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

Abstract

The present invention discloses a computer network-based expert question answering system and construction method, suitable for recommending experts for questioning users according to different fields, wherein the question answering system includes: a knowledge base construction unit, which is used to build a domain knowledge base including concepts and entities The domain expert determination unit is used to determine the experts who belong to the information in the information collection according to the information collection of the field. The information in the information collection is the information associated with concepts or entities obtained from domain-related websites; the question receiving unit uses for receiving the question input by the user; the similarity determining unit is used for determining the first similarity between the expert and the question; the expert matching unit is used for answering the question with the expert with the highest first similarity. The expert question answering system and construction method of the present invention can objectively and accurately recommend experts in the field to which the question belongs to the user, and improve the accuracy rate of the user in solving the question.

Description

Expert question answering system and its construction method based on computer network

technical field

The invention belongs to the field of computer networks, in particular to an expert question answering system based on computer networks and a construction method thereof.

Background technique

With the development of Internet information technology, the ways and objects for Internet users to exchange information are gradually diversified. When a user has a question that needs to be answered, there are many ways to get the answer. The traditional way includes using, for example, telephone, e-mail or other instant messaging tools to ask questions to familiar or acquainted social circles, so as to obtain answers.

A recent common method is that computer network users can also directly submit questions on websites with question-and-answer systems, such as directly inputting the content of questions to be searched on a search website, and the question-and-answer system matches existing question-and-answer databases through keywords. Give search results. The main implementation method is that the question answering system first establishes its own knowledge database, and searches and matches the input keywords through different input methods (text, image), etc., to obtain answers.

However, the above-mentioned traditional methods cannot break through the limitations of familiar or acquainted people. When the question you want to know does not know the answer or cannot be contacted within the existing communication circle, you cannot obtain the answer. Although the above-mentioned second method, which is the most commonly used method recently, overcomes the disadvantages of the traditional method, for example, it can only be matched by keywords, which leads to the fact that the retrieved answer to the question does not match the answer to the question that the user actually wants to obtain.

In view of this, how to obtain answers from experts in the field to which the question belongs when asking a question on a website with a question answering system has become a technical problem that needs to be solved at present.

Contents of the invention

Aiming at the defects in the prior art, the present invention provides an expert question answering system based on computer network and its construction method.

In the first aspect, an embodiment of the present invention provides an expert question answering system based on a computer network, and a knowledge base construction unit is used to construct a domain knowledge base. The domain knowledge base includes: at least one concept in the field, and each concept Corresponding multiple entities;

A field expert determining unit, configured to determine the experts to whom the information in the information set belongs according to the information set in the field, and the information in the information set is information related to the field obtained from websites or comments related to the field The concept or the information associated with the entity, the expert is the sender of the information or the recipient of the information;

a question receiving unit, configured to receive a question input by a user;

a similarity determining unit, configured to determine a first degree of similarity between the expert determined by the domain expert determining unit and the question received by the question receiving unit;

The expert matching unit is used to sort the first similarities determined by the similarity determination unit according to their size, and select the experts corresponding to the first similarities in the top N positions to answer the question, where N is a natural number greater than or equal to 1.

Optionally, the knowledge base construction unit is specifically used for

Carry out directional grabbing to the website corresponding to the field, and set up a form set of two-tuple forms, the forms in the form set include: a navigation word, an element set composed of a plurality of elements corresponding to the navigation word;

Determine whether the navigation word of each form in the form set matches the at least one concept, and if the navigation word in a form matches the at least one concept, use the element in the form to which the navigation word belongs as the The core entity corresponding to the at least one concept, and the core entity corresponding to each concept constitutes the entity set of the concept.

Optionally, the knowledge base construction unit is also used for

When the navigation word of at least one form in the form collection does not match the at least one concept, then obtain the element set in the form to which the navigation word that does not match the at least one concept belongs and each concept the second similarity of the entity set;

For multiple second similarities of each unmatched navigation word, the multiple second similarities of the navigation word are sorted according to size, and the element in the list to which the navigation word belongs is regarded as the first M second similarity. The non-core entity in the concept corresponding to the similarity; M is a natural number greater than or equal to 1.

Optionally, the knowledge base construction unit is also used for

When the concept does not include core entities and non-core entities, supplement the core entities corresponding to the concept;

Wherein, the concept corresponds to multiple entities including: the core entity and/or the non-core entity.

Optionally, the domain expert determines the unit, specifically for

Obtain the information in the social networking site corresponding to the field, and determine whether the information content includes the concept name or entity name in the field knowledge base;

If the information content includes the concept name or entity name, generate a candidate set of experts according to the sender and receiver of the information, and

calculating the third degree of similarity between the information and the field, using the sender of the information, the receiver and the third degree of similarity between the information as a triplet information to generate an information set;

Obtain the ranking of each expert in the expert candidate set according to the experts in the expert candidate set and the information in the information set;

and / or,

The top X experts are selected as the experts to whom the information in the information set belongs, and X is a natural number greater than or equal to 1.

Optionally, the domain expert determination unit is also used for

For each expert in the expert candidate set, obtain all information of each expert in the information set;

According to all information of each expert in the information set and all concepts in the domain knowledge base, the concept similarity vectors of each expert for all concepts are obtained.

Optionally, the similarity determining unit is specifically used for

performing word segmentation processing on the question received by the question receiving unit to obtain a first set of words corresponding to the question;

Obtaining question similarity vectors between the first set and all concepts in the domain knowledge base;

A first degree of similarity between the expert and the question is determined according to the concept similarity vector and the question similarity vector.

In a second aspect, the present invention provides an expert question answering system, comprising:

a receiving unit, configured to receive a question input by a user;

A similarity determination unit, configured to determine the similarity between the question and each expert in the expert question answering system, where the expert is a technically familiar person in the field to which the question belongs;

An expert selection unit, configured to sort the similarities according to their size, and select experts corresponding to the top N similarities, where N is a natural number greater than or equal to 1;

A question answering unit, configured to enable the expert selected by the expert selection unit to answer the question for the user.

