CN105335447A

CN105335447A - Computer network-based expert question-answering system and construction method thereof

Info

Publication number: CN105335447A
Application number: CN201410400821.4A
Authority: CN
Inventors: 白明
Original assignee: Beijing Qihoo Technology Co Ltd; Qizhi Software Beijing Co Ltd
Current assignee: Beijing Qihoo Technology Co Ltd; Qizhi Software Beijing Co Ltd
Priority date: 2014-08-14
Filing date: 2014-08-14
Publication date: 2016-02-17

Abstract

The invention discloses a computer network-based expert question-answering system and a construction method thereof, and aims at providing experts for users according to different domains. The question-answering system comprises a knowledge base construction unit, a domain expert determination unit, a question reception unit, a similarity determination unit and an expert matching unit, wherein the knowledge base construction unit is used for constructing a domain knowledge base comprising concepts and entities; the domain expert determination unit is used for determining experts of information in an information set according to the information set of domains; the information in the information set is information correlated to the concepts or entities, obtained from websites related to the domains; the question reception unit is used for receiving questions input by the users; the similarity determination unit is used for determining first similarities of the experts and the questions; and the expert matching unit is used for recommending the experts with the highest first similarities to answer the questions. According to the expert question-answering system and the construction method, the experts in the domains to which the questions belong are objectively and correctly recommended to the users, and the accuracy for solving problems by the users is improved.

Description

Based on expert's question answering system and the construction method thereof of computer network

Technical field

The invention belongs to computer network field, particularly a kind of expert's question answering system based on computer network and construction method thereof.

Background technology

Along with the development of Internet information technique, Internet user carry out the mode of information interchange and object diversified gradually.When user exists the problem needing answer, answer can be obtained in several ways.Traditional mode comprises and uses such as phone, Email or other immediate communication tools, to be familiar with or the social circle that is in acquaintance ask a question, thus obtain answer.

Nearest a kind of conventional mode is, computer network user directly can also submit a question on the website with question answering system, such as on search website, directly input needs the problem content of search, by question answering system by the existing Q & A database of keyword match, provides result for retrieval.Main implementation is that first question answering system sets up one's own knowledge data base, by different input mode (word, image) etc., carries out retrieval coupling, obtain answer with the key word inputted.

But, above-mentioned traditional approach, cannot break through be familiar with or acquaintance crowd restriction, when institute for understanding problem within existing communication circle, nobody knows the answer or does not contact time, cannot answer be obtained.Although the namely recently conventional mode of the above-mentioned second way overcomes the drawback of traditional approach, such as, can only be mated by key word, cause that the answer of the problem retrieved and user are actual thinks that the answer of acquisition problem does not conform to thus.

Given this, when how to ask a question on the website with question answering system, obtain and become the current technical issues that need to address with the answer of domain experts belonging to this problem.

Summary of the invention

For defect of the prior art, the invention provides a kind of expert's question answering system based on computer network and construction method thereof.

First aspect, the embodiment of the present invention provides a kind of expert's question answering system based on computer network, construction of knowledge base unit, for building domain knowledge base, described domain knowledge base comprises: at least one concept in described field, the multiple entities corresponding with each concept;

Domain expert's determining unit, for the information aggregate according to described field, determine the expert belonging to information described in described information aggregate, information in described information aggregate is that obtain the website or comment of being correlated with from described field with described concept or described entity associated information, and described expert is the person of sending of described information or the recipient of described information;

Problem receiving element, for receiving the problem of user's input;

Similarity determining unit, for determining the first similarity of the expert that described domain expert's determining unit is determined and the problem that described problem receiving element receives;

Expert's matching unit, sorts according to size for the first similarity determined by described similarity determining unit, and the expert choosing the first similarity of coming top N corresponding answers described problem, N be more than or equal to 1 natural number.

Alternatively, described construction of knowledge base unit, specifically for

The website corresponding to described field is carried out orientation and is captured, and set up the list set of two tuple form, the list in described list set comprises: the element set of multiple element compositions that navigation word, described navigation word are corresponding;

Determine whether the navigation word of each list in described list set mates with at least one concept described, word and at least one concept matching described if navigate described in a list, then using the element in list belonging to described navigation word as kernel entity corresponding at least one concept described, and kernel entity corresponding to each concept forms the entity sets of described concept.

Alternatively, described construction of knowledge base unit, also for

When the navigation word that there is at least one list in described list set does not match with at least one concept described, then obtain the second similarity of the entity sets of element set in list belonging to the navigation word that do not match with at least one concept described and each concept respectively;

For multiple second similarities of each navigation word do not mated, multiple described second similarity of this navigation word sorted according to size, the element in list belonging to this navigation word is as the non-core entity in concept corresponding to the second similarity coming front M position; M be more than or equal to 1 natural number.

Alternatively, described construction of knowledge base unit, also for

When described concept does not comprise kernel entity and non-core entity, supplement the kernel entity that described concept is corresponding;

Wherein, the corresponding multiple entity of described concept comprises: described kernel entity and/or described non-core entity.

Alternatively, described domain expert's determining unit, specifically for

Obtain the information in social network sites corresponding to described field, determine whether the described information content comprises concept name in described domain knowledge base or entity title;

If the described information content comprises described concept name or entity title, then generate expert's candidate collection according to the sender of described information, recipient, and

The third phase calculating described information and described field is like spending, and by the sender of described information, the third phase of recipient and described information is seemingly spent as a triplet information, information generated set;

According to the information in the expert of described expert's candidate collection and described information aggregate, obtain the rank of each expert in described expert's candidate collection;

And/or,

Choose the forward X of a rank expert as the expert belonging to information described in described information aggregate, X be more than or equal to 1 natural number.

Alternatively, described domain expert's determining unit, also for

For each expert in described expert's candidate collection, obtain all information of each expert in described information aggregate;

According to all information of each expert in described information aggregate and all concepts in described domain knowledge base, obtain the concept similarity vector of each expert to all concepts.

Alternatively, described similarity determining unit, specifically for

Cut word process to the described problem that described problem receiving element receives, obtain the word corresponding with described problem first gathers;

Obtain the problem similarity vector of all concepts in described first set and described domain knowledge base;

According to described concept similarity vector and described problem similarity vector, determine the first similarity of described expert and described problem.

Second aspect, the invention provides a kind of expert's question answering system, comprising:

Receiving element, for receiving the problem of user's input;

Similarity determining unit, for determining the similarity of each expert in described problem and expert's question answering system, the people that is skillful in described expert field belonging to described problem;

Selection of specialists unit, for described similarity being sorted according to size, chooses the expert that the similarity that comes top N is corresponding, N be more than or equal to 1 natural number;

Answer unit, answers described problem for making the expert of described selection of specialists unit selection for described user.

Alternatively, described similarity determining unit, specifically for

Cut word process to described problem, obtain the word corresponding with described problem first gathers;

Obtain described first set and the problem similarity vector of all concepts in domain knowledge base, described domain knowledge base is the knowledge base comprising multiple entities corresponding at least one concept, at least one concept described obtained in advance in described expert's question answering system;

According to concept similarity vector and the described problem similarity vector of each expert, determine the similarity of expert and described problem in described expert's question answering system; The concept similarity vector of each expert described is that in all information and described domain knowledge base sent according to this expert, all concepts obtain in advance, and all information that described expert sends are that obtain from described field related web site or comment with described concept or described entity associated information.

The third aspect, the embodiment of the present invention provides a kind of construction method of the expert's question answering system based on computer network, comprising:

Build domain knowledge base, described domain knowledge base comprises: at least one concept in described field, the multiple entities corresponding with each concept;

According to the information aggregate in described field, determine the expert belonging to information described in described information aggregate, information in described information aggregate is that obtain the website or comment of being correlated with from described field with described concept or described entity associated information, and described expert is the person of sending of described information or the recipient of described information;

If described expert's question answering system receives problem, then determine the first similarity of described expert and described problem, described first similarity sorted according to size, the expert choosing the first similarity of coming top N corresponding answers described problem, N be more than or equal to 1 natural number.

Alternatively, described structure domain knowledge base, comprising:

Alternatively, described structure domain knowledge base, also comprises:

If the navigation word that there is at least one list in described list set does not match with at least one concept described, then obtain the second similarity of the entity sets of element set in list belonging to the navigation word that do not match with at least one concept described and each concept respectively;

For multiple second similarities of each navigation word do not mated, multiple described second similarity of this navigation word sorted according to size, the element in list belonging to this navigation word is as the non-core entity in concept corresponding to the second similarity coming front M position;

M be more than or equal to 1 natural number.

Alternatively, described structure domain knowledge base, also comprises:

If described concept does not comprise kernel entity and non-core entity, then supplement kernel entity corresponding to described concept;

Wherein, multiple entities that described concept is corresponding comprise: described kernel entity and/or described non-core entity.

Alternatively, the described information aggregate according to described field, determine the expert belonging to information described in described information aggregate, comprising:

The third phase calculating described information and described field, like spend, using the third phase of the sender of described information, recipient and described information like spending as a triplet information, generates described information aggregate;

And/or,

Alternatively, also comprise:

Alternatively, if described expert's question answering system receives problem, determine the first similarity of described expert and described problem, comprising:

Fourth aspect, the invention provides a kind of automatic question-answering method, comprising:

Receive the problem of user's input, determine the similarity of each expert in described problem and expert's question answering system, the people that is skillful in described expert field belonging to described problem;

Described similarity sorted according to size, the expert choosing the similarity that comes top N corresponding answers described problem, N be more than or equal to 1 natural number.

Alternatively, the described similarity determining expert in described problem and expert's question answering system, comprising:

According to concept similarity vector and the described problem similarity vector of each expert, determine the similarity of expert and described problem in described expert's question answering system; The concept similarity vector of each expert described is that all concepts in all information and described domain knowledge base sent according to this expert obtain in advance; All information that described expert sends are that obtain from described field related web site or comment with described concept or described entity associated information.

As shown from the above technical solution, expert's question answering system based on computer network of the present invention and construction method thereof, by building domain knowledge base, and determine the expert relevant to concept in domain knowledge base or entity, and then when receiving the problem that user proposes, the first similarity of each expert and this problem can be determined, the problem that an answer user in the expert that selection and comparison is similar proposes, thus, said method has taken into full account the combination of professional website and social networks, can for user objective and accurate recommend the expert in field belonging to problem to answer the problem of user, improve the accuracy rate solving customer problem.

Above-mentioned explanation is only the general introduction of technical solution of the present invention, in order to better understand technological means of the present invention, and can be implemented according to the content of instructions, and can become apparent, below especially exemplified by the specific embodiment of the present invention to allow above and other object of the present invention, feature and advantage.

Accompanying drawing explanation

In order to be illustrated more clearly in the embodiment of the present invention or technical scheme of the prior art, below the accompanying drawing used required in describing embodiment is briefly described, apparently, accompanying drawing in the following describes is only examples more of the present invention, 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.

The schematic flow sheet of the construction method of expert's question answering system that Fig. 1 provides for one embodiment of the invention;

The schematic flow sheet of the structure domain knowledge base that Fig. 2 A provides for one embodiment of the invention;

The schematic flow sheet of expert belonging to the determination field that Fig. 2 B provides for one embodiment of the invention;

The schematic flow sheet that Fig. 2 C mates with domain experts for the problem that one embodiment of the invention provides;

The schematic flow sheet of the automatic question-answering method that Fig. 3 provides for one embodiment of the invention;

The structural representation of expert's question answering system that Fig. 4 provides for one embodiment of the invention;

The structural representation of expert's question answering system that Fig. 5 provides for another embodiment of the present invention;

The structural representation of expert's question answering system that Fig. 6 provides for another embodiment of the present invention;

The social network diagram that the expert candidate collection P that Fig. 7 provides for one embodiment of the invention and massage set E is formed.

Embodiment

Below with reference to accompanying drawings exemplary embodiment disclosed by the invention is described in more detail.Although show exemplary embodiment disclosed by the invention in accompanying drawing, however should be appreciated that can realize in a variety of manners the present invention open and not should limit by the embodiment set forth here.On the contrary, provide these embodiments to be in order to more thoroughly the present invention can be understood, and complete for scope disclosed by the invention can be conveyed to those skilled in the art.

Fig. 1 shows the schematic flow sheet of the construction method of expert's question answering system that one embodiment of the invention provides, and as shown in Figure 1, the construction method of expert's question answering system of the present embodiment as mentioned below.

101, build domain knowledge base, described domain knowledge base comprises: at least one concept in described field, the multiple entities corresponding with each concept;

102, according to the information aggregate in described field, determine the expert belonging to information described in described information aggregate, information in described information aggregate is that obtain the website or comment of being correlated with from described field with described concept or described entity associated information, and described expert is the person of sending of described information or the recipient of described information;

If 103 described expert's question answering systems receive problem, then determine the first similarity of described expert and described problem, described first similarity sorted according to size, the expert choosing the first similarity of coming top N corresponding answers described problem.

Wherein, N be more than or equal to 1 natural number, the value of N can be arranged according to actual needs.

The expert preferably made number one in the present embodiment answers the problem of user, but the situation that may occur is, the expert made number one, in the problem of other users of answer, now can select to come the problem that deputy expert answers user.

The construction method of expert's question answering system of the present embodiment, by building domain knowledge base, and determine the expert relevant to concept in domain knowledge base or entity, and then when receiving the problem that user proposes, the first similarity of each expert and this problem can be determined, the problem that an answer user in the expert that selection and comparison is similar proposes, take into full account the combination of professional website and social networks, can for user objective and accurate recommend the expert in field belonging to problem to answer the problem of user, improve and solve the accuracy rate of customer problem.

In the present embodiment, aforesaid domain knowledge base can include concept and entity, between concept and entity be wherein set up related.Build the key of domain knowledge base be to obtain concept under entity.

Usually, by manually setting up at least one concept in this field in domain knowledge base, and the multiple entities under each concept can also can be supplemented by manual type.Certainly, because field is very many, and concept in domain knowledge base is also very many, and set up domain knowledge base by manual type more consuming time, cost is high.Thus, obtain the entity in domain knowledge base under concept by automated manner in the present embodiment, such as, realized by the sub-step of subordinate.

Aforesaid step 101 also can comprise the following sub-step 1011 shown in Fig. 2 A and sub-step 1012;

1011, corresponding to described field website is carried out orientation and is captured, and set up the list set of two tuple form, the list in described list set comprises: the element set of multiple element compositions that navigation word, described navigation word are corresponding.

For example, the directed various information captured in professional website, such as, extract the classification navigation conduct navigation word in this professional website, set up the list etc. of set form.

List in the present embodiment can be binary group information: a t=< navigation word, { element set occurred in list } >, multiple two tuple form can form a list set, such as, list set H={t|t=< navigates word, { element set occurred in list } >.It should be noted that the element set at this place can adopt t _srepresent.

1012, determine whether the navigation word of each list in described list set mates with at least one concept described, word and at least one concept matching described if navigate described in a list, then using the element in list belonging to described navigation word as kernel entity corresponding at least one concept described, and kernel entity corresponding to each concept forms the entity sets of described concept.

For example, by extracting each list t in list set H, if the navigation word of list t is relevant with the name of concept in domain knowledge base as identical, under then the element set of list t can being assigned to this concept, as the kernel entity under concept, the similarity of record concept and kernel entity is 1.Now, the kernel entity under concept can form entity sets c _t.

Alternatively, for convenience of subsequent calculations, list t can be deleted from list set H.

In a particular application, also may there is the navigation word in one or more list and the unmatched scene of concept, for this reason, for preventing the entity omitting affiliated field, sufficient, complete during entity in guarantee field under concept, aforesaid step also can comprise the following sub-step 1013 shown in Fig. 2 A and sub-step 1014.

If the navigation word that there is at least one list in 1013 described list set does not match with at least one concept described, then obtain the second similarity of the entity sets of element set in list belonging to the navigation word that do not match with at least one concept described and each concept respectively;

1014, for multiple second similarities of each navigation word do not mated, multiple described second similarity of this navigation word sorted according to size, the element in list belonging to this navigation word is as the non-core entity in concept corresponding to the second similarity coming front M position.

M be more than or equal to 1 natural number.In a particular application, M is preferably 5,6 or 8 etc.

Multiple entities corresponding to each concept in the present embodiment can comprise kernel entity and/or non-core entity.

For example, in sub-step 1013 and sub-step 1014, each list t in list set H is extracted, with the element set t of list t _sthe entity sets c of the kernel entity comprised with each concept in domain knowledge base _tcalculate the second similarity:

Such as, sim=|t _s∩ c _t|/| t _s∪ c _t|

Select n (n=5) the individual concept that the second similarity sim is the highest, element set ts is distributed under these concepts, and record the similarity of this sim as entity and concept.

It should be noted that " second " in second similarity at this place is without special implication, be only used to distinguish the explanation of the similarity occurred at diverse location in the present invention.

Current, aforementioned structure domain knowledge base can use.Each concept in traversal domain knowledge base, determines there is corresponding entity in each concept as kernel entity or non-core entity.

If aforesaid concept does not comprise kernel entity and non-core entity, now, need the kernel entity of artificial supplementation and conceptual dependency.

Such as, if described concept does not comprise kernel entity and non-core entity, then supplement kernel entity corresponding to described concept;

Will be understood that, if do not comprise entity (this entity comprises kernel entity and non-core entity) also there is part concept in domain knowledge base under, then can artificial supplementation a part kernel entity, guarantee that in domain knowledge base, each concept comprises entity.

In concrete use procedure, all sub-steps in above-mentioned steps 101 can be repeated, to obtain stable domain knowledge base.

Certainly, the domain knowledge base that also can build in regular update the present embodiment, such as, can upgrade once for one week, or two days upgrade a domain knowledge base, and the present embodiment does not limit it, can arrange according to actual needs.

Thus, in the concept in the domain knowledge base built in the present embodiment, all include entity, and multiple entities corresponding to described concept comprise: described kernel entity and/or described non-core entity.

In other embodiments, build domain knowledge base also to build by the mode of Domain ontology model (ontology).Such as, there is a large amount of ontology model in the website be correlated with in existing various field, can facilitate and directly capture use.

In a kind of possible implementation, aforesaid step 102 can comprise following sub-step 1021 to the sub-step 1025 shown in Fig. 2 B;

1021, obtain the information in social network sites corresponding to described field, determine whether the described information content comprises concept name in described domain knowledge base or entity title;

If the 1022 described information contents comprise described concept name or entity title, then generate expert's candidate collection according to the sender of described information, recipient, and

1023, the third phase calculating described information and described field, like spend, using the third phase of the sender of described information, recipient and described information like spending as a triplet information, generates described information aggregate;

That is, above-mentioned sub-step 1021 to sub-step 1023 needs the expert candidate collection P determined in field, the massage set in field/information aggregate E.

For example, by capturing the message (microblogging) in social networks, word is cut to message content, if the concept name containing domain knowledge base in this message or entity title, then the sender of message and recipient is added expert candidate collection P; Meanwhile, the third phase in this message and field can also be calculated like spending:

m_sim＝sum{sim(w)}/n,

Wherein, sim (w) represents that the third phase in word w in message and field is like spending, and n represents the number of word in message.

Thus, can by tlv triple < sender of the message, message recipient, the third phase of this message adds massage set E like degree m_sim>.

The message at this place can be the one of information.

1024, according to the information in the expert of described expert's candidate collection and described information aggregate, the rank of each expert in described expert's candidate collection is obtained.

For example, the actual social network diagram defined in a field of aforesaid expert candidate collection P, massage set E, utilizes PeopleRank algorithm can obtain the rank score of each expert in expert candidate collection P.

1025, choose the forward X of a rank expert as the expert belonging to information described in described information aggregate, X be more than or equal to 1 natural number.

Will be understood that, aforementioned selected part expert can be optional step.

That is, if expert's candidate collection comprises more expert, but part expert may seldom occur, for this reason, can choose part expert in expert's candidate collection as the conventional expert in expert's question answering system.

The expert of above-mentioned acquisition belongs to the person of sending sending information/message in the professional website in field, or receive the recipient of information/message, therefore, may occur, after some expert only sent information once or twice or thrice, no longer appear in this professional website, Given this, the rank of all experts in expert's candidate collection can be obtained, and the expert of screening part.

Alternatively, also can obtain the concept similarity vector of each expert to all concepts in advance, such as, be obtained by the mode of unshowned sub-step A01 and sub-step A02 in following Fig. 2 B.

A01, for each expert in described expert's candidate collection, obtain all information of each expert in described information aggregate;

A02, according to all information of each expert in described information aggregate and all concepts in described domain knowledge base, obtain the concept similarity vector of each expert to all concepts.

Will be understood that, for the some expert p in expert's candidate collection, collect the massage set E that field that this expert sends is relevant _p, for each the concept c in domain knowledge base, calculate the concept similarity vector CSVec of expert and all concepts:

pc_sim＝|E _p∩c _t|/|E _p∪c _t|,

Wherein, c _trepresent the entity that concept c comprises.

In the present embodiment, concept similarity vector CSVec is the vector of a multidimensional.

The rank of each expert and the rank score of correspondence in field can be specified by the sub-step of above-mentioned steps 102, and then know expert's influence power in the field;

Further, also the clearer expert of learning which sub-direction can be stressed in the field according to aforesaid concept similarity vector.

In the implementation that another kind is possible, aforesaid step 103 can comprise following sub-step 1031 to the sub-step 1033 shown in Fig. 2 C;

1031, cut word process to described problem, obtain the word corresponding with described problem first gathers;

1032, the problem similarity vector of all concepts in described first set and described domain knowledge base is obtained;

1033, according to described concept similarity vector and described problem similarity vector, the first similarity of described expert and described problem is determined.

Thus, the expert's question answering system set up by said method accurately, objectively for user can provide the expert in field belonging to customer problem, and then can improve the matching degree of expert and problem, improves the accuracy rate of answering a question.

Fig. 3 shows the schematic flow sheet of the automatic question-answering method that one embodiment of the invention provides, and as shown in Figure 3, the automatic question-answering method of the present embodiment is as described below.

301, receive the problem of user's input, determine the similarity of each expert in described problem and expert's question answering system, the people that is skillful in described expert field belonging to described problem;

302, described similarity sorted according to size, the expert choosing the similarity that comes top N corresponding answers described problem, N be more than or equal to 1 natural number.

For example, in problem described in determination in step 301 and expert's question answering system, the similarity of expert can realize especially by following not shown sub-step.

3011, cut word process to described problem, obtain the word corresponding with described problem first gathers;

3012, obtain described first set and the problem similarity vector of all concepts in domain knowledge base, described domain knowledge base is the knowledge base comprising multiple entities corresponding at least one concept, at least one concept described obtained in advance in described expert's question answering system;

3013, according to concept similarity vector and the described problem similarity vector of each expert, the similarity of expert and described problem in described expert's question answering system is determined.

It should be noted that the concept similarity vector of each expert in this sub-step 3013 can be to obtain in advance in expert's question answering system.Such as by the sub-step A01 of the step 102 in an aforesaid embodiment of the method and the concept similarity vector of A02 acquisition.

Will be understood that: the concept similarity vector of each expert is that all concepts in all information and described domain knowledge base sent according to this expert obtain in advance; All information that described expert sends are that obtain from described field related web site or comment with described concept or described entity associated information.

Automatic question-answering method in the present embodiment can realize the expert of this area can the technical matters of on-line automatic answer this area, and then improves the accuracy rate that user obtains problem answers.

Fig. 4 shows the structural representation of expert's question answering system that one embodiment of the invention provides, as shown in Figure 4, expert's question answering system of the present embodiment can comprise: construction of knowledge base unit 41, domain expert's determining unit 42, problem receiving element 43, similarity determining unit 44 and expert's matching unit 45;

Wherein, construction of knowledge base unit 41 is for building domain knowledge base, and described domain knowledge base comprises: at least one concept in described field, the multiple entities corresponding with each concept;

Domain expert's determining unit 42 is for the information aggregate according to described field, determine the expert belonging to information described in described information aggregate, information in described information aggregate is that obtain the website or comment of being correlated with from described field with described concept or described entity associated information, and described expert is the person of sending of described information or the recipient of described information;

Problem receiving element 43 is for receiving the problem of user's input;

Similarity determining unit 44 is for determining the first similarity of the expert that described domain expert's determining unit 42 is determined and the problem that described problem receiving element 43 receives;

Expert's matching unit 45 sorts according to size for the first similarity determined by described similarity determining unit 44, and the expert choosing the first similarity of coming top N corresponding answers described problem, N be more than or equal to 1 natural number.

For example, described construction of knowledge base unit 41 specifically for, the website corresponding to described field is carried out orientation and is captured, and set up the list set of two tuple form, the list in described list set comprises: the element set of multiple element compositions that navigation word, described navigation word are corresponding;

In a kind of possible implementation, above-mentioned construction of knowledge base unit 41 also for, when the navigation word that there is at least one list in described list set does not match with at least one concept described, then obtain the second similarity of the entity sets of element set in list belonging to the navigation word that do not match with at least one concept described and each concept respectively;

Alternatively, described construction of knowledge base unit 41 also for

In the optional implementation of the second, described domain expert's determining unit 42 specifically for, obtain the information in social network sites corresponding to described field, determine whether the described information content comprises concept name in described domain knowledge base or entity title;

And/or,

Alternatively, described domain expert's determining unit 42 also for, for each expert in described expert's candidate collection, obtain all information of each expert in described information aggregate;

In addition, aforementioned similarity determining unit 44 specifically for, cut word process to the described problem that described problem receiving element receives, obtain the word corresponding with described problem first gathers;

Expert's question answering system of the present embodiment can realize when user asks a question, search the expert in field belonging to described problem, with the problem making this expert answer user's proposition, and then improve the accuracy rate of answering a question, improve the efficiency of expert's question answering system simultaneously.

The device of the present embodiment, may be used for the technical scheme performing embodiment of the method shown in Fig. 1 to Fig. 3, it realizes principle and technique effect is similar, and relevant part illustrates see the part of embodiment of the method, repeats no more herein.

Fig. 5 shows the structural representation of expert's question answering system that another embodiment of the present invention provides, and as shown in Figure 5, expert's question answering system of the present embodiment can comprise: receiving element 51, similarity determining unit 52, selection of specialists unit 53 and answer unit 54;

Wherein, receiving element 51 is for receiving the problem of user's input;

Similarity determining unit 52 for determining the similarity of each expert in described problem and expert's question answering system, the people that is skillful in described expert field belonging to described problem;

Selection of specialists unit 53, for described similarity being sorted according to size, chooses the expert that the similarity that comes top N is corresponding, N be more than or equal to 1 natural number;

Answer unit 54 answers described problem for the expert making described selection of specialists unit 53 and choose for described user.

In the present embodiment, described similarity determining unit 52 specifically for, cut word process to described problem, obtain the word corresponding with described problem first gathers;

Expert's question answering system in the present embodiment can realize the expert of this area can the technical matters of on-line automatic answer this area, and then improves the accuracy rate that user obtains problem answers.

In addition, the present invention also can adopt another embodiment to be described expert's question answering system of aforementioned structure.As shown in Figure 6, expert's question answering system builds according to the division of different field, constructed expert's question answering system comprises processed offline part and online processing section, for processed offline part, at each specific area, described system comprises and builds domain knowledge library module, expert's mark module, and module is putd question to for user in online processing section, as shown in Figure 6.

Wherein, structure domain knowledge library module and expert's mark module are built on backstage by expert's question answering system, so that user utilizes user to put question to module to ask a question and obtains answer and/or recommend expert.

Expert's question answering system is towards multiple field, and for each specific area, corresponding structure domain knowledge library module and expert's mark module, build domain knowledge library module and comprise concept, be allocated in the entity under concept and mutual relationship between the two.For the field in domain knowledge base, concept and relation, can be determined by artificial or automanual mode according to the location of expert's question answering system, function etc.

Some Domain ontology models of current existence, also can directly use.For the entity in knowledge base, be the key component building domain knowledge library module, be also improve recommend expert with the basic content of correlativity of asking a question, its content is mainly derived from the professional website of specific area.In a preferred embodiment of the invention, the entity excavating specific area is completed by following not shown step:

S601, the relevant speciality website of specific area carried out to orientation and capture, extract the list t of the navigation of its classification and set form, obtain list set H.

Each list in the present embodiment can be a binary group information, such as: list t=< navigates word, { the element set ts}> occurred in list, finally obtain the list set H={t|t=< navigation word of a list t, { the element set ts}> occurred in list.

The selection of professional website can be determined by artificial or automanual mode, and constantly updates along with the development of technology and society.Orientation in the present embodiment captures as prior art, and the present embodiment does not describe in detail it.

S602, for each list t in list set H, judge whether the name of the concept in the navigation word of list t and domain knowledge base is correlated with, if the navigation word of list and described conceptual dependency, then the element set of this list t is assigned under this concept, as kernel entity.

In addition, in the present embodiment, the kernel entity of each concept can form the entity sets ct of this concept.

In the present embodiment, the similarity that also can record concept and kernel entity is 1, the list t of this and conceptual dependency can be deleted from list set H further.

If also exist in S603 list set H not with the list of conceptual dependency in domain knowledge base, then extract remaining each list t do not mated in set H, the entity sets ct calculating similarity by each concept in element set ts and the S602 of list t:

sim＝|t _s∩c _t|/|t _s∪c _t|,

Thus, n the concept that similarity sim is the highest can be selected, the element set ts of this list is assigned to as non-core entity under these concepts, and using the sim that the calculates similarity as non-core entity and concept.

In the present embodiment, the n in this step determines according to the accuracy of expert's question answering system and considering of operation efficiency, preferred n=5.

If S604 is after aforesaid step S602 and step 603, also there is part concept in domain knowledge base and do not have entity, for guaranteeing that each concept all comprises entity, can entity under this concept of artificial supplementation.

Each concept in above S601 ~ S604 step traversal domain knowledge base, and then form stable knowledge base.

According to the knowledge base that above step is formed, based on professional website, for specific area and each concept wherein, according to similarity algorithm, quantized the degree of correlation of entity under each concept and concept, thus the degree of correlation of asking a question for the objective expert of determination and user in follow-up is laid a good foundation.

It should be noted that the professional website in the present embodiment is not limited to the professional website of technical field, as long as comprise the information of the problem that can solve specific area, the website of content, the no matter size of contained quantity of information, can professional website alternatively.Those skilled in the art reasonably can select according to the coverage rate of question answering system and precise requirements.

In expert's mark module, expert's mark comprises two parts: 1) expert's rank score PR2) expert's concept affinity score CSVec.The former expert's influence power in the field, which sub-direction the latter more specifically quantization means expert bias toward in field, namely with the degree of correlation of each concept in this field.

About expert's rank score PR, first, need to determine expert candidate collection P, massage set E.First expert's question answering system captures the message in social networks, such as microblogging, BBS etc., cuts word to message content, if the concept name containing knowledge base in this message or entity title, then the transmission of message and recipient is added expert candidate collection P; Meanwhile, the similarity in this message and field is calculated:

m_sim＝sum{sim(w)}/n,

Wherein, sim (w) represents the similarity in word w and field, and n represents the number of word in message.By tlv triple < sender of the message, message recipient, message similarity m_sim> adds massage set E.

Expert candidate collection P, the actual social network diagram defined in a field of message E, as shown in Figure 7, in the network for five candidate experts, for P1, P2 two candidate experts, P1, arrow between P2 represents that P1 have issued a piece of news to P2 by social networks, so P1 is as sender of the message, P2 joins in candidate collection P as message recipient, simultaneously, this message is after cutting word, obtain n word, calculate the similarity in each word and the field obtained, thus obtain the similarity m_sim12 in this message and affiliated field, obtain tlv triple <P1 thus, P2, m_sim12>, this tlv triple is added in massage set E.Similar, P1 also sends a piece of news to P3, the similarity in the message sent and affiliated field is m_sim13, P4 also sends a piece of news to P2, the similarity in the message sent and affiliated field is m_sim42, and these senders, recipient and message similarity are also correspondingly joined message in conjunction with in E respectively as tlv triple.

Thus, expert candidate collection P obtained above, massage set E define the social network diagram in affiliated field, and existing PeopleRank algorithm preferably can be utilized can to calculate the rank score of each expert.

About the concept similarity vector CSVec of expert, for the some expert p in expert candidate collection P, collect the massage set Ep of all message in affiliated field that this expert p sends, for each the concept c in field, calculate the similarity of expert and concept c:

pc_sim＝|E _p∩c _t|/|E _p∪c _t|

Wherein, c _trepresent the entity (this entity comprises aforesaid kernel entity and non-core entity) that concept c comprises.

Obtain the concept similarity vector CSVec of expert thus.

In the question answering system of prior art, still do not exist by computer network by calculating the objective expert correlation determined and " importance " index.Expert's mark module in expert's question answering system of the embodiment of the present invention, by cutting word to message in social networks, and correlation analysis is carried out in the field in the word obtained and base module, quantize the expert's rank (message sending in social networks according to expert or receive is determined) determining specific area on the one hand, on the other hand, quantize to determine specific specialists in each notional degree of correlation, thus can when recommending the expert in field belonging to customer problem for user, there is objective recommendation order, the different manifestations of expert in different field can be considered when recommending simultaneously.

It should be noted that due to user submit a question on the computer network time, the expert recommended often that user pays close attention to most provides answer as far as possible rapidly, instead of the expert recommended has how strong academic aptitude or professional knowledge in this field.Therefore, in an embodiment of the present invention, when determining expert's rank score, main influence factor is the active degree of expert in social networks (sending out number of times and the object of message) and pays close attention to and concerned degree, instead of the academic aptitude of expert's reality or how many to the grasp of relevant knowledge.But, generally speaking, the expert that rank score is high in social networks, often its to send out the degree that message extensively approved by society corresponding also higher, therefore, expert's rank mode in the embodiment of this area can guarantee that the possibility that the enquirement of user is answered by expert is large, also has higher objectivity and accuracy simultaneously.

In user's question and answer module, user can submit a question Q, and the word unit of cutting in this module cuts word to problem Q, obtains the set Qw={w} of word, and then set of computations Qw determines the similarity of each concept in field with institute, obtain the vectorial QSVec of problem concept similarity;

For each domain expert p, calculate the similarity of this expert and this problem:

rank_sim＝(CSVec*QSVec)*PR，

Wherein, CSVec*QSVec represents similarity vector inner product, and PR is specialist field rank score, chooses the expert p that rank_sim is the highest, as recommendation expert.

Now, the problem of user, by social networks application programming interface (ApplicationProgrammingInterface is called for short API) interface, is pushed to expert, is answered by expert, thus complete whole question answering process by expert's question answering system.

In user's question and answer module in embodiments of the present invention, calculate by doing inner product to expert's concept similarity CSVec vector sum problem concept similarity vector QSVec, quantize the degree of correlation of the problem that determines, concept and expert, simultaneously in conjunction with the expert's " importance " determined in expert's mark module and correlation metric PR, thus improve most possibly user ask a question and determine field expert between degree of correlation, thus can objectively for user recommends expert.

Illustrate the construction method of a kind of expert's question answering system based on computer network of C10 above by multiple embodiment, the method can comprise:

According to the information aggregate in described field, determine the expert belonging to information described in described information aggregate, information in described information aggregate is that obtain from the website be correlated with in described field with described concept or described entity associated information, and described expert is the person of sending of described information or the recipient of described information;

Other embodiments of the present invention also disclose:

C11, method according to aforementioned C10, described structure domain knowledge base, comprising:

C12, method according to aforementioned C11, described structure domain knowledge base, also comprises:

Wherein, M be more than or equal to 1 natural number.

C13, method according to aforementioned C11 or C12, described structure domain knowledge base, also comprises:

C14, method according to aforementioned C10, the described information aggregate according to described field, determine the expert belonging to information described in described information aggregate, comprise: obtain the information in social network sites corresponding to described field, determine whether the described information content comprises concept name in described domain knowledge base or entity title;

And/or,

C15, method according to aforementioned C14, also comprise:

C16, method according to aforementioned C15, if described expert's question answering system receives problem, determine the first similarity of described expert and described problem, comprising:

Other embodiments of the present invention also disclose:

D17, a kind of automatic question-answering method, comprising:

Receive the problem of user's input;

Determine the similarity of each expert in described problem and expert's question answering system, the people that is skillful in described expert field belonging to described problem;

Described similarity is sorted according to size, chooses the expert that the similarity that comes top N is corresponding, N be more than or equal to 1 natural number;

Expert selected by recommendation answers described problem for described user.

D18, method according to aforementioned D17, the described similarity determining expert in described problem and expert's question answering system, comprising:

Algorithm involved in the embodiment of the present invention or display intrinsic not relevant to any certain computer, virtual system or other equipment.Various general-purpose system also can with use based on together with this teaching.According to description above, the structure constructed required by this type systematic is apparent.In addition, the present invention is not also for any certain programmed language.It should be understood that and various programming language can be utilized to realize content of the present invention described here.

In instructions of the present invention, describe a large amount of detail.But can understand, embodiments of the invention can be put into practice when not having these details.In some instances, be not shown specifically known method, structure and technology, so that not fuzzy understanding of this description.

Similarly, be to be understood that, to disclose and to help to understand in each inventive aspect one or more to simplify the present invention, in the description above to exemplary embodiment of the present invention, each feature of the present invention is grouped together in single embodiment, figure or the description to it sometimes.But, the method for the disclosure should not explained the following intention in reflection: namely the present invention for required protection requires feature more more than the feature clearly recorded in each claim.Or rather, as claims below reflect, all features of disclosed single embodiment before inventive aspect is to be less than.Therefore, the claims following embodiment are incorporated to this embodiment thus clearly, and wherein each claim itself is as independent embodiment of the present invention.

It will be understood by those skilled in the art that adaptively to change the module in the equipment in embodiment and they are arranged and be in one or more equipment that this embodiment is different.Module in embodiment or unit or assembly can be combined into a module or unit or assembly, and multiple submodule or subelement or sub-component can be put them in addition.Except at least some in such feature and/or process or unit is mutually exclusive part, any combination can be adopted to combine all processes of all features disclosed in this instructions (comprising adjoint claim, summary and accompanying drawing) and so disclosed any method or equipment or unit.Unless expressly stated otherwise, each feature disclosed in this instructions (comprising adjoint claim, summary and accompanying drawing) can by providing identical, alternative features that is equivalent or similar object replaces.

In addition, those skilled in the art can understand, although embodiments more described herein to comprise in other embodiment some included feature instead of further feature, the combination of the feature of different embodiment means and to be within scope of the present invention and to form different embodiments.Such as, in the following claims, the one of any of embodiment required for protection can use with arbitrary array mode.

All parts embodiment of the present invention with hardware implementing, or can realize with the software module run on one or more processor, or realizes with their combination.It will be understood by those of skill in the art that the some or all functions that microprocessor or digital signal processor (DSP) can be used in practice to realize according to the some or all parts in the equipment of a kind of browser terminal of the embodiment of the present invention.The present invention can also be embodied as part or all equipment for performing method as described herein or device program (such as, computer program and computer program).Realizing program of the present invention and can store on a computer-readable medium like this, or the form of one or more signal can be had.Such signal can be downloaded from internet website and obtain, or provides on carrier signal, or provides with any other form.

The present invention will be described instead of limit the invention to it should be noted above-described embodiment, and those skilled in the art can design alternative embodiment when not departing from the scope of claims.In the claims, any reference symbol between bracket should be configured to limitations on claims.Word " comprises " not to be got rid of existence and does not arrange element in the claims or step.Word "a" or "an" before being positioned at element is not got rid of and be there is multiple such element.The present invention can by means of including the hardware of some different elements and realizing by means of the computing machine of suitably programming.In the unit claim listing some devices, several in these devices can be carry out imbody by same hardware branch.Word first, second and third-class use do not represent any order.Can be title by these word explanations.

Last it is noted that above each embodiment is only in order to illustrate technical scheme of the present invention, be not intended to limit; Although with reference to foregoing embodiments to invention has been detailed description, those of ordinary skill in the art is to be understood that: it still can be modified to the technical scheme described in foregoing embodiments, or carries out equivalent replacement to wherein some or all of technical characteristic; And these amendments or replacement, do not make the essence of appropriate technical solution depart from the scope of various embodiments of the present invention technical scheme, it all should be encompassed in the middle of the scope of claim of the present invention and instructions.

Claims

1., based on expert's question answering system of computer network, it is characterized in that, comprising:

Construction of knowledge base unit, for building domain knowledge base, described domain knowledge base comprises: at least one concept in described field, the multiple entities corresponding with each concept;

Domain expert's determining unit, for the information aggregate according to described field, determine the expert belonging to information described in described information aggregate, information in described information aggregate is that obtain from the website be correlated with in described field with described concept or described entity associated information, and described expert is the person of sending of described information or the recipient of described information;

Problem receiving element, for receiving the problem of user's input;

2. system according to claim 1, is characterized in that, described construction of knowledge base unit, specifically for

3. system according to claim 2, is characterized in that, described construction of knowledge base unit, also for

4. the system according to Claims 2 or 3, is characterized in that, described construction of knowledge base unit, also for

5. system according to claim 1, is characterized in that, described domain expert's determining unit, specifically for

The third phase calculating described information and described field like spend, using the third phase of the sender of described information, recipient and described information like spending as a triplet information, information generated set;

According to the expert in described expert's candidate collection and the information in described information aggregate, obtain the rank of each expert in described expert's candidate collection;

And/or,

6. system according to claim 5, is characterized in that, described domain expert's determining unit, also for

7. system according to claim 6, is characterized in that, described similarity determining unit, specifically for

8. expert's question answering system, is characterized in that, comprising:

Receiving element, for receiving the problem of user's input;

Answer unit, answers described problem for recommending the expert of described selection of specialists unit selection for described user.

9. system according to claim 8, is characterized in that, described similarity determining unit, specifically for

10., based on a construction method for expert's question answering system of computer network, it is characterized in that, comprising: