CN118113914A - Artificial intelligence treatment method and robot for progressively identifying sensitive language - Google Patents

Artificial intelligence treatment method and robot for progressively identifying sensitive language Download PDF

Info

Publication number
CN118113914A
CN118113914A CN202311734811.XA CN202311734811A CN118113914A CN 118113914 A CN118113914 A CN 118113914A CN 202311734811 A CN202311734811 A CN 202311734811A CN 118113914 A CN118113914 A CN 118113914A
Authority
CN
China
Prior art keywords
type
taking
knowledge
label
network content
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
Application number
CN202311734811.XA
Other languages
Chinese (zh)
Inventor
朱定局
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
South China Normal University
Original Assignee
South China Normal University
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 South China Normal University filed Critical South China Normal University
Priority to CN202311734811.XA priority Critical patent/CN118113914A/en
Publication of CN118113914A publication Critical patent/CN118113914A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90—Details of database functions independent of the retrieved data types
    • G06F16/903—Querying
    • G06F16/90335—Query processing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Software Systems (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Mathematical Physics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Computational Linguistics (AREA)
  • Computing Systems (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The artificial intelligence treatment method and the robot for progressively identifying the sensitive language take object knowledge, language property knowledge and network content as inputs, and can generate correlation in the model training process, so that knowledge can assist in identifying the network content, the accuracy of identifying the sensitive language is improved, and after knowledge is updated, the accuracy of identifying the sensitive language can be improved by performing transfer learning through an original model.

Description

Artificial intelligence treatment method and robot for progressively identifying sensitive language
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to an artificial intelligence treatment method and a robot for progressively identifying sensitive language.
Background
In the process of implementing the present invention, the inventor finds that at least the following problems exist in the prior art: the prior art relates to the examination and the examination of the network content are very strict, but all adopt a manual mode. Because the quantity and time of auditors are limited, omission often occurs, artificial intelligence is needed to assist manual primary audit, so that the advantages of the artificial intelligence can be utilized to perform mass rapid primary audit, suspected network contents are selected, and then the network contents needing to be subjected to manual audit are subjected to manual audit, so that the quantity of the network contents needing to be subjected to manual audit is greatly reduced, key audit can be grasped, and the network contents which are obviously not suspected do not need manual audit.
Accordingly, the prior art is still in need of improvement and development.
Disclosure of Invention
Based on the above, it is necessary to provide an artificial intelligent governance method and a robot for progressively recognizing sensitive language, which can progressively recognize sensitive language, so as to improve the efficiency of intelligently recognizing sensitive language.
In a first aspect, an embodiment of the present invention provides an artificial intelligence governance method, the method comprising:
A knowledge data acquisition step: acquiring the type of a sensitive language to be identified as a first type; acquiring an object tag of the sensitive language of the first type as a first tag; acquiring a property tag of the sensitive language of the first type as a second tag; acquiring object knowledge corresponding to the first type as first knowledge; acquiring the knowledge of the speaking property corresponding to the first type as second knowledge;
The first type sensitive language object identification intelligent model using step: acquiring network content to be identified; taking the network content and the first knowledge as input, predicting through a first type sensitive language object identification intelligent model, taking the obtained output as a first position label of the network content on a first type, and taking the first position label as a set of positions of the content of each first label belonging to the first type in the network content;
The first type sensitive language property identification intelligent model using step: extracting the position and the content in a preset range around the position from the network content according to each position in the set of the positions; and taking the corresponding content and the second knowledge of each position in the set of the positions as input, calculating to obtain an output label through the intelligent model for identifying the properties of the first type sensitive language, and taking the output label as the second label corresponding to each position.
Preferably, the method further comprises:
The updated model using step: acquiring network content to be identified; taking the network content and the updated first knowledge as input, predicting through a first type sensitive language object identification intelligent model, taking the obtained output as a first position label of the network content on a first type, and taking the first position label as a set of positions of the content of each first label belonging to the first type in the network content; extracting the position and the content in a preset range around the position from the network content according to each position in the set of the positions; and taking the corresponding content of each position in the set of the positions and the updated second knowledge as input, calculating to obtain an output label through the intelligent model for identifying the properties of the first type sensitive language, and taking the output label as the second label corresponding to each position.
Preferably, the method further comprises:
Model migration learning step: taking the network content and the third knowledge as input, taking a third position label of the network content as expected output, and training and testing the first type sensitive language object recognition intelligent model to obtain a second type sensitive language object recognition intelligent model; and taking the corresponding content and fourth knowledge of each position in the set of positions as input, taking the fourth label corresponding to each position as expected output, and training and testing the first type sensitive language property identification intelligent model to obtain the second type sensitive language property identification intelligent model.
Preferably, the method further comprises:
sensitive speaking treatment steps: if the network content exists the sensitive language, notify the user that the network content exists the sensitive language, need to deal with in time.
In a second aspect, embodiments of the present invention provide an artificial intelligence abatement system, the system comprising:
Knowledge data acquisition module: acquiring the type of a sensitive language to be identified as a first type; acquiring an object tag of the sensitive language of the first type as a first tag; acquiring a property tag of the sensitive language of the first type as a second tag; acquiring object knowledge corresponding to the first type as first knowledge; acquiring the knowledge of the speaking property corresponding to the first type as second knowledge;
the first type sensitive language object identification intelligent model using module: acquiring network content to be identified; taking the network content and the first knowledge as input, predicting through a first type sensitive language object identification intelligent model, taking the obtained output as a first position label of the network content on a first type, and taking the first position label as a set of positions of the content of each first label belonging to the first type in the network content;
The first type of sensitive language property identification intelligent model using module: extracting the position and the content in a preset range around the position from the network content according to each position in the set of the positions; and taking the corresponding content and the second knowledge of each position in the set of the positions as input, calculating to obtain an output label through the intelligent model for identifying the properties of the first type sensitive language, and taking the output label as the second label corresponding to each position.
Preferably, the system further comprises:
