CN105335400A - Method and apparatus for obtaining answer information for questioning intention of user - Google Patents
Method and apparatus for obtaining answer information for questioning intention of user Download PDFInfo
- Publication number
- CN105335400A CN105335400A CN201410350679.7A CN201410350679A CN105335400A CN 105335400 A CN105335400 A CN 105335400A CN 201410350679 A CN201410350679 A CN 201410350679A CN 105335400 A CN105335400 A CN 105335400A
- Authority
- CN
- China
- Prior art keywords
- language material
- user
- reply
- answer
- target classification
- 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
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application discloses a method and an apparatus for obtaining answer information for questioning intention of a user. The method comprises: obtaining historical dialogue records of a first user and multiple second users in an instant communication process; performing user intention identification on question corpora raised by the second users from the historical dialogue records to obtain question corpora in the same user intention and reply corpora given for the question corpora in the same user intention by the first user; for the same user intention, clustering the reply corpora to obtain a plurality of categories, and in a target category with a reply corpus number greater than a preset threshold, calculating weights of the reply corpora becoming category centers in the target category; and according to the weights, determining a central answer of the user intention. Through the method and the apparatus, the manpower and time costs can be further reduced.
Description
Technical field
The application relates to data mining technology field, particularly relates to the method for the enquirement intention acquisition answer information of user and device.
Background technology
Constantly perfect along with e-commerce user behavior database, and the fast development of the technology such as traditional communication, mobile communication, increasing people obtain the commodity needed for oneself by the mode of shopping online, the kind of commodity can relate to the every aspect of people's daily life, for people's life provides a great convenience.
In the process of shopping online, buyer user often needs to carry out some online communications with seller user, such as, a buyer user, after have received commodity, find that color is not liked, or size is improper etc., need to carry out returning goods or exchanging goods, now, this buyer user just can relate to the contact staff of this seller by online communication instrument, link up goods return and replacement matters with contact staff.
Under traditional implementation; seller user generally needs to employ contact staff to complete above-mentioned online communication service specially; cost of labor can be higher, and when occurring that multiple buyer sends advisory message simultaneously, often can cause that buyer user's wait in line phenomenon.In order to solve this problem, some e-commerce user behavior database is served for seller user provides " intelligent robot ", can automatically answer by computing machine the various problems that buyer user sends by this service, finally reach a kind of quick help businessman and complete the routine work that artificial customer service does.But have a crucial problem to need to solve in implementation procedure in this service, that is exactly the intention how making computer system can identify user exactly, and provides apt reply accordingly.Such as, if user says " this clothes is off color, and I will move back ", computer system needs to be understood as " client needs to return goods ", and then provides correct answer.
In prior art, for how accurately identifying that user view proposes some solutions, such as, by analyzing the chat record between user, set up language model, after receiving the current chat language material of user, carry out semantic analysis, the semantic results analyzed is calculated the theme of maximum probability by topic model, and as the intention of this user.
Although how prior art is for identify that user view gives implementation, but about the answer that concrete user view is corresponding, generally need to carry out manual configuration by user or backstage technician, also be, intelligent robot is in the process replacing customer service and buyer user to engage in the dialogue, after automatically identifying the intention of buyer user, just the answer for this intention good for human configuration in advance can be returned to this buyer user, realize and its dialogue.But, in advance manually for the process of each intention configuration answer may expend more manpower and time cost equally.
Therefore, how make computer system can automatic acquisition to answer corresponding to user view, and automatically replying, to save time further and human cost, is the technical matters solved in the urgent need to those skilled in the art.
Summary of the invention
This application provides the method for the enquirement intention acquisition answer information of user and device, manpower and time cost can be saved further.
This application provides following scheme:
Enquirement intention for user obtains a method for answer information, comprising:
Acquisition first user and multiple second user carry out the dialog history record in instant messaging process;
From described dialog history record, user view identification is carried out to the problem language material that each second user proposes, each problem language material comprised under obtaining same user view, and the reply language material obtaining that this first user provides each problem language material under same user view;
For same user view, language material is replied to each and carries out cluster, draw multiple classification, and comprising reply language material number more than in the target classification of preset threshold value, calculating each respectively and replying the weight that language material becomes class center in described target classification;
According to described weight, determine the center answer of this user view.
Enquirement intention for user obtains a device for answer information, comprising:
Dialog history record acquiring unit, for obtaining the dialog history record that first user and multiple second user carry out in instant messaging process;
Reply language material acquiring unit, for carrying out user view identification to the problem language material that each second user proposes from described dialog history record, each problem language material comprised under obtaining same user view, and the reply language material obtaining that this first user provides each problem language material under same user view;
Weight calculation unit, for for same user view, language material is replied to each and carries out cluster, draw multiple classification, and comprising reply language material number more than in the target classification of preset threshold value, calculating each respectively and replying the weight that language material becomes class center in described target classification;
Center answer determining unit, for according to described weight, determines the center answer of this user view.
According to the specific embodiment that the application provides, this application discloses following technique effect:
Pass through the embodiment of the present application, data mining can be carried out to dialog history record by computer program, for each user view automatic acquisition is to corresponding answer, and no longer need user or the backstage technician answer input manual for each user view performs or work is set etc., manpower and time cost can be saved further.
Certainly, the arbitrary product implementing the application might not need to reach above-described all advantages simultaneously.
Accompanying drawing explanation
In order to be illustrated more clearly in the embodiment of the present application or technical scheme of the prior art, be briefly described to the accompanying drawing used required in embodiment below, apparently, accompanying drawing in the following describes is only some embodiments of the application, for those of ordinary skill in the art, under the prerequisite not paying creative work, other accompanying drawing can also be obtained according to these accompanying drawings.
