CN105893397A - Video recommendation method and apparatus - Google Patents
Video recommendation method and apparatus Download PDFInfo
- Publication number
- CN105893397A CN105893397A CN201510379649.3A CN201510379649A CN105893397A CN 105893397 A CN105893397 A CN 105893397A CN 201510379649 A CN201510379649 A CN 201510379649A CN 105893397 A CN105893397 A CN 105893397A
- Authority
- CN
- China
- Prior art keywords
- participle
- coefficient
- search word
- recommendation
- historical search
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 58
- 238000010606 normalization Methods 0.000 claims abstract description 23
- 238000013507 mapping Methods 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 10
- 230000015572 biosynthetic process Effects 0.000 claims description 8
- 230000008859 change Effects 0.000 claims description 2
- 230000006870 function Effects 0.000 description 10
- 239000000203 mixture Substances 0.000 description 5
- 239000012141 concentrate Substances 0.000 description 2
- 238000005336 cracking Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/3332—Query translation
- G06F16/3334—Selection or weighting of terms from queries, including natural language queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/73—Querying
- G06F16/735—Filtering based on additional data, e.g. user or group profiles
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
Embodiments of the present invention disclose a video recommendation method and apparatus, which are applied to a server. The method comprises: acquiring a text string input by a user, and identifying at least one word segment included in the text string; based on text information of each identified word segment, determining a text coefficient corresponding to each word segment; separately determining whether each word segment is the same as a historical search word in a preset weighted word library, and if a determination result is yes, weighting the text coefficient corresponding to the word segment by using a weighting coefficient corresponding to the historical search word that is the same as the word segment, so as to obtain a recommendation coefficient corresponding to the word segment; if the determination result is no, weighting the text coefficient corresponding to the word segment by using a pre-determined weighting coefficient, so as to obtain a recommendation coefficient corresponding to the word segment; normalizing the recommendation coefficient of each word segment; and based on a standard recommendation coefficient that is obtained through normalization and corresponds to each word segment, recommending a video to the user in a preset manner. According to the method and apparatus disclosed by the embodiments of the present invention, video recommendation is more accurate, and user experience is better.
Description
Technical field
The present invention relates to Video Applications system, particularly to a kind of video recommendation method and device.
Background technology
Video website is set up and is equipped with the function recommending video to user now, and this function can be supplied to user more
The selection of many associated videos, further increases click volume and the attention rate of video website.
When prior art recommends video, it it is self text envelope of the text string utilizing user to input in the search box
Breath is recommended, and concrete a kind of implementation is: first, it is thus achieved that the text string of user's input；Secondly,
Word frequency-reverse document-frequency (TF-IDF) algorithm is utilized to determine the first weight of each participle in text string；
Again, with reference to the part of speech of each participle in text string, to each participle on above-mentioned first weighted basis
Weighting, obtains text coefficient；Finally, by text coefficient directly as the recommendation coefficient of video, and according to
The size of this recommendation coefficient recommends video to user.In above-mentioned specific implementation, general acquiescence noun
Weight coefficient is more than verb or adjectival weight coefficient, but when the most famous in the text string of user's input
Word has again verb, and the expective right of verb is great when noun, such as: " cracking of password " this text string,
Mainly " cracking " desired by user, the expective right that i.e. verb " cracks " is great in noun " password ",
According to above-mentioned specific implementation, part of speech weighting is misfitted with user's expection, is caused the recommendation system finally obtained
Number is inaccurate, and then the most inaccurate to the video of user's recommendation, and user's impression is the best.
Summary of the invention
Based on the problems referred to above, the embodiment of the invention discloses a kind of video recommendation method and device, so that pushing away
The video recommended is more accurate.Technical scheme is as follows:
Embodiments provide a kind of video recommendation method, be applied to server, may include that
Obtain the text string of user's input, and identify at least one participle included in described text string；
Self text message based on each participle identified, determines the text coefficient that each participle is corresponding；
Judge that each participle is the most identical with the historical search word in default weighting dictionary respectively, if it is determined that
Result is yes, utilizes weight coefficient historical search word corresponding to identical with described participle, to described participle
Corresponding text coefficient weighting, obtains the recommendation coefficient that described participle is corresponding；If it is judged that be no, profit
Use predetermined weight coefficient, described participle correspondence text coefficient is weighted, obtains the recommendation system that described participle is corresponding
Number, wherein, the weight coefficient corresponding to described historical search word is more than described predetermined weight coefficient；
The recommendation coefficient of each participle is normalized, obtains the standard recommendation system that each participle is corresponding
Number；
Based on the standard recommendation coefficient corresponding to each participle, recommend video according to predetermined manner to user.
Optionally, the forming process of the weighting dictionary preset, including:
Obtain multiple historical search word and volumes of searches corresponding to each historical search word；
Volumes of searches corresponding for each historical search word is normalized, to obtain each historical search word
Corresponding weight coefficient, using the set of each historical search word and the weight coefficient of correspondence thereof as weighting dictionary.
Optionally, described volumes of searches corresponding for each historical search word is normalized, every to obtain
The weight coefficient that individual historical search word is corresponding, including:
Use min-max standardized method, volumes of searches corresponding for each historical search word is mapped to 0.0～1.0 models
In enclosing, obtain primary weight coefficient；
In the range of using sigmoid function that described primary weight coefficient is mapped to 0.5～1.0, obtain each history
The weight coefficient that search word is corresponding.
Optionally, described predetermined weight coefficient is more than 0 and to be less than the designated value in the range of 0.5.
