CN109902165A - Intelligent interactive answering method, system, device based on Markov Logic Networks - Google Patents
Intelligent interactive answering method, system, device based on Markov Logic Networks Download PDFInfo
- Publication number
- CN109902165A CN109902165A CN201910174742.9A CN201910174742A CN109902165A CN 109902165 A CN109902165 A CN 109902165A CN 201910174742 A CN201910174742 A CN 201910174742A CN 109902165 A CN109902165 A CN 109902165A
- Authority
- CN
- China
- Prior art keywords
- tuple
- network
- evidence
- structural
- domain knowledge
- 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 74
- 230000002452 interceptive effect Effects 0.000 title claims abstract description 41
- 230000004044 response Effects 0.000 claims abstract description 45
- 238000005516 engineering process Methods 0.000 claims abstract description 5
- 239000000284 extract Substances 0.000 claims abstract description 5
- 238000003745 diagnosis Methods 0.000 claims description 14
- 238000013507 mapping Methods 0.000 claims description 13
- 230000004913 activation Effects 0.000 claims description 9
- 230000003213 activating effect Effects 0.000 claims description 7
- 230000003993 interaction Effects 0.000 claims description 6
- 238000005457 optimization Methods 0.000 claims description 6
- 238000013461 design Methods 0.000 claims description 5
- 238000005295 random walk Methods 0.000 claims description 5
- 238000013145 classification model Methods 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 4
- 238000007635 classification algorithm Methods 0.000 claims description 3
- 238000012790 confirmation Methods 0.000 claims description 3
- 238000004891 communication Methods 0.000 abstract description 5
- 230000010354 integration Effects 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 12
- 230000008569 process Effects 0.000 description 11
- 241000700605 Viruses Species 0.000 description 7
- 238000010276 construction Methods 0.000 description 5
- 238000012545 processing Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 125000002015 acyclic group Chemical group 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000000470 constituent Substances 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 230000020169 heat generation Effects 0.000 description 1
- 238000010438 heat treatment Methods 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 231100000566 intoxication Toxicity 0.000 description 1
- 230000035987 intoxication Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000003607 modifier Substances 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 231100000572 poisoning Toxicity 0.000 description 1
- 230000000607 poisoning effect Effects 0.000 description 1
- 238000005381 potential energy Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 238000009827 uniform distribution Methods 0.000 description 1
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention belongs to network communications and field of computer technology, relate to a kind of intelligent interactive answering method, system, device based on Markov Logic Networks, it is intended to solve the problems, such as that intelligent Answer System effectively in conjunction with context and background and cannot be unable to Real-time Feedback, inefficiency in practical applications.The method of the present invention includes: analysis input information, extracts structuring tuple and carries out semantization extension using domain knowledge map;Dependency rule in domain knowledge map is activated, and the posterior probability of candidate answer information is calculated to evidence tuple assignment by the way of approximate resoning and/or data input;Export the high response message of the posterior probability of preset number.The present invention can be with effective integration context operations and field uncertainty knowledge, and approximation, which is deduced, interacts strong combination with user, really provide effective solution, while the automatic sorting of knowledge may be implemented, reduce artificial and data cost.
Description
Technical Field
The invention belongs to the technical field of network communication and computers, and particularly relates to an intelligent interactive question answering method, system and device based on a Markov logic network.
Background
Fault analysis and diagnosis of network devices and communication devices play a crucial role in the operation and maintenance of communication device manufacturers. Due to the complexity of the structure and the components, the types of failures are diverse. Based on the correlation analysis of the large operation and maintenance data of the equipment, it is absolutely necessary to automatically detect the fault reason and give a solution. The intelligent diagnostic problem can be formulated as interactions or influences between various types of objects, the complexity and uncertainty of which are difficult to express through traditional features. Under the traditional artificial intelligence framework, uncertainty and complexity belong to two independent research branches. For the former, a probability representation-based method is generally adopted, while assumptions such as independent equal distributions are applied; while the latter are handled by methods based on logical representations, but limited to deterministic cases, the resistance to noise and uncertainty is weak.
The probabilistic representation model represented by the classifier can be trained more efficiently, but usually only the fault itself can be returned, the logic rule cannot be obtained, so that detailed explanation from the fault to the symptom cannot be obtained, the combination with the existing domain knowledge and system knowledge is difficult, and in the multi-fault diagnosis, an additional selection strategy needs to be added to select a proper fault from a plurality of classes with the probability exceeding the threshold value. The probability representation model represented by the Bayesian network has the inherent advantages that some uncertainties of prediction details can be represented, so that the system has stronger interpretability, artificial knowledge and experience supervision are enabled to play a role, but the parameter is probability, the distribution form which can be factorized is greatly limited, the potential energy function of the model is necessarily conditional probability, and a directed graph is necessarily acyclic, so that the popularization and the application of the model are influenced. The diagnosis model represented by the expert system usually needs domain experts to manually construct the diagnosis model, the diagnosis in the field is very accurate, but is only limited to deterministic situations, the noise resistance and uncertainty are very weak, the knowledge extraction and diagnosis model construction process is very complicated and time-consuming, and the method is completely inapplicable when facing a high-change-rate examined system. However, the workload of constructing the fault tree is heavy, the difficulty is high, the requirement on an analyst is high, and the popularization of the fault tree is limited.
Generally, the current methods in the field have limitations, and cannot effectively feed back the single-fault diagnosis result or the multiple-fault diagnosis result in real time while effectively combining the context and the background field knowledge.
Disclosure of Invention
In order to solve the above problems in the prior art, that is, the problems that the intelligent question-answering system cannot effectively combine the context and the background in practical application, cannot feed back in real time and has low efficiency, the invention provides an intelligent interactive question-answering method based on a Markov logic network, which comprises the following steps:
step S10, analyzing the acquired input information, extracting a structural tuple, and performing semantic expansion on the entity and relationship expression in the structural tuple by using a domain knowledge map;
step S20, activating the domain knowledge graph based on the structural tuple and the semantic expansion result to obtain a rule sub-network and an evidence tuple set;
step S30, based on the rule sub-network, the evidence tuple set is assigned by adopting a mode of approximate reasoning and/or information input, and the posterior probability of the candidate response information is calculated;
step S40, selecting a preset number of response messages as final response messages and outputting the final response messages according to the sequence of the posterior probability of the candidate response messages from high to low;
wherein the domain knowledge graph comprises:
the system comprises a fact library, an ontology library, a rule library, an entity-relationship network corresponding to the superior-inferior relationship between the fact library and the entities in the ontology library, a tuple-relationship network corresponding to the inclusion relationship between the logic rules automatically learned in the rule library and the facts in the ontology library, and an intention-relationship network corresponding to the logic rules manually designed according to the field requirements in the rule library.
In some preferred embodiments, the domain knowledge graph is constructed by:
step B10, performing word meaning disambiguation on the unstructured corpora in the acquired data according to word labels, part of speech labels and concepts, extracting corpora as structured tuples based on sentence patterns and concept templates, and constructing a fact library;
step B20, establishing a hierarchical structure of the upper and lower relations between the corpus concepts and a hierarchical structure of the inclusion relation between the facts in the fact base, predicting the mapping relation between the entities and the hierarchical concepts based on a hierarchical multi-classification algorithm, and constructing an ontology base; logic rules among intents are set based on field requirements, and a rule base is constructed according to the logic rules and through automatic learning of Markov logic networks among structured tuples.
In some preferred embodiments, step B20, "based on the logic rules between the domain requirement design intents, automatically learning the markov logic network between the structural tuples according to the logic rules, and constructing the rule base", includes the steps of:
step B21, finding a semantic path between the structural tuples by combining random walk and the subgraph pattern, wherein the structural tuples appear in the context in the semantic path, and the semantic elements meet one or more of the same, upper and lower positions, components or attribute relations;
and step B22, converting the semantic path into a logic clause based on the logic rule between the domain requirement design intents, performing concept mapping on the semantic elements of each structural tuple in the logic clause, obtaining the weight corresponding to the logic clause with the concept variable by maximizing a certain conditional log-likelihood target, and forming a rule base by the logic clause set with the weight.
In some preferred embodiments, the step S10 "parsing the acquired input information, extracting a tuple with a structure, and semantically expanding the entity and relationship representation in the tuple with a domain knowledge graph" includes the steps of:
step S11, converting the input data into words and sequences of the parts of speech thereof by using Chinese word segmentation technology, and extracting structural tuples by combining the parts of speech and a concept template;
and step S12, expanding the entity and relationship representation in the structural tuple by using an ontology library in the domain knowledge graph and the obtained synonym library.
In some preferred embodiments, in step S20, "activate the domain knowledge graph based on the structural tuple and the semantic expansion result, to obtain a rule sub-network and an evidence tuple set", the method includes:
and activating the logic rules in the rule base and the structural tuples in the fact base in the domain knowledge graph based on the structural tuples and the semantic expansion result thereof to obtain a rule sub-network and an evidence tuple set.
In some preferred embodiments, the step S30 of assigning the evidence tuple set by using approximate inference and/or information entry to calculate posterior probability of candidate response information based on the rule sub-network includes:
step S31, based on the rule sub-network and the evidence tuple set, sequentially selecting non-assigned evidence tuples, generating a diagnosis question, and executing:
obtaining an evidence tuple assigned by the user based on the confirmation information acquired by the human-computer interaction device; and/or reasoning the evidence tuples which are possible to be established through MPE to obtain the evidence tuples of reasoning assignment;
and step S32, calculating the posterior probability of the candidate response information based on the evidence tuple assigned by the user and/or the evidence tuple assigned by inference.
In some preferred embodiments, before "semantically extending the entities and the relationship representations in the structured tuples by using the domain knowledge graph" in step S10, there are further steps of automatic optimization and iterative update, where the method is:
and automatically optimizing and iteratively updating the domain knowledge graph by adopting an online optimization method of a Markov logic network structure and parameters.
On the other hand, the invention provides an intelligent interactive question-answering system based on a Markov logic network, which comprises an input module, a statement analysis module, a semantic activation module, an approximation module and an output module;
the input module is configured to acquire input information;
the statement analysis module is configured to analyze the acquired input information, extract a structural tuple and perform semantic expansion on an entity and a relation expression in the extracted structural tuple by adopting a domain knowledge graph;
the semantic activation module is configured to activate the domain knowledge graph based on the extracted structural module and semantic expansion result to obtain a rule subnetwork and an evidence tuple set;
the approximation deduction module is configured to assign values to the evidence tuple set by adopting an approximation reasoning and/or information input mode based on a rule sub-network, and calculate the posterior probability of the candidate response information;
the output module is configured to select a preset number of response messages as final response messages and output the final response messages according to the sequence of the posterior probability of the candidate response messages from high to low;
the domain knowledge graph is constructed, automatically optimized and iteratively updated by adopting a generalization learning module, wherein the generalization learning module comprises a concept learning module, a relation learning module and a structure learning module;
the concept learning module is configured to learn a hierarchical multi-label classification model of an entity and establish a mapping relation between the entity and a concept;
the relation learning module is configured to discover potential unknown relations between entities and complete a real-estate;
the structure learning module is configured to find semantic paths between facts and learn a Markov logic network structure and parameters under the constraint of intention rules.
In a third aspect of the present invention, a storage device is provided, in which a plurality of programs are stored, the programs being adapted to be loaded and executed by a processor to implement the above-mentioned intelligent interactive question-answering method based on a markov logic network.
In a fourth aspect of the present invention, a processing apparatus is provided, which includes a processor, a storage device; the processor is suitable for executing various programs; the storage device is suitable for storing a plurality of programs; the program is adapted to be loaded and executed by a processor to implement the above-described intelligent interactive question-answering method based on a markov logic network.
The invention has the beneficial effects that:
(1) the intelligent interactive question-answering method based on the Markov logic network can effectively fuse the context operation and the field uncertainty knowledge, and effectively combines the approximation with the instant interaction at the same time, thereby really achieving the purpose of providing a reasonable and effective solution. Meanwhile, automatic knowledge induction can be realized, and labor and data costs are reduced.
(2) In the method, in the field knowledge graph construction, more general structured knowledge representation is adopted, the fact is represented by a triple group of an entity 1-relation-entity 2 and an entity-attribute value without limitation, the implicit dependency relationship between the tuple corresponding to the field intention is mined by combining random walk and a sub-graph mode, useful diagnosis logic of assigned weight is generated, the method is suitable for rapid increase of the scale of the knowledge graph, and online approximate reasoning is strongly supported.
Drawings
Other features, objects and advantages of the present application will become more apparent upon reading of the following detailed description of non-limiting embodiments thereof, made with reference to the accompanying drawings in which:
FIG. 1 is a schematic flow chart of an intelligent interactive question-answering method based on a Markov logic network;
FIG. 2 is a schematic diagram of a construction process of a domain knowledge graph of the intelligent interactive question-answering method based on the Markov logic network;
FIG. 3 is a schematic diagram of a three-layer network model of a domain knowledge graph of the intelligent interactive question-answering method based on a Markov logic network of the present invention;
FIG. 4 is a diagram illustrating an example of semantic path partial order according to an embodiment of an intelligent interactive question-answering method based on a Markov logic network;
FIG. 5 is a schematic diagram of Markov logic network activation in one embodiment of the intelligent interactive question-answering method based on the Markov logic network of the present invention;
fig. 6 is a schematic diagram of the system structure of the intelligent interactive question-answering method based on the markov logic network.
Detailed Description
The present application will be described in further detail with reference to the following drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the relevant invention and not restrictive of the invention. It should be noted that, for convenience of description, only the portions related to the related invention are shown in the drawings.
It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict. The present application will be described in detail below with reference to the embodiments with reference to the attached drawings.
The invention provides an intelligent interactive question-answering method based on a Markov logic network, which has the advantages that on one hand, a huge Markov network is described simply and clearly from the perspective of probability statistics, and the capacity of a modular knowledge domain is flexibly merged into the Markov network; on the other hand, the Markov logic network introduces excellent uncertainty processing capability to the first-order predicate logic from the logic perspective, can tolerate the problems of incompleteness, contradiction and the like existing in a knowledge domain, and is very suitable for fault analysis and diagnosis of network equipment and communication equipment due to the interpretability and the capability of coping with uncertainty.
The invention relates to an intelligent interactive question-answering method based on a Markov logic network, which comprises the following steps:
step S10, analyzing the acquired input information, extracting a structural tuple, and performing semantic expansion on the entity and relationship expression in the structural tuple by using a domain knowledge map;
step S20, activating the domain knowledge graph based on the structural tuple and the semantic expansion result to obtain a rule sub-network and an evidence tuple set;
step S30, based on the rule sub-network, the evidence tuple set is assigned by adopting a mode of approximate reasoning and/or information input, and the posterior probability of the candidate response information is calculated;
step S40, selecting a preset number of response messages as final response messages and outputting the final response messages according to the sequence of the posterior probability of the candidate response messages from high to low;
wherein the domain knowledge graph comprises:
the system comprises a fact library, an ontology library, a rule library, an entity-relationship network corresponding to the superior-inferior relationship between the fact library and the entities in the ontology library, a tuple-relationship network corresponding to the inclusion relationship between the logic rules automatically learned in the rule library and the facts in the ontology library, and an intention-relationship network corresponding to the logic rules manually designed according to the field requirements in the rule library.
In order to more clearly describe the intelligent interactive question-answering method based on the markov logic network, the following describes each step in the embodiment of the method in detail with reference to fig. 1.
The intelligent interactive question-answering method based on the Markov logic network comprises the steps of S10-S40, wherein the steps are described in detail as follows:
the method comprises the following steps of constructing a domain knowledge graph aiming at an intelligent interactive question-answering method, wherein the domain knowledge graph comprises the following steps:
the system comprises a fact library, an ontology library, a rule library, an entity-relationship network corresponding to the superior-inferior relationship between the fact library and the entities in the ontology library, a tuple-relationship network corresponding to the inclusion relationship between the automatically learned logic rules in the rule library and the facts in the ontology library, and an intention-relationship network corresponding to the manually designed logic rules according to the field requirements in the rule library. As shown in fig. 2, a schematic view of a construction process of a knowledge graph in the field of the intelligent interactive question-answering method based on the markov logic network is provided, and the specific construction steps are as follows:
and step B10, performing word meaning disambiguation on the unstructured corpora in the acquired data according to word labels, part of speech labels and concepts, extracting corpora as structured tuples based on the sentence patterns and the concept templates, and constructing a fact library.
The method comprises the steps of constructing a domain knowledge graph, extracting a structured tuple with specific intentions (including phenomena, reasons, states, operations and schemes) based on a corpus, constructing a domain thesaurus specifically comprising ① a predicate thesaurus representing the phenomena, assisting in extracting the structured tuple representing the 'phenomena', such as incapability of logging in, disappearance, unavailable finding, stuck, abnormal and the like, ② a thesaurus representing the reasons, assisting in extracting the structured tuple representing the 'reasons' and the 'phenomena', such as cause, cause and the like, ③ a predicate thesaurus representing the states, such as yes, stay and the like, and obtaining a word sequence by using natural language processing technologies such as Chinese participle tagging and part-of-speech tagging, and extracting the structured tuples with different intentions by combining a basic sentence pattern and a concept template, wherein the structured tuples are shown in table 1:
TABLE 1
Where NP represents a nominal phrase, VP represents a verb phrase, PP represents a prepositive phrase, AP represents an adjective phrase, s represents a subject, v represents a predicate, o represents an object, and a represents a shape.
It should be noted that the above table is only used for better illustrating the technical solution of the present invention, and not for limiting the present invention, and those skilled in the art should understand that any way of automatically extracting the structured tuples combining the schema and the concept template should be included in the scope of the present invention.
Step B20, establishing a hierarchical structure of the upper and lower relations between the corpus concepts and a hierarchical structure of the inclusion relation between the facts in the fact base, predicting the mapping relation between the entities and the hierarchical concepts based on a hierarchical multi-classification algorithm, and constructing an ontology base; and designing logic rules among intents based on the field requirements, automatically learning a Markov logic network among the structured tuples according to the logic rules, and constructing a rule base.
An example hierarchy between concepts is as follows:
router | Equipment | goods | article, gateway | network | System | Whole | Structure, gateway | Equipment | goods | article
Wherein, the entity on the right side of the vertical bar "|" is the upper concept of the entity on the left side.
An example hierarchy between facts is as follows:
(s: open + Individual/application, v: Slow) | (s: open + application, v: Slow) | Catton | phenomenon (v: stuck + immobilized) | Catton | phenomenon
Wherein, the entity on the right side of the slash '/' is a central word, and the entity on the left side is a modifier
The optimal path prediction based hierarchical concept learning method performs automatic identification of entity concept labels, so that entity-concept mapping relations are formed, for example:
SD card | Equipment | articles | article, Hua is Mate9| Mobile phone | articles |
The entity-concept mapping table is obtained based on encyclopedic knowledge offline training, and supports semantic generalization in the extraction and approximate deduction processes of structural tuples.
Step B21, finding semantic paths between the structural tuples by combining random walk and subgraph patterns, wherein the structural tuples appear in the context in the semantic paths, and the semantic elements satisfy one or more of the same, upper and lower positions, components or attribute relations.
For example, the following tuples appear in the context of the fact library as shown in FIG. 4:
a: (s: Hua is USG6101, v: frequent/restart)
B: (s: voltage, v: Stable)
C: (v: not, o: original/Power supply)
D: (v: Replacing, o: original/Power supply)
Based on domain knowledge, it can be concluded that "hua be USG 6101" has a property of "voltage", that "power supply" has a property of "voltage", and that "power supply" is a component of "hua be USG 6101". A semantic path of length 1 (the length of the path is the number of edges) includes a-B, B-C, C-D, B-D, A-C, A-D. If two tuple pairs contain a common tuple, for example, tuple pair A-B and tuple pair B-C both contain tuple B, then the two tuple pairs can be connected to obtain a semantic path with length 2, i.e. A-B-C, and so on, a semantic path with length γ (γ ≧ 1) can be obtained.
To accommodate the rapid growth in data size, an efficient path sampling mechanism is employed to generate a representative subset of paths. For this purpose, a uniform sampling of the discriminant path is obtained based on the markov chain monte carlo algorithm, where the discriminant path refers to a path that contains structured tuples intended for the cause or solution. As shown in fig. 4, which is an exemplary diagram of semantic path partial order according to an embodiment of the intelligent interactive question-answering method based on the markov logic network of the present invention, the partial order diagram of the path is regarded as a state space of the markov chain, a partial order diagram is locally constructed around a current node, and neighbors of the current node include two types of nodes, one type is a hyper-path node, and the other type is a sub-path node. If the current node represents the path p, the neighbor-hyper path is all paths with the edge ratio p larger than 1, the neighbor-sub path is all paths with the edge ratio p smaller than 1, and the empty path has no sub path node. In order to obtain the uniform distribution of all paths, the method utilizes the idea of Metropolis-Hastings algorithm to carry out random walk on a partial sequence diagram, each round selects a neighbor according to a predefined transition probability, and if the node is a new judgment path and has not been visited before, the path is reserved.
And step B22, converting the semantic path into a logic clause based on the logic rule between the domain requirement design intents, performing concept mapping on the semantic elements of each structural tuple in the logic clause, obtaining the weight corresponding to the logic clause with the concept variable through a maximum conditional log-likelihood target, and forming a rule base by the logic clause set with the weight.
The domain-intention network is usually an undirected graph, and a fully-connected subgraph (clique) corresponds to some logical clause expression form, and an example is shown in table 2:
TABLE 2
The intention labels of the tuples in the above example are:
a: (s: gateway, v: frequent/restart) -phenomenon
B: (s: Voltage, v: Stable) -State
C: (v: Yes, o: original/Power) -State
D: (v: Replacing, o: original/Power supply) -scheme
Under the constraint of the domain intention network, the above semantic path can be converted into the following clause form:
(s: Hua is USG6101, v: frequent/restart) ^ (s: voltage, v: stable) ^ (v: not, o: original/power) } (v: change, o: original/power)
Using an ontology library, the concept mapping is performed on instance elements in a tuple (s: Hua USG6101, v: frequent/restart):
hua is USG 6101: gateway
The constituent elements in the other tuples are not examples and need not be notional abstracted.
Thus, the clauses in the above example can be abstracted to a logical clause f as follows:
f: (s: gateway, v: frequent/restart) ^ (s: voltage, v: stable) ^ (v: not, o: original/power) } (v: replacement, o: original/power)
Then, a conditional log-likelihood target is defined, for example, given a sample obtained example network, let X represent an evidence tuple set and Y represent a scheme tuple set, where the conditional log-likelihood of Y is given by the following equation (1):
Pw(Yj=yj| X ═ X) is defined as shown in formula (2):
wherein,representing at least one instantiation containing a query variable YjThe set of logical clauses of (a),representing inclusion of a query variable YjAnd takes the value yjThe ith logical clause of (a) is valued as the number of instantiations, wiRepresenting the weight of the ith logical clause.
And finally, obtaining a weight set corresponding to the logic clause by maximizing the conditional log-likelihood, namely obtaining a rule base expressed in the form of 'weight + rule'.
As shown in fig. 3, the present invention is a schematic diagram of a three-layer network model of a domain knowledge graph of an intelligent interactive question-answering method based on a markov logic network, wherein the first layer is an entity-relationship network, the second layer is a tuple-relationship network, the third layer is an intention relationship network, and linguistic data in and between layers have corresponding relevance, so that the domain knowledge graph is formed.
Before the step S10 of semantically expanding the entity and relationship representation in the structured tuple by using the domain knowledge graph, the method further comprises the steps of automatic optimization and iterative update, and the method comprises the following steps:
and automatically optimizing and iteratively updating the domain knowledge graph by adopting an online optimization method of a Markov logic network structure and parameters.
In a preferred embodiment of the present invention, the input information is "wonderful and continuous restart when the Hua is Rong Yao 6 is used recently", and the interactive question-answering process of the present invention is described in detail below with reference to the schematic flow chart of the intelligent interactive question-answering method based on the Markov logic network and the input of this embodiment.
And step S10, analyzing the acquired input information, extracting a structural tuple, and performing semantic expansion on the entity and relationship expression in the structural tuple by using a domain knowledge graph.
And step S11, converting the input data into words and sequences of parts of speech thereof by using Chinese word segmentation technology, and extracting the structural tuple by combining the parts of speech and the concept template.
For the input information, the results of word segmentation and part-of-speech tagging are as follows:
< Hua is glorious 6# nz, recently # nt, # v, # n, # i, # u, # d, # v >
Wherein nz represents a proper noun, nt represents a time noun, v represents a verb, n represents a general noun, i represents an idiom, u represents an auxiliary word, and d represents an adverb.
Further analysis in combination with templates can yield the structured tuples and their intentions as follows:
(s: Hua is glory 6, v: power off/on) — phenomenon
(v: use, o: Hua is Rong 6) -Wu
The target tuple (s: gorgeous 6, v: power on/off) is output for further analysis.
Step S12, extending the entity and relationship representation in the structural tuple by using the ontology base in the domain knowledge graph and the obtained public synonym base.
For the target tuple (s: Hua Yan Yang 6, v: power on/off), the mapping relation of the entity 'Hua Yan Yang 6' in the hierarchical concept is obtained based on the ontology base as follows:
huanyanyang 6 mobile phone article
Based on the domain synonym library, it is easy to obtain that the semantics of 'continuous' and 'frequent' are similar, and the semantics of 'switching on and switching off' and 'restarting' are similar and are respectively normalized to be expressed in a knowledge graph, namely 'frequent' and 'restarting'. Thus, the target tuple (s: Huan is glory 6, v: Power off/on) can be normalized and generalized to:
(s: Hua is glory 6, v: frequent/restart) and (s: handset, v: frequent/restart).
And step S20, activating the domain knowledge graph based on the structural tuple and the semantic expansion result to obtain a rule sub-network and an evidence tuple set.
And activating the logic rules in the rule base and the structural tuples in the fact base in the domain knowledge graph based on the structural tuples and the semantic expansion result thereof to obtain a rule sub-network and an evidence tuple set.
As shown in fig. 5, which is a schematic diagram of activation of a markov logic network according to an embodiment of the intelligent interactive question-answering method based on the markov logic network of the present invention, a rule obtained through the activation is listed as follows, but not limited to the following contents:
f 1: 0.7(s: cell phone, v: poisoned) ═ s (s: cell phone, v: frequent/restart)
f 2: 1.2(s: cell phone, v: heat) ═ s (s: cell phone, v: frequent/restart)
f 3: 0.9(s: cell phone/battery, v: loose) ═ s (s: cell phone, v: frequent/restart)
f 4: 1.5(s: CPU, v: overload) (s: mobile phone, v: heat generation)
f 5: 2.0(s: cell phone, v: intoxication) Λ (v: install, o: cell phone/housekeeper) (v: update, o: virus/bank) Λ (v: check and kill, o: virus)
f 7: 1.8(s: CPU, v: overload) ^ > (v: clear, o: background/not used/applied/program)
Screening from the fact base based on the entities "Huaqi Rong 6" and "cell phones" yields the following evidence tuple sets, but not limited to the following list:
e 1: (s: Hua Yingya 6, v: accessories, o: battery)
e 2: (s: Hua is glory 6, v: master, o: CPU)
e 3: (s: Hua Yan 6, v: installation, o: cell phone housekeeper)
And step S30, based on the rule sub-network, assigning values to the evidence tuple set by adopting a mode of approximate reasoning and/or information entry, and calculating the posterior probability of the candidate response information.
Step S31, based on the rule sub-network and the evidence tuple set, sequentially selecting non-assigned evidence tuples, generating a diagnosis question, and executing:
obtaining an evidence tuple assigned by the user based on the confirmation information acquired by the human-computer interaction device; and/or reasoning the evidence tuples which are potentially possible to be established through the MPE to obtain the evidence tuples of the reasoning assignment.
The set of logical clauses of the target tuple associated with its generalization result is S1:{f1,f2,f3Selecting a logic clause with a larger weight and a smaller length, i.e. f2And generating a question according to the evidence tuple of the unknown value related in the logic clause, and then generating a diagnosis question as follows:
"do hua is glory 6 generating heat? "
Two user feedback scenarios are considered for this question:
(one) user feedback as 'no heat'
No matter how the value of other evidence tuples is, if the sentence is always true, the sentence is deleted, and other logic clauses in the set, such as f, are selected in sequence3And generating a diagnostic question: is the battery loose? If the user feedback is not loose, the clause is always true, the clause is deleted, and other logic clauses in the set, such as f, are selected in sequence1Then, through MPE reasoning, the value of the evidence tuple which makes the weight sum of the satisfied clauses the maximum is found, and the possible poisoning of the mobile phone is presumed.
(II) user feedback as "heating"
Through MPE reasoning, speculating that the CPU is possibly overloaded and according to f7And directly providing a corresponding solution and clearing the application program which is not used in the background.
And step S32, calculating the posterior probability of the candidate response information based on the evidence tuple assigned by the user and/or the evidence tuple assigned by inference.
In case (one) of the above example, f will be5And f6Instantiation, based on the resulting Markov logic network, the posterior probabilities of the solutions "(v: update, o: Virus + Bank) < lambda > (v: kill, o: Virus)" and "(v: recovery, o: factory + settings)" being found to be 0.65 and 0.35, respectively, were calculated.
And step S40, selecting a preset number of response messages as final response messages according to the sequence of the posterior probability of the candidate response messages from high to low, and outputting the final response messages.
The posterior probability of the candidate response information in the above example is from high to low as the solution "(v: update, o: virus + reservoir) ^ (v: kill, o: virus)" holds the posterior probability of 0.65, and "(v: recovery, o: factory + set)" holds the posterior probability of 0.35, and the present embodiment selects the highest posterior probability of the candidate response information as the final response information, and then returns the best solution as "kill virus with cell phone manager".
The intelligent interactive question-answering system based on the Markov logic network comprises an input module, a statement analysis module, a semantic activation module, an approximate deduction module and an output module, wherein the input module is used for inputting a question and answer sentence;
the input module is configured to acquire input information;
the statement analysis module is configured to analyze the acquired input information, extract a structural tuple and perform semantic expansion on an entity and a relation expression in the extracted structural tuple by adopting a domain knowledge graph;
the semantic activation module is configured to activate the domain knowledge graph based on the extracted structural module and semantic expansion result to obtain a rule subnetwork and an evidence tuple set;
the approximation deduction module is configured to assign values to the evidence tuple set by adopting an approximation reasoning and/or information input mode based on a rule sub-network, and calculate the posterior probability of the candidate response information;
the output module is configured to select a preset number of response messages as final response messages and output the final response messages according to the sequence of the posterior probability of the candidate response messages from high to low;
the domain knowledge graph is constructed, automatically optimized and iteratively updated by adopting a generalization learning module, wherein the generalization learning module comprises a concept learning module, a relation learning module and a structure learning module;
the concept learning module is configured to learn a hierarchical multi-label classification model of an entity and establish a mapping relation between the entity and a concept;
the relation learning module is configured to discover potential unknown relations between entities and complete a real-estate;
the structure learning module is configured to find semantic paths between facts and learn a Markov logic network structure and parameters under the constraint of intention rules.
As shown in fig. 6, which is a schematic diagram of a system structure of the intelligent interactive question-answering method based on the markov logic network, the domain knowledge graph is constructed and then automatically optimized and iteratively updated by using a generalized learning module, wherein the concept learning module learns a hierarchical multi-label classification model of an entity and establishes a relationship between the entity and a concept; the relation learning module learns the potential unknown relation between the entities and completes the fact base; the structure learning module learns semantic paths between facts, and markov logic network structures and parameters. The user inputs problem information, the system analyzes the sentences and activates the sentences, input information is diagnosed by adopting an approximate deduction and user interaction mode, candidate response information and posterior probability thereof are obtained, and finally a preset number of response information is selected as an analysis result and output.
It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working process and related description of the system described above may refer to the corresponding process in the foregoing method embodiments, and will not be described herein again.
It should be noted that, the intelligent interactive question-answering system based on the markov logic network provided in the foregoing embodiment is only illustrated by the division of the foregoing functional modules, and in practical applications, the above functions may be allocated to different functional modules according to needs, that is, the modules or steps in the embodiment of the present invention are further decomposed or combined, for example, the modules in the foregoing embodiment may be combined into one module, or may be further split into a plurality of sub-modules, so as to complete all or part of the functions described above. The names of the modules and steps involved in the embodiments of the present invention are only for distinguishing the modules or steps, and are not to be construed as unduly limiting the present invention.
A storage device according to a third embodiment of the present invention stores a plurality of programs, which are adapted to be loaded and executed by a processor to implement the above-described intelligent interactive question-answering method based on a markov logic network.
A processing apparatus according to a fourth embodiment of the present invention includes a processor, a storage device; a processor adapted to execute various programs; a storage device adapted to store a plurality of programs; the program is adapted to be loaded and executed by a processor to implement the above-described intelligent interactive question-answering method based on a markov logic network.
It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes and related descriptions of the storage device and the processing device described above may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
Those of skill in the art would appreciate that the various illustrative modules, method steps, and modules described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that programs corresponding to the software modules, method steps may be located in Random Access Memory (RAM), memory, Read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. To clearly illustrate this interchangeability of electronic hardware and software, various illustrative components and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as electronic hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
The terms "comprises," "comprising," or any other similar term are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.
So far, the technical solutions of the present invention have been described in connection with the preferred embodiments shown in the drawings, but it is easily understood by those skilled in the art that the scope of the present invention is obviously not limited to these specific embodiments. Equivalent changes or substitutions of related technical features can be made by those skilled in the art without departing from the principle of the invention, and the technical scheme after the changes or substitutions can fall into the protection scope of the invention.
Claims (10)
1. An intelligent interactive question-answering method based on a Markov logic network is characterized by comprising the following steps:
step S10, analyzing the acquired input information, extracting a structural tuple, and performing semantic expansion on the entity and relationship expression in the structural tuple by using a domain knowledge map;
step S20, activating the domain knowledge graph based on the structural tuple and the semantic expansion result to obtain a rule sub-network and an evidence tuple set;
step S30, based on the rule sub-network, the evidence tuple set is assigned by adopting a mode of approximate reasoning and/or information input, and the posterior probability of the candidate response information is calculated;
step S40, selecting a preset number of response messages as final response messages and outputting the final response messages according to the sequence of the posterior probability of the candidate response messages from high to low;
wherein the domain knowledge graph comprises:
the system comprises a fact library, an ontology library, a rule library, an entity-relationship network corresponding to the superior-inferior relationship between the fact library and the entities in the ontology library, a tuple-relationship network corresponding to the inclusion relationship between the logic rules automatically learned in the rule library and the facts in the ontology library, and an intention-relationship network corresponding to the logic rules manually designed according to the field requirements in the rule library.
2. The intelligent interactive question-answering method based on the Markov logic network as claimed in claim 1, wherein the domain knowledge graph is constructed by the following steps:
step B10, performing word meaning disambiguation on the unstructured corpora in the acquired data according to word labels, part of speech labels and concepts, extracting corpora as structured tuples based on sentence patterns and concept templates, and constructing a fact library;
step B20, establishing a hierarchical structure of the upper and lower relations between the corpus concepts and a hierarchical structure of the inclusion relation between the facts in the fact base, predicting the mapping relation between the entities and the hierarchical concepts based on a hierarchical multi-classification algorithm, and constructing an ontology base; and designing logic rules among intents based on the field requirements, automatically learning a Markov logic network among the structured tuples according to the logic rules, and constructing a rule base.
3. The intelligent interactive question-answering method based on the Markov logic network as claimed in claim 2, wherein in step B20 "design logic rules among intents based on the field requirements, automatically learn the Markov logic network among the structured tuples according to the logic rules, and construct the rule base", the steps are:
step B21, finding a semantic path between the structural tuples by combining random walk and the subgraph pattern, wherein the structural tuples appear in the context in the semantic path, and the semantic elements meet one or more of the same, upper and lower positions, components or attribute relations;
and step B22, setting logic rules among intentions based on the field requirements, converting the semantic paths into logic clauses, performing concept mapping on semantic elements of each structural tuple in the logic clauses, obtaining weights corresponding to the logic clauses with concept variables by maximizing a conditional log-likelihood target, and forming a rule base by the logic clause set with the weights.
4. The intelligent interactive question-answering method based on the Markov logic network as claimed in claim 1, wherein in step S10 "analyze the obtained input information, extract the tuple with the structure, and semantically expand the entity and relationship representation in the tuple by using the domain knowledge graph", the steps are:
step S11, converting the input data into words and sequences of the parts of speech thereof by using Chinese word segmentation technology, and extracting structural tuples by combining the parts of speech and a concept template;
and step S12, expanding the entity and relationship representation in the structural tuple by using an ontology library in the domain knowledge graph and the obtained synonym library.
5. The intelligent interactive question-answering method based on Markov logic network as claimed in claim 1, wherein in step S20 "activate the domain knowledge graph based on the structural tuple and semantic expansion result to obtain rule sub-network and evidence tuple set", the method comprises:
and activating the logic rules in the rule base and the structural tuples in the fact base in the domain knowledge graph based on the structural tuples and the semantic expansion result thereof to obtain a rule sub-network and an evidence tuple set.
6. The intelligent interactive question-answering method based on the Markov logic network as claimed in claim 1, wherein in step S30, based on the rule sub-network, the evidence tuple set is assigned by adopting the approximate reasoning and/or information input mode, and the posterior probability of the candidate answer information is calculated, which comprises the following steps:
step S31, based on the rule sub-network and the evidence tuple set, sequentially selecting non-assigned evidence tuples, generating a diagnosis question, and executing:
obtaining an evidence tuple assigned by the user based on the confirmation information acquired by the human-computer interaction device; and/or reasoning the evidence tuples which are possible to be established through MPE to obtain the evidence tuples of reasoning assignment;
and step S32, calculating the posterior probability of the candidate response information based on the evidence tuple assigned by the user and/or the evidence tuple assigned by inference.
7. The intelligent interactive question-answering method based on the Markov logic network as claimed in claim 1, wherein before "semantic expansion of the entity and relationship representation in the structured tuple by using the domain knowledge graph" in step S10, there are further provided steps of automatic optimization and iterative update, and the method comprises:
and automatically optimizing and iteratively updating the domain knowledge graph by adopting an online optimization method of a Markov logic network structure and parameters.
8. An intelligent interactive question-answering system based on a Markov logic network is characterized by comprising an input module, a statement analysis module, a semantic activation module, an approximate deduction module and an output module;
the input module is configured to acquire input information;
the statement analysis module is configured to analyze the acquired input information, extract a structural tuple and perform semantic expansion on an entity and a relation expression in the extracted structural tuple by adopting a domain knowledge graph;
the semantic activation module is configured to activate the domain knowledge graph based on the extracted structural module and semantic expansion result to obtain a rule subnetwork and an evidence tuple set;
the approximation deduction module is configured to assign values to the evidence tuple set by adopting an approximation reasoning and/or information input mode based on a rule sub-network, and calculate the posterior probability of the candidate response information;
the output module is configured to select a preset number of response messages as final response messages and output the final response messages according to the sequence of the posterior probability of the candidate response messages from high to low;
the domain knowledge graph is constructed, automatically optimized and iteratively updated by adopting a generalization learning module, wherein the generalization learning module comprises a concept learning module, a relation learning module and a structure learning module;
the concept learning module is configured to learn a hierarchical multi-label classification model of an entity and establish a mapping relation between the entity and a concept;
the relation learning module is configured to discover potential unknown relations between entities and complete a real-estate;
the structure learning module is configured to find semantic paths between facts and learn a Markov logic network structure and parameters under the constraint of intention rules.
9. A storage device having stored thereon a plurality of programs, wherein the programs are adapted to be loaded and executed by a processor to implement the intelligent interactive question-answering method based on a markov logic network according to any one of claims 1 to 7.
10. A treatment apparatus comprises
A processor adapted to execute various programs; and
a storage device adapted to store a plurality of programs;
wherein the program is adapted to be loaded and executed by a processor to perform:
an intelligent interactive question-answering method based on a markov logic network according to any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910174742.9A CN109902165B (en) | 2019-03-08 | 2019-03-08 | Intelligent interactive question-answering method, system and device based on Markov logic network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910174742.9A CN109902165B (en) | 2019-03-08 | 2019-03-08 | Intelligent interactive question-answering method, system and device based on Markov logic network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109902165A true CN109902165A (en) | 2019-06-18 |
CN109902165B CN109902165B (en) | 2021-02-23 |
Family
ID=66946918
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910174742.9A Active CN109902165B (en) | 2019-03-08 | 2019-03-08 | Intelligent interactive question-answering method, system and device based on Markov logic network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109902165B (en) |
Cited By (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110457485A (en) * | 2019-07-02 | 2019-11-15 | 厦门美域中央信息科技有限公司 | A kind of body of data logic reasoning based on Inductive Logic Programming |
CN110457455A (en) * | 2019-07-25 | 2019-11-15 | 重庆兆光科技股份有限公司 | A kind of three-valued logic question and answer consulting optimization method, system, medium and equipment |
CN110717052A (en) * | 2019-10-15 | 2020-01-21 | 山东大学 | Environment characterization method in service robot intelligent service |
CN110825862A (en) * | 2019-11-06 | 2020-02-21 | 北京诺道认知医学科技有限公司 | Intelligent question-answering method and device based on pharmacy knowledge graph |
CN110837548A (en) * | 2019-11-05 | 2020-02-25 | 泰康保险集团股份有限公司 | Answer matching method and device, electronic equipment and storage medium |
CN111046191A (en) * | 2019-12-25 | 2020-04-21 | 国网江苏省电力有限公司电力科学研究院 | Electric power field semantic enhancement method and device |
CN111737487A (en) * | 2020-06-10 | 2020-10-02 | 深圳数联天下智能科技有限公司 | Method for assisting body construction, electronic equipment and storage medium |
CN112199959A (en) * | 2020-10-15 | 2021-01-08 | 中国科学院自动化研究所 | Semantic culture robot system |
CN112199478A (en) * | 2020-09-11 | 2021-01-08 | 北京三快在线科技有限公司 | Automatic question answering method, device, electronic equipment and computer readable storage medium |
CN112529184A (en) * | 2021-02-18 | 2021-03-19 | 中国科学院自动化研究所 | Industrial process optimization decision method fusing domain knowledge and multi-source data |
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 |
CN112800236A (en) * | 2021-01-14 | 2021-05-14 | 大连东软教育科技集团有限公司 | Method, device and storage medium for generating learning path based on knowledge graph |
CN113012803A (en) * | 2019-12-19 | 2021-06-22 | 京东方科技集团股份有限公司 | Computer device, system, readable storage medium and medical data analysis method |
CN113064969A (en) * | 2021-04-08 | 2021-07-02 | 易联众信息技术股份有限公司 | Query method, system, medium and device for question-answering system |
CN113535871A (en) * | 2021-06-25 | 2021-10-22 | 杨粤湘 | Vehicle destination prediction method, device, equipment and medium based on travel map |
CN113642986A (en) * | 2021-08-02 | 2021-11-12 | 上海示右智能科技有限公司 | Method for constructing digital notarization |
CN113792152A (en) * | 2021-08-23 | 2021-12-14 | 南京信息工程大学 | Method for fusing triangular graph and knowledge graph |
CN116611813A (en) * | 2023-05-08 | 2023-08-18 | 武汉人云智物科技有限公司 | Intelligent operation and maintenance management method and system based on knowledge graph |
CN116842199A (en) * | 2023-09-01 | 2023-10-03 | 东南大学 | Knowledge graph completion method based on multi-granularity hierarchy and dynamic embedding |
CN117076653A (en) * | 2023-10-17 | 2023-11-17 | 安徽农业大学 | Knowledge base question-answering method based on thinking chain and visual lifting context learning |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105095195A (en) * | 2015-07-03 | 2015-11-25 | 北京京东尚科信息技术有限公司 | Method and system for human-machine questioning and answering based on knowledge graph |
CN105528349A (en) * | 2014-09-29 | 2016-04-27 | 华为技术有限公司 | Method and apparatus for analyzing question based on knowledge base |
US20170103013A1 (en) * | 2015-10-09 | 2017-04-13 | The Board Of Trustees Of The University Of Illinois | System and methods for automatically localizing faults |
CN109062939A (en) * | 2018-06-20 | 2018-12-21 | 广东外语外贸大学 | A kind of intelligence towards Chinese international education leads method |
-
2019
- 2019-03-08 CN CN201910174742.9A patent/CN109902165B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105528349A (en) * | 2014-09-29 | 2016-04-27 | 华为技术有限公司 | Method and apparatus for analyzing question based on knowledge base |
CN105095195A (en) * | 2015-07-03 | 2015-11-25 | 北京京东尚科信息技术有限公司 | Method and system for human-machine questioning and answering based on knowledge graph |
US20170103013A1 (en) * | 2015-10-09 | 2017-04-13 | The Board Of Trustees Of The University Of Illinois | System and methods for automatically localizing faults |
CN109062939A (en) * | 2018-06-20 | 2018-12-21 | 广东外语外贸大学 | A kind of intelligence towards Chinese international education leads method |
Non-Patent Citations (1)
Title |
---|
SUN, ZHENGYA 等: "Scalable learning and inference in Markov logic networks", 《INTERNATIONAL JOURNAL OF APPROXIMATE REASONING》 * |
Cited By (34)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110457485A (en) * | 2019-07-02 | 2019-11-15 | 厦门美域中央信息科技有限公司 | A kind of body of data logic reasoning based on Inductive Logic Programming |
CN110457455A (en) * | 2019-07-25 | 2019-11-15 | 重庆兆光科技股份有限公司 | A kind of three-valued logic question and answer consulting optimization method, system, medium and equipment |
CN110457455B (en) * | 2019-07-25 | 2022-02-22 | 重庆兆光科技股份有限公司 | Ternary logic question-answer consultation optimization method, system, medium and equipment |
CN110717052A (en) * | 2019-10-15 | 2020-01-21 | 山东大学 | Environment characterization method in service robot intelligent service |
CN110837548A (en) * | 2019-11-05 | 2020-02-25 | 泰康保险集团股份有限公司 | Answer matching method and device, electronic equipment and storage medium |
CN110837548B (en) * | 2019-11-05 | 2022-11-11 | 泰康保险集团股份有限公司 | Answer matching method and device, electronic equipment and storage medium |
CN110825862A (en) * | 2019-11-06 | 2020-02-21 | 北京诺道认知医学科技有限公司 | Intelligent question-answering method and device based on pharmacy knowledge graph |
CN110825862B (en) * | 2019-11-06 | 2022-12-06 | 北京诺道认知医学科技有限公司 | Intelligent question and answer method and device based on pharmacy knowledge graph |
CN113012803A (en) * | 2019-12-19 | 2021-06-22 | 京东方科技集团股份有限公司 | Computer device, system, readable storage medium and medical data analysis method |
CN111046191B (en) * | 2019-12-25 | 2022-11-01 | 国网江苏省电力有限公司电力科学研究院 | Semantic enhancement method and device in power field |
CN111046191A (en) * | 2019-12-25 | 2020-04-21 | 国网江苏省电力有限公司电力科学研究院 | Electric power field semantic enhancement method and device |
CN111737487A (en) * | 2020-06-10 | 2020-10-02 | 深圳数联天下智能科技有限公司 | Method for assisting body construction, electronic equipment and storage medium |
CN111737487B (en) * | 2020-06-10 | 2024-08-09 | 深圳数联天下智能科技有限公司 | Method for assisting body construction, electronic equipment and storage medium |
CN112199478A (en) * | 2020-09-11 | 2021-01-08 | 北京三快在线科技有限公司 | Automatic question answering method, device, electronic equipment and computer readable storage medium |
CN112199959B (en) * | 2020-10-15 | 2024-04-12 | 中国科学院自动化研究所 | Semantic culture robot system |
CN112199959A (en) * | 2020-10-15 | 2021-01-08 | 中国科学院自动化研究所 | Semantic culture robot system |
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 |
CN112800236A (en) * | 2021-01-14 | 2021-05-14 | 大连东软教育科技集团有限公司 | Method, device and storage medium for generating learning path based on knowledge graph |
CN112529184A (en) * | 2021-02-18 | 2021-03-19 | 中国科学院自动化研究所 | Industrial process optimization decision method fusing domain knowledge and multi-source data |
US11409270B1 (en) | 2021-02-18 | 2022-08-09 | Institute Of Automation, Chinese Academy Of Sciences | Optimization decision-making method of industrial process fusing domain knowledge and multi-source data |
CN112529184B (en) * | 2021-02-18 | 2021-07-02 | 中国科学院自动化研究所 | Industrial process optimization decision method fusing domain knowledge and multi-source data |
CN113064969A (en) * | 2021-04-08 | 2021-07-02 | 易联众信息技术股份有限公司 | Query method, system, medium and device for question-answering system |
CN113535871A (en) * | 2021-06-25 | 2021-10-22 | 杨粤湘 | Vehicle destination prediction method, device, equipment and medium based on travel map |
CN113535871B (en) * | 2021-06-25 | 2024-02-27 | 杨粤湘 | Travel map-based vehicle destination prediction method, device, equipment and medium |
CN113642986A (en) * | 2021-08-02 | 2021-11-12 | 上海示右智能科技有限公司 | Method for constructing digital notarization |
CN113642986B (en) * | 2021-08-02 | 2024-04-16 | 上海示右智能科技有限公司 | Method for constructing digital notarization |
CN113792152A (en) * | 2021-08-23 | 2021-12-14 | 南京信息工程大学 | Method for fusing triangular graph and knowledge graph |
CN116611813B (en) * | 2023-05-08 | 2024-03-29 | 武汉人云智物科技有限公司 | Intelligent operation and maintenance management method and system based on knowledge graph |
CN116611813A (en) * | 2023-05-08 | 2023-08-18 | 武汉人云智物科技有限公司 | Intelligent operation and maintenance management method and system based on knowledge graph |
CN116842199B (en) * | 2023-09-01 | 2023-12-26 | 东南大学 | Knowledge graph completion method based on multi-granularity hierarchy and dynamic embedding |
CN116842199A (en) * | 2023-09-01 | 2023-10-03 | 东南大学 | Knowledge graph completion method based on multi-granularity hierarchy and dynamic embedding |
CN117076653A (en) * | 2023-10-17 | 2023-11-17 | 安徽农业大学 | Knowledge base question-answering method based on thinking chain and visual lifting context learning |
CN117076653B (en) * | 2023-10-17 | 2024-01-02 | 安徽农业大学 | Knowledge base question-answering method based on thinking chain and visual lifting context learning |
Also Published As
Publication number | Publication date |
---|---|
CN109902165B (en) | 2021-02-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109902165B (en) | Intelligent interactive question-answering method, system and device based on Markov logic network | |
Ma et al. | Online active learning of decision trees with evidential data | |
Li et al. | TDEER: An efficient translating decoding schema for joint extraction of entities and relations | |
Wang et al. | RDF2Rules: Learning rules from RDF knowledge bases by mining frequent predicate cycles | |
Carmona et al. | Overview on evolutionary subgroup discovery: analysis of the suitability and potential of the search performed by evolutionary algorithms | |
CN107357757B (en) | Algebraic application problem automatic solver based on deep reinforcement learning | |
CN111930774B (en) | Automatic construction method and system for electric power knowledge graph body | |
KR20050062624A (en) | Learning/thinking machine and learning/thinking method based on structured knowledge, computer system, and information generation method | |
Jo | NTSO (neural text self organizer): a new neural network for text clustering | |
Schwab et al. | Ant colony algorithm for the unsupervised word sense disambiguation of texts: Comparison and evaluation | |
Wątróbski | Ontology learning methods from text-an extensive knowledge-based approach | |
Djeddi et al. | Ontology alignment using artificial neural network for large-scale ontologies | |
Essayeh et al. | Towards ontology matching based system through terminological, structural and semantic level | |
Gasmi et al. | Cold-start cybersecurity ontology population using information extraction with LSTM | |
Xue et al. | Improving the efficiency of NSGA-II based ontology aligning technology | |
Deng et al. | Research on the construction of event logic knowledge graph of supply chain management | |
Proto et al. | Useful ToPIC: Self-tuning strategies to enhance latent Dirichlet allocation | |
Yu et al. | A structured ontology construction by using data clustering and pattern tree mining | |
Kovács et al. | Conceptualization with incremental bron-kerbosch algorithm in big data architecture | |
Wu et al. | Hierarchical topic tree: A hybrid model comprising network analysis and density peak search | |
Ciravegna et al. | LODIE: Linked Open Data for Web-scale Information Extraction. | |
Jaques et al. | Proof and Trust in the OpenAGRIS implementation | |
Saini et al. | Domobot: An ai-empowered bot for automated and interactive domain modelling | |
Chen | English translation template retrieval based on semantic distance ontology knowledge recognition algorithm | |
Vanya et al. | Towards automated evaluation of explanations in graph neural networks |
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 |