Optionally, the similarity determining unit is specifically used for

performing word segmentation processing on the question to obtain a first set of words corresponding to the question;

Obtaining the question similarity vectors between the first set and all the concepts in the domain knowledge base, the domain knowledge base is the pre-acquired knowledge of at least one concept and multiple entities corresponding to the at least one concept in the expert question answering system Library;

According to the conceptual similarity vector of each expert and the similarity vector of the question, determine the similarity between the expert and the question in the expert question answering system; the conceptual similarity vector of each expert is based on all the information sent by the expert and the All the concepts in the domain knowledge base are obtained in advance, and all the information sent by the experts is the information associated with the concepts or the entities obtained from the domain-related websites or comments.

In a third aspect, an embodiment of the present invention provides a method for constructing a computer network-based expert question answering system, including:

Constructing a domain knowledge base, the domain knowledge base includes: at least one concept in the domain, and multiple entities corresponding to each concept;

According to the information collection in the field, determine the experts to whom the information in the information collection belongs, and the information in the information collection is obtained from websites or comments related to the field and associated with the concept or the entity information, the expert is the sender of the information or the recipient of the information;

If the expert question answering system receives a question, determine the first degree of similarity between the expert and the question, sort the first degree of similarity according to size, and select an expert corresponding to the first degree of similarity in the top N positions To answer the question, N is a natural number greater than or equal to 1.

Optionally, the construction of domain knowledge base includes:

Carry out directional grabbing to the website corresponding to the field, and set up a form set of two-tuple forms, the forms in the form set include: a navigation word, an element set composed of a plurality of elements corresponding to the navigation word;

Determine whether the navigation word of each form in the form set matches the at least one concept, and if the navigation word in a form matches the at least one concept, use the element in the form to which the navigation word belongs as the The core entity corresponding to the at least one concept, and the core entity corresponding to each concept constitutes the entity set of the concept.

Optionally, said building domain knowledge base also includes:

If the navigation word of at least one form in the form set does not match the at least one concept, then obtain the element set and the entity of each concept in the form to which the navigation word that does not match the at least one concept belongs the second similarity of the set;

For multiple second similarities of each unmatched navigation word, the multiple second similarities of the navigation word are sorted according to size, and the element in the list to which the navigation word belongs is regarded as the first M second similarity. Non-core entities in the concept corresponding to the similarity;

M is a natural number greater than or equal to 1.

Optionally, said building domain knowledge base also includes:

If the concept does not include core entities and non-core entities, then supplement the core entities corresponding to the concept;

Wherein, the multiple entities corresponding to the concept include: the core entity and/or the non-core entity.

Optionally, the determining, according to the information set in the field, the expert to whom the information in the information set belongs includes:

Obtain the information in the social networking site corresponding to the field, and determine whether the information content includes the concept name or entity name in the field knowledge base;

If the information content includes the concept name or entity name, generate a candidate set of experts according to the sender and receiver of the information, and

calculating a third similarity between the information and the field, and using the sender of the information, the receiver, and the third similarity of the information as a triplet of information to generate the information set;

Obtain the ranking of each expert in the expert candidate set according to the experts in the expert candidate set and the information in the information set;

and / or,

The top X experts are selected as the experts to whom the information in the information set belongs, and X is a natural number greater than or equal to 1.

Optionally, also include:

For each expert in the expert candidate set, obtain all information of each expert in the information set;

According to all information of each expert in the information set and all concepts in the domain knowledge base, the concept similarity vectors of each expert for all concepts are obtained.

Optionally, if the expert question answering system receives a question, determining the first similarity between the expert and the question includes:

performing word segmentation processing on the question to obtain a first set of words corresponding to the question;

Obtaining question similarity vectors between the first set and all concepts in the domain knowledge base;

A first degree of similarity between the expert and the question is determined according to the concept similarity vector and the question similarity vector.

In a fourth aspect, the present invention provides an automatic question answering method, comprising:

Receive a question input by the user, determine the similarity between the question and each expert in the expert question answering system, and the expert is a person familiar with the technology in the field to which the question belongs;

The similarities are sorted by size, and the experts corresponding to the top N similarities are selected to answer the questions, where N is a natural number greater than or equal to 1.

Optionally, the determining the similarity between the question and the experts in the expert question answering system includes:

performing word segmentation processing on the question to obtain a first set of words corresponding to the question;

Obtaining the question similarity vectors between the first set and all the concepts in the domain knowledge base, the domain knowledge base is the pre-acquired knowledge of at least one concept and multiple entities corresponding to the at least one concept in the expert question answering system Library;

According to the conceptual similarity vector of each expert and the similarity vector of the question, determine the similarity between the expert and the question in the expert question answering system; the conceptual similarity vector of each expert is based on all the information sent by the expert and the All the concepts in the domain knowledge base are pre-acquired; all the information sent by the experts is the information associated with the concepts or the entities obtained from the domain-related websites or comments.

It can be seen from the above technical solutions that the computer network-based expert question answering system and its construction method of the present invention, by constructing the domain knowledge base, and determining the experts related to the concepts or entities in the domain knowledge base, and then when receiving the questions raised by the user, , the first similarity between each expert and the question can be determined, and one of the more similar experts can be selected to answer the question raised by the user. Therefore, the above method fully considers the combination of professional websites and social networks, and can provide users with an objective Accurately recommend experts in the field to which the question belongs to answer the user's question, improving the accuracy of solving the user's question.

The above description is only an overview of the technical solution of the present invention. In order to understand the technical means of the present invention more clearly, it can be implemented according to the contents of the description, and in order to make the above and other purposes, features and advantages of the present invention more obvious and understandable, Specific embodiments of the present invention are enumerated below.

Description of drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the accompanying drawings that need to be used in the description of the embodiments. Obviously, the accompanying drawings in the following description are only of the present invention. For some examples, those of ordinary skill in the art can also obtain other drawings based on these drawings on the premise of not paying creative efforts.

Fig. 1 is a schematic flow chart of a construction method of an expert question answering system provided by an embodiment of the present invention;

FIG. 2A is a schematic flow diagram of building a domain knowledge base provided by an embodiment of the present invention;

FIG. 2B is a schematic flow diagram of determining experts in a field provided by an embodiment of the present invention;

FIG. 2C is a schematic flow diagram of matching the questions provided by an embodiment of the present invention with experts in the field;

Fig. 3 is a schematic flow chart of an automatic question answering method provided by an embodiment of the present invention;

FIG. 4 is a schematic structural diagram of an expert question answering system provided by an embodiment of the present invention;

5 is a schematic structural diagram of an expert question answering system provided by another embodiment of the present invention;

FIG. 6 is a schematic structural diagram of an expert question answering system provided by another embodiment of the present invention;

FIG. 7 is a social network diagram formed by an expert candidate set P and a message set E provided by an embodiment of the present invention.

detailed description

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the present invention can be more thoroughly understood and the scope of the present disclosure can be fully conveyed to those skilled in the art.