Optionally, the described recommendation coefficient to each participle is normalized, and obtains each participle corresponding
Standard recommendation coefficient, including:
Using all participles recommend the value of quadratic sum evolution of coefficient as denominator, with the recommendation system of each participle
Number, as molecule, calculates the standard recommendation coefficient that each participle is corresponding.
The embodiment of the present invention additionally provides a kind of video recommendations device, is applied to server, may include that knowledge
Other unit, determine unit, weighted units, normalization unit and recommendation unit；Wherein,
Described recognition unit, for obtaining the text string of user's input, and identifies in described text string and is wrapped
At least one participle included；
Described determine unit, self text message of each participle for identifying based on described recognition unit,
Determine the text coefficient that each participle is corresponding；
Described weighted units, for judging whether each participle is searched with the history in default weighting dictionary respectively
Rope word is identical, if it is judged that be yes, utilizes adding corresponding to the historical search word identical with described participle
Weight coefficient, the text coefficient weighting corresponding to described participle, obtain the recommendation coefficient that described participle is corresponding；As
Really judged result is no, utilizes predetermined weight coefficient, weights described participle correspondence text coefficient, obtains institute
Stating the recommendation coefficient that participle is corresponding, wherein, the weight coefficient corresponding to described historical search word is more than described pre-
Determine weight coefficient；
Described normalization unit, the recommendation coefficient of each participle for obtaining described weighted units is returned
One change processes, and obtains the standard recommendation coefficient that each participle is corresponding；
Described recommendation unit, for based on the standard obtained by the described normalization unit corresponding to each participle
Recommend coefficient, recommend video according to predetermined manner to user.
Optionally, also include: dictionary forms unit, and wherein, described dictionary forms unit, including: obtain
Subelement and formation subelement,
Described acquisition subelement, for obtaining multiple historical search word and each corresponding searching of historical search word
Suo Liang；
Described formation subelement, for volumes of searches corresponding for each historical search word is normalized,
The weight coefficient corresponding to obtain each historical search word, by each historical search word and the weighting system of correspondence thereof
The set of number is as weighting dictionary.
Optionally, described formation subelement, including: the first mapping block, the second mapping block；Wherein,
Described first mapping block, is used for using min-max standardized method, and each historical search word is corresponding
Volumes of searches be mapped to 0.0～1.0 in the range of, obtain primary weight coefficient；
Described second mapping block, is used for using sigmoid function to be mapped to by described primary weight coefficient
In the range of 0.5～1.0, obtain the weight coefficient that each historical search word is corresponding, by each historical search word and
The set of corresponding weight coefficient is as weighting dictionary.
Optionally, described predetermined weight coefficient is more than 0 and to be less than the designated value in the range of 0.5.
Optionally, described normalization unit, specifically for: open with the quadratic sum recommending coefficient of all participles
The value of side as denominator, using the recommendation coefficient of each participle as molecule, calculates each participle corresponding
Standard recommendation coefficient.
In the embodiment of the present invention, server obtains the text string of user's input, and identifies in described text string
At least one included participle；Self text message based on each participle identified, determines each point
The text coefficient that word is corresponding；Judge respectively each participle whether with the historical search word in default weighting dictionary
Identical, if it is judged that be yes, utilize weighting system historical search word corresponding to identical with described participle
Number, the text coefficient weighting corresponding to described participle, obtain the recommendation coefficient that described participle is corresponding；If sentenced
Disconnected result is no, utilizes predetermined weight coefficient, weights described participle correspondence text coefficient, obtains described point
The recommendation coefficient that word is corresponding；The recommendation coefficient of each participle of normalization；Each participle obtained based on normalization
Corresponding standard recommendation coefficient, recommends video according to predetermined manner to user.Compared with prior art, originally
On the text coefficient basis of the part participle that text string that inventive embodiments inputs user is corresponding, according to weighting
Weight coefficient in storehouse, weights text coefficient, and another part participle weights according to pre-determined factor, Jin Ergen
Carry out recommending video according to the recommendation coefficient after weighting.Because the weight coefficient in weighting storehouse is according to numerous users
The factor such as historical search amount determine, so, the text coefficient of participle is directly determined by hinge structure
For recommending coefficient, the recommendation coefficient after weighting is truer, it is possible to reflect the temperature of participle more accurately.Profit
The video recommended with such recommendation coefficient is the most accurate, more conforms to the wish of user, and user experiences
More preferably.
Accompanying drawing explanation
In order to be illustrated more clearly that the embodiment of the present invention or technical scheme of the prior art, below will be to enforcement
In example or description of the prior art, the required accompanying drawing used is briefly described, it should be apparent that, describe below
In accompanying drawing be only some embodiments of the present invention, for those of ordinary skill in the art, do not paying
On the premise of going out creative work, it is also possible to obtain other accompanying drawing according to these accompanying drawings.
The flow chart of a kind of video recommendation method that Fig. 1 is provided by the embodiment of the present invention；
Fig. 2 is the substantially mapping scattergram of primary weight coefficient；
Fig. 3 is the substantially mapping scattergram of weight coefficient；
A kind of structural representation of a kind of video recommendations device that Fig. 4 is provided by the embodiment of the present invention；
The another kind of structural representation of a kind of video recommendations device that Fig. 5 is provided by the embodiment of the present invention；
The another kind of structural representation of a kind of video recommendations device that Fig. 6 is provided by the embodiment of the present invention.
Detailed description of the invention
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clearly
Chu, be fully described by, it is clear that described embodiment be only a part of embodiment of the present invention rather than
Whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art are not making creation
The every other embodiment obtained under property work premise, broadly falls into the scope of protection of the invention.
Embodiments provide a kind of video recommendation method and device, be applied to server, described method
Comprise the following steps:
Obtain the text string of user's input, and identify at least one participle included in described text string；
Self text message based on each participle identified, determines the text coefficient that each participle is corresponding；
Judge that each participle is the most identical with the historical search word in default weighting dictionary respectively, if it is judged that
It is yes, utilizes weight coefficient historical search word corresponding to identical with described participle, corresponding to described participle
Text coefficient weighting, obtain the recommendation coefficient that described participle is corresponding；If it is judged that be no, utilize pre-
Determine weight coefficient, described participle correspondence text coefficient weighted, obtains the recommendation coefficient that described participle is corresponding,
Wherein, the weight coefficient corresponding to described historical search word is more than described predetermined weight coefficient；
The recommendation coefficient of each participle is normalized, obtains the standard recommendation system that each participle is corresponding
Number；
Based on the standard recommendation coefficient corresponding to each participle, recommend video according to predetermined manner to user.
It should be noted that a kind of video recommendation method that the embodiment of the present invention is provided is applied to server,
User is typically by the search box input text string of the software such as search engine, video website, and above-mentioned server is i.e.
For the server that the software such as search engine, website is corresponding, text string can be title or the mark of movie and television play
Sign, it is also possible to be the description to movie and television play.
Compared with prior art, the literary composition of the part participle that the embodiment of the present invention inputs user text string is corresponding
On this coefficient basis, according to the weight coefficient in weighting storehouse, weighting text coefficient, another part participle is pressed
Weight according to pre-determined factor, and then carry out recommending video according to the recommendation coefficient after weighting, because in weighting storehouse
Weight coefficient is that the factors such as the historical search amount according to numerous users determine, so, hinge structure is straight
Connecing and be defined as the text coefficient of participle recommending coefficient, the recommendation coefficient after weighting is truer, it is possible to more accurate
The temperature of true reflection participle, the video utilizing such recommendation coefficient to recommend is the most accurate, more accords with
Share the wish at family, user experiences more preferably.
A kind of flow chart of a kind of video recommendation method that Fig. 1 is provided by the embodiment of the present invention, the method should
For server, may comprise steps of:
S101, it is thus achieved that the text string of user's input, and identify in described text string included at least one
Participle；
It should be noted that the text string of above-mentioned user's input can be individual character, word, it is also possible to be to retouch
Stating one section of word of character, text string can be divided at least one participle, wherein, this at least one participle
Part of speech can be noun, verb or adjective.
Concrete, when user is by the search box input text string of the software with function of search, should be with
The text string of the software receipt user input of function of search, and text string is sent to server, then press
According to corresponding segmenting method, text string is divided into some words.Conventional a kind of segmenting method is: should
Text string goes to compare with the participle in locally stored dictionary, obtains several participles that text string is corresponding；
It is of course also possible to utilize other segmenting methods known in those skilled in the art, the embodiment of the present invention is not
This is defined.If simply an individual character or single word in text string, server identify only
It it is a participle；If text string is one section of description, what server identified is then several participles.
S102, self text message based on each participle identified, determine the text that each participle is corresponding
Coefficient；
It should be noted that self text message above-mentioned refers to the expressing information of the various dimensions for participle,
This dimension of participle part of speech such as can be utilized to express this participle, it is also possible to utilize participle word to anticipate this dimension table
Reach this participle, it is, of course, also possible to utilize this participle of feature representation of other dimension.Self text envelope of participle
Breath is only relevant with participle itself, unrelated with the linguistic context residing for participle, source etc..
Concrete, in S101 step, after identifying at least one participle included by text string, each participle
Self text message can apparent, utilize self text message of each participle, it may be determined that each participle
Corresponding text coefficient.Text coefficient can be that self text message based on participle is precalculated,
Directly invoke when needed, it is also possible to calculate temporarily.Determine that the mode of text coefficient is a lot, letter here
Single one of introducing: utilize word frequency-reverse document-frequency (TF-IDF) algorithm and the method for part of speech weighting, determine
The text coefficient of each participle.
Wherein, word frequency-reverse document-frequency (TF-IDF) algorithm is a kind of statistical method, in order to assess one
Words is for the significance level of a copy of it file in a file set or a corpus.Words important
Property is directly proportional increase along with the number of times that it occurs hereof, but can occur in corpus along with it simultaneously
Frequency is inversely proportional to decline.Part of speech is the character of participle, such as: the participle of noun character, verb character
Participle etc..
Describe for convenience, here clearly to determine the process of its text coefficient, this point for any one participle
Word is relevant with a video content participle.Utilize word frequency-reverse document-frequency (TF-IDF) algorithm and word
Property weighting method, determine the concretely comprising the following steps of text coefficient of any one participle: for the literary composition identified
Any one participle in this string, adds up its occurrence number in a default videotext, then divided by
All words sum of this videotext, obtains this participle probability in this videotext, is TF；System
Count and concentrate at default videotext, the videotext quantity of this participle occurs, this videotext concentrates
Business, divided by there is the videotext quantity of this participle, is then taken the logarithm by all videotexts, the numerical value obtained
For IDF；Last TF is multiplied by IDF and is the TF-IDF coefficient of this participle；According to default part of speech weight coefficient pair
The TF-IDF coefficient weighting obtained, the result obtained is the text coefficient of this participle.The text system of other participles
Number can also be obtained by said method, repeats no more here.
Certainly, the method for above-mentioned word frequency-reverse document-frequency (TF-IDF) algorithm and part of speech weighting only determines that
A kind of common method of participle text coefficient, those skilled in the art can also utilize method known in other to determine
The text coefficient of participle, this is not defined by the embodiment of the present invention.
S103, judges that each participle is the most identical with the historical search word in default weighting dictionary respectively, as
Really judged result is yes, performs S104 step；If it is judged that be no, perform S105 step；
S104, utilizes weight coefficient historical search word corresponding to identical with described participle, to described participle
Corresponding text coefficient weighting, obtains the recommendation coefficient that described participle is corresponding；
S105, utilizes predetermined weight coefficient, weights described participle correspondence text coefficient, obtains described participle
Corresponding recommendation coefficient；
Wherein, the weight coefficient corresponding to described historical search word is more than described predetermined weight coefficient；
It should be noted that above-mentioned weighting dictionary at least includes herein below: historical search word and its correspondence
Weight coefficient, this historical search word is user's search word in historical time section, according to weight coefficient
Coefficient after the volumes of searches normalization of historical search word.This weighting dictionary can real-time update, it is provided that to user
Numerical value accurately, and most of word that user searched for appears in this weighting dictionary, certainly, big portion
The weight coefficient of participle language is different.
Concrete, each participle that S101 step identifies compares with the historical search word in this weighting storehouse,
Judge whether this weighting storehouse exists the historical search word identical with the participle identified, it is judged that after terminating, right
In there is the participle of identical historical search word, according to weight coefficient corresponding to identical historical search word to by
The text coefficient weighting that S102 step determines, obtains the recommendation coefficient of such participle；Identical for not existing
The participle of historical search word, according to the text coefficient weighting to being determined by S102 step of the predetermined weight coefficient,
Recommendation coefficient to such participle.Generally, the historical search word in weighting dictionary is that user searches for
The word that frequency ratio is higher, so weight coefficient corresponding to historical search word is more than predetermined weight coefficient.
Below the forming process in weighted words storehouse is described in detail.The forming process of the weighting dictionary preset,
May comprise steps of:
Obtain multiple historical search word and volumes of searches corresponding to each historical search word；
Volumes of searches corresponding for each historical search word is normalized, to obtain each historical search word
Corresponding weight coefficient, using the set of each historical search word and the weight coefficient of correspondence thereof as weighting dictionary.
It should be noted that the historical search word of user is stored in specific file, corresponding historical search word
Corresponding volumes of searches can be added up and be stored in this specific file, certainly, and the search that historical search word is corresponding
Amount can not also store, but calculated when needed temporarily.Server can directly obtain from this specific file
Take this historical search word, obtain the volumes of searches of correspondence.For Baidu user searches for record, historical search
Word can periodically write in the tables of data of Long Yuan, and server can obtain history in real time from the tables of data of this Long Yuan and search
Rope word, obtains the volumes of searches of correspondence, it is also possible to periodically obtained from the tables of data of this Long Yuan by timed task
Historical search word, obtains the volumes of searches of correspondence.
Concrete, obtain multiple historical search word and volumes of searches corresponding to each historical search word, the most right
The volumes of searches that each historical search word is corresponding is normalized, and obtains each corresponding adding of historical search word
Weight coefficient, the weight coefficient of these all historical search words obtained and correspondence collectively forms weighting dictionary.
In reality, historical search word is probably the word in different field, so the volumes of searches of correspondence just may be used
Can have different dimensions, so be unfavorable for comparing and analyzing, so often by search corresponding for historical search word
Amount is normalized, and normalized method is numerous, and conventional has two kinds: min-max Standardization Act and Z-score
Standardized method, because both approaches is technology known in those skilled in the art, thus the most right
Min-max Standardization Act is simply introduced down, and Z-score standardized method is no longer introduced, and can refer to correlation technique literary composition
Shelves are known.
Utilize min-max Standardization Act, normalized is divided into data prediction and sigmoid function normalization
Two steps are carried out.Then at the second step of the default forming process weighting dictionary: described by each history
The volumes of searches that search word is corresponding is normalized, the weight coefficient corresponding to obtain each historical search word,
May comprise steps of:
A uses min-max standardized method, and volumes of searches corresponding for each historical search word is mapped to 0.0～1.0
In the range of, obtain primary weight coefficient；
In the range of b uses sigmoid function that described primary weight coefficient is mapped to 0.5～1.0, obtain each going through
The weight coefficient that history search word is corresponding.
Concrete, in a step, use formula (1) that the volumes of searches that historical search word is corresponding is calculated,
In the range of volumes of searches is mapped to 0.0～1.0, obtain primary weight coefficient；
Wherein x is the volumes of searches that historical search word is corresponding, and m is the primary weight coefficient obtained.
In b step, the primary weight coefficient using formula (2) to obtain a step calculates, by primary
In the range of weight coefficient is mapped to 0.5～1.0, obtain the weight coefficient that each historical search word is corresponding.
Wherein m is primary weight coefficient, and n is the weight coefficient finally obtained.
Illustrate, in the specific file of server, there is four pairs of historical search words and corresponding volumes of searches, point
The most as follows:
Wherein, left column is historical search word, and the right side is classified as the volumes of searches of correspondence.
