CN112199959A - Semantic culture robot system - Google Patents
Semantic culture robot system Download PDFInfo
- Publication number
- CN112199959A CN112199959A CN202011100729.8A CN202011100729A CN112199959A CN 112199959 A CN112199959 A CN 112199959A CN 202011100729 A CN202011100729 A CN 202011100729A CN 112199959 A CN112199959 A CN 112199959A
- Authority
- CN
- China
- Prior art keywords
- rule
- semantic
- nodes
- node
- module
- 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
- 238000012545 processing Methods 0.000 claims abstract description 52
- 238000000605 extraction Methods 0.000 claims abstract description 30
- 238000005065 mining Methods 0.000 claims abstract description 13
- 238000000034 method Methods 0.000 claims description 52
- 238000004364 calculation method Methods 0.000 claims description 19
- 230000009471 action Effects 0.000 claims description 13
- 238000005457 optimization Methods 0.000 claims description 11
- 238000003058 natural language processing Methods 0.000 claims description 9
- 238000010276 construction Methods 0.000 claims description 7
- 238000000354 decomposition reaction Methods 0.000 claims description 4
- 238000013135 deep learning Methods 0.000 claims description 4
- 230000004927 fusion Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 238000007619 statistical method Methods 0.000 claims description 2
- 238000007781 pre-processing Methods 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 4
- 238000006467 substitution reaction Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 238000013467 fragmentation Methods 0.000 description 1
- 238000006062 fragmentation reaction Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J11/00—Manipulators not otherwise provided for
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/08—Programme-controlled manipulators characterised by modular constructions
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1656—Programme controls characterised by programming, planning systems for manipulators
- B25J9/1661—Programme controls characterised by programming, planning systems for manipulators characterised by task planning, object-oriented languages
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1679—Programme controls characterised by the tasks executed
-
- 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
- G06F16/335—Filtering based on additional data, e.g. user or group profiles
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/237—Lexical tools
- G06F40/247—Thesauruses; Synonyms
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Robotics (AREA)
- Computational Linguistics (AREA)
- Mechanical Engineering (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Machine Translation (AREA)
Abstract
The invention belongs to the field of robot application, particularly relates to a semantic culture robot system, and aims to solve the problems that the existing culture system and equipment are complex to use, insufficient in interpretation capability and incapable of understanding the potential intention of a user. The system comprises: the instruction receiving module is configured to receive user input information and extract corresponding text data; the semantic processing module is configured to perform semantic extraction, semantic processing and knowledge mining on the text data, and take processed semantic information as first information; the rule processing module is configured to take the first information as a node, construct a semantic rule and decompose the semantic rule; the search module is configured to search the decomposed semantic rules in a pre-constructed knowledge graph to obtain a search result; the control module is configured to generate a control instruction according to the search result; a robot configured to execute the control instruction. The invention improves the interpretation capability of the semantic culture robot and simplifies the use complexity.
Description
Technical Field
The invention belongs to the field of robot application, and particularly relates to a semantic culture robot system.
Background
None of the existing cultural systems or equipment has this ability to understand human intent, especially when faced with common cultural user objects such as: the old, children, farmers and the like generally have the conditions of low cultural degree and low knowledge mastering degree. At the moment, due to the deficiency of prior knowledge, various complex culture systems or equipment are difficult to meet the simple requirement, and further development and propagation of culture are hindered. The main reasons for this problem are that the existing cultural systems or equipment are complicated to use, have insufficient interpretability and cannot understand the potential intentions of the user, i.e. the system is difficult to recognize the intentions of "people", so that a huge semantic gap exists between the user and the equipment and the system.
In addition, due to the problems of fragmentation and independence of semantic information, flexible combination cannot be achieved, and therefore complex user requirements cannot be met. Simplification and personification of knowledge acquisition and propagation are effective ways for solving the problem. Therefore, the invention provides a semantic culture robot system which enables a system or equipment to understand the requirements of people in a semantic mode.
Disclosure of Invention
In order to solve the above problems in the prior art, namely, to solve the problems of complicated use, insufficient interpretability and incapability of understanding the potential intention of a user of the existing cultural system and equipment, in a first aspect of the present invention, a semantic cultural robot system is provided, the system comprising: the robot comprises an instruction receiving module, a semantic processing module, a rule processing module, a searching module, a control module and a robot;
the instruction receiving module is configured to receive input information of a user and extract text data corresponding to the input information; after extraction, sending the text data to a semantic processing module;
the semantic processing module is configured to perform semantic extraction, semantic processing and knowledge mining processing on the text data by adopting a natural language processing method, and send processed semantic information serving as first information to the rule processing module;
the rule processing module is configured to take each piece of first information as a node, construct a semantic rule according to a preset semantic structure, and take the semantic rule as a rule node; optimizing each rule node by a preset rule optimization method, judging whether the in-degree or out-degree of each optimized rule node is greater than a set threshold value, and if so, decomposing the corresponding rule node;
the search module is configured to search the decomposed semantic rules in a pre-constructed knowledge graph and send a search result with the minimum time complexity or space complexity to the control module;
the control module is configured to generate a corresponding control instruction according to the received search result and send the control instruction to the robot;
the robot is configured to execute the control instruction.
In some preferred embodiments, the semantic culture robot system further comprises a data acquisition module, a semantic labeling module and a knowledge graph construction module;
the data acquisition module is configured to acquire sample data; the sample data comprises text data, audio data and image data;
the semantic annotation module is configured to perform semantic annotation on the acquired sample data; the semantic annotation method comprises a rule method, a probability statistical method and a deep learning method;
the knowledge graph construction module is configured to extract text data of the labeled sample data, and the text data is processed by the semantic processing module and the rule processing module in sequence to obtain decomposed semantic rules; and constructing a knowledge graph based on the decomposed semantic rules.
In some preferred embodiments, the semantic culture robot system further comprises a knowledge feedback module:
the knowledge feedback module is configured to traverse a pre-constructed knowledge graph, judge whether semantic rules after decomposition of user input information exist, and if not, add the knowledge graph and update.
In some preferred embodiments, the semantic extraction includes entity extraction, relationship extraction, location extraction, and time extraction.
In some preferred embodiments, the semantic processing and knowledge mining processing includes synonym/near-synonym recognition, semantic disambiguation, semantic fusion, semantic mining.
In some preferred embodiments, the rule nodes are classified into non-computation rule nodes and computation rule nodes;
the non-calculation rule nodes comprise rule relation nodes and rule action nodes;
the calculation rule nodes comprise rule selection nodes, rule joint nodes, rule intersection nodes, rule negative calculation nodes, rule connection nodes and rule Cartesian product nodes.
In some preferred embodiments, the rule processing module "decomposes the corresponding rule node", and the method includes:
taking the regular nodes with the in-degree or out-degree larger than a set threshold as large-granularity/medium-granularity regular nodes; taking a regular node with the in-degree or out-degree equal to a set threshold value as a small-granularity regular node;
and decomposing the large/medium granularity rule nodes into small granularity rule nodes.
In some preferred embodiments, the preset rule optimization method includes a rule splitting method, a rule merging method;
the rule splitting method comprises the following steps: if the rule node A is contained by the rule node B, the rule node is decomposed into a rule node A and a rule node C;
the rule merging method comprises the following steps: and if the selection set of the rule nodes is a subset of the selection set of another rule node or the rule nodes and the rule nodes have selection condition coincidence, merging.
In some preferred embodiments, the method of "searching the decomposed rule nodes in the pre-constructed knowledge graph and sending the search result with the minimum time complexity or space complexity to the control module" includes:
step 1: matching the rule relation nodes in the knowledge graph with the decomposed rule nodes, and if the matching is successful, continuing to match all the selected nodes in the corresponding relation node table with the decomposed rule nodes; if the matching is unsuccessful, setting the output flow of the relation table and all rule relations under the relation table as 0, and simultaneously calculating the output flow of the successfully matched rule relation node;
step 2: selecting all rule selection nodes under the rule relationship nodes successfully matched in the step 1 to be matched with the decomposed rule nodes, if the matching is successful, calculating the output flow of each successfully matched rule relationship node, and setting the output flow of all unsuccessfully matched nodes as 0;
and step 3: calculating the out-degree flow of all other rule nodes under the rule selection node successfully matched in the step 2;
step 4, putting the first action nodes under the rule nodes with the output flow rate of more than 0 in the step 3 into a rule execution set; the first action node is a regular action node with the incoming traffic greater than 0;
and 5, taking the execution set as a search result.
The invention has the beneficial effects that:
the invention improves the interpretation capability of the semantic culture robot and simplifies the use complexity. The invention firstly constructs the knowledge graph of multi-granularity semantic culture by performing semantic mapping, entity and relationship extraction and part of speech category analysis on the contents of multi-modal voice, text, image video and other structured, semi-structured and unstructured data.
After the knowledge graph is constructed, natural language processing is carried out in the system according to the input and service requirements of the user, the user requirements are converted into corresponding set rules which are required to be executed by the system and are rich in semantic information, and corresponding semantic calculation can be carried out. According to the constructed semantic rules, with knowledge stored in a pre-constructed knowledge map as assistance, deep information mining and efficient intelligent retrieval are carried out on the intentions of the user, and when the rules are met, corresponding results are returned to be output or corresponding cultural services are provided. The inspiration and activation combination of complex requirements is realized by setting the semantic rules, and finally, the semantic gap is solved efficiently and in high quality.
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.
FIG. 1 is a block diagram of a semantic culture robot system according to an embodiment of the invention;
FIG. 2 is a detailed framework diagram of a semantic culture robot system according to an embodiment of the invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
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.
A semantic culture robot system according to a first embodiment of the present invention includes: the robot comprises an instruction receiving module, a semantic processing module, a rule processing module, a searching module, a control module and a robot;
the instruction receiving module is configured to receive input information of a user and extract text data corresponding to the input information; after extraction, sending the text data to a semantic processing module;
the semantic processing module is configured to perform semantic extraction, semantic processing and knowledge mining processing on the text data by adopting a natural language processing method, and send processed semantic information serving as first information to the rule processing module;
the rule processing module is configured to take each piece of first information as a node, construct a semantic rule according to a preset semantic structure, and take the semantic rule as a rule node; optimizing each rule node by a preset rule optimization method, judging whether the in-degree or out-degree of each optimized rule node is greater than a set threshold value, and if so, decomposing the corresponding rule node;
the search module is configured to search the decomposed semantic rules in a pre-constructed knowledge graph and send a search result with the minimum time complexity or space complexity to the control module;
the control module is configured to generate a corresponding control instruction according to the received search result and send the control instruction to the robot;
the robot is configured to execute the control instruction.
In order to more clearly explain the semantic culture robot system of the present invention, the following describes each module in various embodiments of the system in detail with reference to the attached drawings.
Fig. 1 is a schematic diagram of a semantic culture robot system according to an embodiment of the present disclosure, where the semantic culture robot system includes a data collection module 100, a semantic annotation module 200, a semantic processing module 300, a rule processing module 400, a knowledge graph construction module 500, an instruction receiving module 600, a search module 700, a control module 800, a robot 900, and a knowledge feedback module 1000. Each module is described in detail below.
A data acquisition module 100 configured to acquire sample data; the sample data includes text data, audio data, and image data.
In this embodiment, the application scenario of the semantic culture robot system of the present invention is mainly in a public culture area, and therefore, the collected sample data is multi-modal (multi-source) sample data containing culture information, and the sample data includes text data, audio data, and image data.
The semantic annotation module 200 is configured to perform semantic annotation on the acquired sample data; the semantic annotation method comprises a rule method, a probability statistics method and a deep learning method.
In the embodiment, semantic annotation is performed on the acquired multi-modal sample data based on a probability statistics method, a rule method and a deep learning method.
The semantic processing module 300 is configured to perform semantic extraction, semantic processing, and knowledge mining processing on the text data by using a natural language processing method, and send processed semantic information to the rule processing module as first information.
In this embodiment, text data corresponding to the labeled sample data is extracted, and semantic extraction is performed on the text data, where the semantic extraction includes entity extraction, relationship extraction, location extraction, and time extraction. Extracting multi-granularity semantic information of the sample data, and realizing semantic mapping representation of the data. As shown in the semantic extraction module of fig. 2.
And performing semantic processing and knowledge mining after semantic extraction, namely processing the information after semantic extraction through synonymy/near-synonym identification, semantic disambiguation, semantic fusion, part of speech analysis, semantic mining and the like, taking the processed semantic information as first information, finally realizing the conversion from sample data of multi-source culture to fine-grained semantic information, and sending the processed semantic information to a rule optimization module.
The rule processing module 400 is configured to take each piece of first information as a node, construct a semantic rule according to a preset semantic structure, and take the semantic rule as a rule node; and optimizing each rule node by a preset rule optimization method, judging whether the in-degree or out-degree of each optimized rule node is greater than a set threshold value, and if so, decomposing the corresponding rule node.
In the embodiment, semantic rule formulation, semantic rule parallelization processing and the like are included, so that efficient integrated combination of semantic information is realized, and query of semantic engines such as knowledge maps is met. The method comprises the following specific steps:
semantic rules are constructed based on a preset semantic structure (preferably, a semantic structure form of a principal and a predicate object and an active complement is adopted in the invention) by taking the semantic information after sample data processing as a node, namely the first information as the node, and a semantic rule network is formed. And judging whether the in-degree or out-degree of the constructed semantic rule is larger than a set threshold value, and if so, decomposing the corresponding semantic. The threshold value set by the in-degree or out-degree of the semantic rules is set to be 1, the semantic rules with the in-degree or out-degree larger than 1 are defined as large-granularity/medium-granularity semantic rules, and the semantic rules with the in-degree or out-degree smaller than 1 are defined as small-granularity semantic rules. And decomposing the large/medium granularity semantic rules into small granularity semantic rules. And optimizing the decomposed semantic rules.
Wherein, the degree of the rule node is: the number of edges in the rule network associated with a rule node is called the degree of the rule node;
and (3) regular node in-degree: the number of directed edges in the rule network with the rule nodes as tails is called the degree of entry of the rule nodes;
and (3) regular node out degree: the number of directed edges with the rule nodes as heads in the rule network is called the out degree of the rule nodes;
and (3) regular node incoming flow: the number of records of directed edges taking the rule nodes as tails in the rule network flows into the rule network;
and (3) regular node outgoing flow: the number of outgoing records of the directed edges taking the rule nodes as heads in the rule network is recorded; a rule node may have one or more inbound or outbound flows.
The rule nodes comprise non-calculation rule nodes and calculation rule nodes; the non-calculation rule nodes comprise rule relation nodes and rule action nodes; the calculation rule nodes comprise rule selection nodes, rule joint nodes, rule intersection nodes, rule negative calculation nodes, rule connection nodes and rule Cartesian product nodes.
The out-of-order flow of the regular relation nodes is the total number of the recording pieces of the regular relation nodes;
the out-degree flow c of the rule selection node is equal to the in-degree flow a of the rule selection node multiplied by a selectable parameter upsilon, as shown in formula (1):
c＝a*υ，υ∈[0,1] (1)
the out-of-measure flow of the rule joint (Union) node is equal to the in-measure flow (a, b) of the rule joint node multiplied by the combinable operation parameter, as shown in formula (2):
c＝(a+b)*υ，
the outgoing flow c of the rule Intersection (Intersection) node is equal to the product of the incoming flow a and b of the rule Intersection node, and then multiplied by an Intersection operation parameter upsilon, as shown in formula (3):
c＝(a*b)*υ，
the outgoing flow c of the regular connection (Join) node is equal to the product of the incoming flows a and b of the regular connection node, and then multiplied by a connectable (Join) operation parameter upsilon, as shown in formula (4):
that is, c is (a is b) is upsilon, and upsilon is e [0,1] (4)
The regular negative (Denial) node out-degree traffic is equal to the regular negative (Denial) node in-degree traffic multiplied by the negateable (Denial) selection parameter, as shown in equation (5):
c＝a*υ，υ∈[0,1] (5)
the regular Cartesian product node out-degree flow is equal to the product of the regular Cartesian product node in-degree flow multiplied by a Cartesian product parameter upsilon, as shown in formula (6):
c＝a*b*υ，υ＝1 (6)
according to the flow of each rule node, the calculation method of the calculation cost of the rule selection node, the rule joint node, the rule intersection node, the rule negative calculation node and the rule Cartesian product node in the operations of inquiry, matching, comparison and the like comprises the following steps:
the Cost (rule joint node, rule intersection node, rule negative calculation node and rule cartesian product node) is Cost (traversal) + Cost (comparison and selection judgment);
the computational cost of the regular connection node is: the Cost (regular connection node) is Cost (traversal) + Cost (comparison selection determination) + Cost (connection operation).
Specifically, as shown in formulas (7), (8), (9), (10), (11) and (12):
wherein, a1，a2，……，akRepresenting in-degree flow, c representing out-degree flow, upsilon representing alternative/combinability/availability of rule nodes,
represents the time cost of executing one traversal operation by the regular node, wherein omega isThe cost of time consumed for carrying out one-time rule selection judgment/joint judgment/intersection judgment/connection judgment, lambda is the cost of time consumed for carrying out one-time specific rule connection operation, mu is the cost of time consumed for carrying out one-time rule Cartesian product operation, and cost (S), (U), (I), (J), (D) and (D) respectively represent the calculation costs of a rule selection node, a rule joint node, a rule intersection node, a rule connection node and a rule Cartesian product node.
Because massive rules are stored in the rule network, the calculation amount is large, if the rules are not optimized, huge loads are brought to a processor, and the problem that how to effectively optimize the massive rule network is urgently needed to be solved is solved. At present, a plurality of rule optimization methods such as optimization methods based on state empty shelf diagrams, EQL, rule difference and the like exist, but the method is not suitable for massive rule characteristics. The invention provides a rule optimization method for rule merging and equivalent replacement aiming at the characteristics of massive rules.
The rule merging method comprises the following steps: that is, the selection set of one rule node is a subset of another selection set or one rule node is overlapped with other rule nodes in the selection condition part, then merging is carried out. And when the partial superposition is carried out for combination, the regular nodes with small node flow are selected for combination and modification.
The rule splitting method comprises the following steps: if the semantic node of the rule A is contained by the rule B, the rule B can be decomposed into the rule A and the rule C, and the rule B is deleted;
rule equivalence substitution method: if sigma appears in the rule netθ1(σθ2(N)) and σθ1∧θ2(N)、σθ(N1. U.N 2) and σθ(N1)∪σθ(N2)、σθ(N1N 2) and σθ(N1)∩σθ(N2)、σθ1(N1∞θN2) and σθ1(N1)∞θσθ1(N2)、σθ1(σθ2(C) And σθ2(σθ1(C))、C1∞θC2 and C2 ∞θC1、(C1∞θ1C2)∞θ2C3 and C1 ∞θ1(C2∞θ2C3) C1 ℃ ^ C2 and C2 ℃ ^ C1, C1 ^ C2 and C2 ^ C1, (C1 ^ C2) ^ C3 and C1 ^ CThe regular structures of (C2 ^ C3), (C1 ^ C2) ^ C3 and C1 ^ (C2 ^ C3) can be replaced with each other. See in detail the literature "Abraham Silbersehitz, Henry F.K orth, S.Sudarshan.Database System ConseePs (Fifth Edition),2006.9.PP: P378-P383"
When the sample data is matched with the rule network, the matching algorithm is as follows:
and matching the rule relation nodes in the rule network with the facts in the sample data (namely words processed by the natural language of the sample data), and if the matching is successful, continuously matching all the selected nodes in the corresponding relation node table with all the facts in the sample data. And if the matching is unsuccessful, setting the output flow rates of the relation table and all nodes under the relation table to be 0. Meanwhile, calculating the outbound flow of the successfully matched relationship nodes in the rule network;
and selecting all rule selection nodes under the successfully matched rule relation nodes to be matched with all facts in the sample data, and if the matching is successful, calculating the out-degree flow of each successfully matched rule node. Setting the outgoing flow of all unsuccessfully matched nodes as 0;
selecting outgoing traffic of all other rule nodes (all nodes except the rule relation node and the rule selection node) under the rule selection node which is successfully matched;
and finally, all rules (all conditions in the rules are met) with the incoming flow rates of all rule action nodes being greater than 0 under other rule nodes with the outgoing flow rate being greater than 0 are put into a rule execution set to perform rule triggering.
And judging whether the in-degree or out-degree of each optimized rule node is greater than a set threshold, and if so, decomposing the corresponding rule node. I.e. the large/medium granularity rule nodes are decomposed into small granularity rule nodes.
A knowledge graph construction module 500 configured to construct a knowledge graph based on the decomposed semantic rules.
In this embodiment, the units of the constructed knowledge graph are the same as the units of the rule net, so the knowledge graph is constructed directly based on the decomposed semantic rules.
An instruction receiving module 600 configured to receive input information of a user and extract text data corresponding to the input information; and after extraction, sending the text data to a semantic processing module.
In this embodiment, when the user object makes a corresponding request, the input information of the user is received and the text data corresponding to the input information is extracted, where the input mode includes voice input, text input, and image input. The extracted text data is sent to the semantic processing module 300 and the rule processing module 400, that is, the text data is subjected to semantic extraction, semantic processing and knowledge mining processing by adopting a natural language processing method, the processed semantic information is used as first information, the processed first information is used as a node framework semantic rule, and optimization decomposition is performed.
The searching module 700 is configured to search the decomposed rule nodes in a pre-constructed knowledge graph, and send a search result with the minimum time complexity or space complexity to the control module.
In this embodiment, the knowledge stored in the knowledge graph constructed in the knowledge graph construction module 500 is combined to perform intelligent retrieval and query, so as to realize deep understanding of the user's intention. And (3) formulating semantic rules by combining knowledge graph query results and semantic elements to meet the requirement of a machine for understanding people. For example, the user enters "where is the river map on Qingming? Lead me past? "clear riverside scene, where, collar, past" is output after natural language processing, and semantic rules "time-current, constraint-clear riverside scene, demand-position guidance" are constructed according to the content of natural language processing. And optimizing and decomposing after establishing the rule, directly performing matching retrieval in the knowledge graph after decomposing, and obtaining the position of the Qingming Shang river map, namely a search result, and sending the search result to the robot.
The method of performing matching retrieval in the knowledge-graph is consistent with the matching algorithm in the rule processing module 400. Namely, after natural language processing is carried out on input data input by a user, the input data are matched with rule nodes in a knowledge graph. The method comprises the following specific steps:
step 1: matching the rule relation nodes in the knowledge graph with the decomposed rule nodes, and if the matching is successful, continuing to match all the selected nodes in the corresponding relation node table with the decomposed rule nodes; if the matching is unsuccessful, setting the output flow of the relation table and all rule relations under the relation table as 0, and simultaneously calculating the output flow of the successfully matched rule relation node;
step 2: selecting all rule selection nodes under the rule relationship nodes successfully matched in the step 1 to be matched with the decomposed rule nodes, if the matching is successful, calculating the output flow of each successfully matched rule relationship node, and setting the output flow of all unsuccessfully matched nodes as 0;
and step 3: calculating the out-degree flow of all other rule nodes under the rule selection node successfully matched in the step 2;
step 4, putting the first action nodes under the rule nodes with the output flow rate of more than 0 in the step 3 into a rule execution set; the first action node is a regular action node with the incoming traffic greater than 0;
step 5, taking the execution set as a search result;
and 6, if a plurality of search results are obtained, selecting the search result with the minimum time complexity or space complexity (namely the minimum calculation cost) and sending the search result to the control module.
And the control module 800 is configured to generate a corresponding control instruction according to the received search result, and send the control instruction to the robot.
In this embodiment, the control system of the robot generates a corresponding control instruction according to the received search result. In the invention, the service type of the current user is judged according to the search result, and the service type comprises intelligent question answering, navigation and image display.
A robot 900 configured to execute the control instruction.
In this embodiment, the robot executes the received control instruction. If the service type is navigation, a browsing route is automatically generated, and the robot guides the user to visit a clear riverside scene. And if the question is an intelligent question and answer or the image is displayed, replying the question of the user or displaying the image which the user wants to watch.
In addition, the semantic culture robot of the invention also comprises a knowledge feedback module 1000;
the knowledge feedback module 1000 is configured to traverse a pre-constructed knowledge graph, determine whether a semantic rule after decomposition of user input information exists, and if not, add the knowledge graph and update.
It should be noted that, the semantic culture robot system provided in the foregoing embodiment is only illustrated by the division of the functional modules, and in practical applications, the above functions may be distributed by 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 multiple 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.
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 "first," "second," and the like are used for distinguishing between similar elements and not necessarily for describing or implying a particular order or sequence.
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 (9)
1. A semantic culture robot system, comprising: the robot comprises an instruction receiving module, a semantic processing module, a rule processing module, a searching module, a control module and a robot;
the instruction receiving module is configured to receive input information of a user and extract text data corresponding to the input information; after extraction, sending the text data to a semantic processing module;
the semantic processing module is configured to perform preprocessing of semantic extraction, semantic processing and knowledge mining on the text data by adopting a natural language processing method, take preprocessed semantic information as first information, and send the first information to the rule processing module;
the rule processing module is configured to take each piece of first information as a node, construct a semantic rule according to a preset semantic structure, and take the semantic rule as a rule node; optimizing each rule node by a preset rule optimization method, judging whether the in-degree or out-degree of each optimized rule node is greater than a set threshold value, and if so, decomposing the corresponding rule node;
the search module is configured to search the decomposed rule nodes in a pre-constructed knowledge graph and send a search result with the minimum time complexity or space complexity to the control module;
the control module is configured to generate a corresponding control instruction according to the received search result and send the control instruction to the robot;
the robot is configured to execute the control instruction.
2. The semantic culture robot system according to claim 1, further comprising a data acquisition module, a semantic annotation module, and a knowledge graph construction module;
the data acquisition module is configured to acquire sample data; the sample data comprises text data, audio data and image data;
the semantic annotation module is configured to perform semantic annotation on the acquired sample data; the semantic annotation method comprises a rule method, a probability statistical method and a deep learning method;
the knowledge graph construction module is configured to extract text data of the labeled sample data, and obtain decomposed semantic rules sequentially through the semantic processing module and the rule processing module; and constructing a knowledge graph based on the decomposed semantic rules.
3. The semantic cultural robot system of claim 2, further comprising a knowledge feedback module;
the knowledge feedback module is configured to traverse a pre-constructed knowledge graph, judge whether semantic rules after decomposition of user input information exist, and if not, add the knowledge graph and update.
4. The semantic culture robot system of claim 2, wherein the semantic extraction comprises entity extraction, relationship extraction, location extraction, time extraction.
5. The semantic culture robot system of claim 2, wherein the semantic processing and knowledge mining processes comprise synonym/near-synonym recognition, semantic disambiguation, semantic fusion, semantic mining.
6. The semantic culture robot system of claim 2, wherein the rule nodes are divided into non-computation rule nodes, computation rule nodes;
the non-calculation rule nodes comprise rule relation nodes and rule action nodes;
the calculation rule nodes comprise rule selection nodes, rule joint nodes, rule intersection nodes, rule negative calculation nodes, rule connection nodes and rule Cartesian product nodes.
7. The semantic culture robot system of claim 2, wherein the rule processing module decomposes the corresponding rule node by:
taking the regular nodes with the in-degree or out-degree larger than a set threshold as large-granularity/medium-granularity regular nodes; taking a regular node with the in-degree or out-degree equal to a set threshold value as a small-granularity regular node;
and decomposing the large/medium granularity rule nodes into small granularity rule nodes.
8. The semantic culture robot system of claim 2, wherein the preset rule optimization method comprises a rule splitting method, a rule merging method;
the rule splitting method comprises the following steps: if the rule node A is contained by the rule node B, the rule node is decomposed into a rule node A and a rule node C;
the rule merging method comprises the following steps: and if the selection set of one rule node is a subset of the selection set of another rule node or the rule nodes and the rule nodes have selection condition coincidence, merging.
9. The semantic culture robot system of claim 2, wherein the method of searching the decomposed rule nodes in the pre-constructed knowledge graph comprises:
step 1: matching the rule relation nodes in the knowledge graph with the decomposed rule nodes, and if the matching is successful, continuing to match all the selected nodes in the corresponding relation node table with the decomposed rule nodes; if the matching is unsuccessful, setting the output flow of the relation table and all rule relations under the relation table as 0, and simultaneously calculating the output flow of the successfully matched rule relation node;
step 2: selecting all rule selection nodes under the rule relationship nodes successfully matched in the step 1 to be matched with the decomposed rule nodes, if the matching is successful, calculating the output flow of each successfully matched rule relationship node, and setting the output flow of all unsuccessfully matched nodes as 0;
and step 3: calculating the out-degree flow of all other rule nodes under the rule selection node successfully matched in the step 2;
step 4, adding the first action node under the rule node with the output flow rate of more than 0 in the step 3 into a rule execution set; the first action node is a regular action node with the incoming traffic greater than 0;
and 5, outputting the execution set as a search result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011100729.8A CN112199959B (en) | 2020-10-15 | 2020-10-15 | Semantic culture robot system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011100729.8A CN112199959B (en) | 2020-10-15 | 2020-10-15 | Semantic culture robot system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112199959A true CN112199959A (en) | 2021-01-08 |
| CN112199959B CN112199959B (en) | 2024-04-12 |
Family
ID=74010201
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011100729.8A Active CN112199959B (en) | 2020-10-15 | 2020-10-15 | Semantic culture robot system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112199959B (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100241639A1 (en) * | 2009-03-20 | 2010-09-23 | Yahoo! Inc. | Apparatus and methods for concept-centric information extraction |
| US20110119047A1 (en) * | 2009-11-19 | 2011-05-19 | Tatu Ylonen Oy Ltd | Joint disambiguation of the meaning of a natural language expression |
| CN104750795A (en) * | 2015-03-12 | 2015-07-01 | 北京云知声信息技术有限公司 | Intelligent semantic searching system and method |
| CN106780067A (en) * | 2016-12-14 | 2017-05-31 | 南京工业职业技术学院 | It is a kind of to consider the sign prediction method that node locally marks characteristic |
| US20180060459A1 (en) * | 2016-09-01 | 2018-03-01 | Energid Technologies Corporation | System and method for game theory-based design of robotic systems |
| CN108170734A (en) * | 2017-12-15 | 2018-06-15 | 国网冀北电力有限公司信息通信分公司 | A kind of intelligence O＆M robot |
| CN108334491A (en) * | 2017-09-08 | 2018-07-27 | 腾讯科技（深圳）有限公司 | Text analyzing method, apparatus, computing device and storage medium |
| CN109902165A (en) * | 2019-03-08 | 2019-06-18 | 中国科学院自动化研究所 | Intelligent interactive answering method, system, device based on Markov Logic Networks |
| CN109977228A (en) * | 2019-03-21 | 2019-07-05 | 浙江大学 | The information identification method of grid equipment defect text |
| CN110781313A (en) * | 2019-09-29 | 2020-02-11 | 北京淇瑀信息科技有限公司 | Graph storage optimization method and device and electronic equipment |
| US20200184376A1 (en) * | 2018-12-05 | 2020-06-11 | The Board Of Trustees Of The University Of Illinois | Holistic Optimization for Accelerating Iterative Machine Learning |
-
2020
- 2020-10-15 CN CN202011100729.8A patent/CN112199959B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100241639A1 (en) * | 2009-03-20 | 2010-09-23 | Yahoo! Inc. | Apparatus and methods for concept-centric information extraction |
| US20110119047A1 (en) * | 2009-11-19 | 2011-05-19 | Tatu Ylonen Oy Ltd | Joint disambiguation of the meaning of a natural language expression |
| CN104750795A (en) * | 2015-03-12 | 2015-07-01 | 北京云知声信息技术有限公司 | Intelligent semantic searching system and method |
| US20180060459A1 (en) * | 2016-09-01 | 2018-03-01 | Energid Technologies Corporation | System and method for game theory-based design of robotic systems |
| CN106780067A (en) * | 2016-12-14 | 2017-05-31 | 南京工业职业技术学院 | It is a kind of to consider the sign prediction method that node locally marks characteristic |
| CN108334491A (en) * | 2017-09-08 | 2018-07-27 | 腾讯科技（深圳）有限公司 | Text analyzing method, apparatus, computing device and storage medium |
| CN108170734A (en) * | 2017-12-15 | 2018-06-15 | 国网冀北电力有限公司信息通信分公司 | A kind of intelligence O＆M robot |
| US20200184376A1 (en) * | 2018-12-05 | 2020-06-11 | The Board Of Trustees Of The University Of Illinois | Holistic Optimization for Accelerating Iterative Machine Learning |
| CN109902165A (en) * | 2019-03-08 | 2019-06-18 | 中国科学院自动化研究所 | Intelligent interactive answering method, system, device based on Markov Logic Networks |
| CN109977228A (en) * | 2019-03-21 | 2019-07-05 | 浙江大学 | The information identification method of grid equipment defect text |
| CN110781313A (en) * | 2019-09-29 | 2020-02-11 | 北京淇瑀信息科技有限公司 | Graph storage optimization method and device and electronic equipment |
Non-Patent Citations (5)
| Title |
|---|
| ANJUM MIRZA ET AL: "Constructing Knowledge Graph by Extracting Correlations from Wikipedia Corpus for Optimizing Web Information Retrieval", 《2018 9TH INTERNATIONAL CONFERENCE ON COMPUTING, COMMUNICATION AND NETWORKING TECHNOLOGIES (ICCCNT)》, 18 October 2018 (2018-10-18), pages 1 - 7 * |
| 孙斌: "基于知识图谱的社交网络话题演化及预测", 《中国优秀硕士学位论文全文数据库基础科学辑》, no. 1, 15 January 2020 (2020-01-15), pages 002 - 316 * |
| 杨颐 等: "基于云计算的汉字文化数字化平台的架构研究", 《计算机科学》, vol. 43, no. 7, 15 July 2016 (2016-07-15), pages 28 - 34 * |
| 毛晶晶: "基于可信度向量和文本信息的知识图谱表示学习", 《中国优秀硕士学位论文全文数据库信息科技辑》, no. 7, 15 July 2019 (2019-07-15), pages 138 - 1460 * |
| 黄璨: "面向医疗语义理解的结构化处理方法的研究与实现", 《中国优秀硕士学位论文全文数据库医药卫生科技辑》, no. 2, 15 February 2020 (2020-02-15), pages 080 - 78 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112199959B (en) | 2024-04-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9760642B2 (en) | System and method of reduction of irrelevant information during search | |
| US9262438B2 (en) | Geotagging unstructured text | |
| US9477747B2 (en) | Method and apparatus for acquiring hot topics | |
| Jung | Online named entity recognition method for microtexts in social networking services: A case study of twitter | |
| EP3096246A1 (en) | Method, system and storage medium for realizing intelligent answering of questions | |
| US20110231353A1 (en) | Artificial intelligence application in human machine interface for advanced information processing and task managing | |
| US8972498B2 (en) | Mobile-based realtime location-sensitive social event engine | |
| CN110110038B (en) | Telephone traffic prediction method, device, server and storage medium | |
| CN105518644A (en) | Method for processing and displaying real-time social data on map | |
| CN112507139B (en) | Knowledge graph-based question and answer method, system, equipment and storage medium | |
| CN111881688B (en) | Event causal relationship identification method, system and device based on shielding generalization mechanism | |
| US20200097759A1 (en) | Table Header Detection Using Global Machine Learning Features from Orthogonal Rows and Columns | |
| US20150324455A1 (en) | Method and apparatus for natural language search for variables | |
| KR20190103504A (en) | Continuous Conversation Method and Its System by Automating Conversation Scenario Collection | |
| CN117851610A (en) | Knowledge graph construction method and device based on pre-training large language model | |
| Wachowicz et al. | Developing a streaming data processing workflow for querying space–time activities from geotagged tweets | |
| CN112966076A (en) | Intelligent question and answer generating method and device, computer equipment and storage medium | |
| CN113961678A (en) | Intelligent dialogue method and device based on affair map | |
| CN113886545A (en) | Knowledge question answering method, knowledge question answering device, computer readable medium and electronic equipment | |
| JP5639549B2 (en) | Information retrieval apparatus, method, and program | |
| CN112199959A (en) | Semantic culture robot system | |
| Ediger et al. | Real-time streaming intelligence: Integrating graph and nlp analytics | |
| CN112416922B (en) | Group association data mining method, device, equipment and storage medium | |
| Tolmachova et al. | Visualizing search history in web learning | |
| CN113609268A (en) | Intelligent psychological knowledge question-answering method and device based on knowledge graph |
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 |