Fig. 1 is the process flow diagram of the method for the intention of the enquirement for the user acquisition answer information that the embodiment of the present application provides;
Fig. 2 is the schematic diagram of the device of the intention of the enquirement for the user acquisition answer information that the embodiment of the present application provides.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present application, be clearly and completely described the technical scheme in the embodiment of the present application, obviously, described embodiment is only some embodiments of the present application, instead of whole embodiments.Based on the embodiment in the application, the every other embodiment that those of ordinary skill in the art obtain, all belongs to the scope of the application's protection.
In the embodiment of the present application, can from immediate communication tool user dialog history record in carry out data mining, obtain out the answer of each user view corresponding, set up " intention-answer " database (also namely preserving the corresponding relation between user view and answer), like this, can be that each user view automatic acquisition is to corresponding answer by computer program, and no longer need user or the backstage technician answer input manual for each user view performs or work is set etc., manpower and time cost can be saved further.Below concrete implementation is introduced in detail.
See Fig. 1, the embodiment of the present application provide firstly the method that a kind of enquirement for user intention obtains answer information, and the method can comprise the following steps:
S101: acquisition first user and multiple second user carry out the dialog history record in instant messaging process;
For " intelligent robot " in E-commerce transaction platform, its main function replaces contact staff, help the various problems that seller user's answer buyer user proposes, before this " intelligent robot " occurs, main is exactly engaged in the dialogue by immediate communication tool by between the contact staff of seller user and buyer user, mainly adopt the mode of " question-response " to engage in the dialogue between the two, and a large amount of dialog history records can be produced in the process.Therefore, the embodiment of the present application just can carry out data mining based on these dialog history records, therefrom gets for different buyer's user views, should provide what kind of answer.Certainly, in actual applications, the method also can be applied to the field that other have similar feature and demand, therefore, in the embodiment of the present application, dialogue two parties in instant messaging is called " first user " and " the second user ", wherein, the second user mainly plays the part of the role of quizmaster in dialog procedure, such as, buyer user in transaction platform, first user then mainly plays the part of the role of answerer in dialog procedure, such as, seller user in transaction platform.
And present inventor finds in the process realizing the application, for the same problem that the second user proposes, the answer that different first users provides is generally different.Such as, for different seller users, because the merchandise items type etc. of each self-sales is different, the partner Courier Service business of respective use may be different, respective return of goods address etc. may be all different, therefore, for the problem that buyer user is identical, the answer provided may be different.Such as, certain buyer's user's query seller user " be may I ask and what express delivery can be used to carry out deliver goods ", the answer that seller user A provides may be " express delivery first ", the answer that seller user B provides may be " express delivery second ", the answer that seller user C provides may be " acquiescence sends out express delivery first; also can select express delivery second ", etc.Therefore, in the embodiment of the present application, the automatic extraction of answer can be carried out respectively for different first users, therefore, when obtaining dialog history record, also can extract respectively in units of first user, for same first user, the dialog history record between this first user and each the second user can be extracted.Such as, first user A carried out dialogue in the past and between second user B, C, D, therefore, just can get the dialog history record between this first user A and second user B, C, D.Certainly, in actual applications, the dialog history record that can to extract in a period of time (such as nearest a week interior, in month etc.) is analyzed.Based on these data, for this first user A excavates the answer for various second user view, like this, " intelligent robot " that first user A holds, after the enquirement receiving other the second users, just can make answer based on these answers.
S102: from described dialog history record, user view identification is carried out to the problem language material that each second user proposes, each problem language material comprised under obtaining same user view, and the reply language material obtaining that this first user provides each problem language material under same user view;
As mentioned before, in dialog history record, second user is mainly as enquirement side, therefore, its dialogue language material is all generally problem language material, but the term of the second user in dialog procedure has very high randomness, for same intention, the concrete syntax expression way of asking questions may be diversified.Such as, be want to ask which kind of express delivery mode seller user uses equally, some buyer users may ask " using what express delivery ", and what have may ask " express delivery with which " etc.If all extract corresponding answer for all problems; then the scale of final " problem-answer " database (also namely preserving the corresponding relation between question and answer) generated can be very huge, and is also unfavorable for follow-up use of puing question to other users of answer.If such as follow-up problem changes again the mode of enquirement, just may cannot inquire corresponding answer from this database, accordingly, " intelligent robot " just cannot answer this problem, now, may still need to be answered by artificial mode.
For this reason, in the embodiment of the present application, in the process of carrying out data mining, just can first for the problem language material of the second user in dialog history record, carry out the identification of user view, like this, make to express identical intention but be the use of the problem language material of different language expression-form, be expressed as user view with being structured, multiple problem language material can be comprised under same user view, then be the reply language material that these problem language materials provide based on first user, analyze the best or most suitable answer that can answer this user view, and can add in " intention-answer " database.Such as, dialog history record comprises following problem language material: " using what express delivery ", " express delivery with which " etc., now, these problem language materials can be identified as " inquiry postal " this user view, accordingly, under " inquiry postal " this user view, above-mentioned each problem language material can be comprised.Follow-up " intelligent robot " is when the problem of answer second user, also can first according to the question text of second user's input, identify the user view of this user, then from " intention-answer " database, extract answer corresponding to this user view, and send to this second user.Like this, even if the second user is when puing question to, Expression of language there occurs change again, as long as user view can be identified from its language, just still can automatically provide suitable answer, reducing the probability of manual intervention.It should be noted that, about the identification specifically how carrying out user view, the implementation in prior art can be adopted, do not belong to the emphasis of the embodiment of the present application, therefore, no longer describe in detail here.
User view identification is being carried out to problem language material, and after the problem language material comprised under determining each user view, the reply language material that first user provides for these problem language materials just can be extracted from dialog history record, by replying the analysis of language material to these, excavate the most applicable center answer of answering corresponding user view.Such as, for user view M, the problem language material wherein comprising the second user has problem language material x, y, z, now, just can from user session record, extract the reply language material that first user provides for problem language material x, y, z, by replying the analysis of language material to these several, determine the center answer of the most applicable this user view M of answer.
Wherein, after determining each problem language material comprised under a certain user view, from dialog history record, specifically how to extract the answer language material that first user provides for these problem language materials, multiple implementation can be had.Such as, wherein under a kind of mode, consider that dialogue both sides adopt the mode of question-response to engage in the dialogue, that is, the second user is after proposition problem, and and then first user just can make corresponding reply.Certainly, this belongs to more satisfactory state, also in some cases, when the second user proposes multiple problem continuously, may there are some entanglements in the order of the reply that first user provides, such as, Article 2 is replied may for replying first problem, etc.But be no matter in the ideal situation or nonideality, the reply that first user provides for the problem of the second user is all after first user asks a question generally, and be reply at several closer apart from this problem within.Therefore, in this mode, can according to the rise time of each language material (this information can be on the books in dialog history record) sequencing, dialog history record between first user and the second user is sorted, form language material sequence, certainly, for the situation that there is multiple second user, can form multiple language material sequence, a language material sequence pair answers second user.Afterwards, for each problem language material under same user view, can from language material sequence, to extract after problem language material and the target retro language material of the nearest preset entry of distance problem language material (such as or two etc.), like this, just the reply language material that first user provides the problem language material in this user view can be defined as by this target retro language material.
Such as, the dialog history record between first user A and the second user B, after arranging, obtains following language material sequence according to language material rise time sequencing:
{ problem language material 1, replys language material 1, problem language material 2, replys language material 2, replys language material 3, problem language material 3, problem language material 4, replys language material 4 ...
Suppose that the problem language material under certain user view comprises problem language material 1, pre-defined with problem language material between distance be less than or equal to 1 reply language material can as the target retro language material of this problem language material (when distance be 0, representative and problem language material direct neighbor, when distance is 1, language material is replied across one between representative and problem language material, by that analogy), then, generate after this problem language material 1 owing to replying language material 1, and with this problem language material 1 direct neighbor, also namely distance is 0, replying language material 2 is also generate after problem language material 1, and the distance between this problem language material 1 is 1, reply language material 1 and reply the target retro language material provided language material 2 can be answered this problem language material 1 during as first user and be extracted.Accordingly, if problem language material 2 is the problem language materials under another user view, then according to aforementioned hypothesis relation, the target retro language material that reply language material 2, reply language material 3 provide when also will be answered this problem language material 2 as first user is extracted, by that analogy.
In a word, for each user view, multiple reply language materials that first user provides for this user view can be extracted from dialog history record, what have in these reply language materials may can answer this user view really, some may not answer this user view, belong to noise, the best language material that also may there is the most applicable this user view of answer in the reply language material of this user view can be answered, therefore, will for each user view in follow-up step, from multiple reply language materials corresponding respectively, filter out noise, and filter out the reply language material of the most applicable answer respective user intention, center answer as respective user intention is preserved.
S103: for same user view, language material is replied to each and carries out cluster, draw multiple classification, and comprising reply language material number more than in the target classification of preset threshold value, calculating each respectively and replying the weight that language material becomes class center in described target classification;
Concrete, after getting multiple reply language material for a user view, just can reply the center answer of extracting language material and can answer this user view from these.During specific implementation, first can reply language material to these and carry out cluster, mainly carry out cluster by each text similarity of replying between language material of comparison, many algorithms specifically can be used to carry out this cluster process.Such as, under a kind of implementation, Jie Kade similarity coefficient can be used to calculate the similarity of replying between language material, then use canopy algorithm to carry out cluster, now, suppose the corresponding N number of reply language material of certain user view, and suppose that minimum similarity degree is T1, consistent similarity is T2, and in an initial condition, each replys the center mark position flag=false of language material.Like this, concrete cluster process can comprise the following steps:
Step 1, from N number of reply language material, random selecting one replys language material n, offers a classification 1, and supposes that this reply language material is such other central point;
Step 2, calculates the distance d between other reply language materials and this central point;
Step 3, if certain distance d replied between language material m and this central point is less than T1, then joins in this classification 1 by this reply language material m;
Step 4, if d is less than T2, is then set to true by the zone bit flag of this reply language material m, shows that this reply language material m have found described classification with this, does not need again for this reply language material offers a new classification;
After once traversal terminates, again from all flag be the reply language material of false again random selecting one reply language material, repeated execution of steps 1 to step 4, by that analogy, until the zone bit of all reply language materials is all set to true, the cluster process to replying language material just can be completed.
After completing above-mentioned cluster process, N number of reply language material corresponding to user view can be divided into multiple classification, wherein, the reply language material of some is included in each classification, the reply language material comprised in some classifications may be considerably less, and these are replied language material and generally can be fallen by as noise filtering.Remaining is all comprise to reply the digital classification more than certain preset threshold value of language material, and in these classifications, the similarity that in same classification, each is replied between language material is all higher, and generally can have one can as the reply language material of class center.Therefore, next, the embodiment of the present application just can be excavated from these classifications can as the reply language material of class center, and this reply language material just may become the center answer answering this user view.
When obtaining the class center in certain classification, each reply language material that can calculate respectively in classification becomes the weight of such other class center, and weight soprano can become class center.Also namely, first supposing that wherein any one replys language material is all class center, and when then calculating specific to certain reply language material, it really becomes the weight of class center.Specifically when calculating certain and replying the weight of language material n, following factor can be considered: the distance t between the rise time of other reply language materials in this classification and current time, and other in classification reply the similarity L between language material and this reply language material n.Wherein, for time gap t, because the reply language material the closer to current time more can show real-time, therefore, other reply language material to the reinforcement degree of the weight of current reply language material, are inversely proportional to time gap t.For this reason, first the distance t calculated can be brought in a loss of time function, obtain an output valve y, recycle the calculating that this y value participates in the weight of replying language material n afterwards.Wherein, the loss of time, the concrete manifestation form of function can have multiple, and such as, wherein under a kind of mode, can be an average be 0, variance be the just too distribution function of σ.Reply for the similarity between language material and this reply language material n for other, similarity is higher, more can improve the weight that this reply language material n becomes class center.Therefore, specifically can calculate by following formula the weight that reply language material n becomes the class center of its place classification:
Wherein:
Y
nin the target classification I of its place, the weight of class center is become for replying language material n;
Y
mfor being brought into by other distance t replied between the rise time of language material m and current time in this target classification I in function preset loss of time, obtain output valve;
L
nmfor the similarity of replying language material n and reply between language material m;
N is the number of the reply language material comprised in this target classification I.
In a word, reply for language material for each in certain classification, all can calculate the weight that can become class center separately in the manner described above, wherein namely weight soprano can be used as class center.
S104: according to described weight, determines the center answer of this user view.
By each step above, for certain user view, under this user view can be determined, reply the target classification that language material number is many, and the class center of each classification can be determined respectively, so just can determine the center answer of this user view according to this class center.Wherein, if under certain user view, comprise that to reply language material number more than the target classification of preset threshold value be one, then just directly such other class center can be defined as the center answer of this user view.But, if under certain user view, comprise that to reply language material number more than the target classification of preset threshold value be multiple, then correspondence can obtain multiple class center, now, directly class center the highest for weight can not be defined as the center answer of this user view.This is because, in dialog history record, the user view that some reply language material covers may be wider, also namely namely same a word may occur in the reply language material of user view A, occur in the reply language material of user view B again, now, if the words can answer user view A really, then for user view B, just belong to noise, should be filtered.But, under user view B, with aforesaid way calculate this reply language material become the weight of class center but may be higher, then now, if directly determine the center answer of user view according to the height of weight, then may there is deviation.
For this reason, in the embodiment of the present application, for comprising, to reply language material number more than the target classification of preset threshold value be the situation of at least two, first other class center of each target class can be got, from dialog history record, the total degree that each class center occurs can be obtained afterwards respectively, and class center is used to answer active user the number of times of intention; Then, be used to answer the number of times of active user's intention and the total degree of this class center appearance according to class center, determine that this class center is used to answer active user and is intended to shared ratio; Afterwards, class center the highest for ratio can be defined as the center answer of this current user view.
Such as, comprise two target classifications under certain user view A, wherein, the class center of a classification is for replying language material x, and the class center of another classification is for replying language material y, and wherein, the weight of replying language material y is greater than the weight of replying language material x.Suppose that finding to reply through statistics the total degree that language material x occurs is 100 times, being wherein 50 times for answering the number of times of this user view A, therefore, is 50% for the ratio answered shared by this user view A; Replying the total degree that language material y occurs is 200 times, and being wherein 20 times for answering the number of times of this user view A, therefore, is 10% for the ratio answered shared by this user view A.Now, although the weight of replying language material x will lower than reply language material y, relatively high for the ratio answered shared by this user view A owing to replying language material x, therefore, finally can select to reply the center answer of language material x as this user view A.
In a word, pass through the embodiment of the present application, data mining can be carried out from dialog history record, thus get the center answer that may be used for answering each user view, like this, with regard to no longer needing first user or backstage technician to carry out manual configuration to the answer of user view, be conducive to saving manpower and time cost further.
Corresponding with the method that the enquirement for user that the embodiment of the present application provides is intended to obtain answer information, the embodiment of the present application additionally provides the device that a kind of enquirement for user intention obtains answer information, and see Fig. 2, this device specifically can comprise:
Dialog history record acquiring unit 201, for obtaining the dialog history record that first user and multiple second user carry out in instant messaging process;
Reply language material acquiring unit 202, for carrying out user view identification to the problem language material that each second user proposes from described dialog history record, each problem language material comprised under obtaining same user view, and the reply language material obtaining that this first user provides each problem language material under same user view;
Weight calculation unit 203, for for same user view, language material is replied to each and carries out cluster, draw multiple classification, and comprising reply language material number more than in the target classification of preset threshold value, calculating each respectively and replying the weight that language material becomes class center in described target classification;
Center answer determining unit 204, for according to described weight, determines the center answer of this user view.
Wherein, described reply language material acquiring unit 202 specifically can comprise:
Language material sequence generates subelement, for the rise time sequencing according to each language material, sorts to the dialog history record between first user and the second user, forms language material sequence;
Extract subelement, for for each problem language material under same user view, from described language material sequence, extract the target retro language material of preset entry after described problem language material, nearest apart from described problem language material, described target retro language material is defined as the reply language material that first user provides the problem language material in this user view.
During specific implementation, described weight calculation unit 203 carries out following calculating respectively specifically for replying language material for each under same target classification:
For current reply language material n, according in the target classification of place, other respectively reply the similarity L between language material and this current reply language material n, and other each distance t replied between the rise time of language material and current time described, calculate the weight that this current reply language material n becomes class center in its place target classification.
More specifically, weight calculation can be carried out by the formula in embodiment of the method.
Wherein, if described in comprise that to reply language material number more than the target classification of preset threshold value be one, then center answer determining unit 204 specifically may be used for: center answer reply language material the highest for weight being defined as this user view.
If described in comprise and reply language material number and be at least two more than the target classification of preset threshold value, then center answer determining unit 204 specifically can comprise:
Class center determination subelement, for obtaining the highest target retro language material of weight respectively from each target classification, as other class center of corresponding target class;
Number of times obtains subelement, and for obtaining the total degree that each class center occurs from described dialog history record respectively, and class center is used to answer active user the number of times of intention;
Ratio-dependent subelement, answers the number of times of active user's intention and the total degree of this class center appearance for being used to according to class center, determines that this class center is used to answer active user and is intended to shared ratio;
Answer determination subelement, for being defined as the center answer of this current user view by class center the highest for described ratio.
In a word, pass through above-described embodiment, data mining can be carried out to dialog history record by computer program, for each user view automatic acquisition is to corresponding answer, and no longer need user or the backstage technician answer input manual for each user view performs or work is set etc., manpower and time cost can be saved further.
As seen through the above description of the embodiments, those skilled in the art can be well understood to the mode that the application can add required general hardware platform by software and realizes.Based on such understanding, the technical scheme of the application can embody with the form of software product the part that prior art contributes in essence in other words, this computer software product can be stored in storage medium, as ROM/RAM, magnetic disc, CD etc., comprising some instructions in order to make a computer equipment (can be personal computer, server, or the network equipment etc.) perform the method described in some part of each embodiment of the application or embodiment.
Each embodiment in this instructions all adopts the mode of going forward one by one to describe, between each embodiment identical similar part mutually see, what each embodiment stressed is the difference with other embodiments.Especially, for system or system embodiment, because it is substantially similar to embodiment of the method, so describe fairly simple, relevant part illustrates see the part of embodiment of the method.System described above and system embodiment are only schematic, the wherein said unit illustrated as separating component or can may not be and physically separates, parts as unit display can be or may not be physical location, namely can be positioned at a place, or also can be distributed in multiple network element.Some or all of module wherein can be selected according to the actual needs to realize the object of the present embodiment scheme.Those of ordinary skill in the art, when not paying creative work, are namely appreciated that and implement.
The intention of the enquirement for user provided the application above obtains method and the device of answer information, be described in detail, apply specific case herein to set forth the principle of the application and embodiment, the explanation of above embodiment is just for helping method and the core concept thereof of understanding the application; Meanwhile, for one of ordinary skill in the art, according to the thought of the application, all will change in specific embodiments and applications.In sum, this description should not be construed as the restriction to the application.
Claims (10)
1., for a method for the enquirement intention acquisition answer information of user, it is characterized in that, comprising:
Acquisition first user and multiple second user carry out the dialog history record in instant messaging process;
From described dialog history record, user view identification is carried out to the problem language material that each second user proposes, each problem language material comprised under obtaining same user view, and the reply language material obtaining that this first user provides each problem language material under same user view;
For same user view, language material is replied to each and carries out cluster, draw multiple classification, and comprising reply language material number more than in the target classification of preset threshold value, calculating each respectively and replying the weight that language material becomes class center in described target classification;
According to described weight, determine the center answer of this user view.
2. method according to claim 1, is characterized in that, the reply language material that this first user of described acquisition provides each problem language material in same user view, comprising:
According to the rise time sequencing of each language material, the dialog history record between first user and the second user is sorted, form language material sequence;
For each problem language material under same user view, from described language material sequence, extract the target retro language material of preset entry after described problem language material, nearest apart from described problem language material, described target retro language material is defined as the reply language material that first user provides the problem language material in this user view.
3. method according to claim 1, is characterized in that, each reply language material of described calculating becomes the weight of class center in described target classification, comprising:
Reply language material for each under same target classification and carry out following calculating respectively:
For current reply language material n, according in the target classification of place, other respectively reply the similarity L between language material and this current reply language material n, and other each distance t replied between the rise time of language material and current time described, calculate the weight that this current reply language material n becomes class center in its place target classification.
4. method according to claim 3, it is characterized in that, it is described that according in the target classification of place, other respectively reply the similarity L between language material and this current reply language material n, and other each distance t replied between the rise time of language material and current time described, calculate the weight that this current reply language material n becomes class center in its place target classification, comprising:
Wherein:
Y
nin the target classification I of its place, the weight of class center is become for replying language material n
Y
mfor being brought into by other distance t replied between the rise time of language material m and current time in this target classification I in function preset loss of time, obtain output valve;
L
nmfor the similarity of replying language material n and reply between language material m;
N is the number of the reply language material comprised in this target classification I.
5. method according to claim 1, is characterized in that, if described in comprise that to reply language material number more than the target classification of preset threshold value be one, then described according to described weight, determine the center answer of this user view, comprising:
Reply language material the highest for weight is defined as the center answer of this user view.
6. method according to claim 1, is characterized in that, if described in comprise and reply language material number and be at least two more than the target classification of preset threshold value, then described according to described weight, determine the center answer of this user view, comprising:
The highest target retro language material of weight is obtained respectively, as other class center of corresponding target class from each target classification;
From described dialog history record, obtain the total degree that each class center occurs respectively, and class center is used to answer active user the number of times of intention;
Be used to answer the number of times of active user's intention and the total degree of this class center appearance according to class center, determine that this class center is used to answer active user and is intended to shared ratio;
Class center the highest for described ratio is defined as the center answer of this current user view.