The substantially mapping scattergram of the primary weight coefficient obtained after a step is as in figure 2 it is shown, wherein, horizontal
Axle represents volumes of searches, and the longitudinal axis is the primary weight coefficient obtained after min-max standardization.Through b step
After the substantially mapping scattergram of weight coefficient that obtains as it is shown on figure 3, wherein, transverse axis represents primary weight coefficient,
The longitudinal axis is the weight coefficient obtained after sigmoid function normalization.
Through above-mentioned steps, the weighting dictionary of formation includes herein below:
Wherein, left column is historical search word, and the right side is classified as the weight coefficient of correspondence.Thus by historical search word
In the range of corresponding volumes of searches is normalized to 0.5～1.0, thus obtain weighting number of times.
Generally, the historical search word in weighting dictionary is the word that user's search rate is higher,
So weight coefficient corresponding to historical search word is more than predetermined weight coefficient.Above-mentioned weight coefficient is at 0.5～1.0 models
In enclosing, therefore described predetermined weight coefficient can be set as the designated value in the range of more than 0 and less than 0.5.Assume
Predetermined weight coefficient is 0.4, then be then to the participle identical with historical search word, according to right in weighting dictionary
Answer weight coefficient weighting, this weight coefficient is all the numerical value between 0.5～1.0, and other search less than with
The participle that historical search word is identical, then according to 0.4 weighting.
S106, is normalized the recommendation coefficient of each participle, obtains the standard that each participle is corresponding
Recommend coefficient；
Concrete, S103 step obtains the recommendation coefficient that participle is corresponding, and this recommendation coefficient is at text coefficient base
The numerical value after the weight coefficient of correspondence it is multiplied by plinth.The text coefficient obtained based on participle self text message is not
There is uniformity, accordingly, it is recommended that coefficient does not the most have uniformity, it is impossible to directly compare.Therefore this
They are normalized by Shi Changchang, so that recommending coefficient to be at the same order of magnitude, the most more having can
Ratio property, the standard recommendation coefficient that after normalization, available each participle is corresponding.
Below a kind of method for normalizing is sketched, then, in this step, the recommendation coefficient of each participle is entered
Row normalized, obtains the standard recommendation coefficient that each participle is corresponding, may include that
Using all participles recommend the value of quadratic sum evolution of coefficient as denominator, with the recommendation system of each participle
Number, as molecule, calculates the standard recommendation coefficient that each participle is corresponding.
Concrete, using the value of the quadratic sum evolution recommending coefficient of all participles as denominator, with each participle
Recommendation coefficient as molecule, the standard recommendation coefficient that each participle is corresponding can be drawn.Illustrate, literary composition
The recommendation coefficient of all participles that this string is corresponding is successively: 0.5,1.2, then according to after this method for normalizing
To standard recommendation coefficient be then followed successively by:Approximate the most successively: 0.4,
0.9。
S107, based on the standard recommendation coefficient corresponding to each participle, recommends to regard to user according to predetermined manner
Frequently.
Concrete, after S106 step obtains standard recommendation coefficient, recommend video according to predetermined manner to user.
This predetermined manner is the method chosen standard recommendation coefficient according to predetermined rule, it is recommended that during video, permissible
Choose participle corresponding to maximum standard recommendation coefficient to go to recommend video, it is also possible to the top being sized
The participle that standard recommendation coefficient is corresponding together decides on the video of recommendation.Such as: the mark of the participle that text string is corresponding
Quasi-recommendation coefficient is followed successively by: 0.51,0.49,0.54,0.88,0,78, if predetermined manner is to choose maximum
Standard recommendation coefficient recommends video, then in this example, be then to recommend video according to the participle of 0.88 correspondence.
In the embodiment of the present invention, server obtains the text string of user's input, and identifies in described text string
At least one included participle；Self text message based on each participle identified, determines each point
The text coefficient that word is corresponding；Judge respectively each participle whether with the historical search word in default weighting dictionary
Identical, if it is judged that be yes, utilize weighting system historical search word corresponding to identical with described participle
Number, the text coefficient weighting corresponding to described participle, obtain the recommendation coefficient that described participle is corresponding；If sentenced
Disconnected result is no, utilizes predetermined weight coefficient, weights described participle correspondence text coefficient, obtains described point
The recommendation coefficient that word is corresponding；The recommendation coefficient of each participle of normalization；Each participle obtained based on normalization
Corresponding standard recommendation coefficient, recommends video according to predetermined manner to user.Compared with prior art, originally
On the text coefficient basis of the part participle that text string that inventive embodiments inputs user is corresponding, according to weighting
Weight coefficient in storehouse, weights text coefficient, and another part participle weights according to pre-determined factor, Jin Ergen
Carry out recommending video according to the recommendation coefficient after weighting.Because the weight coefficient in weighting storehouse is according to numerous users
The factor such as historical search amount determine, so, the text coefficient of participle is directly determined by hinge structure
For recommending coefficient, the recommendation coefficient after weighting is truer, it is possible to reflect the temperature of participle more accurately.Profit
The video recommended with such recommendation coefficient is the most accurate, more conforms to the wish of user, and user experiences
More preferably.
Corresponding to said method embodiment, the embodiment of the present invention additionally provides a kind of video recommendations device, such as Fig. 4
Shown in, this device may include that recognition unit 210, determines unit 220, weighted units 230, normalization list
Unit 220 and recommendation unit 250；Wherein,
Described recognition unit 210, for obtaining the text string of user's input, and identifies institute in described text string
Including at least one participle；
Described determine unit 220, self literary composition of each participle for identifying based on described recognition unit 210
This information, determines the text coefficient that each participle is corresponding；
Described weighted units 230, for judge respectively each participle whether with the history in default weighting dictionary
Search word is identical, if it is judged that be yes, utilizes corresponding to the historical search word identical with described participle
Weight coefficient, the text coefficient weighting corresponding to described participle, obtain the recommendation coefficient that described participle is corresponding；
If it is judged that be no, utilize predetermined weight coefficient, described participle correspondence text coefficient is weighted, obtains
The recommendation coefficient that described participle is corresponding, wherein, the weight coefficient corresponding to described historical search word is more than described
Predetermined weight coefficient；
Described normalization unit 240, for the recommendation coefficient to each participle that described weighted units 230 obtains
It is normalized, obtains the standard recommendation coefficient that each participle is corresponding；
Described recommendation unit 250, obtained by based on the described normalization unit 240 corresponding to each participle
Standard recommendation coefficient, recommend video according to predetermined manner to user.
In the embodiment of the present invention, server obtains the text string of user's input, and identifies in described text string
At least one included participle；Self text message based on each participle identified, determines each point
The text coefficient that word is corresponding；Judge respectively each participle whether with the historical search word in default weighting dictionary
Identical, if it is judged that be yes, utilize weighting system historical search word corresponding to identical with described participle
Number, the text coefficient weighting corresponding to described participle, obtain the recommendation coefficient that described participle is corresponding；If sentenced
Disconnected result is no, utilizes predetermined weight coefficient, weights described participle correspondence text coefficient, obtains described point
The recommendation coefficient that word is corresponding；The recommendation coefficient of each participle of normalization；Each participle obtained based on normalization
Corresponding standard recommendation coefficient, recommends video according to predetermined manner to user.Compared with prior art, originally
On the text coefficient basis of the part participle that text string that inventive embodiments inputs user is corresponding, according to weighting
Weight coefficient in storehouse, weights text coefficient, and another part participle weights according to pre-determined factor, Jin Ergen
Carry out recommending video according to the recommendation coefficient after weighting.Because the weight coefficient in weighting storehouse is according to numerous users
The factor such as historical search amount determine, so, the text coefficient of participle is directly determined by hinge structure
For recommending coefficient, the recommendation coefficient after weighting is truer, it is possible to reflect the temperature of participle more accurately.Profit
The video recommended with such recommendation coefficient is the most accurate, more conforms to the wish of user, and user experiences
More preferably.
As it is shown in figure 5, the another kind of structural representation of a kind of video recommendations device provided by the embodiment of the present invention
Figure, compared with Fig. 4 shown device, Fig. 5 shown device also includes: dictionary forms unit, wherein, institute's predicate
Storehouse forms unit, including: obtain subelement 310a and form subelement 310b,
Described acquisition subelement 310a, is used for obtaining multiple historical search word and each historical search word is corresponding
Volumes of searches；
Described formation subelement 310b, for being normalized place by volumes of searches corresponding for each historical search word
Reason, the weight coefficient corresponding to obtain each historical search word, by adding of each historical search word and correspondence thereof
The set of weight coefficient is as weighting dictionary.
As shown in Figure 6, the another kind of structural representation of a kind of video recommendations device provided by the embodiment of the present invention
Figure, in Fig. 6 shown device, the formation subelement 310b in Fig. 5 shown device, may include that first reflects
Penetrate module 310b1, the second mapping block 310b2；Wherein,
Described first mapping block 310b1, is used for using min-max standardized method, by each historical search
In the range of the volumes of searches that word is corresponding is mapped to 0.0～1.0, obtain primary weight coefficient；
Described second mapping block 310b2, is used for using sigmoid function to be mapped to by described primary weight coefficient
In the range of 0.3～1.0, obtain the weight coefficient that each historical search word is corresponding, by each historical search word and
The set of corresponding weight coefficient is as weighting dictionary.
On the basis of embodiment illustrated in fig. 6, described predetermined weight coefficient is more than 0 and to be less than in the range of 0.5
Designated value.
On the basis of embodiment illustrated in fig. 4, described normalization unit 240, specifically for: with all participles
The value of the quadratic sum evolution of recommendation coefficient is as denominator, using the recommendation coefficient of each participle as molecule, through meter
Calculate and obtain the standard recommendation coefficient that each participle is corresponding.
For system or device embodiment, owing to it is substantially similar to embodiment of the method, so describe
Fairly simple, relevant part sees the part of embodiment of the method and illustrates.
It should be noted that in this article, the relational terms of such as first and second or the like be used merely to by
One entity or operation separate with another entity or operating space, and not necessarily require or imply these
Relation or the order of any this reality is there is between entity or operation.And, term " includes ", " comprising "
Or its any other variant is intended to comprising of nonexcludability, so that include the mistake of a series of key element
Journey, method, article or equipment not only include those key elements, but also other including being not expressly set out
Key element, or also include the key element intrinsic for this process, method, article or equipment.Do not having
In the case of more restrictions, statement " including ... " key element limited, it is not excluded that including described wanting
Process, method, article or the equipment of element there is also other identical element.
One of ordinary skill in the art will appreciate that all or part of step realizing in said method embodiment
The program that can be by completes to instruct relevant hardware, and described program can be stored in computer-readable
Take in storage medium, the storage medium obtained designated herein, such as: ROM/RAM, magnetic disc, CD etc..
The foregoing is only presently preferred embodiments of the present invention, be not intended to limit protection scope of the present invention.
