songs to be selected are ranked up, and determine that First ray song is single;The personalized ordering strategy include song temperature preference strategy,
At least one of singer's temperature preference strategy, song quality preference strategy;List is sung according to the First ray, generates and shows
List is sung in recommendation for the target user.
In one embodiment, above-mentioned processor is also configured to: obtaining the user behavior data of the target user
And/or the context of this search;The user behavior data includes the target user to every in the multiple song to be selected
The complete of a song to be selected listens probability and/or the target user to believe the collection of each song to be selected in the multiple song to be selected
Breath;The keyword that the search instruction that the context of this search inputs before including the target user includes;According to institute
The context for stating user behavior data and/or this search carries out multiple songs to be selected that First ray song singly includes
It is single to obtain the second sequence song for minor sort again;List is sung according to second sequence, generates and shows pushing away for the target user
It is single to recommend song.
In one embodiment, above-mentioned processor is also configured to: being obtained each to be selected in the multiple song to be selected
The user of the searching times of song and/or each song to be selected is complete to listen probability;According to the searching times of each song to be selected
And/or the complete multiple songs to be selected for listening probability singly to include to second sequence song of user of each song to be selected are arranged again
Sequence generates and shows the recommendation song list for the target user.
In one embodiment, above-mentioned processor is also configured to: according to preset feature ordering model to described more
A song to be selected is screened and/or is sorted, and it is single to obtain third sequence song;The feature ordering model is according to multiple songs
User is complete to listen probability, and multiple users are directed to the Information on Collection of multiple songs, the initial Preferences of multiple users and multiple use
The machine learning model that the training of at least one characteristic information obtains in the voiceprint at family;Single, life is sung according to the third sequence
At and to show that the recommendation for the target user is sung single.
In one embodiment, above-mentioned processor is also configured to: the history of multiple songs is listened attentively to according to multiple users
Record determines that each song is for each characteristic information at least one selected described characteristic information in the multiple song
Characteristic value;The characteristic value of each characteristic information at least one selected characteristic information is directed to according to each song to described
Multiple songs are screened and/or are sorted, and it is single to obtain standard song;By the input of the multiple song based at least one described feature
The feature ordering model to be trained that information is established, the training song for obtaining the feature ordering model output to be trained are single;Root
The feature ordering model is adjusted according to the single comparison result of the single and described training song of standard song, and the multiple song is defeated
Enter the feature ordering model adjusted, then by the training song of feature ordering model output adjusted it is single again with
Standard song is singly compared, and then adjusts the feature ordering model again according to the comparison result compared again, until
The training song of the feature ordering model output is single to sing single match with the standard.
In one embodiment, above-mentioned processor is also configured to: from multiple songs respectively obtain with it is described at least
The corresponding group of songs of each characteristic information in one characteristic information, each group of songs includes at least one song;According to described every
The corresponding group of songs of a characteristic information is screened and/or is sorted to the multiple song, and it is single to obtain standard song;It will be the multiple
Song inputs the feature ordering model to be trained established based at least one described characteristic information, obtains the spy to be trained
The training song for levying order models output is single;According to the single and described training song of standard song, single comparison result adjusts the feature
Order models, and the multiple song is inputted into the feature ordering model adjusted, then by the feature adjusted
The training song of order models output is single to be singly compared with standard song again, and then again according to the comparison result compared again
The secondary adjustment feature ordering model sings single match with the standard up to the training song of feature ordering model output is single.
In one embodiment, above-mentioned processor is also configured to: obtaining the multiple songs for recommending song singly to include
In each song clicking rate and/or it is described recommend song singly include the corresponding multiple labels of multiple songs in each label
Clicking rate;According to the clicking rate of each song and/or the clicking rate of each label, it is single to adjust the recommendation song.
In one embodiment, above-mentioned processor is also configured to: according to the clicking rate of each song and/or often
The clicking rate of a label obtains the song to be deleted that clicking rate in the multiple song is less than or equal to default clicking rate threshold value,
Or clicking rate is less than or equal to the corresponding song to be deleted of label of default clicking rate threshold value in multiple labels;From the recommendation
The song to be deleted is deleted in song list.