The updated model use module: acquiring network content to be identified; taking the network content and the updated first knowledge as input, predicting through a first type sensitive language object identification intelligent model, taking the obtained output as a first position label of the network content on a first type, and taking the first position label as a set of positions of the content of each first label belonging to the first type in the network content; extracting the position and the content in a preset range around the position from the network content according to each position in the set of the positions; and taking the corresponding content of each position in the set of the positions and the updated second knowledge as input, calculating to obtain an output label through the intelligent model for identifying the properties of the first type sensitive language, and taking the output label as the second label corresponding to each position.
Preferably, the system further comprises:
Model transfer learning module: taking the network content and the third knowledge as input, taking a third position label of the network content as expected output, and training and testing the first type sensitive language object recognition intelligent model to obtain a second type sensitive language object recognition intelligent model; and taking the corresponding content and fourth knowledge of each position in the set of positions as input, taking the fourth label corresponding to each position as expected output, and training and testing the first type sensitive language property identification intelligent model to obtain the second type sensitive language property identification intelligent model.
Preferably, the system further comprises:
Sensitive speech governance module: if the network content exists the sensitive language, notify the user that the network content exists the sensitive language, need to deal with in time.
In a third aspect, embodiments of the present invention provide an artificial intelligence device comprising a module of a system according to any of the embodiments of the second aspect.
In a fourth aspect, embodiments of the present invention provide a computer readable storage medium having stored thereon a computer program which when executed by a processor implements the steps of the method according to any of the embodiments of the first aspect.
In a fifth aspect, an embodiment of the present invention provides a robot, including a memory, a processor, and an artificial intelligence robot program stored on the memory and executable on the processor, the processor implementing the steps of the method according to any one of the embodiments of the first aspect when the processor executes the program.
According to the artificial intelligence treatment method and the robot for progressively recognizing the sensitive language, object knowledge, language property knowledge and network content are taken as inputs, association can be generated between the object knowledge and the language property knowledge in the model training process, further knowledge can assist in recognizing the network content, the accuracy of recognizing the sensitive language is improved, and after knowledge is updated, transfer learning is conducted through an original model, so that the accuracy of recognizing the sensitive language can be improved.
Drawings
FIG. 1 is a flow chart of an artificial intelligence abatement method provided by an embodiment of the invention;
FIG. 2 is a block diagram of an artificial intelligence abatement system in accordance with one embodiment of the invention.
Detailed Description
The following describes the technical scheme in the embodiment of the present invention in detail in connection with the implementation mode of the present invention.
Basic embodiment of the invention
In a first aspect, as shown in fig. 1, an embodiment of the present invention provides an artificial intelligence governance method, the method including a knowledge data acquisition step; a first type sensitive language object identification intelligent model using step; the first type of sensitive language property identifies a smart model use step. In a preferred embodiment, the method further comprises an updated model use step. In a preferred embodiment, the method further comprises a model migration learning step. In a preferred embodiment, the method further comprises sensitive speech governance.
In a second aspect, as shown in FIG. 2, an embodiment of the present invention provides an artificial intelligence abatement system comprising a knowledge data acquisition module; a type sensitive language object identification intelligent model using module; the first type of sensitive language property identifies a smart model usage module. In a preferred embodiment, the system further comprises an updated model usage module. In a preferred embodiment, the system further comprises a model migration learning module. In a preferred embodiment, the system further comprises sensitive speech governance.
PREFERRED EMBODIMENTS OF THE PRESENT INVENTION
Defamation of the specific individual and blackening of the specific group language will necessarily occur the specific individual name and specific group word, so that the network content with the specific individual name and specific group word can be listed as suspicious network content, and the specific individual name and specific group word can be marked out for manual review, so that time is saved, and no need to find out where the specific individual name and specific group word occur. Web content having a specific individual name, a specific community typeface, requires further prediction of its tag. The step-by-step recognition by two models has the advantage of avoiding the problem of insufficient training data, because if the type is directly recognized, recognition effect is necessarily affected because the marked data is insufficient, if only unsupervised learning is utilized, recognition accuracy of specific individual names and specific group word patterns is reduced, and if the recognition accuracy of specific individual names and specific group word patterns is not high, the accuracy of the type recognition of the language is lower. The network content comprises audio content, video content and text content, and the audio and video content can be identified as the text content through a deep learning model. So that the identification is performed step by two steps, much more accurately than by a model directly. The particular individual includes a character.
1. A type acquisition step; acquiring the type of a sensitive language to be identified as a first type; such as a type of utterance involving a particular individual or a type of utterance commenting a particular community;
2. Object tag acquisition: acquiring object labels of the sensitive language of the first type, wherein the object labels can be used as first labels, and a plurality of first labels can be arranged; for example, if the first type is a speaker type of a particular individual, then the tag includes the name of the particular individual, if the first type is a speaker type of a particular group, then the tag includes a particular group regime, etc.;
3. A property label acquisition step: acquiring a property label of the sensitive language of the first type as a second label, wherein a plurality of second labels can be provided; for example, if the first type is a specific individual's speaker type, the tag includes negative comments, non-negative comments, and if the first type is a specific group's speaker type, the tag may also include negative comments, non-negative comments;
4. Object knowledge acquisition: acquiring object knowledge corresponding to the first type as first knowledge; for example, the object knowledge related to the speaker type of the specific individual is a set of specific individual name, specific individual photo, specific individual voiceprint, and the object knowledge related to the speaker type of the specific group is a set of specific group name, specific group definition.
5. A step of obtaining knowledge of the language property: acquiring the knowledge of the speaking property corresponding to the first type as second knowledge; for example, the property knowledge of the type of language related to a particular individual is wise, confusing, etc., the property knowledge of the type of language related to a particular community is superior, lagging, etc.
The technical effects are as follows: the photos of the specific individuals are required to be used as knowledge as well as the voiceprints, so that the recognition accuracy can be improved, and the situation that the photos of the specific individuals are not missed, but the photos and the voiceprints of the specific individuals are stolen is avoided. The labels of the sensitive language are used as knowledge input models, so that the models are favorably associated with the labels when the data are processed, and the principle and the attention mechanism are similar, so that the models pay more attention to and pay more attention to partial data related to the labels, and the recognition accuracy is improved. The first tag and the first knowledge are adopted to identify the data from the network content, so that a large amount of data marked with the object can be utilized; the second label and the second knowledge are adopted to identify the property of the language from the content of the identified object, so that the identification range can be reduced, and the requirement on the data volume can be reduced only by identifying the property without identifying the object.
6. Model training test data set acquisition: acquiring network content for training test, and acquiring a set of positions of content of each first tag belonging to a first type in the network content, for example (Zhang three, position 1, li four, position 2, li four, position 3, wang five, none) as a first position tag; the first type, for example, relates to the type of utterance of a particular individual; extracting the position and the content (for example, the content of 20 sentences before and after the name of a specific individual) in a preset range around the position from the network content according to each position in the set of the positions to serve as the corresponding content of the position; acquiring a second tag on the first type to which the corresponding content of the location belongs as the second tag corresponding to the location (e.g., detracting the speaker of the specific individual, not detracting the speaker of the specific individual); the network content comprises words, pictures, audio and video.
7. The intelligent model training and testing step for identifying the sensitive language object of the first type comprises the following steps: taking the network content and the first knowledge as input, taking a first position label of the network content as expected output, training and testing a deep learning model, and obtaining a first type sensitive language object identification intelligent model; for example, the first type is a speaker related to a particular individual, then the first type sensitive speaker object recognition smart model is a speaker related to a particular individual;
8. The first type sensitive language property recognition intelligent model training test step: taking the corresponding content and the second knowledge of each position in the set of the positions as input, taking the second label corresponding to each position as expected output, training and testing the deep learning model, and obtaining a first type sensitive language property identification intelligent model; for example, the first type is a speaker involving a particular individual, then the first type sensitive speaker property recognition smart model is a speaker involving a particular individual;
The technical effects are as follows: the object knowledge, the language property knowledge and the network content are used as inputs, and can be associated in the model training process, so that knowledge can assist in network content identification, and network content identification accuracy is improved.
The technical effects are as follows: the method has the advantages that the data volume of object recognition is increased, the labeling of the data set of the object recognition is easy, so that the recognition accuracy of the sensitive language object can be improved, the labeling and the recognition of the property are easier after the object is determined, and the recognition accuracy of the property of the sensitive language can be improved.
9. The first type sensitive language object identification intelligent model using step: acquiring network content to be identified; and taking the network content and the first knowledge as input, predicting through a first type sensitive language object identification intelligent model, taking the obtained output as a first position label of the network content on a first type, and taking the first position label as a set of positions of the content of each first label belonging to the first type in the network content.
10. The first type sensitive language property identification intelligent model using step: extracting the position and the content in a preset range around the position from the network content according to each position in the set of the positions; and taking the corresponding content and the second knowledge of each position in the set of the positions as input, calculating to obtain an output label through the intelligent model for identifying the properties of the first type sensitive language, and taking the output label as the second label corresponding to each position.
After the step 10 of the process, the process is carried out,
11. And a model updating step: if the first knowledge is changed, acquiring updated first knowledge; acquiring network content for training test, and acquiring a set of positions of the content of each first label belonging to a first type in the network content; extracting the position and the content in a preset range around the position from the network content according to each position in the set of the positions to serve as the corresponding content of the position; acquiring a second tag on the first type to which the corresponding content of the position belongs as a second tag corresponding to the position; taking the network content for training test and updated first knowledge as input, taking a first position label of the network content as expected output, training and testing the deep learning model to obtain a first type sensitive language object identification intelligent model; for example, the first type is a speaker related to a particular individual, then the first type sensitive speaker object recognition smart model is a speaker related to a particular individual; if the second knowledge is changed, acquiring updated second knowledge; taking the corresponding content of each position in the set of the positions and updated second knowledge as input, taking the second label corresponding to each position as expected output, training and testing the deep learning model, and obtaining a first type sensitive language property identification intelligent model; for example, the first type is a speaker involving a particular individual, then the first type sensitive speaker property recognition smart model is a speaker involving a particular individual;
The technical effects are as follows: based on the existing model, training and testing are performed through the updated first knowledge and the updated second knowledge, so that the training and testing efficiency and speed can be greatly improved compared with training from the initial deep learning model, and the training and testing can be converged more quickly, because only partial first knowledge and second knowledge are changed during training and testing, and the training and testing network data can be unchanged.
12. The updated model using step: acquiring network content to be identified; taking the network content and the updated first knowledge as input, predicting through a first type sensitive language object identification intelligent model, taking the obtained output as a first position label of the network content on a first type, and taking the first position label as a set of positions of the content of each first label belonging to the first type in the network content; extracting the position and the content in a preset range around the position from the network content according to each position in the set of the positions; taking the corresponding content of each position in the set of the positions and updated second knowledge as input, calculating to obtain an output tag through a first type sensitive language property identification intelligent model, and taking the output tag as a second tag corresponding to each position;
After the step 10 of the process, the process is carried out,
13. A step of preparing transfer learning data: acquiring another type of sensitive language to be identified as a second type; acquiring object tags of the second type of sensitive language as third tags, wherein a plurality of third tags can be used; acquiring a property label of the second type of sensitive language as a fourth label, wherein the fourth label can be a plurality of labels; acquiring object knowledge corresponding to the second type as third knowledge; acquiring the speaker property knowledge corresponding to the second type as fourth knowledge;
14. The step of obtaining the training test data of the transfer learning model: acquiring network content for training test, and acquiring a set of positions of the content of each third tag belonging to the second type in the network content as third position tags; extracting the position and the content (for example, the content of 20 sentences before and after the name of a specific individual) in a preset range around the position from the network content according to each position in the set of the positions to serve as the corresponding content of the position; acquiring a fourth tag on a second type to which the corresponding content of the location belongs as the fourth tag corresponding to the location (e.g., detracting the speaker of the specific individual, not detracting the speaker of the specific individual);
15. Model migration learning step: taking the network content and the third knowledge as input, taking a third position label of the network content as expected output, and training and testing the first type sensitive language object recognition intelligent model to obtain a second type sensitive language object recognition intelligent model; taking the corresponding content and the fourth knowledge of each position in the set of the positions as input, taking the fourth label corresponding to each position as expected output, and training and testing the first type sensitive language property identification intelligent model to obtain a second type sensitive language property identification intelligent model;
16. The model obtained by transfer learning comprises the following steps: acquiring network content to be identified; taking the network content and the third knowledge as input, predicting through a second type sensitive language object recognition intelligent model, taking the obtained output as a third position label of the network content on a second type, and taking the third position label as a set of positions of the content of each third label belonging to the second type in the network content; extracting the position and the content in a preset range around the position from the network content according to each position in the set of the positions; and taking the corresponding content and fourth knowledge of each position in the set of the positions as input, calculating to obtain an output label through a second type sensitive language property recognition intelligent model, and taking the output label as a fourth label corresponding to each position.
The technical effects are as follows: because of the commonality between different sensitive language, transfer learning can be utilized to improve the efficiency and effect of model generation. Based on the existing intelligent model for identifying the first type of sensitive language, the intelligent model for identifying the second type of sensitive language is obtained through transfer learning, so that the training and testing efficiency and speed can be improved, and the computing resources and time are saved.
After the step 10 of the process, the process is carried out,
17. Sensitive speaking treatment steps: if the network content exists the sensitive language, notify the user that the network content exists the sensitive language, need to deal with in time.
The foregoing examples illustrate only a few embodiments of the invention and are described in detail herein without thereby limiting the scope of the invention. It should be noted that it will be apparent to those skilled in the art that various modifications and improvements can be made without departing from the spirit of the invention, which are within the scope of the invention. Accordingly, the scope of protection of the present invention is to be determined by the appended claims.

Claims (10)

1. An artificial intelligence governance method, the method comprising:
A knowledge data acquisition step: acquiring the type of a sensitive language to be identified as a first type; acquiring an object tag of the sensitive language of the first type as a first tag; acquiring a property tag of the sensitive language of the first type as a second tag; acquiring object knowledge corresponding to the first type as first knowledge; acquiring the knowledge of the speaking property corresponding to the first type as second knowledge;
The first type sensitive language object identification intelligent model using step: acquiring network content to be identified; taking the network content and the first knowledge as input, predicting through a first type sensitive language object identification intelligent model, taking the obtained output as a first position label of the network content on a first type, and taking the first position label as a set of positions of the content of each first label belonging to the first type in the network content;
The first type sensitive language property identification intelligent model using step: extracting the position and the content in a preset range around the position from the network content according to each position in the set of the positions; and taking the corresponding content and the second knowledge of each position in the set of the positions as input, calculating to obtain an output label through the intelligent model for identifying the properties of the first type sensitive language, and taking the output label as the second label corresponding to each position.
2. The artificial intelligence governance method of claim 1, further comprising:
The updated model using step: acquiring network content to be identified; taking the network content and the updated first knowledge as input, predicting through a first type sensitive language object identification intelligent model, taking the obtained output as a first position label of the network content on a first type, and taking the first position label as a set of positions of the content of each first label belonging to the first type in the network content; extracting the position and the content in a preset range around the position from the network content according to each position in the set of the positions; and taking the corresponding content of each position in the set of the positions and the updated second knowledge as input, calculating to obtain an output label through the intelligent model for identifying the properties of the first type sensitive language, and taking the output label as the second label corresponding to each position.
3. The artificial intelligence governance method of claim 1, further comprising:
Model migration learning step: taking the network content and the third knowledge as input, taking a third position label of the network content as expected output, and training and testing the first type sensitive language object recognition intelligent model to obtain a second type sensitive language object recognition intelligent model; and taking the corresponding content and fourth knowledge of each position in the set of positions as input, taking the fourth label corresponding to each position as expected output, and training and testing the first type sensitive language property identification intelligent model to obtain the second type sensitive language property identification intelligent model.
4. The artificial intelligence governance method of claim 1, further comprising:
sensitive speaking treatment steps: if the network content exists the sensitive language, notify the user that the network content exists the sensitive language, need to deal with in time.
5. An artificial intelligence governance system, the system comprising:
Knowledge data acquisition module: acquiring the type of a sensitive language to be identified as a first type; acquiring an object tag of the sensitive language of the first type as a first tag; acquiring a property tag of the sensitive language of the first type as a second tag; acquiring object knowledge corresponding to the first type as first knowledge; acquiring the knowledge of the speaking property corresponding to the first type as second knowledge;
the first type sensitive language object identification intelligent model using module: acquiring network content to be identified; taking the network content and the first knowledge as input, predicting through a first type sensitive language object identification intelligent model, taking the obtained output as a first position label of the network content on a first type, and taking the first position label as a set of positions of the content of each first label belonging to the first type in the network content;
The first type of sensitive language property identification intelligent model using module: extracting the position and the content in a preset range around the position from the network content according to each position in the set of the positions; and taking the corresponding content and the second knowledge of each position in the set of the positions as input, calculating to obtain an output label through the intelligent model for identifying the properties of the first type sensitive language, and taking the output label as the second label corresponding to each position.
6. The artificial intelligence abatement system of claim 5, further comprising:
The updated model use module: acquiring network content to be identified; taking the network content and the updated first knowledge as input, predicting through a first type sensitive language object identification intelligent model, taking the obtained output as a first position label of the network content on a first type, and taking the first position label as a set of positions of the content of each first label belonging to the first type in the network content; extracting the position and the content in a preset range around the position from the network content according to each position in the set of the positions; and taking the corresponding content of each position in the set of the positions and the updated second knowledge as input, calculating to obtain an output label through the intelligent model for identifying the properties of the first type sensitive language, and taking the output label as the second label corresponding to each position.
7. The artificial intelligence abatement system of claim 5, further comprising:
Model transfer learning module: taking the network content and the third knowledge as input, taking a third position label of the network content as expected output, and training and testing the first type sensitive language object recognition intelligent model to obtain a second type sensitive language object recognition intelligent model; and taking the corresponding content and fourth knowledge of each position in the set of positions as input, taking the fourth label corresponding to each position as expected output, and training and testing the first type sensitive language property identification intelligent model to obtain the second type sensitive language property identification intelligent model.
8. The artificial intelligence abatement system of claim 5, further comprising:
Sensitive speech governance module: if the network content exists the sensitive language, notify the user that the network content exists the sensitive language, need to deal with in time.
9. A robot comprising a memory, a processor and an artificial intelligence robot program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method according to any one of claims 1-4 when the program is executed by the processor.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the steps of the method according to any of claims 1-4.
CN202311734811.XA 2023-12-15 2023-12-15 Artificial intelligence treatment method and robot for progressively identifying sensitive language Pending CN118113914A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311734811.XA CN118113914A (en) 2023-12-15 2023-12-15 Artificial intelligence treatment method and robot for progressively identifying sensitive language

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311734811.XA CN118113914A (en) 2023-12-15 2023-12-15 Artificial intelligence treatment method and robot for progressively identifying sensitive language

Publications (1)

Publication Number Publication Date
CN118113914A true CN118113914A (en) 2024-05-31

Family

ID=91209473

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311734811.XA Pending CN118113914A (en) 2023-12-15 2023-12-15 Artificial intelligence treatment method and robot for progressively identifying sensitive language

Country Status (1)

Country Link
CN (1) CN118113914A (en)

Similar Documents

Publication Publication Date Title
CN110781276B (en) Text extraction method, device, equipment and storage medium
WO2022142041A1 (en) Training method and apparatus for intent recognition model, computer device, and storage medium
US11031009B2 (en) Method for creating a knowledge base of components and their problems from short text utterances
CN116629275B (en) Intelligent decision support system and method based on big data
CN108959474B (en) Entity relation extraction method
CN110222330B (en) Semantic recognition method and device, storage medium and computer equipment
CN113158656B (en) Ironic content recognition method, ironic content recognition device, electronic device, and storage medium
CN110968725B (en) Image content description information generation method, electronic device and storage medium
CN113448843B (en) Image recognition software test data enhancement method and device based on defect analysis
CN110909144A (en) Question-answer dialogue method and device, electronic equipment and computer readable storage medium
CN113990352B (en) User emotion recognition and prediction method, device, equipment and storage medium
CN111931020B (en) Formula labeling method, device, equipment and storage medium
CN112464927B (en) Information extraction method, device and system
CN114416979A (en) Text query method, text query equipment and storage medium
CN110737770B (en) Text data sensitivity identification method and device, electronic equipment and storage medium
CN113420122A (en) Method, device and equipment for analyzing text and storage medium
CN112183102A (en) Named entity identification method based on attention mechanism and graph attention network
CN114218945A (en) Entity identification method, device, server and storage medium
CN115064154A (en) Method and device for generating mixed language voice recognition model
CN115510188A (en) Text keyword association method, device, equipment and storage medium
CN109408175B (en) Real-time interaction method and system in general high-performance deep learning calculation engine
CN114676705B (en) Dialogue relation processing method, computer and readable storage medium
CN118350464A (en) Conversational target positioning method and device based on text input with arbitrary granularity
CN114266252A (en) Named entity recognition method, device, equipment and storage medium
CN113342935A (en) Semantic recognition method and device, electronic equipment and readable storage medium

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