7., for a device for the enquirement intention acquisition answer information of user, it is characterized in that, comprising:
Dialog history record acquiring unit, for obtaining the dialog history record that first user and multiple second user carry out in instant messaging process;
Reply language material acquiring unit, for carrying out user view identification to the problem language material that each second user proposes from described dialog history record, each problem language material comprised under obtaining same user view, and the reply language material obtaining that this first user provides each problem language material under same user view;
Weight calculation unit, for for same user view, language material is replied to each and carries out cluster, draw multiple classification, and comprising reply language material number more than in the target classification of preset threshold value, calculating each respectively and replying the weight that language material becomes class center in described target classification;
Center answer determining unit, for according to described weight, determines the center answer of this user view.
8. device according to claim 7, is characterized in that, described reply language material acquiring unit comprises:
Language material sequence generates subelement, for the rise time sequencing according to each language material, sorts to the dialog history record between first user and the second user, forms language material sequence;
Extract subelement, for for each problem language material under same user view, from described language material sequence, extract the target retro language material of preset entry after described problem language material, nearest apart from described problem language material, described target retro language material is defined as the reply language material that first user provides the problem language material in this user view.
9. device according to claim 7, is characterized in that, described weight calculation unit carries out following calculating respectively specifically for replying language material for each under same target classification:
For current reply language material n, according in the target classification of place, other respectively reply the similarity L between language material and this current reply language material n, and other each distance t replied between the rise time of language material and current time described, calculate the weight that this current reply language material n becomes class center in its place target classification.
10. device according to claim 9, is characterized in that, described weight calculation unit carries out weight calculation in the following manner:
Wherein:
Y
nin the target classification I of its place, the weight of class center is become for replying language material n
Y
mfor being brought into by other distance t replied between the rise time of language material m and current time in this target classification I in function preset loss of time, obtain output valve;
L
nmfor the similarity of replying language material n and reply between language material m;
N is the number of the reply language material comprised in this target classification I.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410350679.7A CN105335400B (en) | 2014-07-22 | 2014-07-22 | Enquirement for user is intended to obtain the method and device of answer information |
| HK16107229.0A HK1219316A1 (en) | 2014-07-22 | 2016-06-22 | Method for obtaining answer information for user question intention and device thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410350679.7A CN105335400B (en) | 2014-07-22 | 2014-07-22 | Enquirement for user is intended to obtain the method and device of answer information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105335400A true CN105335400A (en) | 2016-02-17 |
| CN105335400B CN105335400B (en) | 2018-11-23 |
Family
ID=55285937
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410350679.7A Active CN105335400B (en) | 2014-07-22 | 2014-07-22 | Enquirement for user is intended to obtain the method and device of answer information |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN105335400B (en) |
| HK (1) | HK1219316A1 (en) |
Cited By (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106095872A (en) * | 2016-06-07 | 2016-11-09 | 北京高地信息技术有限公司 | Answer sort method and device for Intelligent Answer System |
| CN106202508A (en) * | 2016-07-20 | 2016-12-07 | 北京小米移动软件有限公司 | The intelligence method of answer problem, Apparatus and system |
| CN106202085A (en) * | 2015-04-30 | 2016-12-07 | 阿里巴巴集团控股有限公司 | The method of information search, device and electronic equipment is carried out according to particular topic |
| CN106202417A (en) * | 2016-07-12 | 2016-12-07 | 北京光年无限科技有限公司 | A kind of man-machine interaction method for intelligent robot and system |
| CN106471502A (en) * | 2016-06-29 | 2017-03-01 | 深圳狗尾草智能科技有限公司 | Intension recognizing method based on water conservancy diversion and system |
| CN106503189A (en) * | 2016-10-31 | 2017-03-15 | 北京百度网讯科技有限公司 | search system optimization method and device based on artificial intelligence |
| CN106658216A (en) * | 2016-12-20 | 2017-05-10 | 天脉聚源(北京)传媒科技有限公司 | Method and device for obtaining information together |
| CN107784033A (en) * | 2016-08-31 | 2018-03-09 | 百度在线网络技术（北京）有限公司 | A kind of dialogue-based method and apparatus recommended |
| CN108153800A (en) * | 2016-12-06 | 2018-06-12 | 松下知识产权经营株式会社 | Information processing method, information processing unit and program |
| CN108345640A (en) * | 2018-01-12 | 2018-07-31 | 上海大学 | A kind of question and answer building of corpus method based on neural network semantic analysis |
| CN109446509A (en) * | 2018-09-06 | 2019-03-08 | 厦门快商通信息技术有限公司 | A kind of dialogue corpus is intended to analysis method, system and electronic equipment |
| CN109522556A (en) * | 2018-11-16 | 2019-03-26 | 北京九狐时代智能科技有限公司 | A kind of intension recognizing method and device |
| CN109597881A (en) * | 2018-12-17 | 2019-04-09 | 北京百度网讯科技有限公司 | Matching degree determines method, apparatus, equipment and medium |
| CN109710941A (en) * | 2018-12-29 | 2019-05-03 | 上海点融信息科技有限责任公司 | User's intension recognizing method and device based on artificial intelligence |
| CN109922070A (en) * | 2019-03-13 | 2019-06-21 | 北京奇艺世纪科技有限公司 | A kind of automatic reply method and device |
| CN110162603A (en) * | 2018-11-30 | 2019-08-23 | 腾讯科技（深圳）有限公司 | A kind of Intelligent dialogue method, dynamic storage method and device |
| CN110457454A (en) * | 2019-07-12 | 2019-11-15 | 卓尔智联(武汉)研究院有限公司 | A kind of dialogue method, server, conversational system and storage medium |
| CN110941710A (en) * | 2019-11-27 | 2020-03-31 | 贝壳技术有限公司 | Method, device, medium and electronic equipment for realizing session |
| CN111611382A (en) * | 2020-05-22 | 2020-09-01 | 贝壳技术有限公司 | Dialect model training method, dialog information generation method, device and system |
| CN112016938A (en) * | 2020-09-01 | 2020-12-01 | 中国银行股份有限公司 | Interaction method and device of robot, electronic equipment and computer storage medium |
| CN112035610A (en) * | 2020-08-28 | 2020-12-04 | 康键信息技术（深圳）有限公司 | Medical field question and answer pair generation method and device, computer equipment and medium |
| CN112131338A (en) * | 2020-06-05 | 2020-12-25 | 支付宝(杭州)信息技术有限公司 | Method and device for establishing question-answer pairs |
| CN112541059A (en) * | 2020-11-05 | 2021-03-23 | 大连中河科技有限公司 | Multi-round intelligent question-answer interaction method applied to tax question-answer system |
| CN113239164A (en) * | 2021-05-13 | 2021-08-10 | 杭州摸象大数据科技有限公司 | Multi-round conversation process construction method and device, computer equipment and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102193973A (en) * | 2010-03-19 | 2011-09-21 | 微软公司 | Presenting answers |
| CN102456060A (en) * | 2010-10-28 | 2012-05-16 | 株式会社日立制作所 | Information processing device and information processing method |
| CN102662952A (en) * | 2012-03-02 | 2012-09-12 | 成都康赛电子科大信息技术有限责任公司 | Chinese text parallel data mining method based on hierarchy |
| US20130211820A1 (en) * | 2012-02-15 | 2013-08-15 | Electronics And Telecommunications Research Institute | Apparatus and method for interpreting korean keyword search phrase |
| CN103823844A (en) * | 2014-01-26 | 2014-05-28 | 北京邮电大学 | Question forwarding system and question forwarding method on the basis of subjective and objective context and in community question-and-answer service |
| CN103902652A (en) * | 2014-02-27 | 2014-07-02 | 深圳市智搜信息技术有限公司 | Automatic question-answering system |
-
2014
- 2014-07-22 CN CN201410350679.7A patent/CN105335400B/en active Active
-
2016
- 2016-06-22 HK HK16107229.0A patent/HK1219316A1/en unknown
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102193973A (en) * | 2010-03-19 | 2011-09-21 | 微软公司 | Presenting answers |
| CN102456060A (en) * | 2010-10-28 | 2012-05-16 | 株式会社日立制作所 | Information processing device and information processing method |
| US20130211820A1 (en) * | 2012-02-15 | 2013-08-15 | Electronics And Telecommunications Research Institute | Apparatus and method for interpreting korean keyword search phrase |
| CN102662952A (en) * | 2012-03-02 | 2012-09-12 | 成都康赛电子科大信息技术有限责任公司 | Chinese text parallel data mining method based on hierarchy |
| CN103823844A (en) * | 2014-01-26 | 2014-05-28 | 北京邮电大学 | Question forwarding system and question forwarding method on the basis of subjective and objective context and in community question-and-answer service |
| CN103902652A (en) * | 2014-02-27 | 2014-07-02 | 深圳市智搜信息技术有限公司 | Automatic question-answering system |
Cited By (34)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106202085B (en) * | 2015-04-30 | 2019-08-20 | 阿里巴巴集团控股有限公司 | The method, apparatus and electronic equipment of information search are carried out according to specific subject |
| CN106202085A (en) * | 2015-04-30 | 2016-12-07 | 阿里巴巴集团控股有限公司 | The method of information search, device and electronic equipment is carried out according to particular topic |
| CN106095872A (en) * | 2016-06-07 | 2016-11-09 | 北京高地信息技术有限公司 | Answer sort method and device for Intelligent Answer System |
| CN106471502A (en) * | 2016-06-29 | 2017-03-01 | 深圳狗尾草智能科技有限公司 | Intension recognizing method based on water conservancy diversion and system |
| CN106202417A (en) * | 2016-07-12 | 2016-12-07 | 北京光年无限科技有限公司 | A kind of man-machine interaction method for intelligent robot and system |
| CN106202508A (en) * | 2016-07-20 | 2016-12-07 | 北京小米移动软件有限公司 | The intelligence method of answer problem, Apparatus and system |
| CN107784033B (en) * | 2016-08-31 | 2021-10-22 | 百度在线网络技术（北京）有限公司 | Method and device for recommending based on session |
| CN107784033A (en) * | 2016-08-31 | 2018-03-09 | 百度在线网络技术（北京）有限公司 | A kind of dialogue-based method and apparatus recommended |
| CN106503189A (en) * | 2016-10-31 | 2017-03-15 | 北京百度网讯科技有限公司 | search system optimization method and device based on artificial intelligence |
| CN106503189B (en) * | 2016-10-31 | 2020-03-03 | 北京百度网讯科技有限公司 | Search system optimization method and device based on artificial intelligence |
| CN108153800A (en) * | 2016-12-06 | 2018-06-12 | 松下知识产权经营株式会社 | Information processing method, information processing unit and program |
| CN108153800B (en) * | 2016-12-06 | 2023-05-23 | 松下知识产权经营株式会社 | Information processing method, information processing apparatus, and recording medium |