Embodiment of the disclosure provides a kind of song recommendations device, and whether which can include key according to search instruction
Word provides different song recommendations methods, meets the individual demand of different user, improves the flexibility of song recommendations
And precision, and then improve user experience.
About the device in above-described embodiment, wherein modules execute the concrete mode of operation in related this method
Embodiment in be described in detail, no detailed explanation will be given here.
Fig. 4 is a kind of structural block diagram for song recommendations device 40 shown according to an exemplary embodiment, the device
Suitable for terminal device.For example, device 40 can be mobile phone, computer, digital broadcasting terminal, messaging device, trip
Play console, tablet device, Medical Devices, body-building equipment, personal digital assistant etc..
Device 40 may include following one or more components: processing component 402, memory 404, and power supply module 406 is more
Media component 408, audio component 410, the interface 412 of input/output (I/O), sensor module 414 and communication component
416。
The integrated operation of the usual control device 40 of processing component 402, such as with display, telephone call, data communication, camera
Operation and record operate associated operation.Processing component 402 may include one or more processors 420 to execute instruction,
To perform all or part of the steps of the methods described above.In addition, processing component 402 may include one or more modules, it is convenient for
Interaction between processing component 402 and other assemblies.For example, processing component 402 may include multi-media module, to facilitate more matchmakers
Interaction between body component 408 and processing component 402.
Memory 404 is configured as storing various types of data to support the operation in device 40.These data are shown
Example includes the instruction of any application or method for operating on device 40, contact data, and telephone book data disappears
Breath, picture, video etc..Memory 404 can be by any kind of volatibility or non-volatile memory device or their group
It closes and realizes, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM) is erasable to compile
Journey read-only memory (EPROM), programmable read only memory (PROM), read-only memory (ROM), magnetic memory, flash
Device, disk or CD.
Power supply module 406 provides electric power for the various assemblies of device 40.Power supply module 406 may include power management system
System, one or more power supplys and other with for device 40 generate, manage, and distribute the associated component of electric power.
Multimedia component 408 includes the screen of one output interface of offer between described device 40 and user.One
In a little embodiments, screen may include liquid crystal display (LCD) and touch panel (TP).If screen includes touch panel, screen
Curtain may be implemented as touch screen, to receive input signal from the user.Touch panel includes one or more touch sensings
Device is to sense the gesture on touch, slide, and touch panel.The touch sensor can not only sense touch or sliding action
Boundary, but also detect duration and pressure associated with the touch or slide operation.In some embodiments, more matchmakers
Body component 408 includes a front camera and/or rear camera.When device 40 is in operation mode, such as screening-mode or
When video mode, front camera and/or rear camera can receive external multi-medium data.Each front camera and
Rear camera can be a fixed optical lens system or have focusing and optical zoom capabilities.
Audio component 410 is configured as output and/or input audio signal.For example, audio component 410 includes a Mike
Wind (MIC), when device 40 is in operation mode, when such as call mode, recording mode, and voice recognition mode, microphone is configured
To receive external audio signal.The received audio signal can be further stored in memory 404 or via communication component
416 send.In some embodiments, audio component 410 further includes a loudspeaker, is used for output audio signal.
I/O interface 412 provides interface between processing component 402 and peripheral interface module, and above-mentioned peripheral interface module can
To be keyboard, click wheel, button etc..These buttons may include, but are not limited to: home button, volume button, start button and lock
Determine button.
Sensor module 414 includes one or more sensors, for providing the status assessment of various aspects for device 40.
For example, sensor module 414 can detecte the state that opens/closes of device 40, the relative positioning of component, such as the component
For the display and keypad of device 40, sensor module 414 can be with the position of 40 1 components of detection device 40 or device
Change, the existence or non-existence that user contacts with device 40, the temperature change in 40 orientation of device or acceleration/deceleration and device 40.
Sensor module 414 may include proximity sensor, be configured to detect object nearby without any physical contact
Presence.Sensor module 414 can also include that optical sensor is used in imaging applications such as CMOS or ccd image sensor
It uses.In some embodiments, which can also include acceleration transducer, gyro sensor, magnetic sensing
Device, pressure sensor or temperature sensor.
Communication component 416 is configured to facilitate the communication of wired or wireless way between device 40 and other equipment.Device
40 can access the wireless network based on communication standard, such as WiFi, 2G or 3G or their combination.In an exemplary implementation
In example, communication component 416 receives broadcast singal or broadcast related information from external broadcasting management system via broadcast channel.
In one exemplary embodiment, the communication component 416 further includes near-field communication (NFC) module, to promote short range communication.Example
Such as, NFC module can be based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra wide band (UWB) technology,
Bluetooth (BT) technology and other technologies are realized.
In the exemplary embodiment, device 40 can be believed by one or more application specific integrated circuit (ASIC), number
Number processor (DSP), digital signal processing appts (DSPD), programmable logic device (PLD), field programmable gate array
(FPGA), controller, microcontroller, microprocessor or other electronic building bricks are realized, for executing the above method.
In the exemplary embodiment, a kind of non-transitorycomputer readable storage medium including instruction, example are additionally provided
It such as include the memory 404 of instruction, above-metioned instruction can be executed by the processor 420 of device 40 to complete the above method.For example, institute
State non-transitorycomputer readable storage medium can be ROM, random access memory (RAM), CD-ROM, tape, floppy disk and
Optical data storage devices etc..
The embodiment of the present disclosure provides a kind of non-transitorycomputer readable storage medium, the instruction in the storage medium
When being executed by the processor of device 40, so that device 40 is able to carry out above-mentioned song recommendations method, which comprises
Search instruction is obtained, it is that target user determines that recommendation song is single that described search instruction, which is used to indicate terminal,;
In response to described search instruction include at least one keyword, from default library obtain with it is described at least one
Multiple songs of keyword match are as multiple songs to be selected;
Do not include keyword or have not been obtained and at least one keyword match in response to described search instruction
Song carries out collaborative filtering acquisition to all songs that the default library includes and meets the multiple of target user's preference
Song is as the multiple song to be selected;
According to the multiple song to be selected, generates and show the recommendation song list for the target user.
Those skilled in the art will readily occur to its of the disclosure after considering specification and practicing disclosure disclosed herein
Its embodiment.This application is intended to cover any variations, uses, or adaptations of the disclosure, these modifications, purposes or
Person's adaptive change follows the general principles of this disclosure and including the undocumented common knowledge in the art of the disclosure
Or conventional techniques.The description and examples are only to be considered as illustrative, and the true scope and spirit of the disclosure are by following
Claim is pointed out.
It should be understood that the present disclosure is not limited to the precise structures that have been described above and shown in the drawings, and
And various modifications and changes may be made without departing from the scope thereof.The scope of the present disclosure is only limited by the accompanying claims.
Claims (20)
1. a kind of song recommendations method characterized by comprising
Search instruction is obtained, it is that target user determines that recommendation song is single that described search instruction, which is used to indicate terminal,;
Include at least one keyword in response to described search instruction, is obtained from default library and at least one described key
The matched multiple songs of word are as multiple songs to be selected;
Do not include keyword in response to described search instruction or have not been obtained song at least one keyword match,
Multiple songs that collaborative filtering acquisition meets target user's preference are carried out to all songs that the default library includes
As the multiple song to be selected;
According to the multiple song to be selected, generates and show the recommendation song list for the target user.
2. generating and showing the method according to claim 1, wherein described according to the multiple song to be selected
List is sung in recommendation for the target user, comprising:
The multiple song to be selected is ranked up based on preset personalized ordering strategy, determines that First ray song is single;It is described
Personalized ordering strategy include song temperature preference strategy, singer's temperature preference strategy, in song quality preference strategy at least
One;
List is sung according to the First ray, generate and shows the recommendation song list for the target user.
3. according to the method described in claim 2, generating and showing it is characterized in that, described sing single according to the First ray
List is sung in recommendation for the target user, comprising:
Obtain the user behavior data of the target user and/or the context of this search;
The user behavior data includes that the target user listens generally the complete of each song to be selected in the multiple song to be selected
The Information on Collection of rate and/or the target user to each song to be selected in the multiple song to be selected;This search
The keyword that the search instruction that context inputs before including the target user includes;
According to the user behavior data and/or this search context, to the First ray song singly include it is multiple to
It selects song to carry out minor sort again, it is single to obtain the second sequence song;
List is sung according to second sequence, generate and shows the recommendation song list for the target user.
4. according to the method described in claim 3, generating and showing it is characterized in that, described sing single according to second sequence
List is sung in recommendation for the target user, comprising:
Obtain the user of the searching times of each song to be selected and/or each song to be selected in the multiple song to be selected it is complete listen it is general
Rate;
Listen probability to second sequence according to the user of the searching times of each song to be selected and/or each song to be selected is complete
Multiple songs to be selected that column song singly includes carry out minor sort again, generate and show the recommendation song list for the target user.
5. generating and showing the method according to claim 1, wherein described according to the multiple song to be selected
List is sung in recommendation for the target user, comprising:
The multiple song to be selected is screened and/or sorted according to preset feature ordering model, obtains third sequence song
It is single;
The feature ordering model listens probability for the user according to multiple songs is complete, and multiple users believe for the collection of multiple songs
It ceases, the training of at least one characteristic information obtains in the initial Preferences of multiple users and the voiceprint of multiple users
Machine learning model;
List is sung according to the third sequence, generate and shows the recommendation song list for the target user.
6. according to the method described in claim 5, it is characterized in that, the method also includes the training of the feature ordering model
Step:
Determine that each song is for selected institute in the multiple song according to the historical record that multiple users listen attentively to multiple songs
State the characteristic value of each characteristic information at least one characteristic information;
The characteristic value of each characteristic information at least one selected characteristic information is directed to according to each song to described more
A song is screened and/or is sorted, and it is single to obtain standard song;
The multiple song is inputted to the feature ordering model to be trained established based at least one described characteristic information, is obtained
The training song of the feature ordering model output to be trained is single;
According to the single and described training song of standard song, single comparison result adjusts the feature ordering model, and will be the multiple
Song inputs the feature ordering model adjusted, then that the training song of the feature ordering model output adjusted is single
It is singly compared with standard song again, and then the feature ordering mould is adjusted according to the comparison result compared again again
Type, until the training song list that the feature ordering model exports is matched with standard song list.
7. according to the method described in claim 5, it is characterized in that, the method also includes the training of the feature ordering model
Step:
From group of songs corresponding with characteristic information each at least one described characteristic information is obtained in multiple songs respectively, each
Group of songs includes at least one song;
The multiple song is screened and/or sorted according to the corresponding group of songs of each characteristic information, obtains standard
Song is single;
The multiple song is inputted to the feature ordering model to be trained established based at least one described characteristic information, is obtained
The training song of the feature ordering model output to be trained is single;
According to the single and described training song of standard song, single comparison result adjusts the feature ordering model, and will be the multiple
Song inputs the feature ordering model adjusted, then that the training song of the feature ordering model output adjusted is single
It is singly compared with standard song again, and then the feature ordering mould is adjusted according to the comparison result compared again again
Type, until the training song list that the feature ordering model exports is matched with standard song list.
8. according to claim 1 to method described in 7 any one claims, which is characterized in that the method also includes:
Obtaining the clicking rate of each song and/or recommendation song in the multiple songs for recommending song singly to include singly includes
The clicking rate of each label in the corresponding multiple labels of multiple songs;
According to the clicking rate of each song and/or the clicking rate of each label, it is single to adjust the recommendation song.
9. according to the method described in claim 8, it is characterized in that, the clicking rate according to each song and/or every
The clicking rate of a label, adjusting the recommendation song singly includes:
According to the clicking rate of each song and/or the clicking rate of each label, it is small to obtain clicking rate in the multiple song
In or equal to default clicking rate threshold value song to be deleted or multiple labels in clicking rate be less than or equal to default clicking rate threshold
The corresponding song to be deleted of the label of value;
The song to be deleted is deleted from recommendation song list.
10. a kind of song recommendations device characterized by comprising
First obtains module, and for obtaining search instruction, it is that target user determines recommendation that described search instruction, which is used to indicate terminal,
Song is single;
Second obtains module, for including at least one keyword in response to described search instruction, obtains from default library
Multiple songs at least one keyword match are as multiple songs to be selected;
Third obtain module, in response to described search instruction do not include keyword or have not been obtained with it is described at least one
The song of keyword match carries out collaborative filtering acquisition to all songs that the default library includes and meets the target use
Multiple songs of family preference are as the multiple song to be selected;
Display module, for according to the multiple song to be selected, generating and showing the recommendation song list for the target user.
11. device according to claim 10, which is characterized in that the display module includes:
First acquisition submodule, for being ranked up based on preset personalized ordering strategy to the multiple song to be selected, really
It is single to determine First ray song;The personalized ordering strategy includes song temperature preference strategy, singer's temperature preference strategy, song matter
Measure at least one of preference strategy;
First shows submodule, single for being sung according to the First ray, generates and shows the recommendation for the target user
Song is single.
12. device according to claim 11, which is characterized in that described first shows that submodule includes:
First acquisition unit, for obtaining the user behavior data of the target user and/or the context of this search;
The user behavior data includes that the target user listens generally the complete of each song to be selected in the multiple song to be selected
The Information on Collection of rate and/or the target user to each song to be selected in the multiple song to be selected;This search
The keyword that the search instruction that context inputs before including the target user includes;
Second acquisition unit, for the context according to the user behavior data and/or this search, to the First ray
It sings the multiple songs to be selected for singly including and carries out minor sort again, it is single to obtain the second sequence song;
Display unit, it is single for being sung according to second sequence, it generates and shows the recommendation song list for the target user.
13. device according to claim 12, which is characterized in that the display unit is for obtaining the multiple song to be selected
The user of the searching times of each song to be selected and/or each song to be selected is complete in song listens probability;According to each song to be selected
The user of bent searching times and/or each song to be selected is complete to listen probability to sing the multiple songs to be selected for singly including to second sequence
Qu Jinhang minor sort again generates and shows that the recommendation song for the target user is single.
14. device according to claim 10, which is characterized in that the display module includes:
Second acquisition submodule, for according to preset feature ordering model to the multiple song to be selected carry out screening and/or
It is single to obtain third sequence song for sequence;
The feature ordering model listens probability for the user according to multiple songs is complete, and multiple users believe for the collection of multiple songs
It ceases, the training of at least one characteristic information obtains in the initial Preferences of multiple users and the voiceprint of multiple users
Machine learning model;
Second shows submodule, single for being sung according to the third sequence, generates and shows the recommendation for the target user
Song is single.
15. device according to claim 14, which is characterized in that described device further include:
Determining module, the historical record for listening attentively to multiple songs according to multiple users determine each song in the multiple song
For the characteristic value of each characteristic information at least one characteristic information described in selected;
4th obtains module, for being directed to each characteristic information at least one selected characteristic information according to each song
Characteristic value the multiple song is screened and/or is sorted, it is single to obtain standard song;
5th obtains module, to be trained for establishing the input of the multiple song based at least one described characteristic information
Feature ordering model, the training song for obtaining the feature ordering model output to be trained are single;
First training module, for adjusting the feature ordering according to the single comparison result of the single and described training song of standard song
Model, and the multiple song is inputted into the feature ordering model adjusted, then by the feature ordering adjusted
The training song of model output is single to be singly compared with standard song again, and then is adjusted again according to the comparison result compared again
The whole feature ordering model, until the training song list that the feature ordering model exports is matched with standard song list.
16. device according to claim 14, which is characterized in that described device further include:
6th obtain module, for from multiple songs respectively obtain at least one described characteristic information in each characteristic information
Corresponding group of songs, each group of songs include at least one song;
7th obtains module, for being screened according to the corresponding group of songs of each characteristic information to the multiple song
And/or sequence, it is single to obtain standard song;
8th obtains module, to be trained for establishing the input of the multiple song based at least one described characteristic information
Feature ordering model, the training song for obtaining the feature ordering model output to be trained are single;
Second training module, for adjusting the feature ordering according to the single comparison result of the single and described training song of standard song
Model, and the multiple song is inputted into the feature ordering model adjusted, then by the feature ordering adjusted
The training song of model output is single to be singly compared with standard song again, and then is adjusted again according to the comparison result compared again
The whole feature ordering model, until the training song list that the feature ordering model exports is matched with standard song list.
17. device described in 0 to 16 any one claim according to claim 1, which is characterized in that described device is also wrapped
It includes:
9th obtains module, for obtaining the clicking rate of each song and/or institute in the multiple songs for recommending song singly to include
State the clicking rate of each label in the corresponding multiple labels of multiple songs for recommending song singly to include;
Module is adjusted, for adjusting the recommendation song according to the clicking rate of each song and/or the clicking rate of each label
It is single.
18. device according to claim 17, which is characterized in that the adjustment module includes:
Third acquisition submodule, for according to the clicking rate of each song and/or the clicking rate of each label, described in acquisition
Clicking rate is less than or equal to clicking rate in the song to be deleted or multiple labels of default clicking rate threshold value and is less than in multiple songs
Or the corresponding song to be deleted of label equal to default clicking rate threshold value;
Submodule is deleted, for deleting the song to be deleted from recommendation song list.
19. a kind of song recommendations device characterized by comprising
Processor;
Memory for storage processor executable instruction;
Wherein, the processor is configured to:
Search instruction is obtained, it is that target user determines that recommendation song is single that described search instruction, which is used to indicate terminal,;
Include at least one keyword in response to described search instruction, is obtained from default library and at least one described key
The matched multiple songs of word are as multiple songs to be selected;
Do not include keyword in response to described search instruction or have not been obtained song at least one keyword match,
Multiple songs that collaborative filtering acquisition meets target user's preference are carried out to all songs that the default library includes
As the multiple song to be selected;
According to the multiple song to be selected, generates and show the recommendation song list for the target user.
20. a kind of computer readable storage medium, is stored thereon with computer instruction, which is characterized in that the instruction is by processor
The step of claim 1 to 9 any one claim the method is realized when execution.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910580213.9A CN110362711A (en) | 2019-06-28 | 2019-06-28 | Song recommendations method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910580213.9A CN110362711A (en) | 2019-06-28 | 2019-06-28 | Song recommendations method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110362711A true CN110362711A (en) | 2019-10-22 |
Family
ID=68217294
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910580213.9A Pending CN110362711A (en) | 2019-06-28 | 2019-06-28 | Song recommendations method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110362711A (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111753126A (en) * | 2020-06-24 | 2020-10-09 | 北京字节跳动网络技术有限公司 | Method and device for video dubbing |
CN111913593A (en) * | 2020-08-06 | 2020-11-10 | 聚好看科技股份有限公司 | Media data searching method and display equipment |
CN111984847A (en) * | 2020-08-21 | 2020-11-24 | 上海风秩科技有限公司 | Information search method, information search device, storage medium and electronic device |
CN112256233A (en) * | 2020-10-22 | 2021-01-22 | 北京字节跳动网络技术有限公司 | Music playing method and device |
CN112364184A (en) * | 2020-11-12 | 2021-02-12 | 北京达佳互联信息技术有限公司 | Method, device, server and storage medium for ordering multimedia data |
CN113010726A (en) * | 2021-03-22 | 2021-06-22 | 平安科技(深圳)有限公司 | Fole song recommendation method, device, equipment and storage medium |
CN113742513A (en) * | 2021-08-09 | 2021-12-03 | 咪咕互动娱乐有限公司 | Song list adjusting method, device, equipment and computer readable storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101828393A (en) * | 2007-08-24 | 2010-09-08 | 谷歌公司 | Recommendation based on medium |
CN107885745A (en) * | 2016-09-29 | 2018-04-06 | 亿览在线网络技术(北京)有限公司 | A kind of song recommendations method and device |
US20180276540A1 (en) * | 2017-03-22 | 2018-09-27 | NextEv USA, Inc. | Modeling of the latent embedding of music using deep neural network |
CN109785859A (en) * | 2019-01-31 | 2019-05-21 | 平安科技(深圳)有限公司 | The method, apparatus and computer equipment of management music based on speech analysis |
-
2019
- 2019-06-28 CN CN201910580213.9A patent/CN110362711A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101828393A (en) * | 2007-08-24 | 2010-09-08 | 谷歌公司 | Recommendation based on medium |
CN107885745A (en) * | 2016-09-29 | 2018-04-06 | 亿览在线网络技术(北京)有限公司 | A kind of song recommendations method and device |
US20180276540A1 (en) * | 2017-03-22 | 2018-09-27 | NextEv USA, Inc. | Modeling of the latent embedding of music using deep neural network |
CN109785859A (en) * | 2019-01-31 | 2019-05-21 | 平安科技(深圳)有限公司 | The method, apparatus and computer equipment of management music based on speech analysis |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111753126A (en) * | 2020-06-24 | 2020-10-09 | 北京字节跳动网络技术有限公司 | Method and device for video dubbing |
CN111753126B (en) * | 2020-06-24 | 2022-07-15 | 北京字节跳动网络技术有限公司 | Method and device for video dubbing |
CN111913593A (en) * | 2020-08-06 | 2020-11-10 | 聚好看科技股份有限公司 | Media data searching method and display equipment |
CN111913593B (en) * | 2020-08-06 | 2023-07-18 | 聚好看科技股份有限公司 | Media data searching method and display device |
CN111984847A (en) * | 2020-08-21 | 2020-11-24 | 上海风秩科技有限公司 | Information search method, information search device, storage medium and electronic device |
CN112256233A (en) * | 2020-10-22 | 2021-01-22 | 北京字节跳动网络技术有限公司 | Music playing method and device |
US11934632B2 (en) | 2020-10-22 | 2024-03-19 | Beijing Bytedance Network Technology Co., Ltd. | Music playing method and apparatus |
CN112364184A (en) * | 2020-11-12 | 2021-02-12 | 北京达佳互联信息技术有限公司 | Method, device, server and storage medium for ordering multimedia data |
CN112364184B (en) * | 2020-11-12 | 2024-04-30 | 北京达佳互联信息技术有限公司 | Method, device, server and storage medium for ordering multimedia data |
CN113010726A (en) * | 2021-03-22 | 2021-06-22 | 平安科技(深圳)有限公司 | Fole song recommendation method, device, equipment and storage medium |
CN113742513A (en) * | 2021-08-09 | 2021-12-03 | 咪咕互动娱乐有限公司 | Song list adjusting method, device, equipment and computer readable storage medium |
CN113742513B (en) * | 2021-08-09 | 2023-08-15 | 咪咕互动娱乐有限公司 | Song list adjusting method, device, equipment and computer readable storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110362711A (en) | Song recommendations method and device | |
CN108197327B (en) | Song recommendation method, device and storage medium | |
CN109800325A (en) | Video recommendation method, device and computer readable storage medium | |
CN106021496A (en) | Video search method and video search device | |
CN107111648A (en) | Information processing system, information processor, control method and program | |
CN104035995B (en) | Group's label generating method and device | |
CN105335414B (en) | Music recommendation method and device and terminal | |
CN110266879A (en) | Broadcast interface display methods, device, terminal and storage medium | |
CN104615663B (en) | File ordering method, apparatus and terminal | |
CN109145213A (en) | Inquiry recommended method and device based on historical information | |
CN109783656A (en) | Recommended method, system and the server and storage medium of audio, video data | |
CN108227950A (en) | A kind of input method and device | |
CN108600818A (en) | Show the method and device of multimedia resource | |
CN109168062A (en) | Methods of exhibiting, device, terminal device and the storage medium of video playing | |
CN110232137A (en) | A kind of data processing method, device and electronic equipment | |
CN109961787A (en) | Determine the method and device of acquisition end time | |
CN109144285A (en) | A kind of input method and device | |
CN108320208A (en) | Vehicle recommends method and device | |
CN112068711A (en) | Information recommendation method and device of input method and electronic equipment | |
CN110110204A (en) | A kind of information recommendation method, device and the device for information recommendation | |
CN109002184A (en) | A kind of association method and device of input method candidate word | |
CN111753126A (en) | Method and device for video dubbing | |
US20220208156A1 (en) | Method for generating song melody and electronic device | |
CN110019885A (en) | A kind of expression data recommended method and device | |
CN110244860A (en) | A kind of input method, device and electronic equipment |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191022 |
|
RJ01 | Rejection of invention patent application after publication |