An online multi-scenario reinforcement learning method

Abstract

The present invention relates to the field of data analysis and processing technology, and in particular to a multi-scenario reinforcement learning method for online learning. The method comprises the following steps: S1, identifying text data in the current online teaching video frame through an OCR algorithm; S2, obtaining named entities that may become knowledge points through entity recognition technology, and obtaining target knowledge points based on the named entities; S3, performing entity recognition on the target knowledge points, matching target review content for the target knowledge points and marking the target knowledge points; S4, performing entity alignment on the target knowledge points and the corresponding target review content, and obtaining target review content that is essentially the same as the target knowledge point content; S5, recording the link of the scene where the corresponding target review content is located and constructing an association link between knowledge and the target knowledge point. The present application can save the course scenes where the knowledge points are located, connect concepts that are essentially the same, and provide them for selection in the scene selection, which can promote learners to consolidate their learning and improve the effect of online learning.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis and processing, and in particular to an online multi-scenario reinforcement learning method. Background Art

[0002] Since 2012, distance learning, such as online video learning, has become increasingly popular. However, due to the lack of in-person teacher interaction, teachers are unable to assess students' progress. Students may also struggle to consolidate their learning due to their lack of independent learning skills. For example, some English learning software currently highlights important words in a text, but due to the shortcomings of online learning, teachers are unable to provide individualized reinforcement learning for each student. Therefore, a reinforcement learning method is needed that can save the course context where the knowledge points are located, allowing for the option to jump back and restore them later during review. This method can connect concepts of similar nature and provide options for context selection. This method can help learners build connections and relationships between knowledge points in their minds, strengthening their memory and consolidating their learning. Summary of the Invention

[0003] The present invention provides an online multi-scenario reinforcement learning method, which can connect concepts of the same nature and provide them for selection in scenario selection. It can promote learners to build context and connections between knowledge in their minds, strengthen memory, and consolidate learning.

[0004] The above technical objectives of the present invention are achieved through the following technical solutions: an online multi-scenario reinforcement learning method, comprising at least the following steps:

[0005] S1, recognize the text data in the current online teaching video frame through the OCR algorithm;

[0006] S2, obtaining named entities that may become knowledge points through entity recognition technology, and obtaining target knowledge points based on the named entities;

[0007] S3, performing entity recognition on the target knowledge point, matching target review content for the target knowledge point and marking the target knowledge point;

[0008] S4, performing entity alignment on the target knowledge point and the corresponding target review content to obtain the target review content that is essentially the same as the target knowledge point content;

[0009] S5, recording the link of the scene where the corresponding target review content is located and building an association link between the knowledge and the target knowledge point.

[0010] Preferably, text recognition is performed on the current online teaching video frame to capture text information contained in the current online teaching video frame.

[0011] Preferably, the named entity obtaining step in S2 includes:

[0012] S2.1, extracting character information from the text information;

[0013] S2.2, perform entity location, entity segmentation and implementation extraction based on the character information to obtain the named entity that may become a knowledge point.

[0014] Preferably, the entity type of the named entity is obtained according to semantic analysis, and the named entity is classified into a corresponding entity type.

[0015] Preferably, entity information of the named entity is extracted, and based on the entity information, it is determined whether historical review content already exists in the named entity. If it is determined that the historical review content already exists, it is marked as a knowledge point named entity.

[0016] Preferably, the learning course library is queried, all the historical review contents are associated with the knowledge point naming entity, and all the historical review contents are obtained for association.

[0017] Preferably, the learner's ID information is obtained, and the learned course information is obtained based on the ID information. The streamlined knowledge point named entities are obtained after the mastered knowledge point named entities are eliminated based on the learned course information.

[0018] Preferably, the simplified knowledge point named entity is highlighted, and the target knowledge point named entity is acquired.

[0019] Preferably, the target review content includes historical learning text information, the address of the historical learning text, and historical learning records that are essentially the same as the target knowledge point content;

[0020] The historical learning records include historical learning time, historical learning times and historical learning time consumption.

[0021] Preferably, the S4 includes establishing a jump channel, and the jump channel includes a plurality of target review content options.

[0022] In summary, the present invention has the following beneficial effects.

[0023] 1. It can save the course scenes where the knowledge points are located, and you can choose to jump and restore them when reviewing later. It can connect concepts of the same nature and provide them for selection in the scene selection. It can help learners build the context and connections between knowledge in their minds, strengthen their memory, and consolidate their learning.

[0024] 2. Based on the learning data contained in the learner ID information, personalized and concise knowledge point named entities are generated to accurately guide the learner's review direction and improve the learner's review efficiency.

[0025] 3. Personalized generation of streamlined knowledge point named entities reduces the server's computing workload and improves scene jump response speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0027] Figure 1 Flowchart of an online multi-scenario reinforcement learning method.

[0028] Figure 2 This is a flowchart of the steps for obtaining the named entities in S2 in an online multi-scenario reinforcement learning method. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present invention are clearly and completely described below with reference to the accompanying drawings.

[0030] Example 1

[0031] like Figures 1 to 2 As shown, an online multi-scenario reinforcement learning method includes at least the following steps:

[0032] S1, identifying text data in the current online teaching video frame through the OCR algorithm, and obtaining knowledge points based on the text data analysis.

[0033] Furthermore, text recognition is performed on the current online teaching video frame to capture text information contained in the current online teaching video frame.

[0034] S2, obtaining named entities that may become knowledge points through entity recognition technology, and obtaining target knowledge points based on the named entities; knowledge points are teaching knowledge points extracted from text content or image content contained in the current teaching video.

[0035] The step of obtaining the named entity in S2 includes:

[0036] S2.1, extracting character information from the text information.

[0037] S2.2, perform entity location, entity segmentation and implementation extraction based on the character information to obtain the named entity that may become a knowledge point.

[0038] The entity type of the named entity is obtained based on semantic analysis, and the named entity is classified into a corresponding entity type, for example, entity types include person, organization, location, time expression, quantity, currency value, percentage, etc.

[0039] Entity information of the named entity is extracted, and based on the entity information, whether historical review content already exists in the named entity is determined; if it is determined that the historical review content already exists, it is marked as a knowledge point named entity.

[0040] The learning course library is queried, all the historical review contents are associated with the knowledge point naming entity, and all the historical review contents are obtained for association.

[0041] Obtain the learner's ID information, obtain the completed course information based on the ID information, and remove the mastered knowledge point named entities from the completed course information to obtain a simplified knowledge point named entity. If the mastered knowledge point named entity is not included in the completed course information, the knowledge point named entity is the simplified knowledge point named entity.

[0042] The simplified knowledge point named entity is highlighted, and the target knowledge point named entity is obtained. The simplified knowledge point named entity is the knowledge point that needs to be reviewed obtained by machine calculation and analysis, and the target knowledge point named entity is the knowledge point that the learner actively selects to review.

[0043] S3, the target knowledge point is subject to entity recognition, target review content is matched for the target knowledge point and the target knowledge point is marked; the target knowledge point on any web page on any platform that has been saved by the learner is marked and highlighted.

[0044] The target review content includes historical learning text information, the address of the historical learning text, and historical learning records that are essentially the same as the target knowledge point;

[0045] The historical learning records include the historical learning time, historical learning frequency, and historical learning time. Later, the historical learning records can be used to further optimize review guidance. For example, in different scenarios with the same amount of text or in the same scenario, the historical learning time of the same knowledge point can be compared twice. If the learning time decreases significantly, it is considered that the knowledge point has been mastered. Later, the simplified knowledge point named entities can be further simplified.

[0046] S4, performing entity alignment on the target knowledge point and the corresponding target review content to obtain the target review content that is essentially the same as the target knowledge point content;

[0047] S5, recording the link of the scene where the corresponding target review content is located and building an association link between the knowledge and the target knowledge point.

[0048] The step S5 includes establishing a jump channel, wherein the jump channel includes a plurality of target review content options.

[0049] The technical solution of this application can save the course scenes where the knowledge points are located, and can be selected to jump and restore during later review. It can connect concepts of the same nature and provide them for selection in the scene selection. It can promote learners to build the context and connection between knowledge in their minds, strengthen memory, and consolidate learning.

[0050] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art based on the technical guidance of the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. An online multi-scenario reinforcement learning method, characterized in that: At least the following steps are included: S1, recognize the text data in the current online teaching video frame through the OCR algorithm; S2, obtaining named entities that may become knowledge points through entity recognition technology, and obtaining target knowledge points based on the named entities; S3, performing entity recognition on the target knowledge point, matching target review content for the target knowledge point and marking the target knowledge point; S4, performing entity alignment on the target knowledge point and the corresponding target review content to obtain the target review content that is essentially the same as the target knowledge point content; S5, recording the link of the scene where the corresponding target review content is located and building an association link between the knowledge and the target knowledge point; Performing text recognition on the current online teaching video frame to capture text information contained in the current online teaching video frame; The step of obtaining the named entity in S2 includes: S2.1, extracting character information from the text information; S2.2, performing entity location, entity segmentation, and entity extraction based on the character information to obtain the named entity that may become a knowledge point; Obtaining entity types of the named entities according to semantic analysis, and classifying the named entities into corresponding entity types; Entity information of the named entity is extracted, and based on the entity information, it is determined whether historical review content already exists in the named entity. If it is determined that the historical review content already exists, it is marked as a knowledge point named entity.

2. The online multi-scenario reinforcement learning method according to claim 1, characterized in that: The learning course library is queried, all the historical review contents are associated with the knowledge point naming entity, and all the historical review contents are obtained for association.

3. The online multi-scenario reinforcement learning method according to claim 2, characterized in that: The learner's ID information is obtained, and the learned course information is obtained according to the ID information. The simplified knowledge point named entities are obtained after the mastered knowledge point named entities are eliminated according to the learned course information.

4. The online multi-scenario reinforcement learning method according to claim 3, characterized in that: The simplified knowledge point named entity is highlighted, and the target knowledge point named entity is obtained.

5. The online multi-scenario reinforcement learning method according to claim 1, characterized in that: The target review content includes historical learning text information, the address of the historical learning text, and historical learning records that are essentially the same as the target knowledge point; The historical learning records include historical learning time, historical learning times and historical learning time consumption.

6. The online multi-scenario reinforcement learning method according to claim 1, characterized in that: The step S5 includes establishing a jump channel, wherein the jump channel includes a plurality of target review content options.

Images