| CN106658216A (en) * | 2016-12-20 | 2017-05-10 | 天脉聚源(北京)传媒科技有限公司 | Method and device for obtaining information together |
| CN108345640A (en) * | 2018-01-12 | 2018-07-31 | 上海大学 | A kind of question and answer building of corpus method based on neural network semantic analysis |
| CN109446509B (en) * | 2018-09-06 | 2023-04-07 | 厦门快商通信息技术有限公司 | Dialogue corpus intention analysis method and system and electronic equipment |
| CN109446509A (en) * | 2018-09-06 | 2019-03-08 | 厦门快商通信息技术有限公司 | A kind of dialogue corpus is intended to analysis method, system and electronic equipment |
| CN109522556A (en) * | 2018-11-16 | 2019-03-26 | 北京九狐时代智能科技有限公司 | A kind of intension recognizing method and device |
| CN109522556B (en) * | 2018-11-16 | 2024-03-12 | 北京九狐时代智能科技有限公司 | Intention recognition method and device |
| CN110162603A (en) * | 2018-11-30 | 2019-08-23 | 腾讯科技（深圳）有限公司 | A kind of Intelligent dialogue method, dynamic storage method and device |
| CN110162603B (en) * | 2018-11-30 | 2023-11-14 | 腾讯科技（深圳）有限公司 | Intelligent dialogue method, dynamic storage method and device |
| CN109597881A (en) * | 2018-12-17 | 2019-04-09 | 北京百度网讯科技有限公司 | Matching degree determines method, apparatus, equipment and medium |
| CN109710941A (en) * | 2018-12-29 | 2019-05-03 | 上海点融信息科技有限责任公司 | User's intension recognizing method and device based on artificial intelligence |
| CN109922070B (en) * | 2019-03-13 | 2021-11-26 | 北京奇艺世纪科技有限公司 | Automatic reply method and device |
| CN109922070A (en) * | 2019-03-13 | 2019-06-21 | 北京奇艺世纪科技有限公司 | A kind of automatic reply method and device |
| CN110457454A (en) * | 2019-07-12 | 2019-11-15 | 卓尔智联(武汉)研究院有限公司 | A kind of dialogue method, server, conversational system and storage medium |
| CN110941710B (en) * | 2019-11-27 | 2020-10-30 | 贝壳找房(北京)科技有限公司 | Method, device, medium and electronic equipment for realizing session |
| CN110941710A (en) * | 2019-11-27 | 2020-03-31 | 贝壳技术有限公司 | Method, device, medium and electronic equipment for realizing session |
| CN111611382A (en) * | 2020-05-22 | 2020-09-01 | 贝壳技术有限公司 | Dialect model training method, dialog information generation method, device and system |
| CN112131338A (en) * | 2020-06-05 | 2020-12-25 | 支付宝(杭州)信息技术有限公司 | Method and device for establishing question-answer pairs |
| CN112131338B (en) * | 2020-06-05 | 2024-02-09 | 支付宝(杭州)信息技术有限公司 | Method and device for establishing question-answer pairs |
| CN112035610A (en) * | 2020-08-28 | 2020-12-04 | 康键信息技术（深圳）有限公司 | Medical field question and answer pair generation method and device, computer equipment and medium |
| CN112016938A (en) * | 2020-09-01 | 2020-12-01 | 中国银行股份有限公司 | Interaction method and device of robot, electronic equipment and computer storage medium |
| CN112541059A (en) * | 2020-11-05 | 2021-03-23 | 大连中河科技有限公司 | Multi-round intelligent question-answer interaction method applied to tax question-answer system |
| CN113239164A (en) * | 2021-05-13 | 2021-08-10 | 杭州摸象大数据科技有限公司 | Multi-round conversation process construction method and device, computer equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| HK1219316A1 (en) | 2017-03-31 |
| CN105335400B (en) | 2018-11-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105335400A (en) | Method and apparatus for obtaining answer information for questioning intention of user | |
| CN106296059B (en) | Method and equipment for determining delivery network points | |
| CN104394118B (en) | A kind of method for identifying ID and system | |
| CN107590688A (en) | The recognition methods of target customer and terminal device | |
| CN107507016A (en) | A kind of information push method and system | |
| CN106445905B (en) | Question and answer data processing, automatic question-answering method and device | |
| CN105573966A (en) | Adaptive Modification of Content Presented in Electronic Forms | |
| CN108920530B (en) | Information processing method and device, storage medium and electronic equipment | |
| CN104077407A (en) | System and method for intelligent data searching | |
| CN103825784A (en) | Non-public protocol field identification method and system | |
| CN112784039A (en) | Method, device and storage medium for distributing online customer service | |
| CN110795471A (en) | Data matching method and device, computer readable storage medium and electronic equipment | |
| CN114266443A (en) | Data evaluation method and device, electronic equipment and storage medium | |
| Haq et al. | Analysis of agile supply chain enablers for Indian food processing industries using analytical hierarchy process | |
| Wu | Applying grey model to prioritise technical measures in quality function deployment | |
| CN112199488B (en) | Incremental knowledge graph entity extraction method and system for power customer service question and answer | |
| CN104965846A (en) | Virtual human establishing method on MapReduce platform | |
| Mastromarco | Foreign capital and efficiency in developing countries | |
| Siswanto et al. | Implementation of decision support system for campus promotion management using fuzzy multiple analytic decision making (FMADM) method (Case study: Universitas multimedia nusantara) | |
| CN110992095B (en) | Consumer portrait generation method and device | |
| CN115329078B (en) | Text data processing method, device, equipment and storage medium | |
| CN104793925A (en) | Microblog function allocating method and device | |
| CN107767278B (en) | Method and device for constructing community hierarchy | |
| CN107705135A (en) | A kind of method that potential commercial value is evaluated based on company's storage contact data | |
| WO2022151524A1 (en) | Method for seeking potential customers by using social software big data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 1219316 Country of ref document: HK |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |