CN111966793A - Intelligent question-answering method and system based on knowledge graph and knowledge graph updating system - Google Patents
Intelligent question-answering method and system based on knowledge graph and knowledge graph updating system Download PDFInfo
- Publication number
- CN111966793A CN111966793A CN201910419946.4A CN201910419946A CN111966793A CN 111966793 A CN111966793 A CN 111966793A CN 201910419946 A CN201910419946 A CN 201910419946A CN 111966793 A CN111966793 A CN 111966793A
- Authority
- CN
- China
- Prior art keywords
- knowledge
- meta
- question
- attributes
- entities
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 60
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 73
- 239000002131 composite material Substances 0.000 claims description 21
- 230000008569 process Effects 0.000 claims description 19
- 239000011159 matrix material Substances 0.000 claims description 14
- 238000000605 extraction Methods 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 9
- 230000004927 fusion Effects 0.000 claims description 5
- 238000000354 decomposition reaction Methods 0.000 claims description 4
- 238000005259 measurement Methods 0.000 claims description 4
- 238000010586 diagram Methods 0.000 description 19
- 238000005516 engineering process Methods 0.000 description 8
- 239000000284 extract Substances 0.000 description 7
- 230000006870 function Effects 0.000 description 5
- 230000014509 gene expression Effects 0.000 description 5
- 230000002452 interceptive effect Effects 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 3
- 238000003491 array Methods 0.000 description 2
- 238000013499 data model Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000000691 measurement method Methods 0.000 description 2
- 238000005192 partition Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000012916 structural analysis Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Human Computer Interaction (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides an intelligent question-answering method and system based on a knowledge graph and a knowledge graph updating system. The method comprises the following steps: receiving a user question; converting the user question into a regularization question; and performing question matching on the regularization question and a template base generated by a knowledge graph based on a heterogeneous information model, and determining a matching result with the highest similarity as an answer. The characteristics of a heterogeneous information network are fully utilized for modeling, richer knowledge and relationship reasoning are provided by combining a ranking collaborative clustering algorithm and a meta-path similarity algorithm, the traditional question-answering system based on the knowledge graph is improved, and the accuracy and the intelligence degree of the question-answering system are improved.
Description
Technical Field
The invention belongs to the technical field of knowledge graphs, and particularly relates to an intelligent question-answering method and system based on a knowledge graph and a knowledge graph updating system.
Background
The question-answering system is a high-level form of information retrieval system that can answer questions posed by users in natural language in accurate and concise natural language. In recent years, a knowledge-graph-based question-answering system has become a hotspot for research and application in academic and industrial fields. The high semantic understanding degree, data precision and efficient retrieval of knowledge graph enable the knowledge graph to be widely applied. However, the rules of the knowledge graph have certain limitations, for example, the number of entities has certain limitations under a specific relationship, which affects the breadth of the question-answering range and the accuracy of the question-answering result.
The existing knowledge graph-based question-answering system is applied more and more, but in practical application, the constructed knowledge graph model is easy to have the problems of data looseness and low coverage rate, so that the knowledge matching coverage in the question-answering process is low, and a satisfactory intelligent answer is often not obtained.
In addition, incremental maintenance of the knowledge-graph is costly.
Disclosure of Invention
The embodiment of the invention provides an intelligent question-answering method and system based on a knowledge graph and a knowledge graph updating system, and the accuracy of answering is improved.
The technical scheme of the embodiment of the invention is as follows:
a wisdom question-answering method based on knowledge graph comprises the following steps:
receiving a user question;
converting the user question into a regularization question;
and performing question matching on the regularization question and a template base generated by a knowledge graph based on a heterogeneous information model, and determining a matching result with the highest similarity as an answer.
In one embodiment, the heterogeneous information model contains entities, attributes, and relationships; the method further comprises a process of pre-generating the knowledge graph based on the heterogeneous information model, wherein the process comprises the following steps:
acquiring unstructured data, semi-structured data and structured data;
extracting entities, attributes, direct relations among the entities, direct relations among the attributes and direct relations among the entities and the attributes from the collected unstructured data, semi-structured data and structured data;
taking the extracted entities and attributes as meta-path node objects, and constructing direct relations between every two meta-path node objects into a composite relation containing a plurality of direct relations based on a multi-level path;
converting direct relations among the entities, direct relations among the attributes, direct relations among the entities and the attributes and the composite relations into meta-paths;
comparing the similarity in the meta-paths based on a ranking co-clustering algorithm and a meta-path similarity algorithm to construct a knowledge ontology;
and generating knowledge based on a knowledge reasoning mode and the knowledge ontology, and representing the knowledge into a knowledge representation format to generate the knowledge graph.
In one embodiment, the meta-path includes a single directed line segment between entities, a single directed line segment between entities and attributes, a single directed line segment between attributes, a plurality of directed line segments between entities and attributes, or a plurality of directed line segments between attributes, wherein the plurality of directed line segments between entities spans an attribute or another entity, the plurality of directed line segments between entities and attributes spans another entity or another attribute, the plurality of directed line segments between attributes spans an entity or another attribute.
In one embodiment, the comparing the similarity among the meta-paths based on the ranking co-clustering algorithm and the meta-path similarity algorithm to construct the ontology comprises:
performing clustering on the meta-paths based on a ranking co-clustering algorithm;
calculating the similarity of the clustered meta-paths based on a meta-path similarity algorithm;
and merging the meta-paths with the similarity larger than a preset threshold value, and constructing a knowledge body based on a merging result.
In one embodiment, the performing clustering on meta-paths based on a ranking co-clustering algorithm includes:
step 1: initializing a cluster division result of the meta-path, and distributing a random cluster to each main target type object;
step 2: obtaining the ranking score of each random cluster;
and step 3: calculating new measurement, and estimating the mixed model coefficient of each main target object;
and 4, step 4: adjusting the random cluster based on the mixed model coefficient;
and 5: repeating the step (2), the step (3) and the step (4) until convergence;
step 6: calculating a ranking distribution matrix according to the ranking distribution model;
and 7, simultaneously clustering rows and columns of the ranking distribution matrix by using a matrix singular value decomposition technology to obtain a clustering result representing a master-slave type, wherein if a null cluster exists, the step (1) is carried out.
An intelligent question-answering device based on a knowledge graph comprises:
the question receiving unit is used for receiving a user question;
the regularization unit is used for converting the user question into a regularization question;
and the answer determining unit is used for matching the regularized question with a template library generated by a knowledge graph based on a heterogeneous information model and determining a matching result with the highest similarity as an answer.
In one embodiment, the heterogeneous information model contains entities, attributes, and relationships; the apparatus also includes a knowledge graph generation unit;
the knowledge graph generating unit is used for acquiring unstructured data, semi-structured data and structured data; extracting entities, attributes, direct relations among the entities, direct relations among the attributes and direct relations among the entities and the attributes from the collected unstructured data, semi-structured data and structured data; taking the extracted entities and attributes as meta-path node objects, and constructing direct relations between every two meta-path node objects into a composite relation containing a plurality of direct relations based on a multi-level path; converting direct relations among the entities, direct relations among the attributes, direct relations among the entities and the attributes and the composite relations into meta-paths; comparing the similarity in the meta-paths based on a ranking co-clustering algorithm and a meta-path similarity algorithm to construct a knowledge ontology; and generating knowledge based on a knowledge reasoning mode and the knowledge ontology, and representing the knowledge into a knowledge representation format to generate the knowledge graph.
In one embodiment, the knowledge graph generating unit is configured to perform clustering on meta paths based on a ranking co-clustering algorithm; calculating the similarity of the clustered meta-paths; and merging the meta-paths with the similarity larger than a preset threshold value, and constructing a knowledge body based on a merging result.
A knowledge-graph update system for a question-answering system, comprising:
a question-answering system for acquiring a data set;
the data acquisition module is used for acquiring unstructured data, semi-structured data and structured data from a database;
the heterogeneous information extraction module is used for extracting the contained entities, attributes, direct relations among the entities, direct relations among the attributes and direct relations among the entities and the attributes from the unstructured data, the semi-structured data and the structured data which are acquired by the data acquisition module;
the heterogeneous knowledge fusion module is used for taking the extracted entities and attributes as meta-path node objects, and constructing the direct relationship between every two meta-path node objects into a composite relationship containing a plurality of direct relationships based on a multi-level path; converting direct relations among the entities, direct relations among the attributes, direct relations among the entities and the attributes and the composite relations into meta-paths;
and the heterogeneous knowledge processing module is used for comparing the similarity in the meta paths based on a ranking collaborative clustering algorithm and a meta path similarity algorithm to construct a knowledge ontology, generating knowledge based on a knowledge reasoning mode and the knowledge ontology, representing the knowledge into a knowledge representation format to generate the knowledge graph, and sending the knowledge graph to the question-answering system.
In one embodiment, the heterogeneous knowledge processing module is configured to perform clustering on meta-paths based on a rank co-clustering algorithm; calculating the similarity of the clustered meta-paths based on a meta-path similarity algorithm; and merging the meta-paths with the similarity larger than a preset threshold value, and constructing a knowledge body based on a merging result.
An intelligent question-answering device based on knowledge graph comprises a processor and a memory;
the memory stores an application program executable by the processor for causing the processor to execute the wisdom knowledge-graph based question-answering method as described in any one of the above.
A computer readable storage medium having stored therein computer readable instructions for performing the wisdom knowledge-graph-based question-answering method according to any one of the preceding claims.
According to the technical scheme, the implementation mode of the invention integrates more types of objects and complex interaction relations among the objects, solves the problems of data looseness and low coverage rate, can enable a knowledge base to be more complete and more convenient and fast to update, and enables an original question-answering system based on the knowledge map to be more accurate and intelligent.
And the extracted entities and attributes are used as meta-path node objects, and the direct relation between every two meta-path node objects is constructed into a composite relation containing a plurality of direct relations based on a multi-level path, so that the semantic extraction capability of the meta-path can be expanded. Because the composite relation of the embodiment of the invention considers more semantic information, the result is refined and more accurate.
In the embodiment of the invention, a knowledge ontology is further constructed by combining the collaborative ranking clustering and the meta-path algorithm. In a heterogeneous information network structure, richer entity, relationship and attribute knowledge are comprehensively extracted by combining a ranking collaborative clustering algorithm and a meta-path similarity algorithm, and a knowledge graph with richer knowledge is helped to be constructed.
Drawings
FIG. 1 is a flow chart of the intellectual question answering method based on knowledge graph of the present invention.
Fig. 2 is an exemplary diagram of a heterogeneous information network in accordance with the present invention.
FIG. 3 is a first exemplary diagram of a meta-path according to the present invention.
FIG. 4 is a second exemplary diagram of a meta-path according to the present invention.
Fig. 5 is a process flow diagram of a question-answering system according to the present invention.
FIG. 6 is a diagram of a knowledge-graph architecture based on heterogeneous information models according to the present invention.
FIG. 7 is a process flow diagram of a ranked co-clustering algorithm according to the present invention.
FIG. 8 is a sequence diagram of a knowledge-graph generation process according to the present invention.
Fig. 9 is a sequence diagram of a user's usage flow of the question-answering system according to the present invention.
FIG. 10 is a block diagram of an intellectual answering system based on knowledge-maps according to the present invention.
Detailed Description
In order to make the technical scheme and advantages of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the detailed description and specific examples, while indicating the scope of the invention, are intended for purposes of illustration only and are not intended to limit the scope of the invention.
For simplicity and clarity of description, the invention will be described below by describing several representative embodiments. Numerous details of the embodiments are set forth to provide an understanding of the principles of the invention. It will be apparent, however, that the invention may be practiced without these specific details. Some embodiments are not described in detail, but rather are merely provided as frameworks, in order to avoid unnecessarily obscuring aspects of the invention. Hereinafter, "including" means "including but not limited to", "according to … …" means "at least according to … …, but not limited to … … only". In view of the language convention of chinese, the following description, when it does not specifically state the number of a component, means that the component may be one or more, or may be understood as at least one.
Most practical systems in real life are composed of a large number of interacting, different types of components, which current analytical methods typically model as Homogeneous information networks (Homogeneous information networks). The modeling method using the homogeneous network usually extracts only part of the information of the actual interactive system, or does not distinguish the differences of the objects and the relations in the interactive system, which causes incomplete information or information loss.
The embodiments of the present invention model these interconnected multi-type networked data as Heterogeneous information (Heterogeneous information) models, and design a structural analysis method by using rich object and relationship information in the network. Compared with the widely researched homogeneous information network, the heterogeneous information network contains comprehensive structural information and rich semantic information, which also provides new opportunities and challenges for data mining.
The question-answering system based on the heterogeneous information model and the knowledge graph fully utilizes the characteristics of the heterogeneous information network to model elements such as scene logic, texts and character relations, improves the traditional question-answering system based on the knowledge graph through richer semantics contained in the heterogeneous information network, and improves the accuracy and the intelligent degree of the question-answering.
The question-answering method for constructing the knowledge graph based on the heterogeneous information model can improve the accuracy and intelligence of the question-answering system and enable the question-answering system to be more flexible. The method is used for modeling elements such as scene logic, texts, character relations and the like in the form of entities, attributes and relations by analyzing multi-source heterogeneous data to generate a heterogeneous information network.
Moreover, a corresponding knowledge graph is constructed on the basis of the heterogeneous information network generated by the invention, the entity set of the knowledge graph is further expanded on the basis of meta-path discovery in the heterogeneous information network, relationship reasoning is further carried out on the basis of a ranking collaborative clustering algorithm and a meta-path similarity algorithm in the heterogeneous information network, and more intelligent answer is carried out through richer semantics contained in the relationship reasoning. The method makes full use of the characteristics of a heterogeneous information network, improves the traditional knowledge-graph-based question-answering system, and has higher coverage rate of matching during question answering and lower maintenance cost after the knowledge base is updated.
FIG. 1 is a flow chart of the intellectual question answering method based on knowledge graph of the present invention.
As shown in fig. 1, the method includes:
step 101: a user question is received.
Step 102: converting the user question into a regularization question.
Step 103: and performing question matching on the regularization question and a template base generated by a knowledge graph based on a heterogeneous information model, and determining a matching result with the highest similarity as an answer.
In one embodiment, the heterogeneous information model contains entities, attributes, and relationships; the method also includes a process of pre-generating a knowledge graph based on heterogeneous information models, the process including: acquiring unstructured data, semi-structured data and structured data; extracting entities, attributes, direct relations among the entities, direct relations among the attributes and direct relations among the entities and the attributes from the collected unstructured data, semi-structured data and structured data; taking the extracted entities and attributes as meta-path node objects, and constructing direct relations between every two meta-path node objects into a composite relation containing a plurality of direct relations based on a multi-level path; converting direct relations among the entities, direct relations among the attributes, direct relations among the entities and the attributes and the composite relations into meta-paths; comparing the similarity in the meta-paths based on a ranking co-clustering algorithm and a meta-path similarity algorithm to construct a knowledge ontology; and generating knowledge based on a knowledge reasoning mode and the knowledge ontology, and representing the knowledge into a knowledge representation format to generate the knowledge graph.
Wherein: structured data, also called row data, is data logically represented and implemented by a two-dimensional table structure, strictly following the data format and length specifications, and mainly stored and managed by a relational database. In contrast to structured data, unstructured data is not suitable for representation by a database two-dimensional table, including office documents of all formats, XML, HTML, various types of reports, pictures and audio, video information, and the like. The data structure of the unstructured data is irregular or incomplete, and the unstructured data has no predefined data model and is inconvenient to represent by a database two-dimensional logic table. Semi-structured data is a form of structured data that does not conform to the structure of a data model in which relational databases or other forms of data tables are associated, but contains relevant tags to separate semantic elements and to stratify records and fields. It is therefore also referred to as a self-describing structure.
Preferably, the meta-path includes a single directed line segment between the entities, a single directed line segment between the entities and the attributes, a single directed line segment between the attributes, a plurality of directed line segments between the entities and the attributes, or a plurality of directed line segments between the attributes, wherein the plurality of directed line segments between the entities span the attributes or the further entities, the plurality of directed line segments between the entities and the attributes span the further entities or the further attributes, and the plurality of directed line segments between the attributes span the entities or the further attributes.
Preferably, comparing the similarity among the meta-paths based on the ranking co-clustering algorithm and the meta-path similarity algorithm to construct the knowledge ontology comprises: performing clustering on the meta-paths based on a ranking co-clustering algorithm; calculating the similarity of the clustered meta-paths based on a meta-path similarity algorithm; and merging the meta-paths with the similarity larger than a preset threshold value, and constructing a knowledge body based on a merging result.
For example, the number of clusters K is given first. Step 0: initial partitions of the target object are generated and initial network clusters are generated from these partitions, i.e., { C0k } Kk ═ 1. Step 1: a ranking-based probability generation model is constructed for each network cluster, i.e., { P (x | Ctk) } Kk ═ 1. Step 2: the posterior probability (p (Ctk | x)) of each target object is calculated, and then the cluster assignment of objects is adjusted according to the new evaluation defined by the cluster posterior probability. And step 3: repeat steps 1 and 2 until there is no significant change in all clusters. That is, { C × k } Kk ═ 1 ═ Ctk } Kk ═ 1 ═ Ct-1k } Kk ═ 1. And 4, step 4: the posterior probability (p (C x k | x)) of each attribute object in each network cluster is calculated.
For example, the meta path similarity algorithm may use the PathCount algorithm or PathSim algorithm, etc.
According to the embodiment of the invention, under the network structure of a heterogeneous information network, a ranking collaborative clustering algorithm and a meta-path similarity algorithm are combined, and richer entity, relationship and attribute knowledge are comprehensively extracted, so that a knowledge graph with richer knowledge is constructed.
In one embodiment, performing clustering on meta-paths based on a ranked co-clustering algorithm comprises: step 1: initializing a cluster division result of the meta-path, and distributing a random cluster to each main target type object; step 2: obtaining the ranking score of each random cluster; and step 3: calculating new measurement, and estimating the mixed model coefficient of each main target object; and 4, step 4: adjusting the random cluster based on the mixed model coefficient; and 5: repeating the step (2), the step (3) and the step (4) until convergence; step 6: calculating a ranking distribution matrix according to the ranking distribution model; and 7, simultaneously clustering rows and columns of the ranking distribution matrix by using a matrix singular value decomposition technology to obtain a clustering result representing a master-slave type, wherein if a null cluster exists, the step (1) is carried out.
The embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
Exemplarily, in the question-answering system, the heterogeneous information model may include four entity types of user, question type, question, and answer. The user proposes that the questions are divided into different types, the questions have corresponding answers, the questions are connected, modeling is carried out on the entities through multi-source data, different relations among different objects are extracted, and therefore the real meaning of the questions can be better judged, and correct answers are given.
Fig. 2 is an exemplary schematic diagram of a heterogeneous information network in accordance with the present invention.
A heterogeneous information network is defined as a directed graph that contains multiple types of objects or relationships, each object belonging to a particular object type and each relationship belonging to a particular relationship type. The network schema is a directed graph defined on object types and relationship types, and is a meta-description of the information network. In addition, the meta path is a path defined on the heterogeneous information network mode and used for linking two types of objects, and the meta path not only describes the semantic relation between the objects, but also can extract the characteristic information between the objects. Compared with the traditional knowledge graph, the heterogeneous information network can fuse more types of objects and complex interaction relations thereof, and can also fuse information of a plurality of social network platforms. Also, heterogeneous information networks contain rich semantics. In heterogeneous information networks, different types of objects and links coexist, with different semantic meanings. Due to the characteristics of the heterogeneous information network, the application of the heterogeneous information network to the improvement of the traditional knowledge graph can enable the question-answering system to be more accurate and intelligent.
FIG. 3 is a first exemplary diagram of a meta-path according to the present invention. FIG. 4 is a second exemplary diagram of a meta-path according to the present invention.
Fig. 3-4 show examples of network modes and meta-paths of heterogeneous information networks, taking a question-answering system as an example. Fig. 3 is a meta path, which includes two entities of a question and an answer, two direct paths respectively include two relations of the answer to the question and the question corresponding to the answer, and finally, the semantic information expressed by the meta path is: a question is found that has the same answer as the specified question. The single directed line segment of the question pointing to the answer in fig. 3 represents the direct relationship of the question to the answer; a single directed line segment with the answer pointing to the question represents the answer's direct relationship to the question.
Fig. 4 is another meta path, which includes attributes of two entities of a user and a question, and includes a type of the question. The four direct paths include four relationships, namely, a question provided by the user, a question type corresponding to the question, a question in the type, a user who provides the question, and the like. The semantic information expressed by the meta path is: users who are of the same type of question as the question posed by the given user are found. It can be seen that the meta path shown in fig. 4 is a composite relationship including a plurality of direct relationships, and the semantic extraction capability of the meta path can be extended. The meta path shown in fig. 4 spans questions and types from user to user.
It can be seen that the meta-path not only describes the semantic relationship between the objects, but also can extract the feature information between the objects. The meta-path has great advantages in describing semantic information, and through the combination of different objects and edges of linked objects, the semantic relationship possibly existing between the two objects is discovered, and the relationship between the objects is comprehensively displayed and the correlation between the two objects is comprehensively measured.
Also, the meta-path may be implemented as a single directed line segment between entities, a single directed line segment between entities and attributes, a single directed line segment between attributes, multiple directed line segments between entities and attributes, or multiple directed line segments between attributes. Wherein: multiple directed line segments between entities span the attribute or another entity, multiple directed line segments between entities and attributes span the other entity or another attribute, and multiple directed line segments between attributes span the entity or another attribute.
For example, in fig. 3, the meta-path is implemented as a plurality of directed line segments between entities (two questions that span another entity (answer)). In fig. 4, the meta-path is implemented as a plurality of directed line segments between entities (two users across additional entities (questions) and attributes (types)).
Fig. 5 is a process flow diagram of a question-answering system according to the present invention. FIG. 6 is a diagram of a knowledge-graph architecture based on heterogeneous information models according to the present invention.
The respective blocks in fig. 6 are explained below.
1. A data acquisition module:
the module collects the multi-source heterogeneous data set according to three types of structured data, semi-structured data and unstructured data, and extracts information contained in the data.
2. Heterogeneous information extraction module:
the module extracts the information collected from the original data set, extracts the entities, attributes and relations in the information, forms an ontology knowledge expression on the basis, and constructs a subsequent meta-path.
3. Heterogeneous knowledge fusion module:
the module processes the extracted entity and attribute, eliminates entity ambiguity, replaces reference nouns, and generates the processed information according to meta-path specification.
4. Heterogeneous knowledge processing module:
the module finds the entity, the relation and the attribute with the highest similarity by calculating the similarity of the generated meta-paths and ranking the collaborative clustering on the generated meta-paths, and better links the candidate entity object in the knowledge base to the correct entity object. And simultaneously, the entity set is expanded through entity discovery of the meta-path, the series of generated expressions construct an ontology in a data-driven automatic mode, so that a prototype of the knowledge graph is built, and finally, further knowledge discovery is completed through a knowledge reasoning technology to generate a final knowledge graph.
The user question conversion in fig. 5 specifically includes: the method comprises the steps of converting the user question into a question text, carrying out question classification on the input natural language, then further carrying out word segmentation, part of speech tagging, entity recognition and entity disambiguation, finishing the discrimination of the user question and generating an expression of the user question according with the template standard.
Therefore, different relationships among different entities are constructed through entity types such as user attributes, users, scenes, scene attributes and the like, a heterogeneous information model is constructed to serve as a knowledge base to provide knowledge sources for questions and answers, and maintenance and incremental updating of knowledge are kept in the using process. And simultaneously, preparing a QA corpus for providing a corpus for model training, and identifying the entity judged by the model and the true answer. Then, combining the two, generating URL in the knowledge map corresponding to the category, resource and entity, extracting knowledge, constructing a template, and generating a template base of the whole question/entity. And the natural language expressions to be linked in the semantic slots are respectively linked to the knowledge maps corresponding to the categories, resources and entities, and are inquired in the template library, and the inquiry result is returned and converted into a form which is easy to understand by the user.
The algorithm design and flow of the present invention is described below. Under the network structure of a heterogeneous information network, the method combines a ranking collaborative clustering algorithm and a meta-path similarity algorithm, comprehensively compares and extracts richer entity, relationship and attribute knowledge, and helps to construct a knowledge map with richer knowledge.
(1) Ranking collaborative clustering algorithm
FIG. 7 is a process flow diagram of a ranked co-clustering algorithm according to the present invention. The ranking clustering and the collaborative clustering are combined by adopting a proper method and are applied to clustering of heterogeneous information networks, and the method fully utilizes the relation between and in the heterogeneous information network types to obtain clustering results of various types, thereby playing a better role. Firstly, a ranking function is used for converting the relation into ranking, secondly, a ranking distribution generation model is built to obtain a matrix of ranking distribution, and finally, a collaborative clustering is used for obtaining an overall knowledge clustering result. In short, the algorithm idea is that firstly, link information mapping between nodes in a heterogeneous information network is converted into ranking information, and on the basis, a global knowledge structure reflecting multiple types of clustering results is mined by using collaborative clustering.
The important steps of calculating the ranking distribution matrix in the ranking cluster are as follows:
conditional ranking and intra-cluster ranking definitions: master type X, slave type Y, master type Cluster
Sub-diagram G ═<{X′∪Y},W′>. Conditional ranking r of YY|X′And intra-cluster ranking r of XX′|X′From the ranking function f of sub-graph G: (r)X′|X′,rY|X′) Given as f (G'), the condition for X is given the nomenclature rX|X′The propagation score on the network G is:
rule 1: different types of high ranked objects tend to occur simultaneously.
According to this, the object ranking of X is determined by the number of its links and the ranking of its associated Y-shaped objects
Rule 2: the same type of high ranked objects tend to appear simultaneously. The formula is modified as follows:
assume clustering result of X { X1,X2…,XKIs known, can be computed at cluster X by means of a ranking functionKRank of Medium Y Condition
And X conditional ranking
Will be provided with
And
notation Pk(Y) and Pk(X) represents. Any object X in Xi(i ═ 1,2, … m), Y ranking distribution
The mixed distribution of the conditional ranking of Y in K clusters is:
upper thetai,kIs the component coefficient of the kth distribution, which can be regarded as x hereiDegree of attribution to the kth cluster. Following the procedure described above, each x is obtainediRank distribution at Y
Thereby obtaining the required matrix PX(Y)。
The knowledge structure contained in the heterogeneous information network comprises more than two types of objects, the ranking distribution matrix is regarded as a two-part weighted graph comprising the two types, and the rows and the columns are clustered simultaneously by using a matrix singular value decomposition technology.
Meta-path similarity algorithm:
PathCount and PathSim are two similarity measurement methods. PathCount measures the similarity between nodes by calculating the number of path instances between two nodes:
PathSim is a normalized version of PathCount, and can find similar peer entities:
the similarity measurement of the paths with the weights is divided into two steps:
1. the similarity of the meta path is calculated by directly applying the traditional similarity measurement method;
2. and accumulating the similarities based on different atom element paths to obtain the similarity based on the path with the weight element.
Based on the meta path and the question-answering system with the weight element path in the heterogeneous information network, the meta path can calculate the problems of the same or similar types of the target problems, and then the answers corresponding to the problems are found out according to the similarity.
FIG. 8 is a sequence diagram of a knowledge-graph generation process according to the present invention. Fig. 9 is a sequence diagram of a user's usage flow of the question-answering system according to the present invention.
When a user proposes a question, the system firstly resolves the question into a certain semantic representation through a relevant semantic interpretation means, and then finds a corresponding answer in a knowledge graph constructed by a heterogeneous information model through a semantic reasoning technology. Each node or edge in the knowledge base subgraph can be used as a candidate answer, the question is observed, information extraction is carried out according to the template, a question feature vector is obtained, and the question and the answer are identified and screened, so that the final answer is obtained.
When a new data set exists in the question-answering system, the unstructured data, the semi-structured data and the structured data are acquired in different modes by calling a data acquisition module, then, entity, relationship, attribute information and the like contained in the acquired data are extracted by calling a heterogeneous information extraction module, the extracted information is processed by a heterogeneous knowledge fusion module to eliminate ambiguity, repetition and the like, the extracted entity and attribute serve as a meta-path node object, the direct relationship between every two objects is further constructed into a compound relationship between different object types through a multi-level path, the effect shown in figure 4 is achieved, and the processing result is generated into a meta-path. And finally, calling a heterogeneous knowledge processing module, comprehensively comparing the similarity in the meta path through a ranking collaborative clustering algorithm and a meta path similarity algorithm, helping to carry out relationship inference, constructing a knowledge body, wherein basic knowledge contained in the knowledge body can be used as a prototype of the knowledge graph, and in order to construct the knowledge graph with richer semantics, acquiring new deeper knowledge or conclusions on the basic knowledge through knowledge inference technologies such as a body inference method and a generative rule inference method, finally generating complete knowledge through the knowledge inference technology, expressing the complete knowledge into a normalized knowledge expression format such as RDF/URI, and writing the knowledge graph into a knowledge graph based on a heterogeneous information model. When new data is generated in the multi-source heterogeneous database, the system initiates a new knowledge generation process of the knowledge map, finally writes the knowledge into the map after the knowledge is generated, and returns a knowledge map updating success signal to the system to complete the calling of the process.
As shown in fig. 9, when a user presents a question in the question-answering system, the question is converted into a regularized analytic question that can be identified by the knowledge graph, then answer reasoning is performed, the regularized question that is parsed in the front is matched with a template library generated by the knowledge graph, and a matching result with the highest similarity is returned to the user as an answer.
In summary, the embodiment of the present invention applies the heterogeneous information network to the question-answering system to improve the accuracy and intelligence of the question-answering system. The specific method comprises the following steps: analyzing multi-source heterogeneous data, modeling elements such as scene logic, texts and character relations, generating a heterogeneous information network, constructing a corresponding knowledge graph on the basis of the generated heterogeneous information network, further expanding an entity set of the knowledge graph based on meta-path discovery in the heterogeneous information network, further performing relation reasoning based on meta-path similarity in the heterogeneous information network, and performing more intelligent answer through richer semantics contained in the relation reasoning. The specific process is as follows: when a user proposes a question, the system firstly resolves the question into a certain semantic representation through a relevant semantic interpretation means, and then finds a corresponding answer in a knowledge graph constructed by a heterogeneous information model through a semantic reasoning technology. Each node or edge in the knowledge base subgraph can be used as a candidate answer, the question is observed, information extraction is carried out according to the template, a question feature vector is obtained, and the question and the answer are identified and screened, so that the final answer is obtained.
First, in the embodiment of the present invention, the knowledge graph is constructed through a heterogeneous information network, and the knowledge graph is improved by using the characteristics of the heterogeneous information network. The main characteristics are: and the semi-structured representation method is convenient for modeling data. And (3) comprehensively modeling and effectively analyzing more complex and diversified data. More types of objects and complex interactive relations among the objects are fused, and information of a plurality of social network platforms is fused, so that the knowledge base is more complete and more convenient to update, and the original knowledge-graph-based question-answering system is more accurate and intelligent. Different types of objects and links coexist, containing rich semantics. The meta path can capture the semantic information, and the limited meta path, the weighted meta path and other methods can expand the semantic extraction capability of the meta path. In the knowledge discovery process, as more semantic information is considered, the result is refined and more accurate.
Moreover, in the embodiment of the invention, a knowledge ontology is further constructed by combining the collaborative ranking clustering and the meta-path algorithm. In a heterogeneous information network structure, richer entity, relationship and attribute knowledge are comprehensively extracted by combining a ranking collaborative clustering algorithm and a meta-path similarity algorithm, and a knowledge graph with richer knowledge is helped to be constructed. Ranking and collaborative clustering algorithm: and converting link information between nodes in the heterogeneous information network into ranking information, and mining a global knowledge structure reflecting various types of clustering results by using a collaborative clustering algorithm. The method makes full use of the relation between and in the heterogeneous information network types and obtains the clustering results of various types, thereby playing a better role. For meta-path similarity calculation algorithm: the similarity of the element paths can be calculated through PathCount and PathSim algorithms, and then the similarity based on different atom element paths is accumulated, so that the similarity based on the paths with the weight elements can be obtained. Therefore, the questions with the same or similar types of the target questions can be obtained, and then the answers corresponding to the questions are found according to the similarity. And comparing and combining the results of the collaborative ranking clustering algorithm and the meta-path algorithm, so that a more complete and rich knowledge graph can be constructed.
FIG. 10 is a block diagram of an intellectual answering system based on knowledge-maps according to the present invention.
As shown in fig. 10, the intellectual property map-based intelligent question answering apparatus includes:
the question receiving unit is used for receiving a user question;
the regularization unit is used for converting the user question into a regularization question;
and the answer determining unit is used for matching the regularized question with a template library generated by a knowledge graph based on a heterogeneous information model and determining a matching result with the highest similarity as an answer.
In one embodiment, the heterogeneous information model contains entities, attributes, and relationships; the apparatus also includes a knowledge graph generation unit;
the knowledge graph generating unit is used for acquiring unstructured data, semi-structured data and structured data; extracting entities, attributes, direct relations among the entities, direct relations among the attributes and direct relations among the entities and the attributes from the collected unstructured data, semi-structured data and structured data; taking the extracted entities and attributes as meta-path node objects, and constructing direct relations between every two meta-path node objects into a composite relation containing a plurality of direct relations based on a multi-level path; converting direct relations among the entities, direct relations among the attributes, direct relations among the entities and the attributes and the composite relations into meta-paths; comparing the similarity in the meta-paths based on a ranking co-clustering algorithm and a meta-path similarity algorithm to construct a knowledge ontology; and generating knowledge based on a knowledge reasoning mode and the knowledge ontology, and representing the knowledge into a knowledge representation format to generate the knowledge graph.
In one embodiment, the knowledge graph generating unit is configured to perform clustering on meta paths based on a ranking co-clustering algorithm; calculating the similarity of the clustered meta-paths; and merging the meta-paths with the similarity larger than a preset threshold value, and constructing a knowledge body based on a merging result.
The embodiment of the invention also provides a knowledge graph updating system of the question answering system. The method comprises the following steps: a question-answering system for acquiring a data set; the data acquisition module is used for acquiring unstructured data, semi-structured data and structured data from a database; the heterogeneous information extraction module is used for extracting the contained entities, attributes, direct relations among the entities, direct relations among the attributes and direct relations among the entities and the attributes from the unstructured data, the semi-structured data and the structured data which are acquired by the data acquisition module; the heterogeneous knowledge fusion module is used for taking the extracted entities and attributes as meta-path node objects, and constructing the direct relationship between every two meta-path node objects into a composite relationship containing a plurality of direct relationships based on a multi-level path; converting direct relations among the entities, direct relations among the attributes, direct relations among the entities and the attributes and the composite relations into meta-paths; and the heterogeneous knowledge processing module is used for comparing the similarity in the meta paths based on a ranking collaborative clustering algorithm and a meta path similarity algorithm to construct a knowledge ontology, generating knowledge based on a knowledge reasoning mode and the knowledge ontology, representing the knowledge into a knowledge representation format to generate the knowledge graph, and sending the knowledge graph to the question-answering system.
In one embodiment, the heterogeneous knowledge processing module is configured to perform clustering on the meta-paths based on a ranking co-clustering algorithm; calculating the similarity of the clustered meta-paths based on a meta-path similarity algorithm; and merging the meta-paths with the similarity larger than a preset threshold value, and constructing a knowledge body based on a merging result.
The embodiment of the invention also provides an intelligent question-answering device based on the knowledge graph and provided with a processor-memory architecture. The intelligent question answering device comprises a processor and a memory; the memory has stored therein an application executable by the processor for causing the processor to execute the wisdom question-answering method based on knowledge-maps as described above.
The memory may be embodied as various storage media such as an Electrically Erasable Programmable Read Only Memory (EEPROM), a Flash memory (Flash memory), and a Programmable Read Only Memory (PROM). The processor may be implemented to include one or more central processors or one or more field programmable gate arrays, wherein the field programmable gate arrays integrate one or more central processor cores. In particular, the central processor or central processor core may be implemented as a CPU or MCU.
It should be noted that not all steps and modules in the above flows and structures are necessary, and some steps or modules may be omitted according to actual needs. The execution order of the steps is not fixed and can be adjusted as required. The division of each module is only for convenience of describing adopted functional division, and in actual implementation, one module may be divided into multiple modules, and the functions of multiple modules may also be implemented by the same module, and these modules may be located in the same device or in different devices.
The hardware modules in the various embodiments may be implemented mechanically or electronically. For example, a hardware module may include a specially designed permanent circuit or logic device (e.g., a special purpose processor such as an FPGA or ASIC) for performing specific operations. A hardware module may also include programmable logic devices or circuits (e.g., including a general-purpose processor or other programmable processor) that are temporarily configured by software to perform certain operations. The implementation of the hardware module in a mechanical manner, or in a dedicated permanent circuit, or in a temporarily configured circuit (e.g., configured by software), may be determined based on cost and time considerations.
The present invention also provides a machine-readable storage medium storing instructions for causing a machine to perform a method as described herein. Specifically, a system or an apparatus equipped with a storage medium on which a software program code that realizes the functions of any of the embodiments described above is stored may be provided, and a computer (or a CPU or MPU) of the system or the apparatus is caused to read out and execute the program code stored in the storage medium. Further, part or all of the actual operations may be performed by an operating system or the like operating on the computer by instructions based on the program code. The functions of any of the above-described embodiments may also be implemented by writing the program code read out from the storage medium to a memory provided in an expansion board inserted into the computer or to a memory provided in an expansion unit connected to the computer, and then causing a CPU or the like mounted on the expansion board or the expansion unit to perform part or all of the actual operations based on the instructions of the program code.
Examples of the storage medium for supplying the program code include floppy disks, hard disks, magneto-optical disks, optical disks (e.g., CD-ROMs, CD-R, CD-RWs, DVD-ROMs, DVD-RAMs, DVD-RWs, DVD + RWs), magnetic tapes, nonvolatile memory cards, and ROMs. Alternatively, the program code may be downloaded from a server computer via a communications network.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (12)
1. An intelligent question-answering method based on a knowledge graph is characterized by comprising the following steps:
receiving a user question;
converting the user question into a regularization question;
and performing question matching on the regularization question and a template base generated by a knowledge graph based on a heterogeneous information model, and determining a matching result with the highest similarity as an answer.
2. The intellectual property graph based question answering method according to claim 1, wherein the heterogeneous information model comprises entities, attributes and relationships; the method further comprises a process of pre-generating the knowledge graph based on the heterogeneous information model, wherein the process comprises the following steps:
acquiring unstructured data, semi-structured data and structured data;
extracting entities, attributes, direct relations among the entities, direct relations among the attributes and direct relations among the entities and the attributes from the collected unstructured data, semi-structured data and structured data;
taking the extracted entities and attributes as meta-path node objects, and constructing direct relations between every two meta-path node objects into a composite relation containing a plurality of direct relations based on a multi-level path;
converting direct relations among the entities, direct relations among the attributes, direct relations among the entities and the attributes and the composite relations into meta-paths;
comparing the similarity in the meta-paths based on a ranking co-clustering algorithm and a meta-path similarity algorithm to construct a knowledge ontology;
and generating knowledge based on a knowledge reasoning mode and the knowledge ontology, and representing the knowledge into a knowledge representation format to generate the knowledge graph.
3. The intellectual question answering method based on the knowledge graph according to claim 2, wherein the meta path includes a single directional line segment between entities, a single directional line segment between entities and attributes, a single directional line segment between attributes, a plurality of directional line segments between entities or attributes, wherein the plurality of directional line segments between the entities span an attribute or another entity, the plurality of directional line segments between the entities and attributes span another entity or another attribute, the plurality of directional line segments between attributes span an entity or another attribute.
4. The intellectual question answering method based on the knowledge graph of claim 2, wherein the comparing the similarity among the meta paths based on the ranking co-clustering algorithm and the meta path similarity algorithm to construct the ontology comprises:
performing clustering on the meta-paths based on a ranking co-clustering algorithm;
calculating the similarity of the clustered meta-paths based on a meta-path similarity algorithm;
and merging the meta-paths with the similarity larger than a preset threshold value, and constructing a knowledge body based on a merging result.
5. The wisdom question-answering method based on knowledge-graph of claim 4, wherein the performing clustering on meta-paths based on ranking co-clustering algorithm comprises:
step 1: initializing a cluster division result of the meta-path, and distributing a random cluster to each main target type object;
step 2: obtaining the ranking score of each random cluster;
and step 3: calculating new measurement, and estimating the mixed model coefficient of each main target object;
and 4, step 4: adjusting the random cluster based on the mixed model coefficient;
and 5: repeating the step (2), the step (3) and the step (4) until convergence;
step 6: calculating a ranking distribution matrix according to the ranking distribution model;
and 7, simultaneously clustering rows and columns of the ranking distribution matrix in a matrix singular value decomposition mode to obtain a clustering result representing a master-slave type, wherein if a hollow cluster exists, the step (1) is carried out.
6. An intelligent question-answering device based on a knowledge graph is characterized by comprising:
the question receiving unit is used for receiving a user question;
the regularization unit is used for converting the user question into a regularization question;
and the answer determining unit is used for matching the regularized question with a template library generated by a knowledge graph based on a heterogeneous information model and determining a matching result with the highest similarity as an answer.
7. The intellectual property graph based question answering device according to claim 6, wherein the heterogeneous information model comprises entities, attributes and relationships; the apparatus also includes a knowledge graph generation unit;
the knowledge graph generating unit is used for acquiring unstructured data, semi-structured data and structured data; extracting entities, attributes, direct relations among the entities, direct relations among the attributes and direct relations among the entities and the attributes from the collected unstructured data, semi-structured data and structured data; taking the extracted entities and attributes as meta-path node objects, and constructing direct relations between every two meta-path node objects into a composite relation containing a plurality of direct relations based on a multi-level path; converting direct relations among the entities, direct relations among the attributes, direct relations among the entities and the attributes and the composite relations into meta-paths; comparing the similarity in the meta-paths based on a ranking co-clustering algorithm and a meta-path similarity algorithm to construct a knowledge ontology; and generating knowledge based on a knowledge reasoning mode and the knowledge ontology, and representing the knowledge into a knowledge representation format to generate the knowledge graph.
8. The intellectual property map based question answering apparatus according to claim 7,
the knowledge graph generating unit is used for clustering the meta-paths based on a ranking collaborative clustering algorithm; calculating the similarity of the clustered meta-paths; and merging the meta-paths with the similarity larger than a preset threshold value, and constructing a knowledge body based on a merging result.
9. A knowledge-graph update system for a question-answering system, comprising:
a question-answering system for acquiring a data set;
the data acquisition module is used for acquiring unstructured data, semi-structured data and structured data from a database;
the heterogeneous information extraction module is used for extracting the contained entities, attributes, direct relations among the entities, direct relations among the attributes and direct relations among the entities and the attributes from the unstructured data, the semi-structured data and the structured data which are acquired by the data acquisition module;
the heterogeneous knowledge fusion module is used for taking the extracted entities and attributes as meta-path node objects, and constructing the direct relationship between every two meta-path node objects into a composite relationship containing a plurality of direct relationships based on a multi-level path; converting direct relations among the entities, direct relations among the attributes, direct relations among the entities and the attributes and the composite relations into meta-paths;
and the heterogeneous knowledge processing module is used for comparing the similarity in the meta paths based on a ranking collaborative clustering algorithm and a meta path similarity algorithm to construct a knowledge ontology, generating knowledge based on a knowledge reasoning mode and the knowledge ontology, representing the knowledge into a knowledge representation format to generate the knowledge graph, and sending the knowledge graph to the question-answering system.
10. The knowledge-graph updating system of a question-answering system according to claim 9,
the heterogeneous knowledge processing module is used for clustering the element paths based on a ranking collaborative clustering algorithm; calculating the similarity of the clustered meta-paths based on a meta-path similarity algorithm; and merging the meta-paths with the similarity larger than a preset threshold value, and constructing a knowledge body based on a merging result.
11. An intelligent question-answering device based on a knowledge graph is characterized by comprising a processor and a memory;
the memory stores an application program executable by the processor for causing the processor to perform the wisdom knowledge-graph-based question-answering method according to any one of claims 1-5.
12. A computer-readable storage medium having stored thereon computer-readable instructions for performing the intellectual property map-based question answering method according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910419946.4A CN111966793B (en) | 2019-05-20 | 2019-05-20 | Intelligent question-answering method and system based on knowledge graph and knowledge graph updating system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910419946.4A CN111966793B (en) | 2019-05-20 | 2019-05-20 | Intelligent question-answering method and system based on knowledge graph and knowledge graph updating system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111966793A true CN111966793A (en) | 2020-11-20 |
| CN111966793B CN111966793B (en) | 2024-03-15 |
Family
ID=73358321
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910419946.4A Active CN111966793B (en) | 2019-05-20 | 2019-05-20 | Intelligent question-answering method and system based on knowledge graph and knowledge graph updating system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111966793B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112527924A (en) * | 2020-12-18 | 2021-03-19 | 清华大学 | Dynamically updated knowledge graph expansion method and device |
| CN112597316A (en) * | 2020-12-30 | 2021-04-02 | 厦门渊亭信息科技有限公司 | Interpretable reasoning question-answering method and device |
| CN114357191A (en) * | 2021-12-31 | 2022-04-15 | 科大讯飞(苏州)科技有限公司 | Knowledge extraction, question answering and recommendation method, related device, equipment and medium |
| CN117077792A (en) * | 2023-10-12 | 2023-11-17 | 支付宝(杭州)信息技术有限公司 | Knowledge graph-based method and device for generating prompt data |
| CN117435749A (en) * | 2023-12-21 | 2024-01-23 | 摩斯智联科技有限公司 | Method, device and storage medium for generating knowledge graph |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106909643A (en) * | 2017-02-20 | 2017-06-30 | 同济大学 | The social media big data motif discovery method of knowledge based collection of illustrative plates |
| CN106997399A (en) * | 2017-05-24 | 2017-08-01 | 海南大学 | A kind of classification question answering system design method that framework is associated based on data collection of illustrative plates, Information Atlas, knowledge mapping and wisdom collection of illustrative plates |
| CN108920556A (en) * | 2018-06-20 | 2018-11-30 | 华东师范大学 | Recommendation expert method based on subject knowledge map |
| US20190012405A1 (en) * | 2017-07-10 | 2019-01-10 | International Business Machines Corporation | Unsupervised generation of knowledge learning graphs |
| CN109255002A (en) * | 2018-09-11 | 2019-01-22 | 浙江大学 | A method of it is excavated using relation path and solves knowledge mapping alignment task |
| CN109255034A (en) * | 2018-08-08 | 2019-01-22 | 数据地平线（广州）科技有限公司 | A kind of domain knowledge map construction method based on industrial chain |
| CN109492077A (en) * | 2018-09-29 | 2019-03-19 | 北明智通（北京）科技有限公司 | The petrochemical field answering method and system of knowledge based map |
| US20190130289A1 (en) * | 2017-10-30 | 2019-05-02 | International Business Machines Corporation | Original idea extraction from written text data |
-
2019
- 2019-05-20 CN CN201910419946.4A patent/CN111966793B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106909643A (en) * | 2017-02-20 | 2017-06-30 | 同济大学 | The social media big data motif discovery method of knowledge based collection of illustrative plates |
| CN106997399A (en) * | 2017-05-24 | 2017-08-01 | 海南大学 | A kind of classification question answering system design method that framework is associated based on data collection of illustrative plates, Information Atlas, knowledge mapping and wisdom collection of illustrative plates |
| US20190012405A1 (en) * | 2017-07-10 | 2019-01-10 | International Business Machines Corporation | Unsupervised generation of knowledge learning graphs |
| US20190130289A1 (en) * | 2017-10-30 | 2019-05-02 | International Business Machines Corporation | Original idea extraction from written text data |
| CN108920556A (en) * | 2018-06-20 | 2018-11-30 | 华东师范大学 | Recommendation expert method based on subject knowledge map |
| CN109255034A (en) * | 2018-08-08 | 2019-01-22 | 数据地平线（广州）科技有限公司 | A kind of domain knowledge map construction method based on industrial chain |
| CN109255002A (en) * | 2018-09-11 | 2019-01-22 | 浙江大学 | A method of it is excavated using relation path and solves knowledge mapping alignment task |
| CN109492077A (en) * | 2018-09-29 | 2019-03-19 | 北明智通（北京）科技有限公司 | The petrochemical field answering method and system of knowledge based map |
Non-Patent Citations (2)
| Title |
|---|
| YANG ZHOU 等: "Integrating Vertex-centric Clustering with Edge-centric Clustering for Meta Path Graph Analysis", 《KDD \'15: PROCEEDINGS OF THE 21TH ACM SIGKDD INTERNATIONAL CONFERENCE ON KNOWLEDGE DISCOVERY AND DATA MINING》, 10 August 2015 (2015-08-10), pages 1563, XP058070978, DOI: 10.1145/2783258.2783328 * |
| 王鹏伟: "基于模板学习和深度学习的知识抽取和表示学习的研究与应用", 《中国博士学位论文全文数据库 信息科技辑》, no. 12, 15 December 2018 (2018-12-15), pages 140 - 15 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112527924A (en) * | 2020-12-18 | 2021-03-19 | 清华大学 | Dynamically updated knowledge graph expansion method and device |
| CN112597316A (en) * | 2020-12-30 | 2021-04-02 | 厦门渊亭信息科技有限公司 | Interpretable reasoning question-answering method and device |
| CN112597316B (en) * | 2020-12-30 | 2023-12-26 | 厦门渊亭信息科技有限公司 | Method and device for interpretive reasoning question-answering |
| CN114357191A (en) * | 2021-12-31 | 2022-04-15 | 科大讯飞(苏州)科技有限公司 | Knowledge extraction, question answering and recommendation method, related device, equipment and medium |
| CN117077792A (en) * | 2023-10-12 | 2023-11-17 | 支付宝(杭州)信息技术有限公司 | Knowledge graph-based method and device for generating prompt data |
| CN117077792B (en) * | 2023-10-12 | 2024-01-09 | 支付宝(杭州)信息技术有限公司 | Knowledge graph-based method and device for generating prompt data |
| CN117435749A (en) * | 2023-12-21 | 2024-01-23 | 摩斯智联科技有限公司 | Method, device and storage medium for generating knowledge graph |
| CN117435749B (en) * | 2023-12-21 | 2024-03-15 | 摩斯智联科技有限公司 | Method, device and storage medium for generating knowledge graph |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111966793B (en) | 2024-03-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110609902B (en) | Text processing method and device based on fusion knowledge graph | |
| CN111966793B (en) | Intelligent question-answering method and system based on knowledge graph and knowledge graph updating system | |
| US10725836B2 (en) | Intent-based organisation of APIs | |
| US10402379B2 (en) | Predictive search and navigation for functional information systems | |
| KR20180041200A (en) | Information processing method and apparatus | |
| CN111539197B (en) | Text matching method and device, computer system and readable storage medium | |
| Zhou et al. | A survey on multi-modal social event detection | |
| Nagy et al. | Multiagent ontology mapping framework for the semantic web | |
| US20220058222A1 (en) | Method and apparatus of processing information, method and apparatus of recommending information, electronic device, and storage medium | |
| CN105393263A (en) | Feature completion in computer-human interactive learning | |
| CN105045875A (en) | Personalized information retrieval method and apparatus | |
| CN111274332A (en) | Intelligent patent retrieval method and system based on knowledge graph | |
| Baglatzi et al. | Semantifying OpenStreetMap. | |
| CN115563313A (en) | Knowledge graph-based document book semantic retrieval system | |
| CN103778206A (en) | Method for providing network service resources | |
| Leonov et al. | Architecture and self-learning concept of knowledge-based systems by use monitoring of internet network | |
| CN113515589A (en) | Data recommendation method, device, equipment and medium | |
| Budikova et al. | ConceptRank for search-based image annotation | |
| Partyka et al. | Enhanced geographically typed semantic schema matching | |
| CN117932022A (en) | Intelligent question-answering method and device, electronic equipment and storage medium | |
| CN113704519B (en) | Data set determining method and device, computer equipment and storage medium | |
| KR20070118154A (en) | Information processing device and method, and program recording medium | |
| CN113076468B (en) | Nested event extraction method based on field pre-training | |
| Yan | Geographic knowledge graph summarization | |
| KR20150066994A (en) | Method and apparatus for text classifier based on open directory project utilizing ancestor descendant categories |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |