CN106919655B

CN106919655B - Answer providing method and device

Info

Publication number: CN106919655B
Application number: CN201710060256.5A
Authority: CN
Inventors: 朱臻
Original assignee: Netease Hangzhou Network Co Ltd
Current assignee: Netease Hangzhou Network Co Ltd
Priority date: 2017-01-24
Filing date: 2017-01-24
Publication date: 2020-05-19
Anticipated expiration: 2037-01-24
Also published as: CN106919655A

Abstract

The embodiment of the invention provides an answer providing method. The method comprises the following steps: receiving a question submitted by a user; constructing a query chain corresponding to the problem according to the entity nodes and the attribute nodes extracted from the problem; searching answers of the questions in a pre-constructed knowledge graph according to the query chain corresponding to the questions, wherein the answers are used as the answers of the questions obtained through searching based on a knowledge graph; and feeding back the answer of the question obtained by searching based on the knowledge graph searching mode to the user. The method and the device can improve the possibility that the retrieved answer is the answer of the question submitted by the user to a certain extent, thereby improving the accuracy of the answer fed back to the user to a certain extent. In addition, the embodiment of the invention provides an answer providing device.

Description

Answer providing method and device

Technical Field

The embodiment of the invention relates to the technical field of automatic question answering, in particular to an answer providing method and device.

Background

This section is intended to provide a background or context to the embodiments of the invention that are recited in the claims. The description herein is not admitted to be prior art by inclusion in this section.

At present, some user intelligent question-answering systems exist, users input questions, and the intelligent question-answering systems search answers of the questions input by the users in a certain searching mode.

The retrieval modes adopted by the commonly used user intelligent question-answering system are a retrieval mode based on a search engine, a retrieval mode based on deep learning, a retrieval mode based on a local question bank and a retrieval mode based on a question-answering community knowledge base.

However, in the current retrieval modes, the problem submitted by the user is not understood and analyzed, but the answer closest to the problem is directly searched according to the preset template or the keyword, so that the problem with low accuracy of the answer fed back to the user is caused, and the user experience is poor.

Disclosure of Invention

In view of the problems of low accuracy of answers fed back to users according to the questions of using the offered prices and poor user experience in the prior art, the invention provides a set of answer providing method and device, thereby improving the accuracy of answer feedback and the user experience.

In this context, embodiments of the present invention are intended to provide an answer providing method and apparatus.

In a first aspect of embodiments of the present invention, there is provided an answer providing method including:

receiving a question submitted by a user;

constructing a query chain corresponding to the problem according to the entity nodes and the attribute nodes extracted from the problem;

searching answers of the questions in a pre-constructed knowledge graph according to the query chain corresponding to the questions, wherein the answers are used as the answers of the questions obtained through searching based on a knowledge graph;

and feeding back the answer of the question obtained by searching based on the knowledge graph searching mode to the user.

Preferably, in the method, constructing a query chain corresponding to the question according to the entity node and the attribute node extracted from the question includes:

performing dependency syntax analysis on the problem to obtain a dependency tree corresponding to the problem;

extracting entity nodes, attribute nodes, question word nodes and core verb nodes in the problem from the dependency tree;

taking attribute nodes from entity nodes to core verb nodes as first class attribute nodes according to the dependency relationship among the nodes in the dependency tree; and the number of the first and second groups,

determining attribute nodes between the query word nodes and the core verb nodes, and determining second-class attribute nodes according to the determined attribute nodes between the query word nodes and the core verb nodes;

and sequentially splicing the entity nodes, the first type of attribute nodes and the second type of attribute nodes to form query chains corresponding to the problem, wherein each query chain comprises one entity node and at least one attribute node, and the at least one attribute node is sequenced in the query chains according to the dependency relationship.

Preferably, in the method, retrieving an answer to the question from a pre-constructed knowledge graph according to a query chain corresponding to the question, the method includes:

extracting entity nodes and adjacent attribute nodes from the query chain corresponding to the problem as an entity attribute key value pair, wherein the adjacent attribute nodes are the attribute nodes adjacent to the entity nodes in the query chain corresponding to the problem;

carrying out synonym expansion on the entity node and the attribute node in the key value pair respectively;

reconstructing an entity attribute key value pair according to the entity nodes and the attribute nodes after the synonym expansion;

for each reconstructed entity attribute key value pair, retrieving a result corresponding to the reconstructed entity attribute key value pair in a pre-constructed knowledge graph spectrum, and storing the result in a result set;

judging whether the adjacent attribute node is the last attribute node in the query chain corresponding to the problem or not;

if yes, determining the answer of the question according to the result set;

if not, forming a new query chain corresponding to the problem by using the result corresponding to the reconstructed entity attribute key value pair and the attribute nodes which are not extracted from the query chain corresponding to the problem, and executing the step of extracting the entity nodes and the adjacent attribute nodes from the query chain corresponding to the problem as an entity attribute key value pair, wherein the result is used as the entity node in the new query chain.

Preferably, in the method, reconstructing the entity attribute key-value pair according to the entity attribute node and the attribute node after the synonym expansion includes:

and aiming at each entity node in the entity node set obtained after the synonym is expanded, forming an entity attribute key value pair by the entity node and each attribute node in the attribute node set after the synonym is expanded, and obtaining a reconstructed entity attribute key value pair set.

Preferably, in the method, determining an answer to the question according to the result set includes:

determining a total number of words included in the question;

determining the word number sum of the entity node and the attribute node in the dependency tree adopted when the result is retrieved as the word number sum corresponding to the result aiming at each different result in the result set;

calculating the ratio of the number of words corresponding to the result to the total number of words as the ratio corresponding to the result;

and determining the result with the maximum ratio in the result set as the answer of the question.

Preferably, in the method, determining a second type of attribute node according to the attribute node from the determined query word node to the core verb node includes:

performing semantic analysis on the problem, and determining whether attribute nodes between the query word nodes and the core verb nodes need to perform antisense word extraction operation or not;

if yes, performing antisense word extraction operation on the attribute nodes between the query word node and the core verb node, and determining the attribute nodes subjected to the antisense word extraction operation as second-class attribute nodes;

and if not, determining the attribute node between the query word node and the core verb node as a second type attribute node.

Preferably, in the method, the answer to the question retrieved by the knowledge graph-based retrieval method is fed back to the user:

and if the answer of the question is not retrieved by adopting a retrieval mode based on a local question bank, feeding back the answer of the question retrieved by adopting a retrieval mode based on a knowledge graph to the user.

Preferably, the answer providing method provided by the embodiment of the present invention further includes:

searching answers of the questions by at least one searching mode among a searching mode based on a search engine, a searching mode based on a question-answer community knowledge base and a searching mode based on deep learning, wherein different searching modes correspond to different priorities;

and if the retrieval mode based on the knowledge graph does not retrieve the obtained answer of the question, feeding back the answer of the question retrieved by the retrieval mode with the highest priority in at least one retrieval mode of the returned answers of the question to the user.

Preferably, in the method, the priority of the at least one retrieval mode is, in order from high to low:

the method comprises a retrieval mode based on a question-answer community knowledge base, a retrieval mode based on deep learning and a retrieval mode based on a search engine.

and if the answer of the question is not retrieved by adopting a retrieval mode based on a local question bank, storing the answer of the question fed back to the user into the local question bank.

In a second aspect of the embodiments of the present invention, there is provided an answer providing apparatus including:

the receiving module is used for receiving the questions submitted by the user;

the building module is used for building a query chain corresponding to the problem according to the entity nodes and the attribute nodes extracted from the problem;

the first retrieval module is used for retrieving answers of the questions in a pre-constructed knowledge graph spectrum according to the query chain corresponding to the questions, and the answers are used as the answers of the questions retrieved based on the retrieval mode of the knowledge graph;

and the feedback module is used for feeding back the answer of the question obtained by searching based on the knowledge graph in a searching mode to the user.

Preferably, the building block comprises:

the analysis unit is used for carrying out dependency syntax analysis on the problem to obtain a dependency tree corresponding to the problem;

the extracting unit is used for extracting entity nodes, attribute nodes, question word nodes and core verb nodes in the problem from the dependency tree;

the first determining unit is used for taking attribute nodes with dependency relationship from the entity nodes to the core verb nodes as first-class attribute nodes according to the dependency relationship among the nodes in the dependency tree;

the second determining unit is used for determining attribute nodes from the query word nodes to the core verb nodes and determining second-class attribute nodes according to the determined attribute nodes from the query word nodes to the core verb nodes;

and the splicing unit is used for sequentially splicing the entity nodes, the first type of attribute nodes and the second type of attribute nodes to form query chains corresponding to the problem, wherein each query chain comprises one entity node and at least one attribute node, and the at least one attribute node is sequenced in the query chains according to the dependency relationship.

Preferably, the first retrieving module includes:

an extracting unit, configured to extract an entity node and an adjacent attribute node from the query chain corresponding to the problem as an entity attribute key-value pair, where the adjacent attribute node is an attribute node adjacent to the entity node in the query chain corresponding to the problem;

the expansion unit is used for carrying out synonym expansion on the entity node and the attribute node in the key value pair respectively;

the first reconstruction unit is used for reconstructing the entity attribute key value pair according to the entity node and the attribute node after the synonym expansion;

the retrieval unit is used for retrieving a result corresponding to each reconstructed entity attribute key value pair from a pre-constructed knowledge graph spectrum and storing the result into a result set;

a judging unit, configured to judge whether the neighboring attribute node is a last attribute node in the query chain corresponding to the problem;

a third determining unit, configured to determine an answer to the question according to the result set if the determining unit determines that the answer is positive;

and a second reconstructing unit, configured to, if the determining unit determines that the query link corresponding to the question is not the same as the query link corresponding to the question, construct a new query link corresponding to the question by using the result corresponding to the reconstructed entity attribute key-value pair and the non-extracted attribute node in the query link corresponding to the question, and perform a step of extracting the entity node and the adjacent attribute node from the query link corresponding to the question as an entity attribute key-value pair, where the result is used as the entity node in the new query link.

Preferably, the first reconstruction unit is specifically configured to:

Preferably, the third determining unit is specifically configured to:

determining a total number of words included in the question;

Preferably, the second determining unit is specifically configured to:

Preferably, the feedback module is specifically configured to:

Preferably, the answer providing apparatus provided in the embodiment of the present invention further includes:

the second retrieval module is used for retrieving answers of the questions by adopting at least one retrieval mode of a search mode based on a search engine, a retrieval mode based on a knowledge base of a question-answer community and a retrieval mode based on deep learning, wherein different retrieval modes correspond to different priorities;

the feedback module is further configured to feed back the answer to the question retrieved by the retrieval manner with the highest priority among the at least one retrieval manner returning the answer to the question to the user if the answer to the question is not retrieved by the retrieval manner based on the knowledge graph.

Preferably, the priority of the at least one retrieval mode is in the order from high to low: the method comprises a retrieval mode based on a question-answer community knowledge base, a retrieval mode based on deep learning and a retrieval mode based on a search engine.

and the storage module is used for storing the answers of the questions fed back to the user into the local question bank if the answers of the questions are not retrieved by adopting a retrieval mode based on the local question bank.

In a third aspect of the embodiments of the present invention, there is provided an answer providing device, for example, which may include a memory and a processor, wherein the processor may be configured to read a program in the memory, and execute the following processes:

receiving a question submitted by a user;

In a fourth aspect of embodiments of the present invention, there is provided a program product comprising program code for performing, when the program product is run, the following:

receiving a question submitted by a user;

According to the answer providing method and device provided by the embodiment of the invention, the questions submitted by the user are understood and analyzed, the entities and the attributes in the questions are extracted from the questions submitted by the user to form the query chain, and the result corresponding to the query chain is searched in the preset knowledge graph, so that the possibility that the searched answers are the answers of the questions submitted by the user is improved to a certain extent, and the accuracy of the answers fed back to the user is improved to a certain extent.

Drawings

The above and other objects, features and advantages of exemplary embodiments of the present invention will become readily apparent from the following detailed description read in conjunction with the accompanying drawings. Several embodiments of the invention are illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings and in which:

FIG. 1 schematically illustrates an application scenario according to an embodiment of the present invention;

FIG. 2 is a flow chart schematically illustrating an embodiment of an answer providing method according to an embodiment of the present invention;

FIG. 3 is a schematic flow chart illustrating an answer providing method according to an embodiment of the present invention;

FIG. 4 is a flow chart diagram schematically illustrating the construction of a query chain corresponding to a question in an embodiment of the present invention;

FIG. 5 is a schematic flow chart illustrating the process of retrieving answers to questions in a pre-constructed knowledge graph according to an embodiment of the present invention;

FIG. 6 is a schematic flow chart illustrating the determination of answers to questions from a result set in an embodiment of the present invention;

fig. 7 is a schematic structural view illustrating an answer providing apparatus according to an embodiment of the present invention;

fig. 8 is a schematic structural view illustrating an answer providing apparatus according to still another embodiment of the present invention;

fig. 9 schematically shows a program product for an answer providing method according to still another embodiment of the present invention.

In the drawings, the same or corresponding reference numerals indicate the same or corresponding parts.

Detailed Description

The principles and spirit of the present invention will be described with reference to a number of exemplary embodiments. It is understood that these embodiments are given solely for the purpose of enabling those skilled in the art to better understand and to practice the invention, and are not intended to limit the scope of the invention in any way. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

As will be appreciated by one skilled in the art, embodiments of the present invention may be embodied as a system, apparatus, device, method, or computer program product. Accordingly, the present disclosure may be embodied in the form of: entirely hardware, entirely software (including firmware, resident software, micro-code, etc.), or a combination of hardware and software.

According to the embodiment of the invention, an answer processing method and device are provided.

In this document, it is to be understood that any number of elements in the figures are provided by way of illustration and not limitation, and any nomenclature is used for differentiation only and not in any limiting sense.

The principles and spirit of the present invention are explained in detail below with reference to several representative embodiments of the invention.

Summary of The Invention

The inventor finds that in the current retrieval mode, the problem submitted by the user cannot be understood and analyzed, but the answer closest to the problem is directly searched according to the preset template or the keyword, so that the problem with low accuracy of the answer fed back to the user is caused, and the user experience is poor.

To this end, the present invention provides an answer providing method and apparatus, the method may include:

receiving a question submitted by a user;

Having described the general principles of the invention, various non-limiting embodiments of the invention are described in detail below.

Application scene overview

Referring to fig. 1, as shown in fig. 1, an application scenario diagram of an answer providing method provided by an embodiment of the present invention includes a user 10, a user terminal 101, and a server 102, where the user 10 may input and submit a question at a client installed in the user terminal 101, and the server 102 receives the question submitted by the user; constructing a query chain corresponding to the problem according to the entity nodes and the attribute nodes extracted from the problem; searching answers of the questions in a pre-constructed knowledge graph according to the query chain corresponding to the questions, wherein the answers are used as the answers of the questions obtained through searching based on a knowledge graph; and feeding back the answer of the question obtained by searching based on the knowledge graph searching mode to the user 10. Wherein the user terminal and the server are communicable via the internet. The user terminal can be a mobile phone, a tablet computer or a desktop computer.

Exemplary method

An answer providing method according to an exemplary embodiment of the present invention is described below with reference to fig. 2 to 6. It should be noted that the embodiments of the present invention can be applied to any applicable scenarios.

Fig. 2 is a schematic flowchart of an embodiment of an answer providing method provided by the present invention, which mainly includes a process of feeding back a corresponding answer to a user according to a question submitted by the user. As shown in fig. 2, an answer providing method provided in the embodiment of the present invention includes the following steps:

step 201, receiving a question submitted by a user.

In this step, a question submitted by a user through a question submitting interface is received. When the problem submitted by the user is received as a voice file, character recognition is carried out on the voice file to obtain a corresponding text, so that the problem can be analyzed subsequently. The specific process of performing text recognition on speech can refer to the prior art, and is not described in detail here.

Step 202, constructing a query chain corresponding to the question according to the entity nodes and the attribute nodes extracted from the question.

In this step, word level preprocessing may be performed on a text corresponding to a problem submitted by a user according to an existing word level processing technology, including word segmentation, part of speech tagging and named entity recognition, where the named entity recognition includes a person name, an organization name, a place name and other entity node recognition using the names as identifiers, and also includes recognition of attribute entities to obtain entity nodes and attribute nodes in the problem, and according to the entity nodes and attribute nodes in the problem, template matching is performed in a pre-constructed graph special template system to obtain a query chain corresponding to the problem submitted by the user, and various types of query chain template libraries such as a person template library, a time template library, a location template library, an organization template library and the like are stored in the pre-constructed graph special template system. For example, a query chain template related to a person is stored in the person template library, where the query chain template may include a person-birth date, a person-gender, a person-occupation, and the like, when it is detected that the entity node in the question is a person and the attribute node is related to the birth date, it may be determined that the query chain template corresponding to the question is the person-birth date, and then a query chain corresponding to the question is constructed according to a specific task in the question submitted by the user, that is, the person is replaced with a specific name. The query chains corresponding to the three different question asking modes can be determined by utilizing a pre-constructed map special system, wherein the different question asking modes for the same answer correspond to the same query chain template, for example, the birth date of Zhang III in 2016 (3 months), and the answer for the birth date of Zhang III can be corresponding to the query chain template, including the birth day of Zhang III, the birth day of Zhang III or the birth day of Zhang III.

The process of template matching in the pre-constructed graph special template system according to the entity node and the attribute node in the problem is the existing process, and is not detailed here.

For example, assume that the user submitted a question: which day is the birthday of Zhang III? After word level processing is carried out on the question, Zhang III is obtained as a human body entity, a birthday is obtained as an attribute entity, template matching is carried out in a pre-constructed special map template library system according to the human body entity and the attribute entity, finally, matching with a birthday template in a human template library is determined, a template query chain corresponding to the birthday template is determined as a human body-birthday, at the moment, Zhang III is extracted from the question and forms a query chain with the birthday, and finally, a query chain corresponding to the question is obtained, namely, Zhang III-birthday.

And extracting entity nodes and attribute nodes in the problem submitted by the user in other manners to construct a query chain corresponding to the problem, which is not limited herein.

The extracted entity nodes are theoretically entity nodes in the knowledge graph constructed in advance in step 203, and the extracted attribute nodes are theoretically attribute nodes in the knowledge graph constructed in advance in step 203.

And 203, retrieving answers of the questions in a pre-constructed knowledge graph according to the query chain corresponding to the questions, wherein the answers are used as the answers of the questions retrieved based on the retrieval mode of the knowledge graph.

In specific implementation, a query chain corresponding to a problem submitted by a user comprises entity nodes and attribute nodes in the problem, and the entity nodes in the query chain are retrieved in a pre-constructed knowledge graph; determining the attributes after searching the entity nodes in the query chain in the pre-constructed knowledge graph, namely the edges connected with the entity nodes searched in the query chain, in the pre-constructed knowledge graph; further, an entity node connected to the other side of the edge connected to the queried entity node is determined as an answer corresponding to the query chain, for example, if the query chain is zhang san-wife (the corresponding user submits a question of who is wife of zhang san.

It should be noted that the knowledge graph may be regarded as a huge graph, where nodes in the graph are entity nodes, edges connected to any entity node in the graph represent attributes of the entity, entities connected to the other side of the edge connected to any entity node are specific attribute values corresponding to any entity node, for example, if any entity node is zhang, an edge connected to any entity node represents a wife, and an entity connected to the other side of the edge connected to zhang is zhang san wife, that is, a specific attribute value is lie four. Specifically, the entities in the knowledge graph may include concepts, formulas, names of people, names of places, organizations, and the like, and the attributes in the knowledge graph may be relationships between two entities connected by an edge corresponding to the attribute or characteristics of the entities themselves. The embodiment of the invention is applied to the question-answering system of the primary and secondary school education, and the knowledge map can be constructed in advance in the following modes:

first, structured data is obtained.

The method for acquiring the structured data comprises the following steps: 1. acquiring structured data related to knowledge in K12 teaching materials from a pre-stored database; 2. acquiring structural data related to K12 teaching materials in encyclopedia, Wikipedia and interactive encyclopedia by a crawler; 3. extracting the structured data in the answer records from the prestored teacher answer records in a mode of combining machine learning and templates, wherein the prestored teacher answer records store manual answers of the teacher to questions asked by students in the education process; 4. education-related knowledge is obtained from unstructured data in some education websites or education question-and-answer communities and structured data is extracted from the unstructured data.

Secondly, the knowledge, concepts, formulas and the like in the acquired structured data are used as entity nodes in the knowledge graph, and specific attribute values of the entity nodes are connected through attribute edges aiming at each entity node, so that the knowledge graph in the education field is obtained.

The attribute of an entity node is represented by an edge connected with the entity node, and an entity node connected to the other side of the edge connected with the entity node is a specific attribute value of the entity node, and can also be understood as that the relationship between any two entity nodes is represented by the edge between the two entity nodes.

It should be noted that the pre-constructed real knowledge graph is used for an entity node list and an attribute node list, the entity node list is used for storing entity nodes in the pre-constructed knowledge graph, and the attribute node list is used for storing attribute nodes in the pre-constructed knowledge graph.

And 204, feeding back answers of the questions retrieved by the retrieval mode based on the knowledge graph to the user.

In specific implementation, when it is determined that answers to questions submitted by a user are not retrieved by using other retrieval methods with higher priority than the retrieval method based on the knowledge graph, the answers to the questions retrieved by the retrieval method based on the knowledge graph can be fed back to the user. Or, under the condition that the problem submitted by the user is determined to be retrieved only by adopting the retrieval mode based on the knowledge graph, the answer of the problem retrieved by the retrieval mode based on the knowledge graph is directly fed back to the user.

The embodiment provided in fig. 2 understands and analyzes the questions submitted by the user, extracts entities and attributes in the questions from the questions submitted by the user to form a query chain, and retrieves the result corresponding to the query chain from the preset knowledge graph, thereby improving the possibility that the retrieved answer is the answer to the questions submitted by the user to a certain extent, and further improving the accuracy of the answer fed back to the user to a certain extent.

It should be noted that, the answers to the questions submitted by the user can be retrieved by using multiple retrieval modes at the same time, wherein, the priority can be set for each retrieval mode in advance, and the answer to the question submitted by the user retrieved by the retrieval mode with the highest priority among the retrieval modes for retrieving the answers to the questions submitted by the user is fed back to the user.

Preferably, the answers to the questions retrieved by the knowledge-graph-based retrieval method are fed back to the user as follows:

The answers of the questions submitted by the user are retrieved by simultaneously adopting a retrieval mode based on the local question bank and a retrieval mode based on the knowledge spectrogram, wherein the priority of the retrieval mode based on the local question bank is higher than that of the retrieval mode based on the knowledge spectrogram, namely, the answers of the questions submitted by the user, which are retrieved by the retrieval mode based on the local question bank, are preferentially fed back to the user.

If the answers to the questions submitted by the user are not retrieved in the local question bank-based retrieval mode and the knowledge graph-based retrieval mode, a situation that the answers to the questions submitted by the user cannot be searched may occur, and in order to further ensure that the answers can be provided for the user, the answer providing method provided by the embodiment of the invention may further include the following steps, as shown in fig. 3:

step 301, at least one of a search mode based on a search engine, a search mode based on a knowledge base of a question-and-answer community, and a search mode based on deep learning is respectively adopted to search answers of the questions, wherein different search modes correspond to different priorities.

In specific implementation, in addition to retrieving answers to questions submitted by users in a retrieval manner based on a local question bank and a retrieval manner based on a knowledge spectrogram, the embodiment of the invention also retrieves answers to questions submitted by users in at least one retrieval manner selected from a search manner based on a search engine, a retrieval manner based on a question-and-answer community knowledge base and a retrieval manner based on deep learning, and preferably retrieves answers to questions submitted by users in a retrieval manner selected from a search manner based on a search engine, a retrieval manner based on a question-and-answer community knowledge base and a retrieval manner based on deep learning. In the search mode based on the search engine, the search mode based on the knowledge base of the question-answering community and the search mode based on the deep learning, each search mode corresponds to different priorities, and answers searched by the search modes with high priorities are preferentially fed back to the user.

The search engine based search method is as follows: directly calling a current mainstream search engine to retrieve answers of questions submitted by a user, and acquiring answers in a first sequence from answers returned by the search engine to serve as answers of the questions submitted by the user; the retrieval mode based on the question-answer community knowledge base is as follows: regularly collecting a large amount of question and answer resources in a network by using a crawler tool, and storing the question and answer resources to a question and answer community knowledge base in a key value pair mode; the retrieval mode based on deep learning is as follows: and constructing a deep learning model in advance, and taking the questions submitted by the user as the input of the deep learning model so as to obtain the output of the deep learning model, wherein the output is used as the answers of the questions submitted by the user. The specific implementation processes of the search mode based on the search engine, the retrieval mode based on the knowledge base of the question-answering community and the retrieval mode based on the deep learning all refer to the prior art, and are not described in detail here.

Step 302, if the answer to the question is not retrieved in the knowledge graph-based retrieval mode, feeding back the answer to the question retrieved in the retrieval mode with the highest priority in at least one retrieval mode of the answers to the question to the user.

In this step, the priority of the knowledge graph-based retrieval mode is higher than the priority of any one of the search engine-based search mode, the question-and-answer community knowledge base-based retrieval mode and the deep learning-based retrieval mode, and when the knowledge graph-based retrieval mode does not retrieve the answer to the question, the answer to the question retrieved by the retrieval mode with the highest priority among at least one retrieval mode returning the answers to the question is fed back to the user. Preferably, the priority of the at least one retrieval mode is in the order from high to low: the method comprises a retrieval mode based on a question-answer community knowledge base, a retrieval mode based on deep learning and a retrieval mode based on a search engine.

For example, assume that in step 301, the answers to the questions are retrieved by using a search method based on a search engine, a search method based on a knowledge base of a question-and-answer community, and a search method based on deep learning, and a search method for returning answers to the questions submitted by the user among the three search methods is determined; and if the retrieval mode based on the knowledge graph does not retrieve the obtained answer of the question, feeding back the answer retrieved by the retrieval mode with higher priority in the retrieval modes returning the answers of the questions submitted by the user among the three modes to the user, and if the retrieval mode returning the answers of the questions submitted by the user in the three retrieval modes is the retrieval mode based on the question-answer community knowledge base and the retrieval mode based on deep learning, returning the answer retrieved by the retrieval mode based on the question-answer community knowledge base to the user.

The order of priority of the various search methods may be: the method comprises a retrieval mode based on a local question bank, a retrieval mode based on a knowledge spectrogram, a retrieval mode based on a knowledge bank of a question and answer community, a retrieval mode based on deep learning and a retrieval mode based on a search engine. When the answer of the question submitted by the user is retrieved by the retrieval mode with higher priority, the retrieval mode with lower priority is instructed to stop the operation of retrieving the answer of the question submitted by the user, and the answer retrieved by the retrieval mode with higher priority is fed back to the user. That is, it is not necessary that all the retrieval modes return answers to the questions submitted by the user, and if the retrieval mode with the high priority already returns answers to the questions submitted by the user and the retrieval mode with the low priority is still used for retrieving answers to the questions submitted by the user, the operation of stopping retrieving answers by using the retrieval mode with the low priority can be instructed, so that the calculation amount can be saved to a certain extent.

Preferably, if the answer to the question is not retrieved by adopting a retrieval mode based on a local question bank, the answer to the question fed back to the user is stored in the local question bank.

The search mode based on the local question bank has the highest priority, and the answer finally fed back to the user can be stored in the local question bank.

Preferably, according to the content provided in fig. 4, a query chain corresponding to the question is constructed according to the entity node and the attribute node extracted from the question:

step 401, performing dependency syntax analysis on the problem to obtain a dependency tree corresponding to the problem.

In this step, the dependency syntax analysis is performed on the text of the problem submitted by the user, so as to obtain a dependency tree corresponding to the problem submitted by the user. After the problem submitted by the user is subjected to dependency syntax analysis, the part of speech of each word and the dependency relationship among the words can be obtained, wherein the dependency tree can also be called a dependency syntax tree, the nodes in the dependency tree are words in the sentence, and the dependency tree is used for reflecting the dependency relationship among the words in the sentence.

Step 402, extracting entity nodes, attribute nodes, question word nodes and core verb nodes in the problem from the dependency tree.

Specifically, an entity node list, an attribute node list, a question word node list and a core verb node list are established for a problem submitted by a user, and the nodes extracted from the dependency tree are stored in the corresponding node lists. And after the answers of the questions submitted by the users are retrieved, deleting the entity node list, the attribute node list, the question word node list and the core verb node list which are established for the questions submitted by the users.

Step 403, according to the dependency relationship between the nodes in the dependency tree, taking the attribute node from the entity node to the core verb node as the first type attribute node.

Step 404, determining attribute nodes between the query word nodes and the core verb nodes, and determining second-class attribute nodes according to the determined attribute nodes between the query word nodes and the core verb nodes.

In specific implementation, the execution sequence of step 403 and step 404 may not be limited.

According to the dependency relationship among the nodes in the dependency tree, the attribute nodes between the entity nodes and the core verb nodes in the problem submitted by the user are used as first-class attribute nodes, the attribute node closest to the entity node in the first-class attribute nodes has the dependency relationship with the entity node, and the first-class attribute nodes have the dependency relationship with each other. And determining attribute nodes from the query word nodes to the core verb nodes according to the dependency relationship between the nodes in the dependency tree, and determining second-class attribute nodes according to the attribute nodes from the query word nodes to the core verb nodes, wherein the second-class attribute nodes are the attribute nodes after processing the attribute nodes from the query word nodes to the core verb nodes, for example, obtaining the second-class attribute nodes after carrying out synonym or antonym operation on the attribute nodes with the dependency relationship from the query word nodes to the core verb nodes.

Preferably, the attribute node of the second type is determined according to the attribute node from the determined query word node to the core verb node in the following way: performing semantic analysis on the problem, and determining whether attribute nodes between the query word nodes and the core verb nodes need to perform antisense word extraction operation or not; if yes, performing antisense word extraction operation on the attribute nodes between the query word node and the core verb node, and determining the attribute nodes subjected to the antisense word extraction operation as second-class attribute nodes; and if not, determining the attribute node between the query word node and the core verb node as a second type attribute node.

For example, assuming that the question submitted by the user is "three of wife is his husband", the entity node is "three of three", the core verb is "yes", the attribute nodes from the entity node to the core verb node are "wife", "brother", and the attribute nodes from the query node to the core verb node are "husband", at which time it can be determined that the attribute node "husband" can be converted to the attribute node "wife" according to the semantics of the question, and thus the second type of attribute node is "wife", that is, the attribute node from the query node to the core verb node is converted to "wife". Wherein the antisense word of "husband" is determined as "wife".

And 405, sequentially splicing the entity nodes, the first type of attribute nodes and the second type of attribute nodes to form query chains corresponding to the problem, wherein each query chain comprises one entity node and at least one attribute node, and the at least one attribute node is sequenced in the query chains according to the dependency relationship.

In specific implementation, according to the dependency relationship of the attribute nodes in the dependency tree, the entity node, the first class of attribute nodes, and the second class of attribute nodes are sequentially spliced to obtain a query chain corresponding to the problem submitted by the user, for example, the problem submitted by the user is that "the brother of the third wife is who is" the husband, "the corresponding query chain is" the third wife-the brother-the wife son ", wherein the first wife in the query chain depends on the entity node, the brother depends on the first wife in the query chain, and the second wife in the query chain depends on the brother.

In the embodiment provided by fig. 4, the purpose of analyzing and understanding the question can be achieved by performing dependency parsing on the question submitted by the user, and the entity node and the attribute node in the question are extracted from the question, so as to construct a query chain which is more relevant and reliable to the question submitted by the user, and thus, the answer of the question is retrieved according to the query chain, and the accuracy of the retrieved answer is higher.

Preferably, according to the provided content of fig. 5, the answer to the question is retrieved from a pre-constructed knowledge graph according to the query chain corresponding to the question:

step 501, extracting an entity node and an adjacent attribute node from the query chain corresponding to the problem as an entity attribute key value pair, where the adjacent attribute node is an attribute node adjacent to the entity node in the query chain corresponding to the problem.

In specific implementation, there may be one or more query chains corresponding to the problem submitted by the user, and for each query chain, the query chain includes an entity node and at least one attribute node, the entity nodes and the attribute nodes are ordered according to the dependency relationship in the problem submitted by the user, and for each query chain corresponding to the problem submitted by the user, the entity node and the attribute node adjacent to the entity node are extracted from the query chain as an entity attribute key-value pair.

Step 502, performing synonym expansion on the entity node and the attribute node in the key value pair respectively.

In specific implementation, different users may use different expression modes to express an entity node and an attribute node in an entity attribute key value pair, for example, "zhang san" in the problem of "husband whose wife is the brother of zhang san" may be expressed as "xiao", etc., and an attribute node "wife" may be expressed as "wife", etc., in order to ensure that the entity node and the attribute node are retrieved in a pre-constructed knowledge graph, synonym expansion is performed on the entity node and the attribute node in the entity attribute key value pair in this step.

In this step, synonym expansion may be performed on the entity nodes and the attribute nodes in the entity attribute key value pair through the general synonym table, synonym expansion may be performed on the attribute nodes in the entity attribute key value pair through the map-specific attribute expansion table corresponding to the pre-constructed knowledge map, and synonym expansion may be performed on the entity nodes in the entity attribute key value pair through the map-specific entity expansion table corresponding to the pre-constructed knowledge map. In the process of constructing the knowledge graph, aiming at each entity node in the knowledge graph, correspondingly storing the expression mode of the entity node in the knowledge graph and the synonym corresponding to the entity node into a graph-dedicated entity expansion table, and aiming at each attribute node in the knowledge graph, correspondingly storing the expression mode of the attribute node in the knowledge graph and the synonym corresponding to the attribute node into a graph-dedicated attribute expansion table.

Step 503, reconstructing the entity attribute key value pair according to the entity node and the attribute node after the synonym expansion.

In specific implementation, each expression mode of the entity node after the synonym expansion forms an entity node set, and each expression mode of the attribute node after the synonym expansion forms an attribute node set, preferably, the entity node set includes an entity node in an entity attribute key value pair before reconstruction, and the attribute node set includes an entity node in an entity attribute key value pair before reconstruction.

Preferably, the entity attribute key-value pairs may be reconstructed as follows:

For example, the entity node set obtained by synonymously expanding the entity node "zhang san" in the question "husband whose wife brother of zhang san is" includes "zhang san" and "xiaoming", the entity node set obtained by synonymously expanding the attribute node "wife" includes "wife" and "wife", and the reconstructed entity attribute key value sets include four entity attribute key value pairs of "zhang san-wife", "xiaoming-wife", "zhang-wife" and "xiaoming-wife".

And step 504, for each reconstructed entity attribute key value pair, retrieving a result corresponding to the reconstructed entity attribute key value pair in a pre-constructed knowledge graph, and storing the result in a result set.

In this step, for each key-value pair in the entity attribute key-value pair set obtained by reconstruction, a result corresponding to the reconstructed entity attribute key-value pair is retrieved from a pre-constructed knowledge graph and stored in a result set.

Step 505, determining whether the adjacent attribute node is the last attribute node in the query chain corresponding to the problem, if yes, executing step 506, otherwise, executing step 507.

In this step, it is determined whether the neighboring attribute node in step 501 is the last attribute node in the query chain corresponding to the problem submitted by the user.

Step 506, according to the result set, determining an answer to the question.

In this step, a plurality of results are stored in the result set, the result obtained by retrieving the entity attribute key value pair corresponding to the last attribute node in the query chain corresponding to the problem submitted by the user can be determined as the answer to the problem submitted by the user, when there are a plurality of query chains corresponding to the problem submitted by the user, the result obtained by retrieving the entity attribute key value pair corresponding to the last attribute node in each query chain can be subjected to deduplication processing, and the result subjected to deduplication processing is determined as the answer to the problem submitted by the user.

Step 507, forming a new query chain corresponding to the problem by using the result corresponding to the reconstructed entity attribute key value pair and the unextracted attribute nodes in the query chain corresponding to the problem, and continuing to execute step 501.

And taking the result corresponding to the reconstructed entity attribute key value pair as an entity node in a new query chain.

In this step, if the adjacent attribute node in step 501 is not the last attribute node in the query chain, a new query chain is formed by the result corresponding to the entity attribute key value pair reconstructed in step 504 and the attribute nodes not extracted in the problem submitted by the user, and step 501 is continuously executed until the adjacent attribute node in step 501 is the last attribute node in the query chain.

For example, assuming that the problem submitted by the user is "whose wife is the brother of the wife of zhang", the query chain corresponding to the problem is "zhang san-wife-brother-wife", the entity node "zhang san" and the attribute node "wife" adjacent to the entity node in the query chain are first extracted to obtain an entity attribute key value pair "zhang san-wife", and after synonym expansion is performed on the entity node and the attribute node in the key value pair, for each reconstructed entity attribute key value pair in the obtained reconstructed entity attribute key value pair set, the result "lie four" corresponding to the reconstructed entity attribute key value pair is retrieved in the pre-constructed indication map; the wife behind the third piece of paper is not the last attribute node in the query chain, so that the result of the fourth item of paper and the attribute node 'brother-wife' left in the query chain are constructed into a new query chain, the fourth-brother item of paper is extracted from the new query chain to be used as an entity attribute key value pair, and after synonym expansion is carried out on the entity node and the attribute node in the key value pair, the result 'third piece' corresponding to the reconstructed entity attribute key value pair is searched in a pre-constructed indication map spectrum for each reconstructed entity attribute key value pair in the obtained reconstructed entity attribute key value pair set; the sibling behind the link four is not the last attribute node in the query chain, so that the result of link three and the attribute node wife remaining in the query chain are constructed into a new query chain, link three-wife is extracted from the new query chain as an entity attribute key value pair, and after synonym expansion is performed on the entity node and the attribute node in the key value pair, the result "wang five" corresponding to the reconstructed entity attribute key value pair is retrieved in a pre-constructed indication graph for each reconstructed entity attribute key value pair in the obtained reconstructed entity attribute key value pair set.

The embodiment provided in fig. 5 may implement a multi-stage query for a question submitted by a user under the condition that a query chain corresponding to the question includes a plurality of attribute nodes, so that the most accurate answer is selected from a result set obtained by the multi-stage query and fed back to the user.

Preferably, the answer to the question is determined from the result set, as may be provided in fig. 6:

step 601, the total number of words included in the question is determined.

In specific implementation, the total word number in the text corresponding to the question submitted by the user is determined. For example, the total number of words in the question "who husband who is the brother of Zhang III" is 12.

Step 602, for each different result in the result set, determining the word sum of the entity node and the attribute node in the dependency tree used for retrieving the result as the word sum corresponding to the result.

In this step, for each result in the result set, the sum of the numbers of words of the entity node and the attribute node in the dependency tree used when the result is retrieved is determined, for example, when the sum of the numbers of words of the entity node "zhang" and the attribute node "wife" in the dependency tree used when the husband who has the wife whose brother of the question "zhang" is who "retrieved the result" lie four "is 4, the sum of the numbers of words of" zhang "and the result is 4, and when the sum of the numbers of words of the entity node" zhang "and the attribute nodes" wife "," brother ", and" wife "in the dependency tree used when the result" wang "i" (i.e., the wife whose wife has the brother of lie four) is retrieved is 8, the sum of the numbers of words of the entity node "zhang" and the attribute nodes "wife" is 8.

Step 603, calculating the ratio of the number of words corresponding to the result and the total number of words as the ratio corresponding to the result.

Step 604, determining the result with the maximum ratio in the result set as the answer to the question.

In specific implementation, the ratio of the number of words corresponding to each result to the total number of words is calculated, the result with the largest ratio is determined as the answer to the question submitted by the user, and continuing to use the example, "lie four" and the sum of the number of words corresponding to the result is 4, the ratio of "lie four" to the result is 4/12, the sum of the number of words corresponding to "wang five" to the result is 8, the ratio of "wang five" to the result is 8/12, and at this time, "wang five" is determined as the answer to the question that the wife's brother of the question "zhang san is whose husband" is.

Exemplary device

Having described the answer providing method according to the exemplary embodiment of the present invention, next, an answer providing apparatus according to an exemplary embodiment of the present invention will be described with reference to fig. 7.

Fig. 7 is a schematic structural diagram of an answer providing apparatus according to an embodiment of the present invention, as shown in fig. 7, the answer providing apparatus may include the following modules:

a receiving module 701, configured to receive a question submitted by a user;

a building module 702, configured to build a query chain corresponding to the question according to the entity nodes and the attribute nodes extracted from the question;

a first retrieval module 703, configured to retrieve an answer to the question in a pre-constructed knowledge graph according to the query chain corresponding to the question, where the answer is used as the answer to the question retrieved in a knowledge graph-based retrieval manner;

a feedback module 704, configured to feed back the answer to the question retrieved in the knowledge graph-based retrieval manner to the user.

Preferably, the building module 702 includes:

an analysis unit 7021, configured to perform dependency syntax analysis on the problem to obtain a dependency tree corresponding to the problem;

an extracting unit 7022, configured to extract an entity node, an attribute node, a question word node, and a core verb node in the problem from the dependency tree;

a first determining unit 7023, configured to use, according to the dependency relationship between the nodes in the dependency tree, an attribute node having a dependency relationship from the entity node to the core verb node as a first-class attribute node;

a second determining unit 7024, configured to determine attribute nodes from the query word node to the core verb node, and determine a second type of attribute node according to the determined attribute nodes from the query word node to the core verb node;

a splicing unit 7025, configured to splice the entity nodes, the first type attribute nodes, and the second type attribute nodes in sequence to form query chains corresponding to the problem, where each query chain includes an entity node and at least one attribute node, and the at least one attribute node is ordered in the query chain according to a dependency relationship.

Preferably, the first retrieving module 703 includes:

an extracting unit 7031, configured to extract an entity node and an adjacent attribute node from the query chain corresponding to the problem as an entity attribute key-value pair, where the adjacent attribute node is an attribute node adjacent to the entity node in the query chain corresponding to the problem;

an expansion unit 7032, configured to perform synonym expansion on the entity node and the attribute node in the key value pair respectively;

a first reconstructing unit 7033, configured to reconstruct the entity attribute key-value pair according to the entity node and the attribute node after the synonym expansion;

a retrieving unit 7034, configured to, for each reconstructed entity attribute key-value pair, retrieve, in a pre-constructed knowledge graph, a result corresponding to the reconstructed entity attribute key-value pair, and store the result in a result set;

a judging unit 7035, configured to judge whether the adjacent attribute node is a last attribute node in the query chain corresponding to the problem;

a third determining unit 7036, configured to determine, when the determining unit 7035 determines that the answer to the question is yes, according to the result set;

a second reconstructing unit 7037, configured to, if the determining unit 7035 determines no, configure a new query chain corresponding to the problem with a result corresponding to the reconstructed entity attribute key-value pair and an attribute node not extracted from the query chain corresponding to the problem, and perform a step of extracting an entity node and an adjacent attribute node from the query chain corresponding to the problem as an entity attribute key-value pair, where the result is used as an entity node in the new query chain.

Preferably, the first reconstructing unit 7033 is specifically configured to:

Preferably, the third determining unit 7036 is specifically configured to:

determining a total number of words included in the question;

Preferably, the second determining unit 7024 is specifically configured to:

Preferably, the feedback module 704 is specifically configured to:

a second retrieval module 705, configured to retrieve answers to the questions in at least one of a search mode based on a search engine, a search mode based on a knowledge base of a question-and-answer community, and a search mode based on deep learning, where different search modes correspond to different priorities;

the feedback module 704 is further configured to, if the answer to the question is not retrieved in the knowledge graph-based retrieval manner, feed back the answer to the question retrieved in the retrieval manner with the highest priority among the at least one retrieval manner returning the answer to the question to the user.

a saving module 706, configured to, if the answer to the question is not retrieved in the local question bank-based retrieval manner, save the answer to the question, which is fed back to the user, in the local question bank.

Exemplary device

After the answer providing method and apparatus according to the exemplary embodiment of the present invention are described, an answer providing apparatus according to another exemplary embodiment of the present invention is described next.

As will be appreciated by one skilled in the art, aspects of the present invention may be embodied as a system, method or program product. Thus, various aspects of the invention may be embodied in the form of: an entirely hardware embodiment, an entirely software embodiment (including firmware, microcode, etc.) or an embodiment combining hardware and software aspects that may all generally be referred to herein as a "circuit," module "or" system.

In some possible embodiments, the answer providing device according to the present invention may comprise at least one processing unit, and at least one memory unit. Wherein the storage unit stores program code that, when executed by the processing unit, causes the processing unit to perform the steps in the answer providing method according to various exemplary embodiments of the present invention described in the above section "exemplary method" of the present specification. For example, the processing unit may execute step 201 shown in fig. 2, receive a question submitted by a user, step 202, construct a query chain corresponding to the question according to an entity node and an attribute node extracted from the question, step 203, retrieve an answer to the question in a pre-constructed knowledge graph according to the query chain corresponding to the question, as an answer to the question retrieved by a knowledge graph-based retrieval method, and step 204, feed back the answer to the question retrieved by the knowledge graph-based retrieval method to the user.

The answer providing apparatus 80 according to this embodiment of the present invention is described below with reference to fig. 8. The answer providing device 80 shown in fig. 8 is only an example, and should not bring any limitation to the function and the scope of the embodiment of the present invention.

As shown in fig. 8, the answer providing apparatus 80 is represented in the form of a general-purpose computing device. The components of the answer providing device 80 may include, but are not limited to: the at least one processing unit 801, the at least one memory unit 802, and a bus 803 that couples various system components including the processing unit 801 and the memory unit 802.

Bus 803 represents one or more of any of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a processor, or a local bus using any of a variety of bus architectures.

The storage unit 802 may include readable media in the form of volatile memory, such as Random Access Memory (RAM)8021 and/or cache memory 8022, and may further include Read Only Memory (ROM) 8023.

Storage unit 802 can also include a program/utility 8025 having a set (at least one) of program modules 8024, such program modules 8024 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may comprise an implementation of a network environment.

The answer providing apparatus 80 may also communicate with one or more external devices 804 (e.g., keyboard, pointing device, etc.), with one or more devices that enable a user to interact with the answer providing apparatus 80, and/or with any device (e.g., router, modem, etc.) that enables the answer providing apparatus 80 to communicate with one or more other computing devices. Such communication may be through input/output (I/O) interfaces 805. Also, the answer providing device 80 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network such as the internet) via the network adapter 806. As shown in fig. 8, the network adapter 806 communicates with other modules of the apparatus 80 for answer provision through the bus 803. It should be understood that although not shown in the figures, other hardware and/or software modules may be used in conjunction with answer providing device 80, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, among others.

Exemplary program product

In some possible embodiments, the aspects of the answer providing method provided by the present invention may also be implemented in the form of a program product, which includes program code for causing a computer device to execute the steps of the answer providing method according to various exemplary embodiments of the present invention described in the above section of "exemplary method" of the present specification when the program product runs on the computer device, for example, the computer device may execute step 201 shown in fig. 2, receive a question submitted by a user, step 202, construct a query chain corresponding to the question according to entity nodes and attribute nodes extracted from the question, step 203, retrieve an answer to the question in a pre-constructed knowledge graph according to the query chain corresponding to the question, as an answer to the question retrieved by a knowledge graph-based retrieval method, and step 204, and feeding back the answer of the question obtained by searching based on the knowledge graph searching mode to the user.

The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

As shown in fig. 9, a program product 90 for answer provision according to an embodiment of the present invention is described, which may employ a portable compact disc read only memory (CD-ROM) and include program codes, and may be executed on a terminal device. However, the program product of the present invention is not limited in this regard and, in the present document, a readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

A readable signal medium may include a propagated data signal with readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A readable signal medium may also be any readable medium that is not a readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

Program code embodied on a readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device over any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., over the internet using an internet service provider).

It should be noted that although several units or sub-units of the apparatus are mentioned in the above detailed description, such division is merely exemplary and not mandatory. Indeed, the features and functions of two or more of the units described above may be embodied in one unit, according to embodiments of the invention. Conversely, the features and functions of one unit described above may be further divided into embodiments by a plurality of units.

Moreover, while the operations of the method of the invention are depicted in the drawings in a particular order, this does not require or imply that the operations must be performed in this particular order, or that all of the illustrated operations must be performed, to achieve desirable results. Additionally or alternatively, certain steps may be omitted, multiple steps combined into one step execution, and/or one step broken down into multiple step executions.

While the spirit and principles of the invention have been described with reference to several particular embodiments, it is to be understood that the invention is not limited to the disclosed embodiments, nor is the division of aspects, which is for convenience only as the features in such aspects may not be combined to benefit. The invention is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.

Claims

1. An answer providing method comprising:

receiving a question submitted by a user;

feeding back answers of the questions retrieved by a retrieval mode based on the knowledge graph to the user;

wherein, according to the entity node and the attribute node extracted from the question, constructing a query chain corresponding to the question comprises:

sequentially splicing the entity nodes, the first type of attribute nodes and the second type of attribute nodes to form query chains corresponding to the problem, wherein each query chain comprises one entity node and at least one attribute node, and the at least one attribute node is sequenced in the query chains according to the dependency relationship;

determining a second type of attribute node according to the attribute node from the determined question word node to the core verb node, wherein the step of determining the second type of attribute node comprises the following steps:

2. The method of claim 1, wherein retrieving answers to the questions in a pre-constructed knowledge graph according to a query chain to which the questions correspond comprises:

if yes, determining the answer of the question according to the result set;

3. The method of claim 2, wherein reconstructing entity attribute key-value pairs from synonym-extended entity attribute nodes and attribute nodes comprises:

4. The method of claim 2, wherein determining an answer to the question from the result set comprises:

determining a total number of words included in the question;

5. The method of claim 1, wherein answers to the questions retrieved by the knowledge-graph based retrieval are fed back to the user:

6. The method of claim 5, further comprising:

7. The method of claim 6, wherein the priority of the at least one retrieval mode is in the order of:

8. The method of any of claims 5-7, further comprising:

9. An answer providing apparatus comprising:

the receiving module is used for receiving the questions submitted by the user;

the feedback module is used for feeding back answers of the questions retrieved by a retrieval mode based on the knowledge graph to the user;

wherein the building block comprises:

the splicing unit is used for sequentially splicing the entity nodes, the first type of attribute nodes and the second type of attribute nodes to form query chains corresponding to the problem, wherein each query chain comprises one entity node and at least one attribute node, and the at least one attribute node is sequenced in the query chains according to the dependency relationship;

wherein the second determining unit is specifically configured to:

10. The apparatus of claim 9, wherein the first retrieving module comprises:

11. The apparatus according to claim 10, wherein the first reconstruction unit is specifically configured to:

12. The apparatus according to claim 10, wherein the third determining unit is specifically configured to:

determining a total number of words included in the question;

13. The apparatus of claim 9, wherein the feedback module is specifically configured to:

14. The apparatus of claim 13, further comprising:

15. The apparatus of claim 14, wherein the priority of the at least one retrieval mode is, in order from high to low: the method comprises a retrieval mode based on a question-answer community knowledge base, a retrieval mode based on deep learning and a retrieval mode based on a search engine.

16. The apparatus of any of claims 13-15, further comprising:

17. An answer providing apparatus comprising at least one processing unit and at least one memory unit, wherein said memory unit has stored program code which, when executed by said processing unit, causes said processing unit to carry out the steps of the method according to any one of claims 1 to 8.

18. A computer readable storage medium storing computer instructions for causing a computer device to perform the steps of the method of any one of claims 1-8 when the computer instructions are executed by the computer device.