Embodiment
As described above, the present inventors have noted that, the feature of itself that releases news that only merely combining information releases newsTo identify the importance of core word, such recognition accuracy is not high.Then the present inventor expects, can combine except information is sent outSuitable auxiliary information outside feature that releases news of cloth user itself optimizes the result of core word importance identification.
The main thought of the application is, in addition to the feature of that releases news of information issue user itself, considers knotThe historical behavior daily record for closing information issue user carrys out the importance that identification information issues the middle core word that releases news of user, so as toImprove the degree of accuracy of importance identification.
The present inventors have noted that the feedback information of information inquiry user is also available important and high quality a letterBreath, the result of core word importance identification can be optimized by such feedback information.Further it is observed that information issue userPersonal information also can be used for being lifted the good foundation of importance recognition accuracy.
To make the purpose, technical scheme and advantage of the application clearer, below in conjunction with drawings and the specific embodiments, to thisApplication is described in further detail.
According to the embodiment of the application one side, there is provided in a kind of issue the releasing news of user for identification informationThe core word importance recognition methods of the importance of core word.
Fig. 1 shows the middle core word that releases news for being used for identification information issue user according to the application one embodimentThe flow chart of the core word importance recognition methods of importance.
As shown in figure 1, at step S110, it is determined that multiple core words in releasing news.
The analysis of such as word segmentation processing and part-of-speech tagging etc, and root can be carried out to information releasing news for user of issueThe core word in releasing news is determined according to pre-defined rule.
In a specific embodiment, word segmentation processing and part-of-speech tagging are carried out for example, can be released news to this, and canSo that the word for being labeled as noun part-of-speech or product word to be defined as to the core word in releasing news.
In one releases news, there may be one or more noun or product words.The application is mainly in oneReleasing news includes the situation of multiple core words.
It is to be herein pointed out it can be determined by any desired manner of known in the art or following exploitationCore word in releasing news, and it is not limited to mode listed above.
It is each in multiple core words according to feature of the core word in releasing news next, at step S120Individual core word assigns corresponding initial weight value Score_initial.
Wherein, the initial weight value can be used for preliminarily identifying importance of the core word in this releases news.
Wherein, the description information for being published the title, attribute and/or details of object etc can be included by releasing news.
Core word can include the frequency that occurs in releasing news of core word in the feature in releasing news（Number）With/Or position feature.
In a specific embodiment, for example, the number that occurs in releasing news of core word is more, weighted value Score_Initial is higher.In addition, for example, if core word is appeared in title description, weighted value Score_initial is high, and such asFruit core word is only present in details description, then weighted value Score_initial is low.These features can be used alone can alsoIt is used in combination.This point is can well to be realized by any desired manner of known in the art or following exploitation, here notRepeat again.
As previously mentioned, the present inventor is exactly it is noted that according only to the spy of itself that releases news in existing schemeLevy to assign each core word weighted value, the importance of multiple core words, such importance are identified by this weighted valueRecognition accuracy is not high, causes the accuracy that this releases news with the correlation calculations of user input query word in information searchIt is not high, so contemplating with reference to other suitable auxiliary informations to adjust the weights of importance of this middle core word that releases newsValue so that improve importance recognition accuracy, be advantageous to search in lifted this release news it is related to user input query wordProperty calculate accuracy.
According to embodiments herein, the historical behavior daily record of user can be issued according to such as information, information inquiry is usedThe feedback information at family, information issue the one or more in the auxiliary information of the personal information of user etc to adjust core wordInitial weight value, so as to improve the degree of accuracy of identification core word importance, as described by with reference to step S130.
At step S130, the historical behavior daily record of user is issued according to information, adjusts each core word accordingly justBeginning weighted value Score_initiali, to obtain corresponding final weight value Score_finali。
In a specific embodiment, for some core word i in an information of information issue user's issue, meterCalculate the number Count_key that the core word occurs in the historical behavior daily record of information issue useriIt is and each in this informationThe number sum ∑ Count_key that core word occurs in the historical behavior daily record of information issue useriRatio Score_keyi, i.e. Score_keyi=Count_keyi/∑Count_keyi, i represent one issue information in i-th of core word.
Wherein, the historical behavior daily record of information issue user can specifically include the keyword purchase day of information issue userWill.The keyword of information issue user's purchase can include participle and participle combination, and participle combination is combined by multiple participles.
In one embodiment, the final weight value Score_final of each core wordiCan be Score_keyiWith it is firstThe weighted sum of beginning weighted value, such as following formula（2）It is shown：
Score_finali=w5*（w7*Score_keyi）+w6*Score_initiali（2）
Wherein, w5、w6And w7Can be the experience weights drawn in an experiment according to experimental result, they can be 0-1 itBetween arbitrary value.
It is described above identifying the importance of core word according to the historical behavior daily record of information issue user.It is actualOn, recognition accuracy can also be improved according further to other appropriate informations of information issue user side.
In one embodiment, the personal information that user can be issued according to information is first accordingly to adjust each core wordBeginning weighted value, to obtain corresponding final weight value.
According to embodiments herein, personal information comprises at least at least one in personal label, summary and regional informationIt is individual.The personal information can be issued in the log-on message of user from information and obtained.For example, log-on message can include such as titleEtc personal label information, the summary infos of remarks etc, the regional information etc. of address etc.
In a specific embodiment, the frequency that can be occurred according to core word in above-mentioned personal information（Number）Or positionPut to adjust the initial weight value of imparting core word.For example, core word is in personal label information, summary info, regional informationOccur more, weighted value can be higher.This situation is similar to situation when considering to release news feature itself.According to hereDisclosure, those skilled in the art can easily realize this point, therefore repeat no more here.
It is described above issuing the historical behavior daily record of user and/or the personal information of information issue user according to informationCome identify release news in each core word importance, it is believed that be to be adjusted according to the information of information issue user side aboveThe whole initial weight value that core word is assigned by the feature of itself that releases news.In fact, can also be according to information inquiry userThe appropriate information of side adjusts the initial weight value.
In one embodiment, can be determined according to the feedback information of information inquiry user in set-up procedure S110 everyThe corresponding initial weight value Score_initial of one core wordi, to obtain corresponding final weight value Score_finali。
Wherein, the feedback information of information inquiry user comprises at least the inquiry and click information, transaction of information inquiry userIt is at least one in information and evaluation behavioural information.The feedback information of these information inquiries user, such as information inquiry userInquiry and click information, click on subsequent transaction information and evaluation behavioural information, can be obtained by network log.Here shouldIt is understood that the different types of feedback information that user can be inquired about with combining information is obtained to adjust using the feature of information issue userThe weighted value gone out, so as to improve the recognition accuracy of core word importance, and then it is accurate to improve the related search to release newsExactness.
In a specific embodiment, can be each to adjust according to the inquiry of information inquiry user and click historical informationThe corresponding initial weight value Score_initial of core wordi, to obtain corresponding final weight value Score_finali.For example,For each core word, can according to the core word in network log certain period of time（Such as 100 days）Interior Query ResultThe number Count_show of middle appearanceiThe number sum ∑ occurred with each in multiple core words in the Query ResultCount_showiRatio Score_showi=Count_showi/∑Count_showiAnd the period（Such as 100 days）The number Count_click that the core word occurs in the Query Result being inside clickediWith each core in the Query Result that is clickedThe number sum ∑ Count_click that heart word occursiRatio Score_clicki=Count_clicki/∑Count_clicki, to adjust initial weight value Score_initialiSo as to obtain final weight value Score_finali, i expressions oneI-th of core word in releasing news.In one embodiment, the final weight value of each core word can be Score_showi、Score_clickiAnd Score_initialiWeighted sum, such as following formula（1）It is shown：
Score_finali=w1*（w3*Score_showi+w4*Score_clicki）+w2*Score_initiali（1）
Wherein, w1、w2、w3And w4Can be preset in an experiment according to experimental result, they can be between 0-1Arbitrary value.
It is described above corresponding to adjust each core word according to the inquiry of information inquiry user and click historical informationInitial weight value.In a similar way, equally can be every to adjust according to subsequent transaction information or evaluation behavioural information is clicked onThe individual corresponding initial weight value of core word, can also be according to the inquiry of information inquiry user and click information, click subsequent transactionInformation, any combination between behavioural information three is evaluated to adjust the corresponding initial weight value of each core word.This area skillArt personnel can realize these schemes according to content disclosed above, therefore for brevity, on their own realization sideFormula, repeat no more here.
Describe in the above embodiments only in conjunction with the historical behavior daily record of information issue user or only in conjunction with letterThe personal information of breath issue user identifies the importance of core word only in conjunction with the feedback information of information inquiry user, thusThe recognition accuracy for the core word importance that releases news can be improved, it is accurate so as to improve the related search to release news and sequenceDegree.It should be understood that the application is not limited to above-described embodiment, but can be known according to any combination in above- mentioned informationThe importance of other core word, the recognition accuracy that so can further improve core word importance are searched with what correlation released newsThe rope degree of accuracy.
For example, in another embodiment, can be inquired about with combining information the inquiry of user and the feedback information of click andInformation issues both historical behavior daily records of user to adjust the corresponding initial weight value Score_ of each core wordinitiali, to obtain corresponding final weight value Score_finali.Such as following formula（3）It is shown：
Score_finali=w1'*（w3'*（w5'*Score_showi+w6'*Score_clicki）+w4'*（Score_keyi））+w2'*Score_initiali（3）
Wherein, w1'、w2'、w3'、w4'、w5' and w6' can be the experience weights drawn in an experiment according to experimental result,They can be between 0-1 arbitrary value.
So far, by according to release news itself feature and obtained the final power of each core word with reference to auxiliary informationWeight values, so as to identify release news in multiple core words each core word importance, it is possible thereby to significantly improve coreThe recognition accuracy of heart word importance.
When current queries user inputs a certain search term, one or more information is searched by the search term, according toThe final weight value of each core word calculates every information and the correlation of the search term in every information, and according to the correlationResult of calculation sorts to described information.
In the embodiment of the present application, when the search term inputted according to information inquiry user scans for, phase can be improvedThe degree of accuracy in the calculating of closing property and irrelevant information filtering, so as to improve the accurate of the related search results ranking to release newsDegree, this point is described in detail with reference to Fig. 2.
Fig. 2 shows the flow chart of the search result ordering method according to the application one embodiment.
As shown in Fig. 2 at step S210, the search term of receive information inquiry user's input.
In one embodiment, the search term of information inquiry user input can be analyzed, to find out the core in the search termHeart word information.In general, search term is shorter character string, by method commonly used in the art or with combining step S110The similar approach of description can well identify core word information therein, for follow-up correlation calculations.
Obviously it will be appreciated that, the application is not limited to above-described embodiment, can not also find out the core word letter in search termBreath, but subsequent step S220 correlation calculations are directly carried out using search term.
Next, at step S220, the importance of the middle core word that releases news based on information issue user, it is determined that hairCloth information and the correlation of search term.
The importance of the wherein middle core word that releases news of information issue user is by above in conjunction with the sheet described by Fig. 1What the method for the identification core word importance of application obtained, its details refers to description above, repeats no more here.
In one embodiment, will can receive in the core word in predetermined release news and step S210Core word information in search term is contrasted, it is determined that the correlation to release news with search term.
Specifically, if one release news in the high core word of final weight value and search term core word informationMatching, it is determined that this releases news higher with search word correlation.If one release news in final weight value it is lowThe core word information matches of core word and search term, it is determined that this releases news relatively low with search word correlation.If oneThe core word information of core word and search term in releasing news all mismatches, it is determined that this release news with search term withoutClose.
Next, at step S230, according to the correlation determined at step S220, correlation is released news and is ranked upAnd show.
It is, can be according to the correlation to release news with search term determined above in step S220, pair with searchingRelated the releasing news of rope word is ranked up and shown.Specifically, can be according to the height of correlation, by the phase with search termBefore higher the releasing news of closing property is shown in, be shown in relatively low the releasing news of the correlation of search term behind, and with searchingUnrelated the releasing news of rope word is not shown.
In other embodiments, can be right according to the final weight value of each core word in obtained above release newsMultiple core words in releasing news carry out importance ranking.Answered for the higher search of correlation requirement or information classification etc.With in scene, only the issue can be just determined with when the core word of importance ranking in releasing news first matches in search termInformation releases news for correlation.
By the above method, according to not only in conjunction with the feature of itself and the history of combining information issue user of releasing newsUser behaviors log identifies the importance for the middle core word that releases news, and is scanned in the search term inputted according to information inquiry userWhen, the degree of accuracy for the correlation calculations that release news can be improved, it is convenient to use so as to improve the related sequence degree of accuracy to release newsThe use at family and the use feeling for lifting user.
Know with the core word importance of the importance of the above-mentioned middle core word that releases news for identification information issue userOther method is similar, and the embodiment of the present application additionally provides the importance of the middle core word that releases news for identification information issue userCore word importance identification equipment.
Fig. 3 shows the middle core word that releases news for being used for identification information issue user according to the application one embodimentThe schematic block diagram of the core word importance identification equipment 300 of importance.
As shown in figure 3, equipment 300 can include core word determining device 310, valuator device 320 and adjusting apparatus 330.
Specifically, core word determining device 310 is determined for multiple core words in releasing news.Valuator device320 can be used for assigning corresponding initial weight according to the feature to release news for each core word in multiple core wordsValue.The historical behavior daily record that adjusting apparatus 330 can be used for issuing user according to information is corresponding to adjust each core wordInitial weight value is to obtain corresponding final weight value.
Set by the importance for being used for the middle core word that releases news that identification information issues user of the embodiment of the present applicationIt is standby, compared to existing technologies, the degree of accuracy of identification core word importance can be significantly improved.
The core word of the importance of the middle core word that releases news described above for identification information issue user is importantProperty identification equipment and describe before be used for identification information issue user the middle core word that releases news importance core wordThe processing of importance recognition methods is corresponding, accordingly, with respect to more detailed ins and outs, may refer to the side described beforeMethod, repeat no more here.
On the other hand, similar with mentioned above searching results sort method, the embodiment of the present application additionally provides search results rankingEquipment, it is described in detail with reference to Fig. 4.
Fig. 4 shows the schematic block diagram of the search results ranking equipment 400 according to the application one embodiment.
As shown in figure 4, equipment 400 can include search term reception device 410, correlation determining device 420 and sequenceWith display device 430.
Specifically, search term reception device 410 can be used for the search term of receive information inquiry user's input.CorrelationDetermining device 420 can be used for based on information issue user the middle core word that releases news importance come determine to release news withThe correlation of search term.The importance of core word wherein in information issue the releasing news of user is by above in conjunction with Fig. 1What the method for the identification core word importance of the application of description obtained.Sequence can be used for according to related to display device 430Property is ranked up and shown to releasing news.
Similarly, by the search results ranking equipment of the embodiment of the present application, the correlation calculations that release news can be improvedThe degree of accuracy, so as to improve the related sequence degree of accuracy to release news, use feeling that is user-friendly and lifting user.
The processing of search result ordering method of the search results ranking equipment described above with describing before be it is corresponding,Accordingly, with respect to more detailed ins and outs, the method described before is may refer to, is repeated no more here.
It will be understood by those skilled in the art that embodiments herein can be provided as method, system or computer program product.Therefore, the application can be using the embodiment in terms of complete hardware embodiment, complete software embodiment or combination software and hardwareForm.Deposited moreover, the application can use to can use in one or more computers for wherein including computer usable program codeStorage media（Including but not limited to magnetic disk storage, CD-ROM, optical memory etc.）The shape of the computer program product of upper implementationFormula.
Embodiments herein is the foregoing is only, is not limited to the application.For those skilled in the artFor, the application can have various modifications and variations.All any modifications made within spirit herein and principle, it is equalReplace, improve etc., it should be included within the scope of claims hereof.