Fig. 1 shows a schematic flowchart of a method for constructing an expert question answering system provided by an embodiment of the present invention. As shown in Fig. 1 , the method for constructing an expert question answering system in this embodiment is as follows.

101. Build a domain knowledge base, where the domain knowledge base includes: at least one concept in the domain, and multiple entities corresponding to each concept;

102. According to the information collection in the field, determine the experts to whom the information in the information collection belongs, and the information in the information collection is obtained from websites or comments related to the field and related to the concept or the information associated with entities, said expert being the sender of said information or the recipient of said information;

103. If the expert question answering system receives a question, determine the first degree of similarity between the expert and the question, sort the first degree of similarity according to size, and select the top N positions corresponding to the first degree of similarity experts to answer the questions.

Wherein, N is a natural number greater than or equal to 1, and the value of N can be set according to actual needs.

In this embodiment, it is preferred that the expert ranked first answers the user's questions, but it may happen that the expert ranked first is answering the questions of other users, and at this time, the expert ranked second can be selected to answer User's question.

The construction method of the expert question answering system in this embodiment, by constructing the domain knowledge base, and determining the experts related to the concepts or entities in the domain knowledge base, and then when receiving the question raised by the user, can determine the relationship between each expert and the question The first degree of similarity selects one of the similar experts to answer the questions raised by the user, fully considering the combination of professional websites and social networks, and can objectively and accurately recommend experts in the field to which the question belongs to answer the user's question, improving the Accuracy of solving user problems.

In this embodiment, the foregoing domain knowledge base may include concepts and entities, wherein a relationship is established between concepts and entities. The key to building a domain knowledge base is to obtain entities under concepts.

Usually, at least one concept in the domain can be manually established in the domain knowledge base, and multiple entities under each concept can also be supplemented manually. Of course, since there are many fields and concepts in the domain knowledge base, it is time-consuming and costly to build the domain knowledge base manually. Therefore, in this embodiment, entities under concepts in the domain knowledge base can be obtained automatically, for example, through subordinate sub-steps.

The aforementioned step 101 may also include the following sub-steps 1011 and 1012 shown in FIG. 2A;

1011. Perform directional capture to the website corresponding to the field, and establish a form set of two-tuple forms, where the forms in the form set include: a navigation word, and an element set composed of multiple elements corresponding to the navigation word.

For example, directional capture of various information in a professional website, for example, extracting category navigation in the professional website as a navigation word, creating a form with a fixed format, etc.

The form in this embodiment can be a two-tuple information: t=<navigation word, {element set that appears in the form}>, and multiple two-tuple forms can form a form set, for example, form set H={t |t=<navigation word, {set of elements appearing in the form}>. It should be noted that the set of elements here can be represented by t _s .

1012. Determine whether the navigation word of each form in the form set matches the at least one concept, and if the navigation word in a form matches the at least one concept, set the element in the form to which the navigation word belongs As the core entity corresponding to the at least one concept, and the core entity corresponding to each concept constitutes an entity set of the concept.

For example, by extracting each form t in the form set H, if the navigation word of the form t is related to the name of the concept in the domain knowledge base, the element set of the form t can be assigned to the concept, as The core entity under the concept, record the similarity between the concept and the core entity as 1. At this point, the core entities under the concept can form the entity set c _t .

Optionally, for the convenience of subsequent calculations, the form t can be deleted from the form set H.

In a specific application, there may also be a scenario where the navigation words in one or more forms do not match the concepts. Therefore, in order to prevent the entities in the domain from being missed and ensure that the entities under the concepts in the domain are sufficient and complete, the aforementioned The step may also include sub-step 1013 and sub-step 1014 shown in FIG. 2A described below.

1013. If there is at least one navigation word in the form set that does not match the at least one concept, obtain the element set and each concept in the form that the navigation word that does not match the at least one concept belongs to The second similarity of the entity set;

1014. For multiple second similarities of each unmatched navigation word, sort the multiple second similarities of the navigation word according to size, and the element in the list to which the navigation word belongs is taken as the top M Non-core entities in the concept corresponding to the second similarity.

M is a natural number greater than or equal to 1. In specific applications, M is preferably 5, 6 or 8, etc.

Multiple entities corresponding to each concept in this embodiment may include core entities and/or non-core entities.

For example, in sub-step 1013 and sub-step 1014, extract each form t in the form set H, use the element set t _s of form t and the entity set c _t of core entities contained in each concept in the domain knowledge base Calculate the second similarity:

For example, sim=|t _s ∩c _t |/|t _s ∪c _t |

Select n (n=5) concepts with the highest second similarity sim, assign the element set ts to these concepts, and record the similarity between the sim as an entity and the concept.

It should be noted that the "second" in the second similarity here has no special meaning, and is only used to distinguish the description of the similarity appearing in different positions in the present invention.

Currently, the aforementioned construction of the domain knowledge base is already available. Traverse each concept in the domain knowledge base, and make sure that each concept has a corresponding entity such as a core entity or a non-core entity.

If the aforementioned concepts do not include core entities and non-core entities, at this time, it is necessary to manually supplement the core entities related to the concept.

For example, if the concept does not include core entities and non-core entities, supplement the core entities corresponding to the concept;

It is understandable that if there are some concepts in the domain knowledge base that do not contain entities (the entities include core entities and non-core entities), some core entities can be manually supplemented to ensure that each concept in the domain knowledge base contains entities .

In a specific use process, all the sub-steps in the above step 101 can be repeated to obtain a stable domain knowledge base.

Of course, the domain knowledge base built in this embodiment can also be updated regularly, for example, once a week, or once every two days, which is not limited in this embodiment and can be set according to actual needs.

Therefore, the concepts in the domain knowledge base constructed in this embodiment all include entities, and the multiple entities corresponding to the concepts include: the core entity and/or the non-core entity.

In other embodiments, the construction of the domain knowledge base can also be constructed in the form of a domain ontology model (ontology). For example, there are already a large number of ontology models in various domain-related websites, which can be easily grabbed and used directly.

In a possible implementation manner, the aforementioned step 102 may include the following sub-steps 1021 to 1025 shown in FIG. 2B;

1021. Obtain the information in the social networking site corresponding to the domain, and determine whether the information content includes the concept name or entity name in the domain knowledge base;

1022. If the information content includes the concept name or entity name, generate an expert candidate set according to the sender and receiver of the information, and

1023. Calculate the third degree of similarity between the information and the field, and use the sender and receiver of the information and the third degree of similarity between the information as a triplet information to generate the information set;

That is to say, the above sub-steps 1021 to 1023 need to determine the expert candidate set P in the field and the message set/information set E in the field.

For example, by grabbing messages (microblogging) in social networks, word segmentation is performed on the message content, and if the message contains the concept name or entity name of the domain knowledge base, the sender and receiver of the message are added to the expert Candidate set P; at the same time, the third similarity between the message and the field can also be calculated:

m_sim=sum{sim(w)}/n,

Among them, sim(w) represents the third similarity between the word w in the message and the domain, and n represents the number of words in the message.

Thus, the triplet <message sender, message receiver, third similarity m_sim of the message> can be added to the message set E.

The message here can be a kind of information.

1024. Obtain the ranking of each expert in the expert candidate set according to the experts in the expert candidate set and the information in the information set.

For example, the aforementioned expert candidate set P and message set E actually form a social network graph in a field, and the ranking score of each expert in the expert candidate set P can be obtained by using the PeopleRank algorithm.

1025. Select top X experts as the experts to whom the information in the information set belongs, where X is a natural number greater than or equal to 1.

It can be understood that the aforementioned selection of some experts may be an optional step.

That is to say, if there are many experts in the expert candidate set, some experts may not appear frequently. Therefore, some experts in the expert candidate set can be selected as common experts in the expert question answering system.

The experts obtained above belong to the senders of information/messages in the professional websites in the field, or the recipients of information/messages, so it may appear that some experts only send out information once or two or three times, and then no longer appear In this professional website, in view of this, the rankings of all experts in the expert candidate set can be obtained, and part of the experts can be screened.

Optionally, the conceptual similarity vectors of each expert for all concepts can also be obtained in advance, for example, by means of sub-step A01 and sub-step A02 not shown in FIG. 2B below.

A01. For each expert in the expert candidate set, obtain all information of each expert in the information set;

A02. According to all information of each expert in the information set and all concepts in the domain knowledge base, obtain concept similarity vectors of each expert for all concepts.

It is understandable that, for an expert p in the expert candidate set, collect the domain-related message set E _p sent by the expert, and for each concept c in the domain knowledge base, calculate the concept similarity vector CSVec between the expert and all concepts :

pc_sim＝|E _p ∩c _t |/|E _p ∪c _t |,

Among them, c _t represents the entity contained in the concept c.

In this embodiment, the conceptual similarity vector CSVec is a multidimensional vector.

Through the sub-steps of step 102 above, the ranking and corresponding ranking scores of experts in the field can be clarified, and then the influence of the experts in the field can be known;

Furthermore, according to the aforementioned conceptual similarity vectors, it can be more clearly known which sub-directions experts focus on in this field.

In another possible implementation, the foregoing step 103 may include the following sub-steps 1031 to 1033 shown in FIG. 2C ;

1031. Perform word segmentation processing on the question to obtain a first set of words corresponding to the question;

1032. Obtain question similarity vectors between the first set and all concepts in the domain knowledge base;

1033. Determine a first degree of similarity between the expert and the question according to the concept similarity vector and the question similarity vector.

Therefore, the expert question answering system established by the above method can accurately and objectively provide users with experts in the field to which the user's question belongs, thereby improving the matching degree between experts and questions, and improving the accuracy of answering questions.

FIG. 3 shows a schematic flowchart of an automatic question answering method provided by an embodiment of the present invention. As shown in FIG. 3 , the automatic question answering method in this embodiment is as follows.

301. Receive a question input by the user, and determine the similarity between the question and each expert in the expert question answering system, where the expert is a technically familiar person in the field to which the question belongs;

302. Sort the similarities according to their size, and select experts corresponding to the top N similarities to answer the question, where N is a natural number greater than or equal to 1.

For example, determining the similarity between the question and the experts in the expert question answering system in step 301 may be specifically implemented through sub-steps not shown in the figure below.

3011. Perform word segmentation processing on the question to obtain a first set of words corresponding to the question;

3012. Obtain question similarity vectors between the first set and all concepts in the domain knowledge base, where the domain knowledge base is pre-acquired in the expert question answering system including at least one concept and multiple entities corresponding to the at least one concept knowledge base;

3013. According to the concept similarity vector of each expert and the question similarity vector, determine the similarity between the expert and the question in the expert question answering system.

It should be noted that the conceptual similarity vector of each expert in this sub-step 3013 may be pre-acquired in the expert question answering system. For example, the conceptual similarity vectors obtained through sub-steps A01 and A02 of step 102 in one of the aforementioned method embodiments.

It can be understood that: the concept similarity vector of each expert is pre-acquired according to all the information sent by the expert and all the concepts in the domain knowledge base; The information associated with the concept or the entity acquired in .

The automatic question answering method in this embodiment can enable experts in the field to automatically answer technical questions in the field online, thereby improving the accuracy of the user's answer to the question.

Figure 4 shows a schematic structural diagram of an expert question answering system provided by an embodiment of the present invention. As shown in Figure 4, the expert question answering system of this embodiment may include: a knowledge base construction unit 41, a domain expert determination unit 42, and a question receiving unit 43. Similarity determination unit 44 and expert matching unit 45;

Wherein, the knowledge base construction unit 41 is used to construct a domain knowledge base, and the domain knowledge base includes: at least one concept in the domain, and multiple entities corresponding to each concept;

The domain expert determining unit 42 is configured to determine the experts to whom the information in the information collection belongs according to the information collection in the field, and the information in the information collection is obtained from websites or comments related to the field. The concept or the information associated with the entity, the expert is the sender of the information or the recipient of the information;

The question receiving unit 43 is used for receiving the question input by the user;

The similarity determining unit 44 is used to determine the first similarity between the expert determined by the domain expert determining unit 42 and the question received by the question receiving unit 43;

The expert matching unit 45 is used to sort the first similarities determined by the similarity determining unit 44 according to size, and select the experts corresponding to the first N similarities to answer the questions, and N is a natural number greater than or equal to 1 .

For example, the knowledge base construction unit 41 is specifically configured to perform directional crawling to the website corresponding to the field, and establish a form set of two-tuple forms, and the forms in the form set include: navigation words, the An element set composed of multiple elements corresponding to the navigation word;

Determine whether the navigation word of each form in the form set matches the at least one concept, and if the navigation word in a form matches the at least one concept, use the element in the form to which the navigation word belongs as the The core entity corresponding to the at least one concept, and the core entity corresponding to each concept constitutes the entity set of the concept.

In a possible implementation, the above-mentioned knowledge base construction unit 41 is further configured to, when there is at least one navigation word in a form in the form set that does not match the at least one concept, respectively acquire The second similarity between the element set in the form to which the at least one concept matches the navigation word and the entity set of each concept;

For multiple second similarities of each unmatched navigation word, the multiple second similarities of the navigation word are sorted according to size, and the element in the list to which the navigation word belongs is regarded as the first M second similarity. The non-core entity in the concept corresponding to the similarity; M is a natural number greater than or equal to 1.

Optionally, the knowledge base construction unit 41 is also used to

When the concept does not include core entities and non-core entities, supplement the core entities corresponding to the concept;

Wherein, the concept corresponds to multiple entities including: the core entity and/or the non-core entity.

In a second optional implementation manner, the domain expert determination unit 42 is specifically configured to acquire information in social networking sites corresponding to the domain, and determine whether the information content includes concept names in the domain knowledge base or entity name;

If the information content includes the concept name or entity name, generate a candidate set of experts according to the sender and receiver of the information, and

calculating the third degree of similarity between the information and the field, using the sender of the information, the receiver and the third degree of similarity between the information as a triplet information to generate an information set;

Obtain the ranking of each expert in the expert candidate set according to the experts in the expert candidate set and the information in the information set;

and / or,

The top X experts are selected as the experts to whom the information in the information set belongs, and X is a natural number greater than or equal to 1.

Optionally, the domain expert determination unit 42 is further configured to, for each expert in the expert candidate set, obtain all information of each expert in the information set;

According to all information of each expert in the information set and all concepts in the domain knowledge base, the concept similarity vectors of each expert for all concepts are obtained.

In addition, the aforementioned similarity determination unit 44 is specifically configured to perform word segmentation processing on the question received by the question receiving unit to obtain a first set of words corresponding to the question;

Obtaining question similarity vectors between the first set and all concepts in the domain knowledge base;

A first degree of similarity between the expert and the question is determined according to the concept similarity vector and the question similarity vector.

The expert question answering system in this embodiment can realize that when a user asks a question, find an expert in the field to which the question belongs, so that the expert can answer the question raised by the user, thereby improving the accuracy of answering the question, and improving the performance of the expert question answering system at the same time. efficiency.

The device of this embodiment can be used to implement the technical solutions of the method embodiments shown in Figures 1 to 3, and its implementation principles and technical effects are similar, and for relevant parts, please refer to the part of the description of the method embodiments, and will not repeat them here .

Figure 5 shows a schematic structural diagram of an expert question answering system provided by another embodiment of the present invention, as shown in Figure 5, the expert question answering system of this embodiment may include: a receiving unit 51, a similarity determining unit 52, and an expert selecting unit 53 and question answering unit 54;

Wherein, the receiving unit 51 is used for receiving the question input by the user;

The similarity determining unit 52 is used to determine the similarity between the question and each expert in the expert question answering system, and the expert is a technical person familiar with the field to which the question belongs;

The expert selection unit 53 is used to sort the similarity according to the size, and select the expert corresponding to the similarity in the top N positions, where N is a natural number greater than or equal to 1;

The question answering unit 54 is configured to make the expert selected by the expert selecting unit 53 answer the question for the user.

In this embodiment, the similarity determining unit 52 is specifically configured to perform word segmentation processing on the question to obtain a first set of words corresponding to the question;

Obtaining the question similarity vectors between the first set and all the concepts in the domain knowledge base, the domain knowledge base is the pre-acquired knowledge of at least one concept and multiple entities corresponding to the at least one concept in the expert question answering system library;

According to the conceptual similarity vector of each expert and the similarity vector of the question, determine the similarity between the expert and the question in the expert question answering system; the conceptual similarity vector of each expert is based on all the information sent by the expert and the All the concepts in the domain knowledge base are obtained in advance, and all the information sent by the experts is the information associated with the concepts or the entities obtained from the domain-related websites or comments.

The expert question answering system in this embodiment can enable experts in the field to automatically answer technical questions in the field online, thereby improving the accuracy rate of users obtaining answers to questions.

The device of this embodiment can be used to implement the technical solutions of the method embodiments shown in Figures 1 to 3, and its implementation principles and technical effects are similar, and for relevant parts, please refer to the part of the description of the method embodiments, and will not repeat them here .

In addition, another embodiment of the present invention may be used to describe the expert question answering system constructed above. As shown in Figure 6, the expert question answering system is constructed according to the division of different fields. The constructed expert question answering system includes an offline processing part and an online processing part. For the offline processing part, in each specific field, the system includes building domain knowledge The library module, the expert score module, and the online processing part are the user question module, as shown in Figure 6.

Among them, the building domain knowledge base module and the expert score module are built by the expert question answering system in the background, so that users can use the user question module to ask questions and obtain answers and/or recommend experts.

The expert question answering system is oriented to multiple fields. For each specific field, a domain knowledge base module and an expert score module are correspondingly constructed. The domain knowledge base module includes concepts, entities assigned under concepts, and the relationship between the two. The domain, concept and relationship in the domain knowledge base can be determined manually or semi-automatically according to the positioning and function of the expert question answering system.

Some existing domain ontology models can also be used directly. For the entities in the knowledge base, it is a key part of building the domain knowledge base module, and it is also the basic content to improve the relevance of the recommended experts and the questions asked. The content mainly comes from professional websites in specific fields. In a preferred embodiment of the present invention, mining entities in a specific field is accomplished through the following steps not shown in the figure:

S601. Perform directional capture on related professional websites in a specific field, extract their category navigation and form t in a fixed format, and obtain form set H.

Each form in this embodiment can be a two-tuple information, for example: form t=<navigation word, {element set ts appearing in the form}>, and finally obtain a form set H={t|t of form t ＝<navigation word, {set of elements ts appearing in the form}>.

The selection of professional websites can be determined manually or semi-automatically, and is constantly updated with the development of technology and society. The directional grabbing in this embodiment is a prior art, which is not described in detail in this embodiment.

S602. For each form t in the form set H, determine whether the navigation word of the form t is related to the name of the concept in the domain knowledge base, if the navigation word of the form is related to the concept, set the elements of the form t Assigned under this concept as a core entity.

In addition, the core entities of each concept in this embodiment may form the entity set ct of the concept.

In this embodiment, it can also be recorded that the similarity between the concept and the core entity is 1, and further, the form t related to the concept can be deleted from the form set H.

S603. If there are forms not related to concepts in the domain knowledge base in the form set H, then extract each unmatched form t remaining in the set H, and use the element set ts of the form t and the entity of each concept in S602 Set ct to calculate similarity:

sim＝|t _s ∩c _t |/|t _s ∪c _t |,

Therefore, the n concepts with the highest similarity sim can be selected, the element set ts of the form can be assigned to these concepts as non-core entities, and the calculated sim can be used as the similarity between non-core entities and concepts.

In this embodiment, n in this step is determined according to the comprehensive consideration of the accuracy and operation efficiency of the expert question answering system, preferably n=5.

S604. If after the aforementioned steps S602 and 603, there are still some concepts without entities in the domain knowledge base, in order to ensure that each concept contains entities, the entities under the concept can be supplemented manually.

Through the above steps S601-S604, each concept in the domain knowledge base is traversed to form a stable knowledge base.

Based on the knowledge base formed by the above steps, based on the professional website, for the specific field and each concept in it, according to the similarity algorithm, the degree of correlation between the entity and the concept under each concept is quantified, so as to objectively determine the expert in the follow-up The degree of relevance to the questions asked by users has laid the foundation.

It should be noted that the professional websites in this embodiment are not limited to professional websites in the technical field, as long as the websites contain information and content that can solve problems in specific fields, regardless of the amount of information contained, they can be used as candidates Professional website. Those skilled in the art can make a reasonable choice according to the coverage and accuracy requirements of the question answering system.

In the expert score module, the expert score includes two parts: 1) expert ranking score PR2) expert concept similarity score CSVec. The former indicates the influence of experts in the field, and the latter quantifies which sub-directions experts focus on in the field more specifically, that is, the degree of relevance to each concept in the field.

Regarding the expert ranking score PR, firstly, it is necessary to determine the expert candidate set P and the message set E. The expert question answering system first captures messages in social networks, such as Weibo, BBS, etc., and segments the message content. If the message contains the concept name or entity name of the knowledge base, the sender and receiver of the message are added to the expert list. Candidate set P; at the same time, calculate the similarity between the message and the domain:

m_sim=sum{sim(w)}/n,

Among them, sim(w) represents the similarity between word w and domain, and n represents the number of words in the message. Add the triple <message sender, message receiver, message similarity m_sim> to the message set E.

Expert candidate set P and message E actually form a social network graph in a field, as shown in Figure 7. In this network, five candidate experts are taken as an example. For two candidate experts P1 and P2, P1 and P2 The arrows in between indicate that P1 sends a message to P2 through the social network, then P1 as the message sender and P2 as the message receiver are added to the candidate set P. At the same time, after the message is word-cut, n words are obtained. Calculate the similarity between each word and the field, so as to obtain the similarity m_sim12 between the message and the field to which it belongs, and thus obtain the triplet <P1, P2, m_sim12>, and add the triplet to the message set E. Similarly, P1 also sends a message to P3, and the similarity between the message and the domain it belongs to is m_sim13, and P4 also sends a message to P2, and the similarity between the message it sends and the domain it belongs to is m_sim42. The sender, receiver and message similarity are also added to the message combination E as triplets.

Thus, the above obtained expert candidate set P and message set E form a social network graph in the field, and preferably the existing PeopleRank algorithm can be used to calculate the ranking score of each expert.

Regarding the concept similarity vector CSVec of experts, for an expert p in the expert candidate set P, collect the message set Ep of all the messages in the field sent by the expert p, and for each concept c in the field, calculate the expert and concept Similarity of c:

pc_sim＝|E _p ∩c _t |/|E _p ∪c _t |

Among them, c _t represents the entity contained in the concept c (the entity includes the aforementioned core entity and non-core entity).

From this, the concept similarity vector CSVec of the expert is obtained.

In the question answering system of the prior art, there is no expert correlation and "importance" index objectively determined by computer network through calculation. The expert score module in the expert question answering system of the embodiment of the present invention cuts the words of the message in the social network, and performs correlation analysis between the obtained words and the fields in the knowledge base module. Expert ranking (determined according to the messages sent or received by experts in social networks), on the other hand, quantitatively determines the degree of relevance of a specific expert on each concept, so that when recommending experts in the field to which the user's question belongs, It has an objective recommendation order, and at the same time, the different performances of experts in different fields can be considered when recommending.

It is worth noting that when users submit questions on the computer network, what users pay most attention to is that the recommended experts give answers as quickly as possible, rather than how strong the recommended experts are in this field. expertise. Therefore, in the embodiment of the present invention, when determining the expert ranking score, the main influencing factors are the activity degree of the expert in the social network (the number of times and the object of the message) and the degree of attention and attention, rather than the expert's actual academic ability or mastery of relevant knowledge. However, generally speaking, experts with high ranking scores in social networks tend to have a higher degree of social acceptance of their messages. Therefore, the expert ranking method in this field embodiment can ensure that users' questions Experts are more likely to answer, but also have a high degree of objectivity and accuracy.

In the user question and answer module, the user can submit a question Q, and the word segmentation unit in this module performs word segmentation on the question Q to obtain a word set Qw={w}, and then calculate the similarity between the set Qw and each concept in the determined field , get the question concept similarity vector QSVec;

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

rank_sim=(CSVec*QSVec)*PR,

Among them, CSVec*QSVec represents the inner product of similarity vectors, PR is the ranking score of the expert field, and the expert p with the highest rank_sim is selected as the recommended expert.

At this time, the expert question answering system pushes the user's question to the expert through the social network application programming interface (Application Programming Interface, API for short), and the expert answers, thereby completing the entire question answering process.

In the user question answering module in the embodiment of the present invention, by doing inner product calculation on the expert concept similarity CSVec vector and the question concept similarity vector QSVec, the degree of correlation between the question, the concept and the expert is quantified and determined, and at the same time combined with the expert score module The determined "importance" of experts and the correlation index PR can maximize the degree of correlation between the questions raised by users and the experts in the identified field, so that experts can be objectively recommended to users.

The above has illustrated a method for constructing a computer network-based expert question answering system of C10 through multiple embodiments, and the method may include:

Constructing a domain knowledge base, the domain knowledge base includes: at least one concept in the domain, and multiple entities corresponding to each concept;

According to the information set in the field, determine the expert to whom the information in the information set belongs, and the information in the information set is the information associated with the concept or the entity obtained from a website related to the field , the expert is the sender of the information or the receiver of the information;

If the expert question answering system receives a question, determine the first degree of similarity between the expert and the question, sort the first degree of similarity according to size, and select an expert corresponding to the first degree of similarity in the top N places To answer the question, N is a natural number greater than or equal to 1.

Other embodiments of the present invention also disclose:

C11. According to the method described in the aforementioned C10, the construction of domain knowledge base includes:

Carry out directional grabbing to the website corresponding to the field, and set up a form set of two-tuple forms, the forms in the form set include: a navigation word, an element set composed of a plurality of elements corresponding to the navigation word;

Determine whether the navigation word of each form in the form set matches the at least one concept, and if the navigation word in a form matches the at least one concept, use the element in the form to which the navigation word belongs as the The core entity corresponding to the at least one concept, and the core entity corresponding to each concept constitutes the entity set of the concept.

C12. According to the method described in the aforementioned C11, the construction of the domain knowledge base also includes:

If the navigation word of at least one form in the form set does not match the at least one concept, then obtain the element set and the entity of each concept in the form to which the navigation word that does not match the at least one concept belongs the second similarity of the set;

For multiple second similarities of each unmatched navigation word, the multiple second similarities of the navigation word are sorted according to size, and the element in the list to which the navigation word belongs is regarded as the first M second similarity. Non-core entities in the concept corresponding to the similarity;

Wherein, M is a natural number greater than or equal to 1.

C13. According to the method described in the aforementioned C11 or C12, the construction of the domain knowledge base also includes:

If the concept does not include core entities and non-core entities, supplement the core entities corresponding to the concept;

Wherein, the multiple entities corresponding to the concept include: the core entity and/or the non-core entity.

C14. According to the method described in the aforementioned C10, the determining the expert to whom the information in the information set belongs according to the information set of the field includes: obtaining the information in the social networking site corresponding to the field, and determining the Whether the information content includes the concept name or entity name in the domain knowledge base;

If the information content includes the concept name or entity name, generate a candidate set of experts according to the sender and receiver of the information, and

calculating a third similarity between the information and the field, and using the sender of the information, the receiver, and the third similarity of the information as a triplet of information to generate the information set;

Obtain the ranking of each expert in the expert candidate set according to the experts in the expert candidate set and the information in the information set;

and / or,

The top X experts are selected as the experts to whom the information in the information set belongs, and X is a natural number greater than or equal to 1.

C15. The method according to the aforementioned C14, further comprising:

For each expert in the expert candidate set, obtain all information of each expert in the information set;

According to all information of each expert in the information set and all concepts in the domain knowledge base, the concept similarity vectors of each expert for all concepts are obtained.

C16. According to the method described in C15 above, if the expert question answering system receives a question, determine the first similarity between the expert and the question, including:

performing word segmentation processing on the question to obtain a first set of words corresponding to the question;

Obtaining question similarity vectors between the first set and all concepts in the domain knowledge base;

A first degree of similarity between the expert and the question is determined according to the concept similarity vector and the question similarity vector.

Other embodiments of the present invention also disclose:

D17, an automatic question answering method, comprising:

Receive user input questions;

determining the degree of similarity between the question and each expert in the expert question answering system, the expert being a technically familiar person in the field to which the question belongs;

Sorting the degree of similarity according to size, selecting experts corresponding to the degree of similarity in the top N places, where N is a natural number greater than or equal to 1;

Recommending the selected expert to answer the question for the user.

D18. According to the method described in D17 above, the determination of the similarity between the question and the experts in the expert question answering system includes:

performing word segmentation processing on the question to obtain a first set of words corresponding to the question;

Obtaining the question similarity vectors between the first set and all the concepts in the domain knowledge base, the domain knowledge base is the pre-acquired knowledge of at least one concept and multiple entities corresponding to the at least one concept in the expert question answering system library;

According to the conceptual similarity vector of each expert and the similarity vector of the question, determine the similarity between the expert and the question in the expert question answering system; the conceptual similarity vector of each expert is based on all the information sent by the expert and the All the concepts in the domain knowledge base are pre-acquired; all the information sent by the experts is the information associated with the concepts or the entities obtained from the domain-related websites or comments.

The algorithms or displays involved in the embodiments of the present invention are not inherently related to any particular computer, virtual system, or other device. Various generic systems can also be used with the teachings based on this. The structure required to construct such a system is apparent from the above description. Furthermore, the present invention is not specific to any particular programming language. It should be appreciated that various programming languages can be utilized to implement the inventive concepts described herein.

In the description of the invention, numerous specific details are set forth. However, it is understood that embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure the understanding of this description.

Similarly, it should be appreciated that in the above description of exemplary embodiments of the invention, in order to streamline the present disclosure and to facilitate understanding of one or more of the various inventive aspects, various features of the invention are sometimes grouped together into a single embodiment , figure, or description of it. This method of disclosure, however, is not to be interpreted as reflecting an intention that the claimed invention requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the Detailed Description are hereby expressly incorporated into this Detailed Description, with each claim standing on its own as a separate embodiment of this invention.

Those skilled in the art can understand that the modules in the device in the embodiment can be adaptively changed and installed in one or more devices different from the embodiment. Modules or units or components in the embodiments may be combined into one module or unit or component, and furthermore may be divided into a plurality of sub-modules or sub-units or sub-assemblies. All features disclosed in this specification (including accompanying claims, abstract and drawings) and any method or method so disclosed may be used in any combination, except where at least some of such features and/or processes or units are mutually exclusive. All processes or units of equipment are combined. Each feature disclosed in this specification (including accompanying claims, abstract and drawings) may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.

Furthermore, those skilled in the art will understand that although some embodiments described herein include some features included in other embodiments but not others, combinations of features from different embodiments are meant to be within the scope of the invention. and form different embodiments. For example, in the following claims, any of the claimed embodiments may be used in any combination.

The various component embodiments of the present invention may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art should understand that a microprocessor or a digital signal processor (DSP) can be used in practice to implement some or all functions of some or all components in a browser terminal device according to an embodiment of the present invention . The present invention can also be implemented as an apparatus or an apparatus program (for example, a computer program and a computer program product) for performing a part or all of the methods described herein. Such a program for realizing the present invention may be stored on a computer-readable medium, or may be in the form of one or more signals. Such a signal may be downloaded from an Internet site, or provided on a carrier signal, or provided in any other form.

It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means can be embodied by one and the same item of hardware. The use of the words first, second, and third, etc. does not indicate any order. These words can be interpreted as names.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: It is still possible to modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the technical solutions of the various embodiments of the present invention. All of them should be covered by the scope of the claims and description of the present invention.

Claims (10)

1. A computer network-based expert question answering system, characterized in that it comprises: A knowledge base construction unit, configured to construct a domain knowledge base, the domain knowledge base includes: at least one concept in the domain, and multiple entities corresponding to each concept; A field expert determining unit, configured to determine the experts to whom the information in the information set belongs according to the information set in the field, where the information in the information set is obtained from websites related to the field and related to the concept or information associated with said entity, said expert being the sender of said information or the recipient of said information; a question receiving unit, configured to receive a question input by a user; a similarity determining unit, configured to determine a first degree of similarity between the expert determined by the domain expert determining unit and the question received by the question receiving unit; The expert matching unit is used to sort the first similarities determined by the similarity determination unit according to their size, and select the experts corresponding to the first similarities in the top N positions to answer the question, where N is a natural number greater than or equal to 1. 2. The system according to claim 1, wherein the knowledge base construction unit is specifically used for Carry out directional grabbing to the website corresponding to the field, and set up a form set of two-tuple forms, the forms in the form set include: a navigation word, an element set composed of a plurality of elements corresponding to the navigation word; Determine whether the navigation word of each form in the form set matches the at least one concept, and if the navigation word in a form matches the at least one concept, use the element in the form to which the navigation word belongs as the The core entity corresponding to the at least one concept, and the core entity corresponding to each concept constitutes the entity set of the concept. 3. The system according to claim 2, wherein the knowledge base construction unit is also used for When the navigation word of at least one form in the form collection does not match the at least one concept, then obtain the element set in the form to which the navigation word that does not match the at least one concept belongs and each concept the second similarity of the entity set; For multiple second similarities of each unmatched navigation word, the multiple second similarities of the navigation word are sorted according to size, and the element in the list to which the navigation word belongs is regarded as the first M second similarity. The non-core entity in the concept corresponding to the similarity; M is a natural number greater than or equal to 1. 4. The system according to claim 2 or 3, wherein the knowledge base construction unit is also used for When the concept does not include core entities and non-core entities, supplement the core entities corresponding to the concept; Wherein, the concept corresponds to multiple entities including: the core entity and/or the non-core entity. 5. The system according to claim 1, wherein the domain expert determination unit is specifically used for Obtain the information in the social networking site corresponding to the field, and determine whether the information content includes the concept name or entity name in the field knowledge base; If the information content includes the concept name or entity name, generate a candidate set of experts according to the sender and receiver of the information, and calculating the third similarity between the information and the field, using the sender of the information, the receiver and the third similarity of the information as a triplet information to generate an information set; Obtain the ranking of each expert in the expert candidate set according to the experts in the expert candidate set and the information in the information set; and / or, The top X experts are selected as the experts to whom the information in the information set belongs, and X is a natural number greater than or equal to 1. 6. The system according to claim 5, wherein the domain expert determination unit is also used for For each expert in the expert candidate set, obtain all information of each expert in the information set; According to all information of each expert in the information set and all concepts in the domain knowledge base, the concept similarity vectors of each expert for all concepts are obtained. 7. The system according to claim 6, wherein the similarity determining unit is specifically used for performing word segmentation processing on the question received by the question receiving unit to obtain a first set of words corresponding to the question; Obtaining question similarity vectors between the first set and all concepts in the domain knowledge base; A first degree of similarity between the expert and the question is determined according to the concept similarity vector and the question similarity vector. 8. An expert question answering system, characterized in that, comprising: a receiving unit, configured to receive a question input by a user; A similarity determination unit, configured to determine the similarity between the question and each expert in the expert question answering system, where the expert is a technically familiar person in the field to which the question belongs; An expert selection unit, configured to sort the similarities according to their size, and select experts corresponding to the top N similarities, where N is a natural number greater than or equal to 1; A question answering unit, configured to recommend an expert selected by the expert selection unit to answer the question for the user. 9. The system according to claim 8, wherein the similarity determining unit is specifically used for performing word segmentation processing on the question to obtain a first set of words corresponding to the question; Obtaining the question similarity vectors between the first set and all the concepts in the domain knowledge base, the domain knowledge base is the pre-acquired knowledge of at least one concept and multiple entities corresponding to the at least one concept in the expert question answering system library; According to the conceptual similarity vector of each expert and the similarity vector of the question, determine the similarity between the expert and the question in the expert question answering system; the conceptual similarity vector of each expert is based on all the information sent by the expert and the All the concepts in the domain knowledge base are obtained in advance, and all the information sent by the experts is the information associated with the concepts or the entities obtained from the domain-related websites or comments. 10. A method for constructing an expert question answering system based on a computer network, characterized in that it comprises: Constructing a domain knowledge base, the domain knowledge base includes: at least one concept in the domain, and multiple entities corresponding to each concept; According to the information set in the field, determine the expert to whom the information in the information set belongs, and the information in the information set is the information associated with the concept or the entity obtained from a website related to the field , the expert is the sender of the information or the recipient of the information; If the expert question answering system receives a question, determine the first degree of similarity between the expert and the question, sort the first degree of similarity according to size, and select an expert corresponding to the first degree of similarity in the top N places To answer the question, N is a natural number greater than or equal to 1.