WO2023134057A1 - 事务信息查询方法、装置、计算机设备及存储介质 - Google Patents

事务信息查询方法、装置、计算机设备及存储介质 Download PDF

Info

Publication number
WO2023134057A1
WO2023134057A1 PCT/CN2022/089316 CN2022089316W WO2023134057A1 WO 2023134057 A1 WO2023134057 A1 WO 2023134057A1 CN 2022089316 W CN2022089316 W CN 2022089316W WO 2023134057 A1 WO2023134057 A1 WO 2023134057A1
Authority
WO
WIPO (PCT)
Prior art keywords
transaction
ontology
text
transaction information
query
Prior art date
Application number
PCT/CN2022/089316
Other languages
English (en)
French (fr)
Inventor
舒畅
陈又新
Original Assignee
平安科技(深圳)有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 平安科技(深圳)有限公司 filed Critical 平安科技(深圳)有限公司
Publication of WO2023134057A1 publication Critical patent/WO2023134057A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33Querying
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36Creation of semantic tools, e.g. ontology or thesauri
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00Handling natural language data
    • G06F40/20Natural language analysis
    • G06F40/279Recognition of textual entities
    • G06F40/289Phrasal analysis, e.g. finite state techniques or chunking
    • G06F40/295Named entity recognition
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00Handling natural language data
    • G06F40/30Semantic analysis
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods

Definitions

  • the present application relates to the technical field of artificial intelligence, and in particular to a transaction information query method, device, computer equipment and storage medium.
  • the purpose of the embodiments of the present application is to provide a transaction information query method, device, computer equipment, and storage medium to solve the problem of low transaction information acquisition efficiency.
  • the embodiment of the present application provides a transaction information query method, which adopts the following technical solution:
  • the embodiment of the present application also provides a transaction information query device, which adopts the following technical solutions:
  • a text acquisition module configured to acquire transaction information text
  • a text identification module configured to identify the events in the transaction information text and the relationship between each event, and identify the ontology in the event and the ontology relationship between each ontology;
  • a framework building module configured to construct a Semantic Web framework according to the event, the event relationship, the ontology and the ontology relationship;
  • a map generation module configured to generate a transaction knowledge map corresponding to the transaction information text based on the semantic web framework
  • a query acquisition module configured to acquire transaction query text
  • a transaction query module configured to calculate the semantic similarity between the transaction query text and each node in the transaction knowledge graph, and determine the transaction information query result according to the semantic similarity and the transaction knowledge graph.
  • an embodiment of the present application further provides a computer device, including a memory and a processor, wherein computer-readable instructions are stored in the memory, and the processor implements the following steps when executing the computer-readable instructions:
  • the embodiment of the present application also provides a computer-readable storage medium, the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions are executed by a processor to implement the following steps:
  • the embodiment of the present application mainly has the following beneficial effects: the transaction query result can be accurately determined from the transaction knowledge map, the required transaction information can be automatically matched, and the transaction information acquisition efficiency is improved.
  • FIG. 1 is an exemplary system architecture diagram to which the present application can be applied;
  • Fig. 2 is a flowchart of an embodiment of the transaction information query method according to the present application.
  • FIG. 3 is a schematic structural diagram of an embodiment of a transaction information query device according to the present application.
  • Fig. 4 is a schematic structural diagram of an embodiment of a computer device according to the present application.
  • a system architecture 100 may include terminal devices 101 , 102 , 103 , a network 104 and a server 105 .
  • the network 104 is used as a medium for providing communication links between the terminal devices 101 , 102 , 103 and the server 105 .
  • Network 104 may include various connection types, such as wires, wireless communication links, or fiber optic cables, among others.
  • Terminal devices 101 , 102 , 103 Users can use terminal devices 101 , 102 , 103 to interact with server 105 via network 104 to receive or send messages and the like.
  • Various communication client applications can be installed on the terminal devices 101, 102, and 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, and the like.
  • Terminal devices 101, 102, 103 can be various electronic devices with display screens and support web browsing, including but not limited to smartphones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III, dynamic Video experts compress standard audio layer 3), MP4 (Moving Picture Experts Group Audio Layer IV, moving picture experts compress standard audio layer 4) player, laptop portable computer and desktop computer, etc.
  • MP3 players Moving Picture Experts Group Audio Layer III, dynamic Video experts compress standard audio layer 3
  • MP4 Moving Picture Experts Group Audio Layer IV, moving picture experts compress standard audio layer 4
  • laptop portable computer and desktop computer etc.
  • the server 105 may be a server that provides various services, such as a background server that provides support for pages displayed on the terminal devices 101 , 102 , 103 .
  • the transaction information query method provided in the embodiment of the present application is generally executed by a server, and correspondingly, the transaction information query device is generally set in the server.
  • terminal devices, networks and servers in Fig. 1 are only illustrative. According to the implementation needs, there can be any number of terminal devices, networks and servers.
  • FIG. 2 shows a flow chart of an embodiment of the transaction information query method according to the present application.
  • the transaction information query method includes the following steps:
  • Step S201 acquire transaction information text.
  • the electronic device on which the transaction information query method runs can communicate with the terminal through a wired connection or a wireless connection.
  • the above-mentioned wireless connection methods may include but not limited to 3G/4G/5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connections known or developed in the future. connection method.
  • the server first needs to obtain the transaction information text.
  • the transaction information text records transaction flow information.
  • the transaction information text can be a government official document, which records the established business process. At least one transaction process is recorded in the transaction information text.
  • the above-mentioned transaction information text can also be stored in a node of a block chain.
  • Blockchain essentially a decentralized database, is a series of data blocks associated with each other using cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify its Validity of information (anti-counterfeiting) and generation of the next block.
  • the blockchain can include the underlying platform of the blockchain, the platform product service layer, and the application service layer.
  • Step S202 identifying the events in the transaction information text and the relationship between each event, and identifying the ontology in the event and the ontology relationship between each ontology.
  • natural language processing is performed on the transactional information text to identify events and affairs relations recorded in the transactional information text.
  • an event is a whole formed by the process of achieving a goal, and an affair relationship refers to the relationship between different events.
  • "how fresh graduates settle in city S” and "how fresh graduates apply for housing subsidies in city S” are two events, and the two events can be conditional relationships.
  • the prerequisite for "fresh graduates applying for housing subsidies in city S” is "fresh graduates Settled in City S".
  • the ontology is the entity involved in the event.
  • the possible ontology includes "fresh graduates, household registration, human resources bureau, telephone, 137XXXXXXX (phone number)", etc., among which, The attribute value of the ontology "phone” is "137XXXXXXX".
  • Step S203 constructing a semantic web framework according to events, event relations, ontology and ontology relations.
  • a Semantic Web framework is generated according to the recognized events, event relations, ontology and ontology relations.
  • the Semantic Web framework is used to describe events, event relations, ontologies and ontology relations in a structured way.
  • the Semantic Web framework can be described in OWL language, which is a description system constructed in OWL.
  • OWL language Web Ontology Language
  • OWL language is a network ontology language and one of the cores of the Semantic Web technology stack. It can quickly and flexibly perform data modeling and describe knowledge graphs at the semantic level.
  • RDF Resource Description Framework
  • RDFS Resource Description Framework Schema
  • GovOffice which belongs to the subclass of organization, rdfs: subClassOf Organization, such as Human Resources and Social Security Bureau, police station; each GovOffice will be associated with contact information, which is a compound knowledge structure CVT, including address, telephone, etc.;
  • Government affairs that is, events, rdfs:Class.Business, including settlement, certification and other affairs, government affairs is a composite knowledge structure CVT, which can contain sub-transactions, rdfs:Class.SubBusiness, which belongs to the subclass of government affairs, rdfs:subClassOf Business, For example, handling the matter of settlement, including preliminary examination, moving out of the original household registration, registration at the police station and other sub-tasks, as well as corresponding departments and clerks, etc.;
  • each transaction and sub-transaction may involve documents and materials, such as the initial review process of settlement, may require the submission of various documents and materials;
  • transaction A is the pre-process of transaction B, represented by rdfs:before, and the inverse relationship rdfs:before owl:inverseOf rdfs: is defined through the OWL semantic network: After, automatically infer that transaction B is the post-process rdfs:after of transaction A.
  • Step S204 generating a transactional knowledge graph corresponding to the transactional information text based on the semantic web framework.
  • the semantic web framework based on OWL is the semantic layer description of the knowledge graph.
  • the transaction knowledge graph can be generated, and the transaction knowledge graph shows the event flow in the transaction information text in the form of a graph.
  • the transactional knowledge graph includes nodes and connection edges, and the connection edges are used to connect nodes. Both nodes and connection edges have representation vectors.
  • Step S205 acquiring transaction query text.
  • the transaction query text is generated based on user operations. For example, the user can input a question and use the user-input question as the transaction query text, or perform speech recognition on the user's voice question and convert it into a transaction query text.
  • Step S206 calculating the semantic similarity between the transaction query text and each node in the transaction knowledge graph, and determining the transaction information query result according to the semantic similarity and the transaction knowledge graph.
  • the transaction query text is converted into a representation vector, and the nodes and edges in the knowledge graph also have representation vectors.
  • the semantic similarity is calculated between the transaction query text and the characterization vectors of each node in the knowledge graph, where the cosine similarity between the characterization vectors can be used as the semantic similarity.
  • Select several nodes according to the semantic similarity For example, select a node whose semantic similarity is greater than the preset similarity threshold, and then start from this node in the transaction knowledge graph, and use all possible paths as transaction information query results, return them to the querying user, and complete the transaction information Inquire. Since the transaction knowledge graph reflects the event flow, the transaction information query results intercepted from the transaction knowledge graph according to the transaction query text can also reflect the event flow required by the user.
  • the event in the transaction information text recording the transaction process and the relationship between the events are identified, and the ontology in the event and the ontology relationship between the ontology are identified, and then the semantic web framework is constructed.
  • the semantic web framework has rich Semantic expression ability, which can accurately and comprehensively describe transaction information; build a transaction knowledge map according to the semantic web framework, so as to display the transaction process in the form of a map; convert the transaction query text into a representation vector, and the calculation is similar to the semantics of each node in the transaction knowledge map Degree, so as to determine the position of the transaction query text in the transaction knowledge graph, so that the transaction query result can be accurately determined from the transaction knowledge graph, and the required transaction information can be automatically matched, which improves the efficiency of transaction information acquisition.
  • the above-mentioned step of identifying events in the transaction information text and the relationship between events may include: splitting the transaction information text into multiple subtexts; for each subtext, determining the subtext according to the TF-IDF information of the subtext Keywords in the text; input the keywords into the event recognition model to obtain the events recorded in the sub-text; input the sub-text with events into the event relationship recognition model to obtain the event relationship between events.
  • split the transaction information text for example, split the transaction information text according to the paragraphs in the transaction information text, or split the transaction information text in units of sentences, or split the transaction information text according to the number in the transaction information text, etc. Split to get multiple subtexts.
  • TF-IDF (term frequency–inverse document frequency) is a commonly used weighting technique for information retrieval and data mining.
  • TF is Term Frequency, which refers to the frequency with which a given word appears in a file.
  • IDF is the Inverse Document Frequency Index (Inverse Document Frequency), which can be obtained by dividing the total number of documents by the number of documents containing a word, and then calculating the logarithm.
  • TF-IDF is used to evaluate the importance of word segmentation to subtext. Keywords can be determined in the subtext according to the TF-IDF value. For example, the TF-IDF value of the word segment can be compared with a preset threshold, and if the TF-IDF value is greater than the preset threshold, the word segment can be used as a keyword in the subtext.
  • the event recognition model can judge whether an event is recorded in the subtext.
  • the keywords need to meet a certain number, and then the subtext is marked according to the keywords, and then input into the event recognition model, and the event recognition model can also output event tags for describing the event.
  • the event recognition model can be constructed based on a bag of words model (bag of words, BOW), and keywords are input into the trained bag of words model to obtain recognition results. For example, key words such as "settled, fresh graduate, registered permanent residence” are input into the bag of words model, and the recognition result "settled event” is obtained.
  • relationship recognition model can be constructed based on the bert model.
  • keywords are identified according to the IF-IDF information, and the keywords are input into the event identification model for event identification, and then the subtext of existing events is input into the event relationship identification model to obtain the event relationship between events, thereby mining the event Event-level information in infotext.
  • the above step of identifying the ontology in the event and the ontology relationship between the ontology includes: processing the subtext where the event is located according to the entity recognition algorithm to obtain the ontology in the subtext and the corresponding ontology type; based on the ontology type and The preset ontology type relationship table determines the ontology relationship between each ontology.
  • the subtext containing the event is obtained, and the entity recognition algorithm is used to perform entity recognition on the subtext to obtain the entities contained in the subtext and the entity types of the entities.
  • entity recognition algorithm is used to perform entity recognition on the subtext to obtain the entities contained in the subtext and the entity types of the entities.
  • ontology andontology type are used here to refer to the "entity” and "entity type” in the subtext.
  • the ontology type is used to express the type attribute of the ontology, for example, for the ontology "S city”, its ontology type is "city”.
  • the ontology relationships can be recorded in a pre-set ontology type relationship table.
  • the ontology relationship between the ontology can be obtained by querying the ontology type relationship table. For example, through the ontology type relationship table, it is determined that "declaration material" is associated with "clerk”.
  • the entity recognition algorithm is used to identify the ontology and the ontology type, and the ontology relationship between the ontology is determined through the ontology type relationship table, so as to dig out the information at the ontology level inside the event.
  • step S204 may include: constructing an initial transaction knowledge graph corresponding to the transaction information text according to the semantic web framework; determining the representation vectors of nodes and connection edges in the initial transaction knowledge graph based on the distance model, and obtaining the transaction knowledge graph.
  • the semantic web framework is a semantic layer description of the knowledge graph, and a knowledge graph can be constructed based on the semantic web framework.
  • the construction of the transactional knowledge graph can include two steps. First, the initial transactional knowledge graph corresponding to the transactional information text is constructed according to the semantic web framework.
  • the initial transactional knowledge graph includes nodes and connection edges, but the representation vectors of nodes and connection edges have not yet been determined.
  • each element node and connection edge
  • the head entity, relationship and tail entity all have representation vectors, which are h, r, t in turn.
  • the distance model can be used to determine the representation vector of each element and measure whether the representation vector of each element is reasonable.
  • the distance model is a scoring function based on distance. For the triplet (h, r, t), when h+r is very close to t, it can be considered that the representation vector can better represent the elements in the initial practical knowledge map, so that Get the transaction knowledge graph.
  • the initial transactional knowledge graph is first constructed according to the semantic web framework, and then the representation vectors of nodes and connection edges in the initial transactional knowledge graph are determined to complete the construction of the transactional knowledge graph.
  • step S206 may include: performing word segmentation processing on the transaction query text to obtain multiple subwords; mapping each subword to a word vector; encoding the word vector through a long-short-term memory network to obtain a text representation of the transaction query text Vector; calculate the semantic similarity between the text representation vector and the representation vector of each node in the transaction knowledge graph; determine the transaction information query result in the transaction knowledge graph according to the semantic similarity, and return the transaction information query result to the query terminal.
  • the transactional query text is firstly subjected to word segmentation processing to obtain multiple word segmentations.
  • Each subword is mapped to a word vector, and according to the order of each subword in the transaction query text, the word vector corresponding to each subword is input into the long-term short-term memory network for encoding, and the long-short-term memory network (Long-Short Term Memory, LSTM , which is a time-recurrent neural network) can encode the position information of each subword in the transaction query text, thereby adding context information and better semantic expression.
  • LSTM Long-Short Term Memory
  • the long short-term memory network encodes and outputs the text representation vector of the transaction query text. Then calculate the semantic similarity between the text representation vector and the representation vector of each node in the practical knowledge graph. Select nodes in the transaction knowledge graph according to the semantic similarity, and then use all the paths derived from the nodes, that is, part of the transaction knowledge graph, as the transaction information novelty search results.
  • the transaction information query results can be sent to the query terminal for display.
  • the above-mentioned step of determining the transaction information query result in the transaction knowledge map according to the semantic similarity may include: sorting each semantic similarity in descending order to obtain a similarity sequence; selecting at least one semantic similarity from the similarity sequence; In the transaction knowledge graph, determine the sub-graph where the node corresponding to the semantic similarity is located; determine the sub-graph as the transaction information query result.
  • N is a preset value, and N is a positive integer
  • the text can also be generated according to the sub-graph derived from the node, and then the text can be used as the transaction information query result.
  • the semantic similarity is sorted in descending order, at least one semantic similarity is selected from the similarity queue according to the size of the semantic similarity, and the sub-map where the node corresponding to the semantic similarity is located is used as the transaction information query result, which can ensure The accuracy of transaction information query results.
  • the transaction query text is segmented, and the semantic and position information of each segmented word vector is encoded through the long-short-term memory network, so that a more accurate text representation vector of the transaction query text can be generated, thereby improving the quality of representation based on the text. Accuracy of vectors for transactional information queries in transactional knowledge graphs.
  • step S206 it may also include: identifying the ontology in the transaction query text and its corresponding ontology type; matching the corresponding sub-graph from the transaction knowledge graph according to the ontology and its corresponding ontology type; matching the matched The sub-graph is determined as the transaction information query result.
  • querying transaction information can also be performed through query templates. Identify the ontology and its corresponding ontology type in the transaction query text through the entity recognition algorithm, fill the query template according to the ontology and ontology type, and then match the query template with the transaction knowledge graph. Ontology and ontology type, the sub-graph can be determined as the transaction information query result.
  • identifying the ontology in the transaction query text and its corresponding ontology type can be directly matched in the transaction knowledge graph according to the ontology and ontology type, which enriches the way of transaction information query.
  • AI artificial intelligence
  • the embodiments of the present application may acquire and process relevant data based on artificial intelligence technology.
  • artificial intelligence is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. .
  • Artificial intelligence basic technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation/interaction systems, and mechatronics.
  • Artificial intelligence software technology mainly includes computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning/deep learning.
  • This application relates to machine learning, knowledge representation and reasoning, and natural language processing in the field of artificial intelligence.
  • the application can be applied in the field of smart government affairs, so as to promote the construction of smart cities.
  • the aforementioned storage medium may be a nonvolatile storage medium such as a magnetic disk, an optical disk, a read-only memory (Read-Only Memory, ROM), or a random access memory (Random Access Memory, RAM).
  • the present application provides an embodiment of a device for querying transaction information.
  • the device embodiment corresponds to the method embodiment shown in FIG. 2 , and the device specifically It can be applied to various electronic devices.
  • the transaction information query device 300 described in this embodiment includes: a text acquisition module 301, a text recognition module 302, a frame construction module 303, a map generation module 304, a query acquisition module 305 and a transaction query module 306, wherein :
  • a text acquisition module 301 configured to acquire transaction information text.
  • the text identification module 302 is configured to identify the events in the transaction information text and the relationship between each event, and identify the ontology in the event and the ontology relationship among the ontologies.
  • the frame building module 303 is used to build a semantic web frame according to events, event relations, ontology and ontology relations.
  • the map generation module 304 is configured to generate a transaction knowledge map corresponding to the transaction information text based on the semantic web framework.
  • a query obtaining module 305 configured to obtain transaction query text.
  • the transaction query module 306 is configured to calculate the semantic similarity between the transaction query text and each node in the transaction knowledge graph, and determine the transaction information query result according to the semantic similarity and the transaction knowledge graph.
  • the event in the transaction information text recording the transaction process and the relationship between the events are identified, and the ontology in the event and the ontology relationship between the ontology are identified, and then the semantic web framework is constructed.
  • the semantic web framework has rich Semantic expression ability, which can accurately and comprehensively describe transaction information; build a transaction knowledge map according to the semantic web framework, so as to display the transaction process in the form of a map; convert the transaction query text into a representation vector, and the calculation is similar to the semantics of each node in the transaction knowledge map Degree, so as to determine the position of the transaction query text in the transaction knowledge graph, so that the transaction query result can be accurately determined from the transaction knowledge graph, and the required transaction information can be automatically matched, which improves the efficiency of transaction information acquisition.
  • the text recognition module 302 may include: a text splitting submodule, a keyword determination submodule, an event identification submodule, and an affair relationship identification submodule, wherein:
  • the text splitting submodule is used to split the transaction information text into multiple subtexts.
  • the keyword determining submodule is used for determining keywords in the subtext according to the TF-IDF information of the subtext for each subtext.
  • the event recognition sub-module is used to input keywords into the event recognition model to obtain events recorded by the subtext.
  • the affair relationship recognition sub-module is used for inputting the subtext of existing events into the affairs relationship recognition model to obtain the affairs relations among the events.
  • keywords are identified according to the IF-IDF information, and the keywords are input into the event identification model for event identification, and then the subtext of existing events is input into the event relationship identification model to obtain the event relationship between events, thereby mining the event Event-level information in infotext.
  • the text recognition module 302 may also include: an ontology recognition submodule and an ontology relationship determination submodule, wherein:
  • the ontology recognition sub-module is used to process the subtext where the event is located according to the entity recognition algorithm to obtain the ontology in the subtext and the corresponding ontology type.
  • the ontology relationship determining sub-module is used to determine the ontology relationship between the ontology based on the ontology type and the preset ontology type relationship table.
  • the entity recognition algorithm is used to identify the ontology and the ontology type, and the ontology relationship between the ontology is determined through the ontology type relationship table, so as to dig out the information at the ontology level inside the event.
  • the map generation module 304 may include: an initial construction submodule and a vector determination submodule, wherein:
  • the initial construction sub-module is used to construct the initial transaction knowledge map corresponding to the transaction information text according to the semantic web framework.
  • the vector determination sub-module is used to determine the representation vectors of nodes and connection edges in the initial transaction knowledge graph based on the distance model to obtain the transaction knowledge graph.
  • the initial transactional knowledge graph is first constructed according to the semantic web framework, and then the representation vectors of nodes and connection edges in the initial transactional knowledge graph are determined to complete the construction of the transactional knowledge graph.
  • the transaction query module 306 may include a text segmentation submodule, a subword mapping submodule, a representation acquisition submodule, a similarity calculation submodule, and a result determination submodule, wherein:
  • the text word segmentation sub-module is used to perform word segmentation processing on the transaction query text to obtain multiple subwords.
  • the subword mapping submodule is used to map each subword into a word vector.
  • the representation acquisition sub-module is used to encode the word vector through the long short-term memory network to obtain the text representation vector of the transaction query text.
  • the similarity calculation sub-module is used to calculate the semantic similarity between the text representation vector and the representation vectors of each node in the transaction knowledge map.
  • the result determination sub-module is used to determine the transaction information query result in the transaction knowledge map according to the semantic similarity, and return the transaction information query result to the query terminal.
  • the transaction query text is segmented, and the semantic and position information of each segmented word vector is encoded through the long-short-term memory network, so that a more accurate text representation vector of the transaction query text can be generated, thereby improving the quality of representation based on the text. Accuracy of vectors for transactional information queries in transactional knowledge graphs.
  • the result determination submodule may include: a similarity ranking unit, a similarity selection unit, a sub-map determination unit, and a result determination unit, wherein:
  • the similarity sorting unit is configured to sort the semantic similarities in descending order to obtain a similarity sequence.
  • the similarity selection unit is configured to select at least one semantic similarity from the similarity sequence.
  • the sub-graph determining unit is configured to determine the sub-graph in which the node corresponding to the semantic similarity is located in the transaction knowledge graph.
  • a result determining unit configured to determine the sub-graph as the transaction information query result.
  • the semantic similarity is sorted in descending order, at least one semantic similarity is selected from the similarity queue according to the size of the semantic similarity, and the sub-map where the node corresponding to the semantic similarity is located is used as the transaction information query result, which can ensure The accuracy of transaction information query results.
  • the transaction information query device 300 may also include: a query identification module, a sub-graph matching module, and a result determination module, wherein:
  • the query identification module is used to identify the ontology and its corresponding ontology type in the transaction query text.
  • the sub-graph matching module is used to match the corresponding sub-graph from the transaction knowledge graph according to the ontology and its corresponding ontology type.
  • the result determination module is configured to determine the matched sub-graph as the transaction information query result.
  • identifying the ontology in the transaction query text and its corresponding ontology type can be directly matched in the transaction knowledge graph according to the ontology and ontology type, which enriches the way of transaction information query.
  • FIG. 4 is a block diagram of the basic structure of the computer device in this embodiment.
  • the computer device 4 includes a memory 41 , a processor 42 and a network interface 43 connected to each other through a system bus. It should be noted that only the computer device 4 with components 41-43 is shown in the figure, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculation and/or information processing according to preset or stored instructions, and its hardware includes but is not limited to microprocessors, dedicated Integrated circuit (Application Specific Integrated Circuit, ASIC), programmable gate array (Field-Programmable Gate Array, FPGA), digital processor (Digital Signal Processor, DSP), embedded devices, etc.
  • ASIC Application Specific Integrated Circuit
  • FPGA Field-Programmable Gate Array
  • DSP Digital Signal Processor
  • the computer equipment may be computing equipment such as a desktop computer, a notebook, a palmtop computer, and a cloud server.
  • the computer device can perform human-computer interaction with the user through keyboard, mouse, remote controller, touch panel or voice control device.
  • the memory 41 includes at least one type of computer-readable storage medium, the computer-readable storage medium can be non-volatile or volatile, and the computer-readable storage medium includes flash memory, hard disk, multimedia card , card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable Program read-only memory (PROM), magnetic memory, magnetic disks, optical disks, etc.
  • the memory 41 may be an internal storage unit of the computer device 4 , such as a hard disk or memory of the computer device 4 .
  • the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk equipped on the computer device 4, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, flash memory card (Flash Card), etc.
  • the memory 41 may also include both an internal storage unit of the computer device 4 and an external storage device thereof.
  • the memory 41 is generally used to store the operating system and various application software installed in the computer device 4, such as computer-readable instructions of transaction information query methods and the like. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or will be output.
  • the processor 42 may be a central processing unit (Central Processing Unit, CPU), controller, microcontroller, microprocessor, or other data processing chips in some embodiments. This processor 42 is generally used to control the general operation of said computer device 4 .
  • the processor 42 is configured to execute computer-readable instructions stored in the memory 41 or process data, for example, execute computer-readable instructions of the transaction information query method.
  • the network interface 43 may include a wireless network interface or a wired network interface, and the network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.
  • the computer device provided in this embodiment can execute the above transaction information query method.
  • the transaction information query method may be the transaction information query method in each of the foregoing embodiments.
  • the event in the transaction information text recording the transaction process and the relationship between the events are identified, and the ontology in the event and the ontology relationship between the ontology are identified, and then the semantic web framework is constructed.
  • the semantic web framework has rich Semantic expression ability, which can accurately and comprehensively describe transaction information; build a transaction knowledge map according to the semantic web framework, so as to display the transaction process in the form of a map; convert the transaction query text into a representation vector, and the calculation is similar to the semantics of each node in the transaction knowledge map Degree, so as to determine the position of the transaction query text in the transaction knowledge graph, so that the transaction query result can be accurately determined from the transaction knowledge graph, and the required transaction information can be automatically matched, which improves the efficiency of transaction information acquisition.
  • the present application also provides another implementation manner, which is to provide a computer-readable storage medium, the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to The at least one processor is made to execute the steps of the above transaction information query method.
  • the event in the transaction information text recording the transaction process and the relationship between the events are identified, and the ontology in the event and the ontology relationship between the ontology are identified, and then the semantic web framework is constructed.
  • the semantic web framework has rich Semantic expression ability, which can accurately and comprehensively describe transaction information; build a transaction knowledge map according to the semantic web framework, so as to display the transaction process in the form of a map; convert the transaction query text into a representation vector, and the calculation is similar to the semantics of each node in the transaction knowledge map Degree, so as to determine the position of the transaction query text in the transaction knowledge graph, so that the transaction query result can be accurately determined from the transaction knowledge graph, and the required transaction information can be automatically matched, which improves the efficiency of transaction information acquisition.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Databases & Information Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

本申请实施例属于人工智能领域,应用于智慧政务领域中,涉及一种事务信息查询方法,方法包括:获取事务信息文本;识别事务信息文本中的事件以及各事件之间的事理关系,并识别事件中的本体以及各本体之间的本体关系;根据事件、事理关系、本体和本体关系,构建语义网框架;基于语义网框架生成事务信息文本所对应的事务知识图谱;获取事务查询文本;计算事务查询文本与事务知识图谱中各节点的语义相似度,并根据语义相似度和事务知识图谱确定事务信息查询结果。本申请还提供一种事务信息查询装置、计算机设备及存储介质。此外,本申请还涉及区块链技术,事务信息文本可存储于区块链中。本申请提高了事务信息获取效率。

Description

事务信息查询方法、装置、计算机设备及存储介质
本申请要求于2022年01月11日提交中国专利局、申请号为202210028520.8,发明名称为“事务信息查询方法、装置、计算机设备及存储介质”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
技术领域
本申请涉及人工智能技术领域,尤其涉及一种事务信息查询方法、装置、计算机设备及存储介质。
背景技术
在日常生活中,常常会遇到各种事务流程相关的公文,这些公文记录了既定的事务流程。人们在办理事务时,需要按照公文的规定去办理。这些条文常常显得冗长繁杂,难以直接获取到有用的信息。发明人意识到,虽然可以在相关机构中安排经办人员帮忙解读公文、辅助办理,但由于公文定义的事务流程可能存在变动、经办人员专业程度存在差异,用户还是存在难以便利地获取到有用信息的情况。
发明内容
本申请实施例的目的在于提出一种事务信息查询方法、装置、计算机设备及存储介质,以解决事务信息获取效率较低的问题。
为了解决上述技术问题,本申请实施例提供一种事务信息查询方法,采用了如下所述的技术方案:
获取事务信息文本;
识别所述事务信息文本中的事件以及各事件之间的事理关系,并识别所述事件中的本体以及各本体之间的本体关系;
根据所述事件、所述事理关系、所述本体和所述本体关系,构建语义网框架;
基于所述语义网框架生成所述事务信息文本所对应的事务知识图谱;
获取事务查询文本;
计算所述事务查询文本与所述事务知识图谱中各节点的语义相似度,并根据所述语义相似度和所述事务知识图谱确定事务信息查询结果。
为了解决上述技术问题,本申请实施例还提供一种事务信息查询装置,采用了如下所述的技术方案:
文本获取模块,用于获取事务信息文本;
文本识别模块,用于识别所述事务信息文本中的事件以及各事件之间的事理关系,并识别所述事件中的本体以及各本体之间的本体关系;
框架构建模块,用于根据所述事件、所述事理关系、所述本体和所述本体关系,构建语义网框架;
图谱生成模块,用于基于所述语义网框架生成所述事务信息文本所对应的事务知识图谱;
查询获取模块,用于获取事务查询文本;
事务查询模块,用于计算所述事务查询文本与所述事务知识图谱中各节点的语义相似度,并根据所述语义相似度和所述事务知识图谱确定事务信息查询结果。
为了解决上述技术问题,本申请实施例还提供一种计算机设备,包括存储器和处理器,所述存储器中存储有计算机可读指令,所述处理器执行所述计算机可读指令时实现如下步骤:
获取事务信息文本;
识别所述事务信息文本中的事件以及各事件之间的事理关系,并识别所述事件中的本体以及各本体之间的本体关系;
根据所述事件、所述事理关系、所述本体和所述本体关系,构建语义网框架;
基于所述语义网框架生成所述事务信息文本所对应的事务知识图谱;
获取事务查询文本;
计算所述事务查询文本与所述事务知识图谱中各节点的语义相似度,并根据所述语义相似度和所述事务知识图谱确定事务信息查询结果。
为了解决上述技术问题,本申请实施例还提供一种计算机可读存储介质,所述计算机可读存储介质存储有计算机可读指令,所述计算机可读指令被处理器执行时实现如下步骤:
获取事务信息文本;
识别所述事务信息文本中的事件以及各事件之间的事理关系,并识别所述事件中的本体以及各本体之间的本体关系;
根据所述事件、所述事理关系、所述本体和所述本体关系,构建语义网框架;
基于所述语义网框架生成所述事务信息文本所对应的事务知识图谱;
获取事务查询文本;
计算所述事务查询文本与所述事务知识图谱中各节点的语义相似度,并根据所述语义相似度和所述事务知识图谱确定事务信息查询结果。
与现有技术相比,本申请实施例主要有以下有益效果:可以准确从事务知识图谱中确定事务查询结果,自动匹配出所需的事务信息,提高了事务信息获取效率。
附图说明
下面将对本申请实施例描述中所需要使用的附图作一个简单介绍。
图1是本申请可以应用于其中的示例性系统架构图;
图2是根据本申请的事务信息查询方法的一个实施例的流程图;
图3是根据本申请的事务信息查询装置的一个实施例的结构示意图;
图4是根据本申请的计算机设备的一个实施例的结构示意图。
具体实施方式
除非另有定义,本文所使用的所有的技术和科学术语与属于本申请的技术领域的技术人员通常理解的含义相同;本文中在申请的说明书中所使用的术语只是为了描述具体的实施例的目的,不是旨在于限制本申请;本申请的说明书和权利要求书及上述附图说明中的术语“包括”和“具有”以及它们的任何变形,意图在于覆盖不排他的包含。本申请的说明书和权利要求书或上述附图中的术语“第一”、“第二”等是用于区别不同对象,而不是用于描述特定顺序。
在本文中提及“实施例”意味着,结合实施例描述的特定特征、结构或特性可以包含在本申请的至少一个实施例中。在说明书中的各个位置出现该短语并不一定均是指相同的实施例,也不是与其它实施例互斥的独立的或备选的实施例。本领域技术人员显式地和隐式地理解的是,本文所描述的实施例可以与其它实施例相结合。
为了使本技术领域的人员更好地理解本申请方案,下面将结合附图,对本申请实施例中的技术方案进行清楚、完整地描述。
如图1所示,系统架构100可以包括终端设备101、102、103,网络104和服务器105。网络104用以在终端设备101、102、103和服务器105之间提供通信链路的介质。网络104可以包括各种连接类型,例如有线、无线通信链路或者光纤电缆等等。
用户可以使用终端设备101、102、103通过网络104与服务器105交互,以接收或发送消息等。终端设备101、102、103上可以安装有各种通讯客户端应用,例如网页浏览器 应用、购物类应用、搜索类应用、即时通信工具、邮箱客户端、社交平台软件等。
终端设备101、102、103可以是具有显示屏并且支持网页浏览的各种电子设备,包括但不限于智能手机、平板电脑、电子书阅读器、MP3播放器(Moving Picture Experts Group Audio Layer III,动态影像专家压缩标准音频层面3)、MP4(Moving Picture Experts Group Audio Layer IV,动态影像专家压缩标准音频层面4)播放器、膝上型便携计算机和台式计算机等等。
服务器105可以是提供各种服务的服务器,例如对终端设备101、102、103上显示的页面提供支持的后台服务器。
需要说明的是,本申请实施例所提供的事务信息查询方法一般由服务器执行,相应地,事务信息查询装置一般设置于服务器中。
应该理解,图1中的终端设备、网络和服务器的数目仅仅是示意性的。根据实现需要,可以具有任意数目的终端设备、网络和服务器。
继续参考图2,示出了根据本申请的事务信息查询方法的一个实施例的流程图。所述的事务信息查询方法,包括以下步骤:
步骤S201,获取事务信息文本。
在本实施例中,事务信息查询方法运行于其上的电子设备(例如图1所示的服务器)可以通过有线连接方式或者无线连接方式与终端进行通信。需要指出的是,上述无线连接方式可以包括但不限于3G/4G/5G连接、WiFi连接、蓝牙连接、WiMAX连接、Zigbee连接、UWB(ultra wideband)连接、以及其他现在已知或将来开发的无线连接方式。
具体地,服务器首先需要获取事务信息文本。事务信息文本记录了事务流程信息。例如,事务信息文本可以是政务公文,记录了既定的办事流程。事务信息文本中记录了至少一个事务流程。
需要强调的是,为进一步保证上述事务信息文本的私密和安全性,上述事务信息文本还可以存储于一区块链的节点中。
本申请所指区块链是分布式数据存储、点对点传输、共识机制、加密算法等计算机技术的新型应用模式。区块链(Blockchain),本质上是一个去中心化的数据库,是一串使用密码学方法相关联产生的数据块,每一个数据块中包含了一批次网络交易的信息,用于验证其信息的有效性(防伪)和生成下一个区块。区块链可以包括区块链底层平台、平台产品服务层以及应用服务层等。
步骤S202,识别事务信息文本中的事件以及各事件之间的事理关系,并识别事件中的本体以及各本体之间的本体关系。
具体地,对事务信息文本进行自然语言处理,以识别事务信息文本中所记录的事件以及事理关系。其中,事件是达成一个目标的流程所构成的整体,事理关系是指不同事件之间的关系。例如,“应届生如何落户S市”以及“应届生如何申请S市住房补贴”就是两个事件,两个事件可以是条件关系,“应届生申请S市住房补贴”的前提条件是“应届生落户S市”。
同时,还要识别事件中的本体以及各本体之间的本体关系。本体是事件中涉及到的实体,在前边“应届生如何落户S市”的例子中,可能出现的本体包括“应届生、户口、人力资源局、电话、137XXXXXXXX(电话号码)”等,其中,本体“电话”的属性值为“137XXXXXXXX”。
步骤S203,根据事件、事理关系、本体和本体关系,构建语义网框架。
具体地,根据识别到的事件、事理关系、本体和本体关系生成语义网框架。语义网框架用于结构化地描述事件、事理关系、本体和本体关系。在一个实施例中,语义网框架可以用OWL语言描述,是用OWL构建的描述体系。OWL语言(Web Ontology Language)是一种网络本体语言,是语义网技术栈的核心之一,可以快速、灵活地进行数据建模,对知识图谱进行语义层地描述。
知识图谱的基石是RDF(Resource Description Framework),即资源描述框架,它的本质是一个数据模型,提供了一个统一的标准,用于描述实体或资源,是表示事物的一种方法和手段。在RDF中,RDFS(Resource Description Framework Schema)是基础的模式语言。然而,RDFS的表达能力依旧有限,因此后来发展出了OWL进行数据建模,对知识图谱进行语义层的描述。
根据识别到的事件、事理关系、本体和本体关系,按照OWL中的预设逻辑进行填充,即可得到基于OWL的语义网框架。
以事务信息文本为政务公文为例,说明OWL本体描述体系中的主要本体以及关系:
组织rdfs:Class.Organization;
企业:Company,属于组织子类,rdfs:subClassOf Organization;
投资基金:Fund,属于组织子类,rdfs:subClassOf Organization;
政府部门:GovOffice,属于组织子类,rdfs:subClassOf Organization,如人社局,派出所;每个GovOffice会和联系信息进行关联,联系信息是复合知识结构CVT,包括地址Address,电话Telephone等;
人:rdfs:Class.Person;
办事员:Officer,属于人的子类,rdfs:subClassOf Person;
政务事务:即事件,rdfs:Class.Business,包括落户,办证等事务,政务事务是复合知识结构CVT,可以包含子事务,rdfs:Class.SubBusiness,属于政务事务的子类,rdfs:subClassOf Business,如办理落户这件事,包含初审,原户口迁出,派出所登记等子事务,以及对应的部门和办事员等;
文件材料,rdfs:Class.File,每个事务和子事务可能会涉及文件和材料,如落户的初审流程中,可能需要各种文件材料的提交;
各个事务之间有事理关系,代表事件之间的顺承、因果、条件和上下位等事理逻辑关系。如果一个办事流程中,需要先办事务A,才能到事务B,则事务A是事务B的前置流程,以rdfs:before进行表示,通过OWL语义网络定义反关系rdfs:before owl:inverseOf rdfs:after,自动推理事务B是事务A的后置流程rdfs:after。
步骤S204,基于语义网框架生成事务信息文本所对应的事务知识图谱。
具体地,基于OWL构建的语义网框架是知识图谱的语义层描述,根据语义网框架可以生成事务知识图谱,事物知识图谱以图谱的形式展示了事务信息文本中的事件流程。事务知识图谱包括节点和连接边,连接边用于连接节点。节点和连接边都具有表征向量。
步骤S205,获取事务查询文本。
具体地,在进行应用时,需要获取事物查询文本。事物查询文本基于用户操作生成,例如,用户可以输入问题,将用户输入的问题作为事务查询文本,或者,对用户的语音问句进行语音识别,转换成事务查询文本。
步骤S206,计算事务查询文本与事务知识图谱中各节点的语义相似度,并根据语义相似度和事务知识图谱确定事务信息查询结果。
具体地,将事务查询文本转换为表征向量,知识图谱中的节点和边也具有表征向量。将事务查询文本与知识图谱中各节点的表征向量进行语义相似度的计算,其中,可以将表征向量之间的余弦相似度作为语义相似度。
根据语义相似度选取若干个节点。例如,选取语义相似度大于预设的相似度阈值的节点,然后在事务知识图谱中,从该节点出发,将所有可能的路径作为事务信息查询结果,将其返回给查询的用户,完成事务信息查询。由于事务知识图谱反应了事件流程,根据事务查询文本从事务知识图谱中截取到的事务信息查询结果也可以反应用户所需要的事件流程。
本实施例中,识别记录事务流程的事务信息文本中的事件以及事件之间的事理关系,并识别事件中的本体以及本体之间的本体关系,然后构建语义网框架,语义网框架具有丰 富的语义表达能力,可以准确全面地描述事务信息;根据语义网框架构建事务知识图谱,从而以图谱的形式展示事务流程;将事务查询文本转化为表征向量,计算与事务知识图谱中各节点的语义相似度,从而确定事务查询文本在事务知识图谱中所在的位置,从而可以准确从事务知识图谱中确定事务查询结果,自动匹配出所需的事务信息,提高了事务信息获取效率。
进一步的,上述识别事务信息文本中的事件以及各事件之间的事理关系的步骤可以包括:将事务信息文本拆分为多个子文本;对于每个子文本,根据子文本的TF-IDF信息确定子文本中的关键词;将关键词输入事件识别模型,以获取子文本记录的事件;将存在事件的子文本输入事理关系识别模型,得到各事件间的事理关系。
具体地,对事务信息文本进行拆分,例如,根据事务信息文本中的段落对事务信息文本进行拆分,或者以句子为单位对事务信息文本进行拆分,或者根据事务信息文本中的编号等进行拆分,得到多个子文本。
对于每个子文本,对子文本进行分词得到多个分词,然后计算每个分词的TF-IDF值,从而得到子文本的TF-IDFF信息。其中,TF-IDF(term frequency–inverse document frequency)是一种用于信息检索与数据挖掘的常用加权技术。TF是词频(TermFrequency),是指某一个给定的词在文件中出现的频率。IDF是逆文本频率指数(Inverse Document Frequency),可以由总文件数目除以包含某词语之文件的数目,再求对数得到。
TF-IDF用以评估分词对子文本的重要程度。可以根据TF-IDF值,在子文本中确定关键词。例如,可以将分词的TF-IDF值与预设阈值相比较,如果TF-IDF值大于预设阈值,则该分词可以作为子文本中的关键词。
然后将关键词输入预先训练完毕的事件识别模型,事件识别模型可以判断子文本中是否记录了事件。在一个实施例中,关键词需要满足一定的数量,然后根据关键词对子文本进行标注,再输入事件识别模型,事件识别模型还可以输出事件标签,用于对事件进行描述。在一个实施例中,事件识别模型可以基于词袋模型(bag of words,BOW)构建,将关键词输入训练完毕的词袋模型,得到识别结果。例如,将“落户、应届生、户口”等关键词输入词袋模型,得到识别结果“落户事件”。
事件之间具有事理关系,例如顺承、因果、条件和上下位等。事件时间识别结果,将存在事件的子文本输入预先训练完毕的事理关系识别模型,由事理关系识别模型判断子文本中事件之间的关系,从而得到事理关系。在一个实施例中,事理关系识别模型可以基于bert模型构建。
本实施例中,根据IF-IDF信息识别关键词,将关键词输入事件识别模型进行事件识别,然后将存在事件的子文本输入事理关系识别模型,得到事件之间的事理关系,从而挖掘出事务信息文本中事件层面的信息。
进一步的,上述识别事件中的本体以及各本体之间的本体关系的步骤包括:根据实体识别算法对事件所在的子文本进行处理,得到子文本中的本体以及对应的本体类型;基于本体类型以及预设的本体类型关系表,确定各本体之间的本体关系。
具体地,获取包含事件的子文本,通过实体识别算法对子文本进行实体识别,得到子文本中包含的实体,以及实体的实体类型。为了与基于OWL语言的语义网框架相对应,这里用“本体”以及“本体类型”指代子文本中的“实体”以及“实体类型”。其中,本体类型用于表述本体的类型属性,例如,对于本体“S市”,其本体类型为“城市”。
不同类型的本体之间具有本体关系,本体关系可以记录在预先设置好的本体类型关系表中。根据本体的本体类型在本体类型关系表中进行查询,即可得到本体之间的本体关系。例如,通过本体类型关系表,确定“申报材料”关联于“办事员”。
本实施例中,通过实体识别算法识别本体以及本体类型,并通过本体类型关系表确定本体之间的本体关系,从而挖掘出事件内部本体层面的信息。
进一步的,上述步骤S204可以包括:根据语义网框架构建事务信息文本所对应的初 始事务知识图谱;基于距离模型确定初始事务知识图谱中节点和连接边的表征向量,得到事务知识图谱。
具体地,语义网框架是对知识图谱的语义层描述,基于语义网框架可以构建知识图谱。构建事务知识图谱的可以包括两步,首先根据语义网框架构建事务信息文本所对应的初始事务知识图谱,初始事务知识图谱中包括节点和连接边,但是尚未确定节点和连接边的表征向量。
还需要通过向量表示初始事务知识图谱中的各元素(节点和连接边)。在图谱中,两个节点及其之间的连接边又可以叫头实体-关系-尾实体的三元组,头实体、关系和尾实体都具有表征向量,依次为h,r,t。距离模型可以用来确定各元素的表征向量,并衡量各元素的表征向量是否合理。距离模型是基于距离的评分函数,对于三元组(h,r,t),当h+r与t很接近时,可以认为表征向量可以较好地表征初始实务知识图谱中的各元素,从而得到事务知识图谱。
本实施例中,先根据语义网框架构建初始事务知识图谱,再确定初始事务知识图谱中节点和连接边的表征向量,完成事务知识图谱的构建。
进一步的,上述步骤S206可以包括:对事务查询文本进行分词处理,得到多个子词;将各子词分别映射为词向量;通过长短期记忆网络对词向量进行编码,得到事务查询文本的文本表征向量;计算文本表征向量与事务知识图谱中各节点的表征向量之间的语义相似度;根据语义相似度,在事务知识图谱中确定事务信息查询结果,并将事务信息查询结果返回至查询终端。
具体地,在事务知识图谱中查询时,先对事务查询文本进行分词处理,得到多个分词。将各子词映射为词向量,按照各子词在事务查询文本中的顺序,将各子词所对应的词向量输入长短期记忆网络进行编码,长短期记忆网络(Long-Short Term Memory,LSTM,是一种时间循环神经网络)可以将各子词在事务查询文本中的位置信息加入编码,从而加入上下文信息,可以更好地进行语义表达。
长短期记忆网络编码后输出事务查询文本的文本表征向量。然后计算文本表征向量与实务知识图谱中各节点的表征向量计算语义相似度。根据语义相似度在事务知识图谱中选取节点,然后将节点引申出的全部路径,即部分事务知识图谱作为事务信息查新结果。事务信息查询结果可以被发送至查询终端进行展示。
进一步的,上述根据语义相似度,在事务知识图谱中确定事务信息查询结果的步骤可以包括:对各语义相似度进行降序排列,得到相似度序列;从相似度序列中选取至少一个语义相似度;在事务知识图谱中,确定语义相似度所对应节点所在的子图谱;将子图谱确定为事务信息查询结果。
具体地,将计算得到的多个语义相似度进行降序排列得到语义相似度序列,可以从相似度序列中选取排在前N位(N为预设数值,且N为正整数)的语义相似度。在事务知识图谱中确定选取到的相似度所对应的节点,然后将该节点引申出的子图谱作为事务信息查询结果。
在一个实施例中,还可以根据节点引申出的子图谱生成文本,然后将文本作为事务信息查询结果。
本实施例中,对语义相似度进行降序排列,根据语义相似度的大小从相似度队列中选取至少一个语义相似度,将语义相似度所对应节点所在的子图谱作为事务信息查询结果,可以确保事务信息查询结果的准确性。
本实施例中,对事务查询文本进行分词处理,通过长短期记忆网络对各分词的词向量进行语义和位置信息的编码,可以生成事务查询文本更加准确的文本表征向量,从而提高了依据文本表征向量在事务知识图谱中进行事务信息查询的准确性。
进一步的,上述步骤S206之后,还可以包括:识别事务查询文本中的本体及其对应的本体类型;根据本体及其对应的本体类型,从事务知识图谱中匹配对应的子图谱;将匹 配到的子图谱确定为事务信息查询结果。
具体地,除了通过向量计算进行事务信息查询,还可以通过查询模板进行事务信息的查询。通过实体识别算法识别事务查询文本中的本体及其对应的本体类型,根据本体和本体类型填充查询模板,然后将查询模板与事务知识图谱进行匹配,如果查询到子图谱中具有与查询模板中相同的本体以及本体类型,则可以将子图谱确定为事务信息查询结果。
本实施例中,识别事务查询文本中的本体及其对应的本体类型,可以直接根据本体以及本体类型在事务知识图谱中进行匹配,丰富了事务信息查询的方式。
本申请实施例可以基于人工智能技术对相关的数据进行获取和处理。其中,人工智能(Artificial Intelligence,AI)是利用数字计算机或者数字计算机控制的机器模拟、延伸和扩展人的智能,感知环境、获取知识并使用知识获得最佳结果的理论、方法、技术及应用系统。
人工智能基础技术一般包括如传感器、专用人工智能芯片、云计算、分布式存储、大数据处理技术、操作/交互系统、机电一体化等技术。人工智能软件技术主要包括计算机视觉技术、机器人技术、生物识别技术、语音处理技术、自然语言处理技术以及机器学习/深度学习等几大方向。
本申请涉及人工智能领域中的机器学习、知识表示与推理以及自然语言处理。此外,本申请可应用于智慧政务领域中,从而推动智慧城市的建设。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程,是可以通过计算机可读指令来指令相关的硬件来完成,该计算机可读指令可存储于一计算机可读取存储介质中,该计算机可读指令在执行时,可包括如上述各方法的实施例的流程。其中,前述的存储介质可为磁碟、光盘、只读存储记忆体(Read-Only Memory,ROM)等非易失性存储介质,或随机存储记忆体(Random Access Memory,RAM)等。
应该理解的是,虽然附图的流程图中的各个步骤按照箭头的指示依次显示,但是这些步骤并不是必然按照箭头指示的顺序依次执行。除非本文中有明确的说明,这些步骤的执行并没有严格的顺序限制,其可以以其他的顺序执行。而且,附图的流程图中的至少一部分步骤可以包括多个子步骤或者多个阶段,这些子步骤或者阶段并不必然是在同一时刻执行完成,而是可以在不同的时刻执行,其执行顺序也不必然是依次进行,而是可以与其他步骤或者其他步骤的子步骤或者阶段的至少一部分轮流或者交替地执行。
进一步参考图3,作为对上述图2所示方法的实现,本申请提供了一种事务信息查询装置的一个实施例,该装置实施例与图2所示的方法实施例相对应,该装置具体可以应用于各种电子设备中。
如图3所示,本实施例所述的事务信息查询装置300包括:文本获取模块301、文本识别模块302、框架构建模块303、图谱生成模块304、查询获取模块305以及事务查询模块306,其中:
文本获取模块301,用于获取事务信息文本。
文本识别模块302,用于识别事务信息文本中的事件以及各事件之间的事理关系,并识别事件中的本体以及各本体之间的本体关系。
框架构建模块303,用于根据事件、事理关系、本体和本体关系,构建语义网框架。
图谱生成模块304,用于基于语义网框架生成事务信息文本所对应的事务知识图谱。
查询获取模块305,用于获取事务查询文本。
事务查询模块306,用于计算事务查询文本与事务知识图谱中各节点的语义相似度,并根据语义相似度和事务知识图谱确定事务信息查询结果。
本实施例中,识别记录事务流程的事务信息文本中的事件以及事件之间的事理关系,并识别事件中的本体以及本体之间的本体关系,然后构建语义网框架,语义网框架具有丰富的语义表达能力,可以准确全面地描述事务信息;根据语义网框架构建事务知识图谱,从而以图谱的形式展示事务流程;将事务查询文本转化为表征向量,计算与事务知识图谱 中各节点的语义相似度,从而确定事务查询文本在事务知识图谱中所在的位置,从而可以准确从事务知识图谱中确定事务查询结果,自动匹配出所需的事务信息,提高了事务信息获取效率。
在本实施例的一些可选的实现方式中,文本识别模块302可以包括:文本拆分子模块、关键词确定子模块、事件识别子模块以及事理关系识别子模块,其中:
文本拆分子模块,用于将事务信息文本拆分为多个子文本。
关键词确定子模块,用于对于每个子文本,根据子文本的TF-IDF信息确定子文本中的关键词。
事件识别子模块,用于将关键词输入事件识别模型,以获取子文本记录的事件。
事理关系识别子模块,用于将存在事件的子文本输入事理关系识别模型,得到各事件间的事理关系。
本实施例中,根据IF-IDF信息识别关键词,将关键词输入事件识别模型进行事件识别,然后将存在事件的子文本输入事理关系识别模型,得到事件之间的事理关系,从而挖掘出事务信息文本中事件层面的信息。
在本实施例的一些可选的实现方式中,文本识别模块302还可以包括:本体识别子模块以及本体关系确定子模块,其中:
本体识别子模块,用于根据实体识别算法对事件所在的子文本进行处理,得到子文本中的本体以及对应的本体类型。
本体关系确定子模块,用于基于本体类型以及预设的本体类型关系表,确定各本体之间的本体关系。
本实施例中,通过实体识别算法识别本体以及本体类型,并通过本体类型关系表确定本体之间的本体关系,从而挖掘出事件内部本体层面的信息。
在本实施例的一些可选的实现方式中,图谱生成模块304可以包括:初始构建子模块以及向量确定子模块,其中:
初始构建子模块,用于根据语义网框架构建事务信息文本所对应的初始事务知识图谱。
向量确定子模块,用于基于距离模型确定初始事务知识图谱中节点和连接边的表征向量,得到事务知识图谱。
本实施例中,先根据语义网框架构建初始事务知识图谱,再确定初始事务知识图谱中节点和连接边的表征向量,完成事务知识图谱的构建。
在本实施例的一些可选的实现方式中,事务查询模块306可以包括文本分词子模块、子词映射子模块、表征获取子模块、相似度计算子模块以及结果确定子模块,其中:
文本分词子模块,用于对事务查询文本进行分词处理,得到多个子词。
子词映射子模块,用于将各子词分别映射为词向量。
表征获取子模块,用于通过长短期记忆网络对词向量进行编码,得到事务查询文本的文本表征向量。
相似度计算子模块,用于计算文本表征向量与事务知识图谱中各节点的表征向量之间的语义相似度。
结果确定子模块,用于根据语义相似度,在事务知识图谱中确定事务信息查询结果,并将事务信息查询结果返回至查询终端。
本实施例中,对事务查询文本进行分词处理,通过长短期记忆网络对各分词的词向量进行语义和位置信息的编码,可以生成事务查询文本更加准确的文本表征向量,从而提高了依据文本表征向量在事务知识图谱中进行事务信息查询的准确性。
在本实施例的一些可选的实现方式中,结果确定子模块可以包括:相似度排列单元、相似度选取单元、子图谱确定单元以及结果确定单元,其中:
相似度排列单元,用于对各语义相似度进行降序排列,得到相似度序列。
相似度选取单元,用于从相似度序列中选取至少一个语义相似度。
子图谱确定单元,用于在事务知识图谱中,确定语义相似度所对应节点所在的子图谱。
结果确定单元,用于将子图谱确定为事务信息查询结果。
本实施例中,对语义相似度进行降序排列,根据语义相似度的大小从相似度队列中选取至少一个语义相似度,将语义相似度所对应节点所在的子图谱作为事务信息查询结果,可以确保事务信息查询结果的准确性。
在本实施例的一些可选的实现方式中,事务信息查询装置300还可以包括:查询识别模块、子图谱匹配模块以及结果确定模块,其中:
查询识别模块,用于识别事务查询文本中的本体及其对应的本体类型。
子图谱匹配模块,用于根据本体及其对应的本体类型,从事务知识图谱中匹配对应的子图谱。
结果确定模块,用于将匹配到的子图谱确定为事务信息查询结果。
本实施例中,识别事务查询文本中的本体及其对应的本体类型,可以直接根据本体以及本体类型在事务知识图谱中进行匹配,丰富了事务信息查询的方式。
为解决上述技术问题,本申请实施例还提供计算机设备。具体请参阅图4,图4为本实施例计算机设备基本结构框图。
所述计算机设备4包括通过系统总线相互通信连接存储器41、处理器42、网络接口43。需要指出的是,图中仅示出了具有组件41-43的计算机设备4,但是应理解的是,并不要求实施所有示出的组件,可以替代的实施更多或者更少的组件。其中,本技术领域技术人员可以理解,这里的计算机设备是一种能够按照事先设定或存储的指令,自动进行数值计算和/或信息处理的设备,其硬件包括但不限于微处理器、专用集成电路(Application Specific Integrated Circuit,ASIC)、可编程门阵列(Field-Programmable Gate Array,FPGA)、数字处理器(Digital Signal Processor,DSP)、嵌入式设备等。
所述计算机设备可以是桌上型计算机、笔记本、掌上电脑及云端服务器等计算设备。所述计算机设备可以与用户通过键盘、鼠标、遥控器、触摸板或声控设备等方式进行人机交互。
所述存储器41至少包括一种类型的计算机可读存储介质,所述计算机可读存储介质可以是非易失性,也可以是易失性,所述计算机可读存储介质包括闪存、硬盘、多媒体卡、卡型存储器(例如,SD或DX存储器等)、随机访问存储器(RAM)、静态随机访问存储器(SRAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、可编程只读存储器(PROM)、磁性存储器、磁盘、光盘等。在一些实施例中,所述存储器41可以是所述计算机设备4的内部存储单元,例如该计算机设备4的硬盘或内存。在另一些实施例中,所述存储器41也可以是所述计算机设备4的外部存储设备,例如该计算机设备4上配备的插接式硬盘,智能存储卡(Smart Media Card,SMC),安全数字(Secure Digital,SD)卡,闪存卡(Flash Card)等。当然,所述存储器41还可以既包括所述计算机设备4的内部存储单元也包括其外部存储设备。
本实施例中,所述存储器41通常用于存储安装于所述计算机设备4的操作系统和各类应用软件,例如事务信息查询方法的计算机可读指令等。此外,所述存储器41还可以用于暂时地存储已经输出或者将要输出的各类数据。所述处理器42在一些实施例中可以是中央处理器(Central Processing Unit,CPU)、控制器、微控制器、微处理器、或其他数据处理芯片。该处理器42通常用于控制所述计算机设备4的总体操作。
本实施例中,所述处理器42用于运行所述存储器41中存储的计算机可读指令或者处理数据,例如运行所述事务信息查询方法的计算机可读指令。所述网络接口43可包括无线网络接口或有线网络接口,该网络接口43通常用于在所述计算机设备4与其他电子设备之间建立通信连接。
本实施例中提供的计算机设备可以执行上述事务信息查询方法。此处事务信息查询方法可以是上述各个实施例的事务信息查询方法。
本实施例中,识别记录事务流程的事务信息文本中的事件以及事件之间的事理关系,并识别事件中的本体以及本体之间的本体关系,然后构建语义网框架,语义网框架具有丰富的语义表达能力,可以准确全面地描述事务信息;根据语义网框架构建事务知识图谱,从而以图谱的形式展示事务流程;将事务查询文本转化为表征向量,计算与事务知识图谱中各节点的语义相似度,从而确定事务查询文本在事务知识图谱中所在的位置,从而可以准确从事务知识图谱中确定事务查询结果,自动匹配出所需的事务信息,提高了事务信息获取效率。
本申请还提供了另一种实施方式,即提供一种计算机可读存储介质,所述计算机可读存储介质存储有计算机可读指令,所述计算机可读指令可被至少一个处理器执行,以使所述至少一个处理器执行如上述的事务信息查询方法的步骤。
本实施例中,识别记录事务流程的事务信息文本中的事件以及事件之间的事理关系,并识别事件中的本体以及本体之间的本体关系,然后构建语义网框架,语义网框架具有丰富的语义表达能力,可以准确全面地描述事务信息;根据语义网框架构建事务知识图谱,从而以图谱的形式展示事务流程;将事务查询文本转化为表征向量,计算与事务知识图谱中各节点的语义相似度,从而确定事务查询文本在事务知识图谱中所在的位置,从而可以准确从事务知识图谱中确定事务查询结果,自动匹配出所需的事务信息,提高了事务信息获取效率。
通过以上的实施方式的描述,本领域的技术人员可以清楚地了解到上述实施例方法可借助软件加必需的通用硬件平台的方式来实现,当然也可以通过硬件,但很多情况下前者是更佳的实施方式。基于这样的理解,本申请的技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来,该计算机软件产品存储在一个存储介质(如ROM/RAM、磁碟、光盘)中,包括若干指令用以使得一台终端设备(可以是手机,计算机,服务器,空调器,或者网络设备等)执行本申请各个实施例所述的方法。显然,以上所描述的实施例仅仅是本申请一部分实施例,而不是全部的实施例,附图中给出了本申请的较佳实施例,但并不限制本申请的专利范围。
本申请可以以许多不同的形式来实现,相反地,提供这些实施例的目的是使对本申请的公开内容的理解更加透彻全面。尽管参照前述实施例对本申请进行了详细的说明,对于本领域的技术人员来而言,其依然可以对前述各具体实施方式所记载的技术方案进行修改,或者对其中部分技术特征进行等效替换。凡是利用本申请说明书及附图内容所做的等效结构,直接或间接运用在其他相关的技术领域,均同理在本申请专利保护范围之内。

Claims (20)

  1. 一种事务信息查询方法,包括下述步骤:
    获取事务信息文本;
    识别所述事务信息文本中的事件以及各事件之间的事理关系,并识别所述事件中的本体以及各本体之间的本体关系;
    根据所述事件、所述事理关系、所述本体和所述本体关系,构建语义网框架;
    基于所述语义网框架生成所述事务信息文本所对应的事务知识图谱;
    获取事务查询文本;
    计算所述事务查询文本与所述事务知识图谱中各节点的语义相似度,并根据所述语义相似度和所述事务知识图谱确定事务信息查询结果。
  2. 根据权利要求1所述的事务信息查询方法,其中,所述识别所述事务信息文本中的事件以及各事件之间的事理关系的步骤包括:
    将所述事务信息文本拆分为多个子文本;
    对于每个子文本,根据所述子文本的TF-IDF信息确定所述子文本中的关键词;
    将所述关键词输入事件识别模型,以获取所述子文本记录的事件;
    将存在事件的子文本输入事理关系识别模型,得到各事件间的事理关系。
  3. 根据权利要求2所述的事务信息查询方法,其中,所述识别所述事件中的本体以及各本体之间的本体关系的步骤包括:
    根据实体识别算法对事件所在的子文本进行处理,得到所述子文本中的本体以及对应的本体类型;
    基于所述本体类型以及预设的本体类型关系表,确定各本体之间的本体关系。
  4. 根据权利要求1所述的事务信息查询方法,其中,所述基于所述语义网框架生成所述事务信息文本所对应的事务知识图谱的步骤包括:
    根据所述语义网框架构建所述事务信息文本所对应的初始事务知识图谱;
    基于距离模型确定所述初始事务知识图谱中节点和连接边的表征向量,得到事务知识图谱。
  5. 根据权利要求1所述的事务信息查询方法,其中,所述计算所述事务查询文本与所述事务知识图谱中各节点的语义相似度,并根据所述语义相似度和所述事务知识图谱确定事务信息查询结果的步骤包括:
    对所述事务查询文本进行分词处理,得到多个子词;
    将各子词分别映射为词向量;
    通过长短期记忆网络对所述词向量进行编码,得到所述事务查询文本的文本表征向量;
    计算所述文本表征向量与所述事务知识图谱中各节点的表征向量之间的语义相似度;
    根据所述语义相似度,在所述事务知识图谱中确定事务信息查询结果,并将所述事务信息查询结果返回至查询终端。
  6. 根据权利要求5所述的事务信息查询方法,其中,所述根据所述语义相似度,在所述事务知识图谱中确定事务信息查询结果的步骤包括:
    对各语义相似度进行降序排列,得到相似度序列;
    从所述相似度序列中选取至少一个语义相似度;
    在所述事务知识图谱中,确定所述语义相似度所对应节点所在的子图谱;
    将所述子图谱确定为事务信息查询结果。
  7. 根据权利要求1所述的事务信息查询方法,其中,在所述获取事务查询文本的步骤之后,还包括:
    识别所述事务查询文本中的本体及其对应的本体类型;
    根据所述本体及其对应的本体类型,从所述事务知识图谱中匹配对应的子图谱;
    将匹配到的子图谱确定为事务信息查询结果。
  8. 一种事务信息查询装置,包括:
    文本获取模块,用于获取事务信息文本;
    文本识别模块,用于识别所述事务信息文本中的事件以及各事件之间的事理关系,并识别所述事件中的本体以及各本体之间的本体关系;
    框架构建模块,用于根据所述事件、所述事理关系、所述本体和所述本体关系,构建语义网框架;
    图谱生成模块,用于基于所述语义网框架生成所述事务信息文本所对应的事务知识图谱;
    查询获取模块,用于获取事务查询文本;
    事务查询模块,用于计算所述事务查询文本与所述事务知识图谱中各节点的语义相似度,并根据所述语义相似度和所述事务知识图谱确定事务信息查询结果。
  9. 一种计算机设备,包括存储器和处理器,所述存储器中存储有计算机可读指令,所述处理器执行所述计算机可读指令时实现如下步骤:
    获取事务信息文本;
    识别所述事务信息文本中的事件以及各事件之间的事理关系,并识别所述事件中的本体以及各本体之间的本体关系;
    根据所述事件、所述事理关系、所述本体和所述本体关系,构建语义网框架;
    基于所述语义网框架生成所述事务信息文本所对应的事务知识图谱;
    获取事务查询文本;
    计算所述事务查询文本与所述事务知识图谱中各节点的语义相似度,并根据所述语义相似度和所述事务知识图谱确定事务信息查询结果。
  10. 根据权利要求9所述的计算机设备,其中,所述识别所述事务信息文本中的事件以及各事件之间的事理关系的步骤包括:
    将所述事务信息文本拆分为多个子文本;
    对于每个子文本,根据所述子文本的TF-IDF信息确定所述子文本中的关键词;
    将所述关键词输入事件识别模型,以获取所述子文本记录的事件;
    将存在事件的子文本输入事理关系识别模型,得到各事件间的事理关系。
  11. 根据权利要求10所述的计算机设备,其中,所述识别所述事件中的本体以及各本体之间的本体关系的步骤包括:
    根据实体识别算法对事件所在的子文本进行处理,得到所述子文本中的本体以及对应的本体类型;
    基于所述本体类型以及预设的本体类型关系表,确定各本体之间的本体关系。
  12. 根据权利要求9所述的计算机设备,其中,所述基于所述语义网框架生成所述事务信息文本所对应的事务知识图谱的步骤包括:
    根据所述语义网框架构建所述事务信息文本所对应的初始事务知识图谱;
    基于距离模型确定所述初始事务知识图谱中节点和连接边的表征向量,得到事务知识图谱。
  13. 根据权利要求9所述的计算机设备,其中,所述计算所述事务查询文本与所述事务知识图谱中各节点的语义相似度,并根据所述语义相似度和所述事务知识图谱确定事务信息查询结果的步骤包括:
    对所述事务查询文本进行分词处理,得到多个子词;
    将各子词分别映射为词向量;
    通过长短期记忆网络对所述词向量进行编码,得到所述事务查询文本的文本表征向量;
    计算所述文本表征向量与所述事务知识图谱中各节点的表征向量之间的语义相似度;
    根据所述语义相似度,在所述事务知识图谱中确定事务信息查询结果,并将所述事务信息查询结果返回至查询终端。
  14. 根据权利要求13所述的计算机设备,其中,所述根据所述语义相似度,在所述事 务知识图谱中确定事务信息查询结果的步骤包括:
    对各语义相似度进行降序排列,得到相似度序列;
    从所述相似度序列中选取至少一个语义相似度;
    在所述事务知识图谱中,确定所述语义相似度所对应节点所在的子图谱;
    将所述子图谱确定为事务信息查询结果。
  15. 一种计算机可读存储介质,所述计算机可读存储介质上存储有计算机可读指令;所述计算机可读指令被处理器执行时实现如下步骤:
    获取事务信息文本;
    识别所述事务信息文本中的事件以及各事件之间的事理关系,并识别所述事件中的本体以及各本体之间的本体关系;
    根据所述事件、所述事理关系、所述本体和所述本体关系,构建语义网框架;
    基于所述语义网框架生成所述事务信息文本所对应的事务知识图谱;
    获取事务查询文本;
    计算所述事务查询文本与所述事务知识图谱中各节点的语义相似度,并根据所述语义相似度和所述事务知识图谱确定事务信息查询结果。
  16. 根据权利要求15所述的计算机可读存储介质,其中,所述识别所述事务信息文本中的事件以及各事件之间的事理关系的步骤包括:
    将所述事务信息文本拆分为多个子文本;
    对于每个子文本,根据所述子文本的TF-IDF信息确定所述子文本中的关键词;
    将所述关键词输入事件识别模型,以获取所述子文本记录的事件;
    将存在事件的子文本输入事理关系识别模型,得到各事件间的事理关系。
  17. 根据权利要求16所述的计算机可读存储介质,其中,所述识别所述事件中的本体以及各本体之间的本体关系的步骤包括:
    根据实体识别算法对事件所在的子文本进行处理,得到所述子文本中的本体以及对应的本体类型;
    基于所述本体类型以及预设的本体类型关系表,确定各本体之间的本体关系。
  18. 根据权利要求15所述的计算机可读存储介质,其中,所述基于所述语义网框架生成所述事务信息文本所对应的事务知识图谱的步骤包括:
    根据所述语义网框架构建所述事务信息文本所对应的初始事务知识图谱;
    基于距离模型确定所述初始事务知识图谱中节点和连接边的表征向量,得到事务知识图谱。
  19. 根据权利要求15所述的计算机可读存储介质,其中,所述计算所述事务查询文本与所述事务知识图谱中各节点的语义相似度,并根据所述语义相似度和所述事务知识图谱确定事务信息查询结果的步骤包括:
    对所述事务查询文本进行分词处理,得到多个子词;
    将各子词分别映射为词向量;
    通过长短期记忆网络对所述词向量进行编码,得到所述事务查询文本的文本表征向量;
    计算所述文本表征向量与所述事务知识图谱中各节点的表征向量之间的语义相似度;
    根据所述语义相似度,在所述事务知识图谱中确定事务信息查询结果,并将所述事务信息查询结果返回至查询终端。
  20. 根据权利要求19所述的计算机可读存储介质,其中,所述根据所述语义相似度,在所述事务知识图谱中确定事务信息查询结果的步骤包括:
    对各语义相似度进行降序排列,得到相似度序列;
    从所述相似度序列中选取至少一个语义相似度;
    在所述事务知识图谱中,确定所述语义相似度所对应节点所在的子图谱;
    将所述子图谱确定为事务信息查询结果。
PCT/CN2022/089316 2022-01-11 2022-04-26 事务信息查询方法、装置、计算机设备及存储介质 WO2023134057A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210028520.8 2022-01-11
CN202210028520.8A CN114357117A (zh) 2022-01-11 2022-01-11 事务信息查询方法、装置、计算机设备及存储介质

Publications (1)

Publication Number Publication Date
WO2023134057A1 true WO2023134057A1 (zh) 2023-07-20

Family

ID=81109588

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/089316 WO2023134057A1 (zh) 2022-01-11 2022-04-26 事务信息查询方法、装置、计算机设备及存储介质

Country Status (2)

Country Link
CN (1) CN114357117A (zh)
WO (1) WO2023134057A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116991977A (zh) * 2023-09-25 2023-11-03 成都不烦智能科技有限责任公司 一种基于大语言模型的领域向量知识精准检索方法及装置
CN117151429A (zh) * 2023-10-27 2023-12-01 中电科大数据研究院有限公司 一种基于知识图谱的政务服务流程编排方法及装置
CN117332282A (zh) * 2023-11-29 2024-01-02 之江实验室 一种基于知识图谱的事件匹配的方法及装置

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114357117A (zh) * 2022-01-11 2022-04-15 平安科技(深圳)有限公司 事务信息查询方法、装置、计算机设备及存储介质
CN116561288B (zh) * 2023-07-12 2024-01-05 腾讯科技(深圳)有限公司 事件查询方法、装置、计算机设备、存储介质及程序产品

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200293517A1 (en) * 2019-02-06 2020-09-17 Parsons Corporation Goal-directed semantic search
CN112559762A (zh) * 2020-12-09 2021-03-26 中电科新型智慧城市研究院有限公司 一种公共安全知识图谱构建方法、装置及可读存储介质
CN113722611A (zh) * 2021-08-23 2021-11-30 讯飞智元信息科技有限公司 政务服务的推荐方法、装置、设备及计算机可读存储介质
CN114357117A (zh) * 2022-01-11 2022-04-15 平安科技(深圳)有限公司 事务信息查询方法、装置、计算机设备及存储介质

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200293517A1 (en) * 2019-02-06 2020-09-17 Parsons Corporation Goal-directed semantic search
CN112559762A (zh) * 2020-12-09 2021-03-26 中电科新型智慧城市研究院有限公司 一种公共安全知识图谱构建方法、装置及可读存储介质
CN113722611A (zh) * 2021-08-23 2021-11-30 讯飞智元信息科技有限公司 政务服务的推荐方法、装置、设备及计算机可读存储介质
CN114357117A (zh) * 2022-01-11 2022-04-15 平安科技(深圳)有限公司 事务信息查询方法、装置、计算机设备及存储介质

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116991977A (zh) * 2023-09-25 2023-11-03 成都不烦智能科技有限责任公司 一种基于大语言模型的领域向量知识精准检索方法及装置
CN116991977B (zh) * 2023-09-25 2023-12-05 成都不烦智能科技有限责任公司 一种基于大语言模型的领域向量知识精准检索方法及装置
CN117151429A (zh) * 2023-10-27 2023-12-01 中电科大数据研究院有限公司 一种基于知识图谱的政务服务流程编排方法及装置
CN117151429B (zh) * 2023-10-27 2024-01-26 中电科大数据研究院有限公司 一种基于知识图谱的政务服务流程编排方法及装置
CN117332282A (zh) * 2023-11-29 2024-01-02 之江实验室 一种基于知识图谱的事件匹配的方法及装置
CN117332282B (zh) * 2023-11-29 2024-03-08 之江实验室 一种基于知识图谱的事件匹配的方法及装置

Also Published As

Publication number Publication date
CN114357117A (zh) 2022-04-15

Similar Documents

Publication Publication Date Title
WO2023134057A1 (zh) 事务信息查询方法、装置、计算机设备及存储介质
CN107735804B (zh) 用于不同标记集合的转移学习技术的系统和方法
US11847113B2 (en) Method and system for supporting inductive reasoning queries over multi-modal data from relational databases
CN113610239B (zh) 针对机器学习的特征处理方法及特征处理系统
US11580119B2 (en) System and method for automatic persona generation using small text components
US11361030B2 (en) Positive/negative facet identification in similar documents to search context
WO2021135439A1 (zh) 基于语音语义的信息检索方法、及其相关设备
CN112231569A (zh) 新闻推荐方法、装置、计算机设备及存储介质
US11573994B2 (en) Encoding entity representations for cross-document coreference
WO2022105119A1 (zh) 意图识别模型的训练语料生成方法及其相关设备
WO2021063089A1 (zh) 规则匹配方法、规则匹配装置、存储介质及电子设备
US11361031B2 (en) Dynamic linguistic assessment and measurement
JP2023517518A (ja) ヌル値又は同等の値を有するリレーショナル・テーブルのためのベクトル埋込モデル
CN112598039A (zh) 获取nlp分类领域阳性样本方法及相关设备
CN111209403A (zh) 数据处理方法、装置、介质及电子设备
WO2022073341A1 (zh) 基于语音语义的疾病实体匹配方法、装置及计算机设备
CN115238009A (zh) 基于血缘分析的元数据管理方法、装置、设备及存储介质
US11664998B2 (en) Intelligent hashing of sensitive information
US20230267277A1 (en) Systems and methods for using document activity logs to train machine-learned models for determining document relevance
CN111597453B (zh) 用户画像方法、装置、计算机设备及计算机可读存储介质
CN114281990A (zh) 文档分类方法及装置、电子设备和介质
CN112507170A (zh) 基于智能决策的数据资产目录构建方法、及其相关设备
JP2024518458A (ja) テキスト内の自動トピック検出のシステム及び方法
US11836176B2 (en) System and method for automatic profile segmentation using small text variations
CN117312535A (zh) 基于人工智能的问题数据处理方法、装置、设备及介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22919699

Country of ref document: EP

Kind code of ref document: A1