CN117421390A - Target group determination method, system, device and storage medium - Google Patents
Target group determination method, system, device and storage medium Download PDFInfo
- Publication number
- CN117421390A CN117421390A CN202311337060.8A CN202311337060A CN117421390A CN 117421390 A CN117421390 A CN 117421390A CN 202311337060 A CN202311337060 A CN 202311337060A CN 117421390 A CN117421390 A CN 117421390A
- Authority
- CN
- China
- Prior art keywords
- text
- entity
- bitmap
- target group
- relation
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 83
- 230000006870 function Effects 0.000 claims abstract description 71
- 238000004364 calculation method Methods 0.000 claims abstract description 30
- 238000012545 processing Methods 0.000 claims abstract description 24
- 230000003993 interaction Effects 0.000 claims abstract description 18
- 239000011159 matrix material Substances 0.000 claims description 20
- 238000004422 calculation algorithm Methods 0.000 claims description 18
- 238000012512 characterization method Methods 0.000 claims description 14
- 238000005457 optimization Methods 0.000 claims description 7
- 230000004913 activation Effects 0.000 claims description 3
- 238000012795 verification Methods 0.000 claims description 3
- 230000009286 beneficial effect Effects 0.000 abstract description 6
- 230000008569 process Effects 0.000 description 11
- 238000010586 diagram Methods 0.000 description 10
- 238000012216 screening Methods 0.000 description 9
- 238000006243 chemical reaction Methods 0.000 description 8
- 238000000605 extraction Methods 0.000 description 8
- 238000012937 correction Methods 0.000 description 4
- 238000012986 modification Methods 0.000 description 4
- 230000004048 modification Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000006467 substitution reaction Methods 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000003491 array Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000009191 jumping Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 230000003213 activating effect Effects 0.000 description 1
- 238000007792 addition Methods 0.000 description 1
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/22—Indexing; Data structures therefor; Storage structures
- G06F16/2228—Indexing structures
- G06F16/2237—Vectors, bitmaps or matrices
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/22—Indexing; Data structures therefor; Storage structures
- G06F16/2228—Indexing structures
- G06F16/2246—Trees, e.g. B+trees
-
- 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
- G06F16/31—Indexing; Data structures therefor; Storage structures
- G06F16/316—Indexing structures
- G06F16/322—Trees
-
- 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
- G06F16/35—Clustering; Classification
-
- 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
- G06F40/10—Text processing
- G06F40/12—Use of codes for handling textual entities
- G06F40/151—Transformation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
- G06F40/295—Named entity recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The invention discloses a target group determining method, a target group determining system, a target group determining device and a target group determining storage medium. The method comprises the following steps: acquiring a first text; processing the first text through a natural language model to obtain a relation triplet; the natural language model is used for carrying out unified representation and interaction on the entities and relations involved in the first text, and carries out prediction learning through an indication function, wherein the indication function comprises a first function of a host entity and a guest entity and a second function of the guest entity and the host entity; converting the relation triplet into a binary tree, and generating bitmap script statements according to the binary tree; and extracting a group list of bitmap script sentences through a bitmap calculation engine to obtain a target group. The embodiment of the invention is beneficial to relieving the dependence of label determination on the skill of an operator; the accuracy of natural language model output is improved, and the accuracy of target group determination is further improved; can be widely applied to the technical field of big data.
Description
Technical Field
The invention relates to the technical field of big data, in particular to a target group determining method, a target group determining system, a target group determining device and a target group determining storage medium.
Background
Along with the development of big data technology, in order to master the business changes of different clients and accurately position newly-added clients, enterprises better know the demands of the clients, need to construct data labels of client data, comprehensively perform client service and prevent clients from losing. In the use stage, by combining the multidimensional labels, a subdivided target customer group can be quickly selected. Therefore, constructing the client tag is one of the most effective modes in the accurate marketing of the big data age, and the client group selection is rapidly and effectively carried out, so that the method has great significance in improving the accurate marketing efficiency and realizing the cost reduction and efficiency enhancement of enterprise marketing.
Due to the combined action of the efficient support of computer hardware, parallel processing capability, hardware optimization, and the data type for which the bit operation is directed, the computer is able to perform the bit operation in a very efficient manner. Meanwhile, the bitmap is high in storage efficiency and easy to process and edit, so that screening of target client groups is realized based on the bitmap, and the method becomes a mainstream mode of screening of the current target client groups. In the related art, when screening a target group through a bitmap, an operator is required to accurately give out label information of the group to be screened and a logical relationship between labels. The process requires an operator to have certain relevant basic knowledge; meanwhile, manual screening is needed for the label information, and the label information is easy to make mistakes by relying on manpower, so that the label screening accuracy is low.
Disclosure of Invention
The present invention aims to solve at least one of the technical problems existing in the prior art to a certain extent.
Therefore, the invention aims to provide an efficient target group determination method, system, device and storage medium.
In order to achieve the technical purpose, the technical scheme adopted by the embodiment of the invention comprises the following steps:
in one aspect, an embodiment of the present invention provides a target group determining method, including the following steps:
the target group determining method of the embodiment of the invention comprises the following steps: acquiring a first text; the first text comprises label information of a target group and logic relation information among labels; processing the first text through a natural language model to obtain a relation triplet; the natural language model is used for carrying out unified representation and interaction on the entities and relations involved in the first text, the natural language model carries out prediction learning through an indication function, and the indication function comprises a first function of a host entity and a guest entity and a second function of the guest entity and the host entity; converting the relation triplet into a binary tree, and generating a bitmap script statement according to the binary tree; and extracting a group list from the bitmap script statement through a bitmap calculation engine to obtain a target group. According to the embodiment of the invention, the relation triples in the first text are obtained by processing the first text given by an operator; the selection of the tags by the operator is not required, and the operator is not required to determine the logical relationship between the tags. The embodiment of the invention is beneficial to relieving the dependence of label determination on the skill of an operator. Meanwhile, according to the embodiment of the invention, the first function and the second function are established in the natural language model, so that the relative relation between the host entity and the guest entity is distinguished, the accuracy of the output of the natural language model is improved, and the accuracy of the determination of the target group is further improved.
In addition, the target group determination method according to the above embodiment of the present invention may further have the following additional technical features:
further, in the target group determining method according to the embodiment of the present invention, the processing the first text through a natural language model to obtain a relationship triplet includes:
splicing the first text, the logic triples and the sub-logic triples to obtain an input sequence; the logical triples are used for representing the logical relations among the host entity, the guest entity and the relation; the sub-logical triples comprise a first logical relationship between a host entity and a guest entity, a second logical relationship between the guest entity and the host entity, a third logical relationship between the entities and the relationship, and a fourth logical relationship between the entities;
performing characterization learning processing on the input sequence through a natural language model to obtain a characterization matrix; each position of the characterization matrix is used for characterizing a defined relationship;
determining an indication function according to the characterization matrix; and predicting the result according to the indication function to obtain a relation triplet.
Further, in one embodiment of the present invention, the natural language model comprises a bert model, and the method further comprises:
Adding the Q matrix and the K matrix of the last layer in the bert model to obtain an interaction table;
processing the interaction table through an activation function to obtain a relation triplet;
and optimizing the relation triplet through an optimization function to obtain an optimized relation triplet.
Further, in one embodiment of the present invention, the method further comprises the steps of:
performing host-guest identification processing on the first text to obtain an entity and a relationship;
verifying the entity involved in the first text through a greedy algorithm of a regular expression; and if the verification result is not passed, correcting the first text to obtain a corrected first text.
Further, in one embodiment of the present invention, the step of converting the relationship triplet into a binary tree includes:
traversing each triplet in the relation triples to generate a corresponding partial binary tree;
and inserting the partial binary tree into the existing binary tree to obtain a converted binary tree.
Further, in one embodiment of the present invention, the traversing each of the relationship triples generates a corresponding partial binary tree comprising:
Traversing the relation triples to obtain a first triples;
taking the relation in the first triplet as a root node of a partial binary tree;
if the first main entity of the first triplet is not null, using the first main entity as a left child node of the root node;
or if the first guest entity of the first triplet is not null, using the first guest entity as the right child node of the root node, and generating a partial binary tree.
Further, in one embodiment of the present invention, the method further comprises:
establishing a group label architecture, wherein the group label architecture comprises a label data layer, a bitmap layer and a label inquiry application layer;
the label data layer is used for storing label information of a group, the bitmap layer is used for establishing a bitmap according to the label information, and the label inquiry application layer is used for inquiring a target group according to the first text.
In another aspect, an embodiment of the present invention provides a target group determining system, including:
the first module is used for acquiring a first text; the first text comprises label information of a target group and logic relation information among labels;
the second module is used for processing the first text through a natural language model to obtain a relation triplet; the natural language model is used for carrying out unified representation and interaction on the entities and relations involved in the first text, the natural language model carries out prediction learning through an indication function, and the indication function comprises a first function of a host entity and a guest entity and a second function of the guest entity and the host entity;
The third module is used for converting the relation triplet into a binary tree and generating a bitmap script statement according to the binary tree;
and the fourth module is used for extracting the group list of the bitmap script statement through a bitmap calculation engine to obtain a target group.
In another aspect, an embodiment of the present invention provides a target group determining apparatus, including:
at least one processor;
at least one memory for storing at least one program;
the at least one program, when executed by the at least one processor, causes the at least one processor to implement the target population determination method described above.
In another aspect, an embodiment of the present invention provides a storage medium in which a processor-executable program is stored, which when executed by a processor is configured to implement the above-described target population determination method.
The target group determining method provided by the embodiment of the invention comprises the following steps: acquiring a first text; the first text comprises label information of a target group and logic relation information among labels; processing the first text through a natural language model to obtain a relation triplet; the natural language model is used for carrying out unified representation and interaction on the entities and relations involved in the first text, the natural language model carries out prediction learning through an indication function, and the indication function comprises a first function of a host entity and a guest entity and a second function of the guest entity and the host entity; converting the relation triplet into a binary tree, and generating a bitmap script statement according to the binary tree; and extracting a group list from the bitmap script statement through a bitmap calculation engine to obtain a target group. According to the embodiment of the invention, the relation triples in the first text are obtained by processing the first text given by an operator; the selection of the tags by the operator is not required, and the operator is not required to determine the logical relationship between the tags. The embodiment of the invention is beneficial to relieving the dependence of label determination on the skill of an operator. Meanwhile, according to the embodiment of the invention, the first function and the second function are established in the natural language model, so that the relative relation between the host entity and the guest entity is distinguished, the accuracy of the output of the natural language model is improved, and the accuracy of the determination of the target group is further improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following description is made with reference to the accompanying drawings of the embodiments of the present invention or the related technical solutions in the prior art, and it should be understood that the drawings in the following description are only for convenience and clarity of describing some embodiments in the technical solutions of the present invention, and other drawings may be obtained according to these drawings without the need of inventive labor for those skilled in the art.
FIG. 1 is a schematic flow chart of an embodiment of a target group determination method provided by the present invention;
FIG. 2 is a schematic diagram illustrating a group tag architecture according to an embodiment of the present invention;
FIG. 3 is a flow chart of one embodiment of a method for determining a customer group based on a bitmap calculation engine according to the present invention;
FIG. 4 is a schematic diagram illustrating one embodiment of a client group bitmap storage based on a bitmap calculation engine according to the present invention;
FIG. 5 is a flow diagram of one embodiment of a bitmap calculation engine-based customer group selection provided by the present invention;
FIG. 6 is a schematic diagram of one embodiment of a binary tree provided by the present invention;
FIG. 7 is a schematic diagram of one embodiment of a multi-labeled binary tree provided by the present invention;
FIG. 8 is a flow chart of an embodiment of a natural language model process provided by the present invention;
FIG. 9 is a schematic diagram illustrating a comparison of a flow of an embodiment of the present application and an embodiment of the related art provided by the present invention;
FIG. 10 is a flow chart of another embodiment of a target group determination method provided by the present invention;
FIG. 11 is a schematic diagram of an embodiment of a target group determination system provided by the present invention;
fig. 12 is a schematic structural diagram of an embodiment of the target group determining device provided by the invention.
Detailed Description
Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to like or similar elements or elements having like or similar functions throughout. The embodiments described below by referring to the drawings are illustrative only and are not to be construed as limiting the invention. The step numbers in the following embodiments are set for convenience of illustration only, and the order between the steps is not limited in any way, and the execution order of the steps in the embodiments may be adaptively adjusted according to the understanding of those skilled in the art.
First, terms involved in the embodiments of the present application will be explained:
bitmap calculation engine: bitMap is a computational engine dedicated to processing bitmaps. It provides a series of functions and algorithms for performing various operations and calculations on the bitmap. Some common bitmap functions include bitmap-body transformations, bitmap logical operations, bitmap value additions, modifications, deletions, and the like. Common bitmap computation engines are ClickHouse, polarDB, postgreSQL, redis, etc. In the embodiment of the invention, clickHouse is adopted as a BitMap calculation engine. It should be noted that, instead of using clickHouse as the BitMap calculation engine, the person skilled in the art may select the BitMap calculation engine according to the actual requirement, and may use the calculation engine in the related technical solutions, models, etc. mentioned in the embodiments of the present invention.
Illustratively, characteristics of a clickHouse include:
and (5) column type storage: the ClickHouse uses a columnar storage structure, so that the query performance can be greatly improved, and particularly in a scene of needing to quickly query a large amount of data.
High performance: clickHouse has high performance query and insert speeds and can support millions or even tens of millions of query and insert operations.
Scalability: the clickHouse may achieve higher performance and reliability through horizontal expansion, and may expand data storage capacity without affecting existing performance.
Support SQL: the clickHouse supports a standard SQL query language that can be integrated with existing SQL tools and applications.
And (3) open source: the ClickHouse is an open source item, has higher customization and expandability, and can be customized and expanded according to different application scenes and requirements.
The ClickHouse is suitable for scenes which need to process a large amount of data and carry out complex analysis, such as the fields of data warehouse, log analysis, advertisement delivery, internet of things and the like. Because of its high performance and scalability, clickHouse has found wide application in big data fields.
Along with the development of big data and other technologies, enterprises need to construct data labels of client data to comprehensively perform client service in order to master service changes of different clients and accurately position newly added clients, so that the demands of the clients are better known, and the clients are prevented from losing. In the use stage, by combining the multidimensional labels, a subdivided target customer group can be quickly selected. Therefore, constructing the client tag is one of the most effective modes in the accurate marketing of the big data age, and the client group selection is rapidly and effectively carried out, so that the method has great significance in improving the accurate marketing efficiency and realizing the cost reduction and efficiency enhancement of enterprise marketing.
Due to the combined action of the efficient support of computer hardware, parallel processing capability, hardware optimization, and the data type for which the bit operation is directed, the computer is able to perform the bit operation in a very efficient manner. Meanwhile, the bitmap is high in storage efficiency and easy to process and edit, so that screening of target client groups is realized based on the bitmap, and the method becomes a mainstream mode of screening of the current target client groups.
In the application scene of selecting the customer group according to the labels, in order to screen out all the customer groups conforming to the specific label characteristics, a bitmap is required to be established for the customer group corresponding to each label, and when the target group is required to be screened, only the designated bitmap is required to be selected to carry out bitwise logical operation according to the logical relationship. Common bitwise logical operations include and, or, not, exclusive or, nor, or, and the like. In addition, when the bit map is logically related, the logical relationship operation mainly involves the collective operations such as AND, OR, NOT and the like. Because the bitmap represents a client set represented by a label, logical operations among the sets can be converted into bitwise logical operations of the bitmap, and the performance of target group selection is improved to a great extent. The method comprises the steps of converting a target searching customer group into logical operation among sets, designing an operation interface by the application system in a conventional mode, manually selecting a label group needing to be selected by a customer on the interface, manually selecting a logical relation among labels, submitting the label, converting the logical operation into bit operation by the application system according to submitted tasks, and finally finding out the target group for output. However, the operation of the set involves partial mathematical knowledge, strict logic thinking is needed between logic calculation, and certain misoperation is possible for the manual operation of a client; meanwhile, the complex operation interface can improve the use threshold of the client, and greatly reduce the use experience of the client. For this reason, according to the embodiment of the invention, an NLP model is designed, an operator only needs to input the text description of the target group to be selected, the model automatically identifies the labels contained in the target group and the logical relationship among the labels, the labels are converted into corresponding bitmap calculation in the background, and the target group is output after the calculation is completed.
The target group determining method and system according to the embodiment of the present invention will be described in detail below with reference to the accompanying drawings, and first, the target group determining method according to the embodiment of the present invention will be described with reference to the accompanying drawings.
Referring to fig. 1, a target group determining method is provided in an embodiment of the present invention, and the target group determining method in the embodiment of the present invention may be applied to a terminal, a server, or software running in the terminal or the server. The terminal may be, but is not limited to, a tablet computer, a notebook computer, a desktop computer, etc. The server may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs, basic cloud computing services such as big data and artificial intelligent platforms. The target group determination method in the embodiment of the invention mainly comprises the following steps:
s100: acquiring a first text; the first text comprises label information of the target group and logic relation information among labels;
S200: processing the first text through a natural language model to obtain a relation triplet; the natural language model is used for carrying out unified representation and interaction on the entities and relations involved in the first text, and carries out prediction learning through an indication function, wherein the indication function comprises a first function of a host entity and a guest entity and a second function of the guest entity and the host entity;
s300: converting the relation triplet into a binary tree, and generating bitmap script statements according to the binary tree;
s400: and extracting a group list of bitmap script sentences through a bitmap calculation engine to obtain a target group.
In some possible implementations, as shown in fig. 2, the target group determining method provided in the embodiments of the present application may be implemented based on the architecture shown in fig. 2, and the client tag system based on the bitmap calculation engine is divided into three parts, that is, a tag data table, a bitmap, and a tag query application. Based on the three parts, the target group is queried through the bitmap by converting the label into the bitmap. Specifically, the structure of the tag data table may be as shown in table 1, and the structure of the bit map may be as shown in table 2.
| Fields | Type(s) | Meaning of |
| userid | Int | Customer id |
| tag_code | String | Label coding |
| tag_value | String | Tag value |
| period | String | Accounting period |
| event_time | DateTime | Event time |
TABLE 1
| Fields | Type(s) | Meaning of |
| tag_code | String | Label coding |
| tag_value | String | Tag value |
| period | String | Accounting period |
| userids | GroupBitmap | Client list bitmap |
| event_time | DateTime | Event time |
TABLE 2
According to the examples given in tables 1 and 2, it can be seen that the tag data table stores the original data of the client tag, the bitmap table is calculated according to the tag data table, and when the bitmap is generated, the bitmap can be obtained by only grouping the tag table according to the tag code, the tag value and the accounting period and then performing aggregation conversion on the client ID.
It will be appreciated that the creation of the bit map provides efficient computation for multi-dimensional tag selection. A bitmap is a data structure that is associated with certain specific values by array indices. In the bitmap, each element occupies 1 bit. When the bit is 1, it indicates that the corresponding element has the specific value. Otherwise, it indicates no. When the bitmaps are built, the client group can be selected by multi-label combination through bitwise logical operation among bitmaps. The specific steps are shown in fig. 3.
Illustratively, as shown in FIG. 4, assuming that client group C1 all possesses tag A, whose set of IDs is [0,1,4,5, 13, 14], and client group C2 all possesses tag B, whose set of IDs is [1,2,3,4, 13, 15], client groups C1 and C2 may be represented in the bitmap as B1 and B2. At this time, if the customer who wants to extract the label a and simultaneously satisfies the label B, the following steps are required to be performed to complete the label selection:
S41: converting the requirements of customers into label combination conditions and inputting the label combination conditions;
s42: finding out a bitmap B1 of the client group corresponding to the label A and a bitmap B2 of the client group corresponding to the label B;
s43: performing operation according to the logical relation in the label combination condition to obtain a bitmap B3 (shown in FIG. 5) of the target client group;
s44: the bitmap B3 is converted into a client ID list and output, and the conversion process can be regarded as the inverse process in fig. 4.
Thus, the memory space occupation of the label bitmap is very small, and the speed of label calculation is very high.
In the process of step S41, a label combination condition is input, which has a complexity of: logical representations between tags and transitions between customer intent and tag computation logic. In general, a set of simple logical relationships can be represented by a triplet, namely: (subject, object, symbol). Wherein: the object represents a subject, the object represents an object, and symbol represents a logical relationship between the subject and the object. In the context of tag lookups, both objects refer to a particular tag, or to the result of multiple tag logical operations. symbol represents a logical relationship between the host and the object, and may be represented by keywords, as shown in table 3, where a represents a subject and B represents an object.
| symbol | Meaning of | Label meaning |
| ∩ | Intersection set | Customer base meeting both label A and label B |
| ∪ | Union set | Customer base satisfying A or B |
| - | Difference set | Customer base meeting a but not meeting B |
| △ | Difference of symmetry | Non-overlapping customer base among A and B |
TABLE 3 Table 3
It will be appreciated that converting a triplet to a bitmap calculation is a relatively simple process, since the triplet may be represented by a simple binary tree, as shown in fig. 6, which shows the logical relationship between tag a and tag B. By means of binary tree, the mixed logic calculation among a plurality of tags can be simply described, and as shown in fig. 7, the meaning of the logic relationship is: clients that satisfy both A and B but do not satisfy C or that are not in E and do not overlap F in D are found.
By representing a triplet as a binary tree structure, the logical computation of the bitmap between ClickHouse can be conveniently converted, as shown in FIG. 6, the leaf nodes are labels, and the non-leaf nodes are logical relations. A simple logical relational operation of a triplet to a bitmap in a clickHouse may be expressed as the following formulas (1) to (4):
subject∩object＝>bitmapAnd(subject,object) (1)
subject∪object＝>bitmapOr(subject,object) (2)
subject-object＝>bitmapAndNot(subject,object) (3)
subjectΔobject＝>bitmapAndNot(bitmapOr(subject,object),bitmapAnd(subject,object)) (4)
the embodiment of the application can be conveniently converted into bitmap logic calculation between ClickHouse by representing one triplet as a binary tree structure. After the conversion of the key logic is completed, only the front and rear sentences are supplemented when the sql sentences of the clickhouse are executed, so that the inquiry can be completed.
And how to translate the customer's needs into triples that can represent logical relationships becomes critical. The common mode is that a plurality of labels are selected by a drop-down box in an application system, logic relations among the labels are designated, the conversion of triples is completed through the rear end of the system, the triples are converted into clickhouse execution sentences through a recursion algorithm, and then target group circle selection and output are completed.
However, this approach has the following 3 drawbacks:
(1) Manually selecting the labels, wherein the labels are easy to be selected by mistake due to the fact that names of the labels are similar;
(2) When a logic relationship is designated for a tag, the logic relationship is easy to be selected by mistake due to the complexity and similarity of the logic relationship;
(3) The clients using the system need to have professional collective operation knowledge and use a certain threshold.
In this regard, the embodiment of the invention provides a target group determining method for implementing target group circling according to client labels. Specifically, a bitmap condition generating method based on NLP algorithm improvement is provided. The industry has a theory of a relation extraction method for relation extraction, and the core idea of the model is shown in fig. 8: in the task of extracting a triplet (object, symbol) in a relation extraction model, the object is represented by s, the object is represented by o, symbol is represented by r because of the conforming description of the relation, the entity and the relation are characterized and interacted together, the extraction mode still adopts the current mainstream method based on table filtering, and the three interaction relation predictions are only split into relation (s, o) between the entities, relation (s|o, r) between the entity and the relation type and relation (r, s|o) between the relation type and the entity when predicting, respectively. Based on this, the embodiment of the application adds a relationship prediction on the basis of fig. 8, and divides the relationship between entities into the relationship between the host entity and the guest entity, and the relationship between the guest entity and the host entity. According to the embodiment of the application, the recognition capability of the model in a common logic relationship scene is improved by increasing the recognition of the logic relationship. Specifically, the user only needs to give out related description information of the target group, and the embodiment of the application extracts tag information from the description information through the target group determination method, generates triples according to the tag information, further generates a binary tree, and extracts the target group through the bitmap calculation engine.
Aiming at the problems that the manual operation is easy to make mistakes and the complex logic selection is not friendly to the customer experience when the target group screening is carried out by carrying out label selection on the target customer group through the BitMap calculation engine, the embodiment of the invention provides an improved NLP logic relation extraction algorithm, and provides an improved BitMap condition generation algorithm based on the algorithm, thereby realizing the determination of the target group. And automatically identifying the label main body contained in the text by utilizing the NLP and automatically judging the logic relationship among the labels, so as to quickly generate the label bitmap calculation condition. Optimizing the use experience of customer group selection, especially reducing the use threshold for target customer group screening by tags.
Optionally, in one embodiment of the present invention, processing the first text through a natural language model to obtain a relationship triplet includes:
splicing the first text, the logic triples and the sub-logic triples to obtain an input sequence; the logic triples are used for representing the logic relations among the host entity, the guest entity and the relation; the sub-logical triples comprise a first logical relationship between a host entity and a guest entity, a second logical relationship between the guest entity and the host entity, a third logical relationship between the entities and the relationship, and a fourth logical relationship between the entities;
Carrying out characterization learning processing on an input sequence through a natural language model to obtain a characterization matrix; each location of the characterization matrix is used to characterize a defined relationship;
determining an indication function according to the characterization matrix; and predicting the result according to the indication function to obtain the relation triplet.
In some possible implementations, referring to fig. 9, compared to the prior art, in the embodiment of the present application, the extracting step of the model for the relationship is:
s91: the text and the relation category are spliced together to be used as an input sequence, and the sequence can be obtained through X= { X 1 ,x 2 ,…x N Expressed by N being the length of the sequence, the extraction task is to identify all relation triples from the textWherein L is triplet input, s l ,r l ,o l The first group of corresponding main entities, relations and objects respectively. An entity set E= { E is formed by a host entity and a guest entity 1 ,e 2 ,…,e k -k is the number of entities; the relation set is r= { R 1 ,R 2 ,…,R M M is the number of relationship types.
S92: the names of all relationship types, such as the relationship "/business/company/counters" are denoted by "counters", and the "is captial of" is denoted by "captial"; and then splicing the predicted text and core words of all relation categories to form a new sequence, and inputting the new sequence into a characterization model, such as a bert model, to perform characterization learning. A representation table matrix is formed, and each position (cell) in the table matrix can represent whether a defined relationship exists (as shown in a formula (5) and a formula (6)). Because of the complexity and nestability of the logical relationships, in order to satisfy the complete recognition of the logical relationships, the embodiment of the invention adjusts the sequence input to be: text, final logical triples, individual sub-logical triples.
T＝Concat(T s ,T p ) (5)
H＝E[T] (6)
S93: judging whether each position in the table matrix has a defined relation, if yes, the position is 1, if not, the position is 0, so that the prediction task is changed into a table-filled two-class task, the model breaks the triplet relation (s, r, o) into three sub-relations to represent and predict learning, and I is defined for the relation between entities respectively e (e a ,e b ) Definition of entities and relationships I r (e, r) and I r (r, e), if considering entity relationship symmetry, I e (e a ,e b )＝I e (e b ,e a ) As shown in the formulas (7) to (9).
S94: when decoding, the triples can be analyzed according to the result of form prediction and the rule, wherein the decoding mode is to take the Q, K matrix in the last layer of multi-head attention mechanism characterized by bert to sum up to obtain an Interaction Map;
s95: the final relationship prediction result is obtained by adding a sigmoid function to the interaction table, see equation (10).
S96: in order to make the result more robust, the model uses an optimization function (as shown in formula (11)) of a two-class cross entropy, wherein I * Is the truth matrix of the interaction table:
it can be understood that when setting the indication function, since there is a certain correspondence between the logical relationship between the host and the guest, in order to enhance the identification of the relationship (i.e. the first function and the second function are used to identify and distinguish the relationship between the host entity and the guest entity in the embodiment of the present application, the first logical relationship and the second logical relationship are used to characterize the different relationship between the host entity and the guest entity in the embodiment of the present application, an indication function is added based on the existing model in the embodiment of the present application, I e (e b ,e a ) At this time, I e (e a ,e b )≠I e (e b ,e a ) The specific function is expressed as formula (12):
the logical correspondence between the hosts is shown in table 4.
| I e (e a ,e b ) | I e (e b ,e a ) |
| s∩o | o∩s |
| s∪o | o∪s |
| s-o | o ∈ (o ∈s), where a represents the complement |
| sΔo | oΔs |
TABLE 4 Table 4
Through the model optimization, when a client uses the tag system to perform target client group circle selection, a text can be input, a tag needing circle selection and a logic relation among the tags are described in the text, and the tag can be automatically converted into a binary tree of a triplet and a logic operation among bitmaps and converted into an execution statement of clickhouse to complete target group circle selection. Finally, the client and the system complete the target group selection in a text dialogue mode.
Optionally, in an embodiment of the present invention, the natural language model includes a bert model, and the method further includes:
adding the Q matrix and the K matrix of the last layer in the bert model to obtain an interaction table;
processing the interaction table through an activation function to obtain a relation triplet;
and optimizing the relation triplet through an optimization function to obtain the optimized relation triplet.
In some possible implementations, the embodiments of the present application improve the prediction accuracy of the model by activating the function and optimizing the function.
Optionally, in one embodiment of the present invention, the method further comprises:
performing host-guest identification processing on the first text to obtain an entity and a relationship;
verifying the entity involved in the first text by a greedy algorithm of the regular expression; and if the verification result is not passed, correcting the first text to obtain a corrected first text.
In some possible embodiments, after the recognition of the host and guest entities is completed, a layer of tag entity matching algorithm is added to check whether the tag input by the customer is correct, and if the error exists in the tag input by the customer, the tag can be automatically corrected. The matching algorithm may employ a greedy algorithm in a regular expression. Of course, embodiments of the present application are not limited to specific implementations of the matching algorithm.
Optionally, in one embodiment of the present invention, the step of converting the relationship triplet into a binary tree includes:
traversing each triplet in the relation triples to generate a corresponding partial binary tree;
and inserting part of the binary tree into the existing binary tree to obtain a converted binary tree.
Optionally, in one embodiment of the present invention, traversing each of the relationship triples generates a corresponding partial binary tree, including:
Traversing the relation triples to obtain a first triples;
taking the relation in the first triplet as a root node of a partial binary tree;
if the first main entity of the first triplet is not null, taking the first main entity as a left child node of the root node;
or if the first guest entity of the first triplet is not null, using the first guest entity as the right child node of the root node, and generating a partial binary tree.
In some possible implementations, the algorithm for converting the triplet into a binary tree may be performed as follows:
s21: creating an empty binary tree;
s22: traversing each triplet;
s23: for each triplet, converting it into a partial binary tree;
s231: taking the relation (symbol) as the data of the root node, and storing the root node into a binary tree;
s232: if the subject is not empty, taking the subject as a left child node of the root node and storing the left child node in the binary tree;
s233: if the object (object) is not empty, the object is taken as the right child node of the root node and the right child node is stored in the binary tree.
S24: inserting the partial binary tree into an existing binary tree at a suitable location;
s241: if the existing binary tree is empty, setting the part of binary tree as a root node;
S242: if the existing binary tree is not empty, inserting the partial binary tree into the existing binary tree at a proper position according to a certain rule (for example, selecting a left subtree or a right subtree according to the nature of the binary tree);
s25: repeating the steps S22-S24 until all the triples are processed;
s26: returning the complete binary tree.
Thus, the embodiment of the application can realize bitmap logic conversion through a simple recursive algorithm. The method comprises the following steps:
s31: starting from the root node, marking the root node as symbol, if the tree on the left side of the root node contains a plurality of nodes, converting the whole left side of the root node, and jumping to S34; otherwise, marking the node on the left as a subject;
s32: starting from the root node, if the tree on the right of the root node contains a plurality of nodes, converting the whole right of the root node, and jumping to S35; otherwise, marking the node on the right as an object;
s33: converting subject, object, symbol;
s34: repeating the steps of S31, S32 and S33 by taking the left child node as the root node until all nodes on the left have completed conversion;
s35: the steps of S31, S32, S33 are repeated with the right child node as the root node until all nodes on the right have completed the conversion.
After the conversion of the key logic is completed, only the front and rear sentences are supplemented when the sql sentences of the clickhouse are executed, so that the inquiry can be completed.
Optionally, in one embodiment of the present invention, the method further comprises:
establishing a group label architecture, wherein the group label architecture comprises a label data layer, a bitmap layer and a label inquiry application layer;
wherein the label data layer is used for storing label information of the group, the bitmap layer is used for establishing a bitmap according to the label information,
the label inquiry application layer is used for inquiring the target group according to the first text.
In some possible embodiments, referring to FIG. 2, the ClickHuse-based customer label system is split into three parts, namely a label data table (i.e., label data layer), a bitmap table (i.e., bitmap layer), and a label query application (i.e., label query application layer). The label data table is used for storing labels of clients; the bitmap is used for establishing a bitmap according to the label, so that the customer selection is facilitated; the label inquiry application is connected with an external system, so that the detailed labels of customers or the customer groups meeting the conditions can be inquired conveniently according to the logic inquiry of a plurality of label combinations.
The following describes in detail a method for determining a target group according to an embodiment of the present application by taking a specific embodiment as an example, and referring to fig. 10, the technical scheme of the method for obtaining a text input by a client, identifying an NLP logical relationship, calculating a bitmap, and converting the bitmap is shown in the left part, and the right part is a target group circling process, where the method mainly includes the following implementation steps:
S51, acquiring a text input by a customer and recording the text as data;
s52, converting the relationship triples into relationship triples through an NLP model, and marking the relationship triples as SORs;
s53, converting the triplet into a logic relation binary Tree, and marking the binary Tree as Tree;
s54, generating a ClickHouse bitmap logic calculation statement Lang through a recursion algorithm;
s55, splicing Lang into script sentences which can be executed in the ClickHouse;
s56, extracting a client list through ClickHouse calculation, and recording the client list as a UserList;
and S57, outputting the UserList to finish the task.
According to the method and the device for selecting the target group, after the target group selecting conditions are improved through the natural language model, efficiency of a client in selecting labels is improved, threshold used by the client is reduced, and the target group selecting of the client can be completed through a text dialogue mode. According to the embodiment of the application, the indication function between the object and the entity is added in the original UniRel model, so that the identification of the logical relationship is increased, and the identification capability of the model in a common logical relationship scene is improved. According to the embodiment of the application, the sequence input of the model is adjusted, the text, the final logic triples and all the sub logic triples are input together, and the extraction capability of the nested logic relationship is improved. After the identification of the host and guest entities is completed, a layer of tag entity matching algorithm is added for checking whether the tags input by the clients are correct or not, and automatic correction is performed when errors are found. The error correction capability of the model for customer error input is enhanced. According to the embodiment of the application, the input of the model is defined as one triplet, so that the application scene of the model is enlarged.
Meanwhile, the embodiment of the application enhances the error correction capability of the model on the error input of the client: after the identification of the host and guest entities is completed, a layer of tag entity matching algorithm is added for checking whether the tags input by the clients are correct or not, and automatic correction is performed when errors are found. The embodiment of the application realizes model universality: by defining the input of the model as a triplet, the prediction range of the model is simplified, and the application scene of the model is enlarged. And after the triplets are identified, converting into corresponding logical binary tree. The expression capacity and consistency of the logic relationship are improved. The embodiment of the application realizes flexible application: the method can be used as a preprocessing flow for target group circle selection, can also be used as any logic relation extraction task for independent use, and has high universality.
Next, a target group determination system according to an embodiment of the present invention will be described with reference to fig. 11.
FIG. 11 is a schematic diagram of a target group determination system according to an embodiment of the invention, the system specifically comprising:
a first module 110, configured to obtain a first text; the first text comprises label information of the target group and logic relation information among labels;
A second module 120, configured to process the first text through a natural language model to obtain a relationship triplet; the natural language model is used for carrying out unified representation and interaction on the entities and relations involved in the first text, and carries out prediction learning through an indication function, wherein the indication function comprises a first function of a host entity and a guest entity and a second function of the guest entity and the host entity;
a third module 130, configured to convert the relationship triplet into a binary tree, and generate a bitmap script statement according to the binary tree;
and a fourth module 140, configured to extract, by using the bitmap calculation engine, a group list of bitmap script statements, so as to obtain a target group.
It can be seen that the content in the above method embodiment is applicable to the system embodiment, and the functions specifically implemented by the system embodiment are the same as those of the method embodiment, and the beneficial effects achieved by the method embodiment are the same as those achieved by the method embodiment.
Referring to fig. 12, an embodiment of the present invention provides a target group determining apparatus, including:
at least one processor 210;
at least one memory 220 for storing at least one program;
the at least one program, when executed by the at least one processor 210, causes the at least one processor 210 to implement the target population determination method.
Similarly, the content in the above method embodiment is applicable to the embodiment of the present device, and the functions specifically implemented by the embodiment of the present device are the same as those of the embodiment of the above method, and the beneficial effects achieved by the embodiment of the above method are the same as those achieved by the embodiment of the above method.
The embodiment of the present invention also provides a computer-readable storage medium in which a processor-executable program is stored, which when executed by a processor is configured to perform the above-described target population determination method.
Similarly, the content in the above method embodiment is applicable to the present storage medium embodiment, and the specific functions of the present storage medium embodiment are the same as those of the above method embodiment, and the achieved beneficial effects are the same as those of the above method embodiment.
In some alternative embodiments, the functions/acts noted in the block diagrams may occur out of the order noted in the operational illustrations. For example, two blocks shown in succession may in fact be executed substantially concurrently or the blocks may sometimes be executed in the reverse order, depending upon the functionality/acts involved. Furthermore, the embodiments presented and described in the flowcharts of the present invention are provided by way of example in order to provide a more thorough understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed, and in which sub-operations described as part of a larger operation are performed independently.
Furthermore, while the invention is described in the context of functional modules, it should be appreciated that, unless otherwise indicated, one or more of the functions and/or features may be integrated in a single physical device and/or software module or may be implemented in separate physical devices or software modules. It will also be appreciated that a detailed discussion of the actual implementation of each module is not necessary to an understanding of the present invention. Rather, the actual implementation of the various functional modules in the apparatus disclosed herein will be apparent to those skilled in the art from consideration of their attributes, functions and internal relationships. Accordingly, one of ordinary skill in the art can implement the invention as set forth in the claims without undue experimentation. It is also to be understood that the specific concepts disclosed are merely illustrative and are not intended to be limiting upon the scope of the invention, which is to be defined in the appended claims and their full scope of equivalents.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in the form of a software product stored in a storage medium, including several programs for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
Logic and/or steps represented in the flowcharts or otherwise described herein, e.g., a ordered listing of executable programs for implementing logical functions, can be embodied in any computer-readable medium for use by or in connection with a program execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the programs from the program execution system, apparatus, or device and execute the programs. For the purposes of this description, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the program execution system, apparatus, or device.
More specific examples (a non-exhaustive list) of the computer-readable medium would include the following: an electrical connection (electronic device) having one or more wires, a portable computer diskette (magnetic device), a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). In addition, the computer readable medium may even be paper or other suitable medium on which the program is printed, as the program may be electronically captured, via, for instance, optical scanning of the paper or other medium, then compiled, interpreted or otherwise processed in a suitable manner, if necessary, and then stored in a computer memory.
It is to be understood that portions of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the various steps or methods may be implemented in software or firmware stored in a memory and executed by a suitable program execution system. For example, if implemented in hardware, as in another embodiment, may be implemented using any one or combination of the following techniques, as is well known in the art: discrete logic circuits having logic gates for implementing logic functions on data signals, application specific integrated circuits having suitable combinational logic gates, programmable Gate Arrays (PGAs), field Programmable Gate Arrays (FPGAs), and the like.
In the foregoing description of the present specification, reference has been made to the terms "one embodiment/example", "another embodiment/example", "certain embodiments/examples", and the like, means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
While embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that: many changes, modifications, substitutions and variations may be made to the embodiments without departing from the spirit and principles of the invention, the scope of which is defined by the claims and their equivalents.
While the preferred embodiment of the present invention has been described in detail, the present invention is not limited to the embodiments described above, and various equivalent modifications and substitutions can be made by those skilled in the art without departing from the spirit of the present invention, and these equivalent modifications and substitutions are intended to be included in the scope of the present invention as defined in the appended claims.
Claims (10)
1. A method of target population determination, comprising the steps of:
acquiring a first text; the first text comprises label information of a target group and logic relation information among labels;
processing the first text through a natural language model to obtain a relation triplet; the natural language model is used for carrying out unified representation and interaction on the entities and relations involved in the first text, the natural language model carries out prediction learning through an indication function, and the indication function comprises a first function of a host entity and a guest entity and a second function of the guest entity and the host entity;
Converting the relation triplet into a binary tree, and generating a bitmap script statement according to the binary tree;
and extracting a group list from the bitmap script statement through a bitmap calculation engine to obtain a target group.
2. The method of claim 1, wherein the processing the first text through a natural language model to obtain a relationship triplet comprises:
splicing the first text, the logic triples and the sub-logic triples to obtain an input sequence; the logical triples are used for representing the logical relations among the host entity, the guest entity and the relation; the sub-logical triples comprise a first logical relationship between a host entity and a guest entity, a second logical relationship between the guest entity and the host entity, a third logical relationship between the entities and the relationship, and a fourth logical relationship between the entities;
performing characterization learning processing on the input sequence through a natural language model to obtain a characterization matrix; each position of the characterization matrix is used for characterizing a defined relationship;
determining an indication function according to the characterization matrix; and predicting the result according to the indication function to obtain a relation triplet.
3. The target group determination method of claim 2, wherein the natural language model comprises a bert model, the method further comprising:
adding the Q matrix and the K matrix of the last layer in the bert model to obtain an interaction table;
processing the interaction table through an activation function to obtain a relation triplet;
and optimizing the relation triplet through an optimization function to obtain an optimized relation triplet.
4. The target group determination method according to claim 2, characterized in that the method further comprises the steps of:
performing host-guest identification processing on the first text to obtain an entity and a relationship;
verifying the entity involved in the first text through a greedy algorithm of a regular expression; and if the verification result is not passed, correcting the first text to obtain a corrected first text.
5. The target group determination method according to claim 1, wherein the step of converting the relation triplet into a binary tree comprises:
traversing each triplet in the relation triples to generate a corresponding partial binary tree;
and inserting the partial binary tree into the existing binary tree to obtain a converted binary tree.
6. The target population determination method of claim 5, wherein traversing each of the relationship triples generates a corresponding partial binary tree comprising:
traversing the relation triples to obtain a first triples;
taking the relation in the first triplet as a root node of a partial binary tree;
if the first main entity of the first triplet is not null, using the first main entity as a left child node of the root node;
or if the first guest entity of the first triplet is not null, using the first guest entity as the right child node of the root node, and generating a partial binary tree.
7. The target group determination method according to claim 1, wherein the method further comprises:
establishing a group label architecture, wherein the group label architecture comprises a label data layer, a bitmap layer and a label inquiry application layer;
the label data layer is used for storing label information of a group, the bitmap layer is used for establishing a bitmap according to the label information, and the label inquiry application layer is used for inquiring a target group according to the first text.
8. A target group determination system, comprising:
The first module is used for acquiring a first text; the first text comprises label information of a target group and logic relation information among labels;
the second module is used for processing the first text through a natural language model to obtain a relation triplet; the natural language model is used for carrying out unified representation and interaction on the entities and relations involved in the first text, the natural language model carries out prediction learning through an indication function, and the indication function comprises a first function of a host entity and a guest entity and a second function of the guest entity and the host entity;
the third module is used for converting the relation triplet into a binary tree and generating a bitmap script statement according to the binary tree;
and the fourth module is used for extracting the group list of the bitmap script statement through a bitmap calculation engine to obtain a target group.
9. A target group determination apparatus, comprising:
at least one processor;
at least one memory for storing at least one program;
the at least one program, when executed by the at least one processor, causes the at least one processor to implement the target population determination method of any one of claims 1 to 7.
10. A computer-readable storage medium in which a processor-executable program is stored, characterized in that the processor-executable program is for implementing the target population determination method according to any one of claims 1 to 7 when being executed by a processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311337060.8A CN117421390A (en) | 2023-10-16 | 2023-10-16 | Target group determination method, system, device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311337060.8A CN117421390A (en) | 2023-10-16 | 2023-10-16 | Target group determination method, system, device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117421390A true CN117421390A (en) | 2024-01-19 |
Family
ID=89522064
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311337060.8A Pending CN117421390A (en) | 2023-10-16 | 2023-10-16 | Target group determination method, system, device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117421390A (en) |
-
2023
- 2023-10-16 CN CN202311337060.8A patent/CN117421390A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7169369B2 (en) | Method, system for generating data for machine learning algorithms | |
| US9646262B2 (en) | Data intelligence using machine learning | |
| US9552551B2 (en) | Pattern detection feedback loop for spatial and temporal memory systems | |
| US8504570B2 (en) | Automated search for detecting patterns and sequences in data using a spatial and temporal memory system | |
| US8645291B2 (en) | Encoding of data for processing in a spatial and temporal memory system | |
| WO2019060010A1 (en) | Content pattern based automatic document classification | |
| US11373117B1 (en) | Artificial intelligence service for scalable classification using features of unlabeled data and class descriptors | |
| US11238111B2 (en) | Response generation | |
| CN112395487A (en) | Information recommendation method and device, computer-readable storage medium and electronic equipment | |
| JP2020061136A (en) | Accessible machine learning backend | |
| Zhang et al. | CapsNet-based supervised hashing | |
| US11741101B2 (en) | Estimating execution time for batch queries | |
| US20210034977A1 (en) | Interpretable Tabular Data Learning Using Sequential Sparse Attention | |
| CN112988698A (en) | Data processing method and device | |
| US11514233B2 (en) | Automated nonparametric content analysis for information management and retrieval | |
| CN117421390A (en) | Target group determination method, system, device and storage medium | |
| US11556514B2 (en) | Semantic data type classification in rectangular datasets | |
| US11681689B2 (en) | Automatic generation of a matching algorithm in master data management | |
| CN115841365A (en) | Model selection and quotation method, system, equipment and medium based on natural language processing | |
| US20220172075A1 (en) | Decoding random forest problem solving through node labeling and subtree distributions | |
| US11782918B2 (en) | Selecting access flow path in complex queries | |
| US20220222265A1 (en) | Insight expansion in smart data retention systems | |
| AU2020101842A4 (en) | DAI- Dataset Discovery: DATASET DISCOVERY IN DATA ANALYTICS USING AI- BASED PROGRAMMING. | |
| US20200167387A1 (en) | Method and system for streamlined auditing | |
| AU2020104034A4 (en) | IML-Cloud Data Performance: Cloud Data Performance Improved using Machine Learning. |
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 | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20240318 Address after: Unit 1, Building 1, China Telecom Zhejiang Innovation Park, No. 8 Xiqin Street, Wuchang Street, Yuhang District, Hangzhou City, Zhejiang Province, 311100 Applicant after: Tianyi Shilian Technology Co.,Ltd. Country or region after: China Address before: 200000 room 1423, No. 1256 and 1258, Wanrong Road, Jing'an District, Shanghai Applicant before: Tianyi Digital Life Technology Co.,Ltd. Country or region before: China |
|
| TA01 | Transfer of patent application right |