All any modification, equivalent substitution and improvement etc. made within the spirit and principles in the present invention, are all contained in
In protection scope of the present invention.
Claims (10)
1. a video recommendation method, it is characterised in that be applied to server, including:
Obtain the text string of user's input, and identify at least one participle included in described text string；
Self text message based on each participle identified, determines the text coefficient that each participle is corresponding；
Judge that each participle is the most identical with the historical search word in default weighting dictionary respectively, if it is determined that
Result is yes, utilizes weight coefficient historical search word corresponding to identical with described participle, to described participle
Corresponding text coefficient weighting, obtains the recommendation coefficient that described participle is corresponding；If it is judged that be no, profit
Use predetermined weight coefficient, described participle correspondence text coefficient is weighted, obtains the recommendation system that described participle is corresponding
Number, wherein, the weight coefficient corresponding to described historical search word is more than described predetermined weight coefficient；
The recommendation coefficient of each participle is normalized, obtains the standard recommendation system that each participle is corresponding
Number；
Based on the standard recommendation coefficient corresponding to each participle, recommend video according to predetermined manner to user.
Method the most according to claim 1, it is characterised in that the forming process of default weighting dictionary,
Including:
Obtain multiple historical search word and volumes of searches corresponding to each historical search word；
Volumes of searches corresponding for each historical search word is normalized, to obtain each historical search word
Corresponding weight coefficient, searches the set of weight coefficient of element word and correspondence thereof as weighting dictionary using each history.
Method the most according to claim 2, it is characterised in that described that each historical search word is corresponding
Volumes of searches be normalized, the weight coefficient corresponding to obtain each historical search word, including:
Use min-max standardized method, volumes of searches corresponding for each historical search word is mapped to 0.0～1.0 models
In enclosing, obtain primary weight coefficient；
In the range of using sigmoid function that described primary weight coefficient is mapped to 0.5～1.0, obtain each history
The weight coefficient that search word is corresponding.
Method the most according to claim 3, it is characterised in that described predetermined weight coefficient be more than 0 and
Designated value in the range of less than 0.5.
Method the most according to claim 1, it is characterised in that the described recommendation coefficient to each participle
It is normalized, obtains the standard recommendation coefficient that each participle is corresponding, including:
Using all participles recommend the value of quadratic sum evolution of coefficient as denominator, with the recommendation system of each participle
Number, as molecule, calculates the standard recommendation coefficient that each participle is corresponding.
6. a video recommendations device, it is characterised in that be applied to server, including: recognition unit, really
Cell, weighted units, normalization unit and recommendation unit；Wherein,
Described recognition unit, for obtaining the text string of user's input, and identifies in described text string and is wrapped
At least one participle included；
Described determine unit, self text message of each participle for identifying based on described recognition unit,
Determine the text coefficient that each participle is corresponding；
Described weighted units, for judging whether each participle is searched with the history in default weighting dictionary respectively
Rope word is identical, if it is judged that be yes, utilizes adding corresponding to the historical search word identical with described participle
Weight coefficient, the text coefficient weighting corresponding to described participle, obtain the recommendation coefficient that described participle is corresponding；As
Really judged result is no, utilizes predetermined weight coefficient, weights described participle correspondence text coefficient, obtains institute
Stating the recommendation coefficient that participle is corresponding, wherein, the weight coefficient corresponding to described historical search word is more than described pre-
Determine weight coefficient；
Described normalization unit, the recommendation coefficient of each participle for obtaining described weighted units is returned
One change processes, and obtains the standard recommendation coefficient that each participle is corresponding；
Described recommendation unit, for based on the standard obtained by the described normalization unit corresponding to each participle
Recommend coefficient, recommend video according to predetermined manner to user.
Device the most according to claim 6, it is characterised in that also include: dictionary forms unit, its
In, described dictionary forms unit, including: obtain subelement and form subelement,
Described acquisition subelement, for obtaining multiple historical search word and each corresponding searching of historical search word
Suo Liang；
Described formation subelement, for volumes of searches corresponding for each historical search word is normalized,
The weight coefficient corresponding to obtain each historical search word, by each historical search word and the weighting system of correspondence thereof
The set of number is as weighting dictionary.
Device the most according to claim 7, it is characterised in that described formation subelement, including: the
One mapping block, the second mapping block；Wherein,
Described first mapping block, is used for using min-max standardized method, and each historical search word is corresponding
Volumes of searches be mapped to 0.0～1.0 in the range of, obtain primary weight coefficient；
Described second mapping block, is used for using sigmoid function to be mapped to by described primary weight coefficient
In the range of 0.5～1.0, obtain the weight coefficient that each historical search word is corresponding, by each historical search word and
The set of corresponding weight coefficient is as weighting dictionary.
Device the most according to claim 8, it is characterised in that described predetermined weight coefficient be more than 0 and
Designated value in the range of less than 0.5.
Device the most according to claim 6, it is characterised in that described normalization unit, specifically for:
Using the value of the quadratic sum evolution recommending coefficient of all participles as denominator, make with the recommendation coefficient of each participle
For molecule, calculate the standard recommendation coefficient that each participle is corresponding.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510379649.3A CN105893397B (en) | 2015-06-30 | 2015-06-30 | A kind of video recommendation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510379649.3A CN105893397B (en) | 2015-06-30 | 2015-06-30 | A kind of video recommendation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105893397A true CN105893397A (en) | 2016-08-24 |
| CN105893397B CN105893397B (en) | 2019-03-15 |
Family
ID=57001745
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510379649.3A Active CN105893397B (en) | 2015-06-30 | 2015-06-30 | A kind of video recommendation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105893397B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111475725A (en) * | 2020-04-01 | 2020-07-31 | 百度在线网络技术（北京）有限公司 | Method, apparatus, device, and computer-readable storage medium for searching for content |
| CN111858564A (en) * | 2019-04-28 | 2020-10-30 | 腾讯科技（深圳）有限公司 | Data processing method, service processing method, device, terminal and storage medium |
| CN113807090A (en) * | 2021-08-10 | 2021-12-17 | 三峡大学 | Word sense weighting-based TF-IDF disease characterization word extraction method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102246169A (en) * | 2008-12-15 | 2011-11-16 | 摩托罗拉移动公司 | Assigning an indexing weight to a search term |
| CN103020164A (en) * | 2012-11-26 | 2013-04-03 | 华北电力大学 | Semantic search method based on multi-semantic analysis and personalized sequencing |
| CN103729359A (en) * | 2012-10-12 | 2014-04-16 | 阿里巴巴集团控股有限公司 | Method and system for recommending search terms |
| CN104077327A (en) * | 2013-03-29 | 2014-10-01 | 阿里巴巴集团控股有限公司 | Core word importance recognition method and equipment and search result sorting method and equipment |
| CN104123332A (en) * | 2014-01-24 | 2014-10-29 | 腾讯科技（深圳）有限公司 | Search result display method and device |
| CN104376065A (en) * | 2014-11-05 | 2015-02-25 | 百度在线网络技术（北京）有限公司 | Determination method and device for importance degree of search word |
-
2015
- 2015-06-30 CN CN201510379649.3A patent/CN105893397B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102246169A (en) * | 2008-12-15 | 2011-11-16 | 摩托罗拉移动公司 | Assigning an indexing weight to a search term |
| CN103729359A (en) * | 2012-10-12 | 2014-04-16 | 阿里巴巴集团控股有限公司 | Method and system for recommending search terms |
| CN103020164A (en) * | 2012-11-26 | 2013-04-03 | 华北电力大学 | Semantic search method based on multi-semantic analysis and personalized sequencing |
| CN104077327A (en) * | 2013-03-29 | 2014-10-01 | 阿里巴巴集团控股有限公司 | Core word importance recognition method and equipment and search result sorting method and equipment |
| CN104123332A (en) * | 2014-01-24 | 2014-10-29 | 腾讯科技（深圳）有限公司 | Search result display method and device |
| CN104376065A (en) * | 2014-11-05 | 2015-02-25 | 百度在线网络技术（北京）有限公司 | Determination method and device for importance degree of search word |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111858564A (en) * | 2019-04-28 | 2020-10-30 | 腾讯科技（深圳）有限公司 | Data processing method, service processing method, device, terminal and storage medium |
| CN111475725A (en) * | 2020-04-01 | 2020-07-31 | 百度在线网络技术（北京）有限公司 | Method, apparatus, device, and computer-readable storage medium for searching for content |
| CN111475725B (en) * | 2020-04-01 | 2023-11-07 | 百度在线网络技术（北京）有限公司 | Method, apparatus, device and computer readable storage medium for searching content |
| CN113807090A (en) * | 2021-08-10 | 2021-12-17 | 三峡大学 | Word sense weighting-based TF-IDF disease characterization word extraction method |
| CN113807090B (en) * | 2021-08-10 | 2024-04-30 | 三峡大学 | TF-IDF disease characterization word extraction method based on word sense weighting |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105893397B (en) | 2019-03-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10489399B2 (en) | Query language identification | |
| US7783629B2 (en) | Training a ranking component | |
| CN109815314B (en) | Intent recognition method, recognition device and computer readable storage medium | |
| CN105893476B (en) | Intelligent answer method, knowledge base optimization method and device, Intelligence repository | |
| US8386240B2 (en) | Domain dictionary creation by detection of new topic words using divergence value comparison | |
| CN111797214A (en) | FAQ database-based problem screening method and device, computer equipment and medium | |
| JP5379138B2 (en) | Creating an area dictionary | |
| CN110334356B (en) | Article quality determining method, article screening method and corresponding device | |
| CN106407420B (en) | Multimedia resource recommendation method and system | |
| CN107291939B (en) | Hotel information clustering matching method and system | |
| CN109597874B (en) | Information recommendation method, device and server | |
| CN110968800A (en) | Information recommendation method and device, electronic equipment and readable storage medium | |
| CN110276009B (en) | Association word recommendation method and device, electronic equipment and storage medium | |
| KR101541306B1 (en) | Computer enabled method of important keyword extraction, server performing the same and storage media storing the same | |
| CN110032622B (en) | Keyword determination method, keyword determination device, keyword determination equipment and computer readable storage medium | |
| CN111767393A (en) | Text core content extraction method and device | |
| US20140289260A1 (en) | Keyword Determination | |
| CN111144112A (en) | Text similarity analysis method and device and storage medium | |
| CN107085568A (en) | A kind of text similarity method of discrimination and device | |
| Widiyaningtyas et al. | Sentiment Analysis Of Hotel Review Using N-Gram And Naive Bayes Methods | |
| CN105893397A (en) | Video recommendation method and apparatus | |
| CN105677664A (en) | Compactness determination method and device based on web search | |
| US8005775B2 (en) | System and method for detecting human judgment drift and variation control | |
| CN110909532B (en) | User name matching method and device, computer equipment and storage medium | |
| CN109213937B (en) | Intelligent search method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |