Auxiliary reading method, device, equipment and storage medium

Abstract

The embodiment of the application discloses a kind of auxiliary reading method, device, equipment and storage medium, output reading content corresponding first target question;Receive the first reply content input;If the first reply content does not cover all the reply points corresponding to the first target question determined based on the above reading content, at least one round of interaction process is triggered corresponding to the reply point not covered by the first reply content, so that the first reply content and the input reply content received in at least one round of interaction process cover all the reply points corresponding to the first target question.The application outputs the target question corresponding to the reading content, after obtaining the first reply content input by the user, if the first reply content does not cover all the reply points corresponding to the first target question, all the content related to the target question in the reading content is guided to be spoken by the user through at least one round of interaction, to provide a kind of scheme for guiding and assisting user reading learning through human-computer interaction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and more particularly, to an auxiliary reading method and device, equipment and a storage medium. BACKGROUND

[0002] In a traditional after-school self-study scenario of students, when students read and learn, they can only read by themselves, lack effective guidance and assistance, and thus the learning effect is poor. SUMMARY

[0003] Therefore, the present application provides an auxiliary reading method, device, equipment and storage medium to guide and assist users in reading and learning and improve the intelligence of auxiliary reading.

[0004] To achieve the above-mentioned purpose, the present application provides the following scheme:

[0005] An auxiliary reading method comprises:

[0006] outputting a first target question corresponding to the reading content;

[0007] receiving input first reply content;

[0008] If the first reply content does not cover all the reply points corresponding to the first target question, triggering at least one round of interaction process corresponding to the reply points not covered by the first reply content, so that the first reply content and the input reply content received in the at least one round of interaction process cover all the reply points corresponding to the first target question.

[0009] All the reply points corresponding to the first target question are determined based on the reading content.

[0010] The above-mentioned method, optionally, further comprises:

[0011] If the first reply content covers all the reply points corresponding to the first target question, outputting a second target question corresponding to the reading content;

[0012] Alternatively,

[0013] After the first reply content and the input reply content received in the at least one round of interaction process cover all the reply points corresponding to the first target question, outputting a second target question corresponding to the reading content.

[0014] The above-mentioned method, optionally, the at least one round of interaction process comprises:

[0015] In each round of interaction process, a historical interaction record starting from the reading content is obtained;

[0016] generate a to-be-answered question by processing the historical interaction record, all answer points corresponding to the first target question, and the degree of coincidence of each historical reply in the historical interaction record with each answer point corresponding to the first target question by using a natural language processing model, wherein if the historical reply content in the historical interaction record does not cover all answer points corresponding to the first target question, an answer corresponding to the to-be-answered question covers at least part of the answer points that are not covered by the first reply content among all answer points corresponding to the first target question;

[0017] output the to-be-answered question;

[0018] receive reply content input for the to-be-answered question.

[0019] The method can further include:

[0020] In each round of interaction, obtain the degree of coincidence of the reply content in the previous round of interaction with each answer point corresponding to the first target question;

[0021] obtain the historical interaction record starting from the reading content;

[0022] generate a to-be-answered question by processing the historical interaction record, all answer points corresponding to the first target question, and the degree of coincidence of each historical reply in the historical interaction record with each answer point corresponding to the first target question by using a natural language processing model, wherein if the historical reply content in the historical interaction record does not cover all answer points corresponding to the first target question, an answer corresponding to the to-be-answered question covers at least part of the answer points that are not covered by the first reply content among all answer points corresponding to the first target question;

[0023] output the to-be-answered question;

[0024] receive reply content input for the to-be-answered question.

[0025] The method can further include:

[0026] obtain the degree of coincidence of the reply content in the previous round of interaction with each answer point corresponding to the first target question by processing the reply content in the previous round of interaction, the first target question, and each answer point corresponding to the first target question by using a point correction model.

[0027] The method can further include: if the historical reply content in the historical interaction record covers all the reply points corresponding to the first target question, the question to be replied to includes a second target question corresponding to the reading content.

[0028] The method can further include, before outputting the first target question corresponding to the reading content, the following steps:

[0029] processing the reading content to generate a plurality of target questions corresponding to the reading content and reply points corresponding to each target question;

[0030] The first target question is one of the plurality of target questions.

[0031] An auxiliary reading device, the device comprising:

[0032] an output module configured to output a first target question corresponding to reading content;

[0033] a receiving module configured to receive input first reply content;

[0034] a processing module configured to, if the first reply content does not cover all the reply points corresponding to the first target question, trigger at least one round of interaction process corresponding to the reply points not covered by the first reply content, so that the first reply content and the input reply content received in the at least one round of interaction process cover all the reply points corresponding to the first target question.

[0035] All the reply points corresponding to the first target question are determined based on the reading content.

[0036] An auxiliary reading device comprising a memory and a processor;

[0037] The memory is configured to store a program.

[0038] The processor is configured to execute the program to implement each step of the auxiliary reading method according to any one of the above.

[0039] A computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement each step of the auxiliary reading method according to any one of the above.

[0040] From the above technical solution can be seen, the auxiliary reading method, device, equipment and storage medium provided by the embodiment of the application, output the first target question corresponding to the reading content; receive the input first reply content; if the first reply content does not cover all the reply points corresponding to the first target question determined based on the above reading content, trigger at least one round of interaction process corresponding to the reply points not covered by the first reply content, so that the first reply content and the input reply content received in at least one round of interaction process cover all the reply points corresponding to the first target question. The application outputs the target question corresponding to the reading content, and after obtaining the first reply content input by the user, if the first reply content does not cover all the reply points corresponding to the first target question, the user is guided to speak all the content related to the target question in the reading content through at least one round of interaction, providing a scheme for guiding and assisting users to read and learn through human-computer interaction, and improving the intelligence of auxiliary reading. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the provided drawings.

[0042] Figure 1 An implementation flowchart of the auxiliary reading method disclosed by the embodiment of the application;

[0043] Figure 2 An example of the auxiliary reading interface disclosed by the embodiment of the application;

[0044] Figure 3 A structural schematic diagram of the auxiliary reading device disclosed by the embodiment of the application;

[0045] Figure 4 A hardware structure block diagram of the auxiliary reading device disclosed by the embodiment of the application. DETAILED DESCRIPTION

[0046] Before the scheme of the application is described, related concepts are explained and described.

[0047] Prompt: When having a conversation with AI (such as a large language model), instructions need to be sent to AI, which can be a text description, such as "please help me recommend a popular song" input when you have a conversation with AI, or a parameter description in a certain format, such as asking AI to draw according to a certain format, and describing related drawing parameters.

[0048] With reference to the drawings and embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0049] In order to guide and assist users in reading and learning, and improve the intelligence of the auxiliary reading, the present application provides the scheme.

[0050] As shown in Figure 1 An implementation flowchart of the auxiliary reading method provided by the embodiments of the present application can include the following steps.

[0051] Step S101: output the first target question corresponding to the reading content.

[0052] The reading content can be part of the content (for example, a chapter or a paragraph of a reading material) in a reading material (for example, an article, a story, a novel, a picture book, a picture, a video, etc.), or the reading content can be the entire content of a reading material.

[0053] Optionally, the first target question corresponding to the reading content can be output when it is determined that the reading content is read.

[0054] As an example, when a reading completion instruction for the reading content is received, it is determined that the reading content has been read. Based on this, an interactive control can be set for the reading content, and the user can operate the interactive control corresponding to the reading content after reading the reading content to indicate that the auxiliary reading system reading content has been completed. When the auxiliary reading system monitors a preset operation event (for example, a single-click operation, a double-click operation, or a circle selection operation, etc.) of the interactive control, a reading completion instruction is generated, so that it is determined that the reading content has been read.

[0055] Optionally, it can be determined that the reading content is read when it is monitored that the output of the reading content is ended. As an example, the reading content is text, and the current page displays the complete reading content or the last part of the reading content. If a page turning instruction is received, it is determined that the output of the reading content is ended. As an example, the reading content is audio or video, and when it is monitored that the audio or video playback is ended, it is determined that the output of the reading content is ended.

[0056] Optionally, in the case that the reading material is audio or video, when the output progress of the audio or video reaches any one of the question points corresponding to the reading material (at least one question point is preset in the output progress of the reading material, and any one question point corresponds to an audio segment or a video segment between the any one question point and the previous question point in the reading material), it is determined that the user has read the audio segment or the video segment between the any one question point and the previous question point, at which time, the output of the reading material can be paused, and the first target question corresponding to the reading content (i.e., the audio segment or the video segment between the any one question point and the previous question point) can be output.

[0057] In the present application, the reading content corresponds to at least one target question, and each target question corresponds to at least one reply point. The first target question is any one of the at least one target question corresponding to the reading content.

[0058] The at least one target question corresponding to the reading content and the reply point corresponding to each target question can be pre-configured, or can be generated by processing the reading content.

[0059] Step S102: receiving the input first reply content.

[0060] The first reply content can cover all the reply points corresponding to the first target question, can cover part of the reply points corresponding to the first target question, or can not cover any of the reply points corresponding to the first target question.

[0061] Step S103: if the first reply content does not cover all the reply points corresponding to the first target question, triggering at least one round of interactive process corresponding to the reply points not covered by the first reply content, so that the first reply content and the input reply content received in the at least one round of interactive process cover all the reply points corresponding to the first target question.

[0062] All the reply points corresponding to the first target question are determined based on the reading content. As an example, all the reply points corresponding to the first target question can be determined according to a preset correspondence between the target question and the reply point.

[0063] That is, in the case that the first reply content does not cover all the reply points corresponding to the first target question, the present application automatically triggers at least one round of human-computer interaction process to guide the user to input information covering all the reply points corresponding to the first target question, so as to help the user better understand and master the reading content.

[0064] The assisted reading method provided in this application outputs a target question corresponding to the reading content. After obtaining the first response content input by the user, if the first response content does not cover all the answer points corresponding to the first target question, the user is guided to say all the content related to the target question in the reading content through at least one round of human-computer interaction. This provides a solution for guiding and assisting users in reading and learning through human-computer interaction, thereby improving the intelligence of assisted reading.

[0065] In an optional embodiment, the assisted reading method provided in this application may further include:

[0066] If the first response covers all the key points of the answer to the first target question, output the second target question corresponding to the reading content.

[0067] The second objective question is different from the first objective question among the multiple objective questions related to the reading content.

[0068] In other words, if the user's first response covers all the key points of the answer to the first target question, it means that the user understands and has mastered all the key points of the answer to the first target question. There is no need for multiple rounds of guidance. You can directly output the next target question corresponding to the reading content to the user.

[0069] Of course, if all target questions corresponding to the reading content have been output after the first target question is output, it means that the learning guidance for the reading content has ended. If the user's first response covers all the key points of the answer to the first target question, then the assisted reading for the reading content ends, that is, the intelligent guidance process for the reading content ends.

[0070] After outputting the second target question, the system can receive the input of the second response content. If the second response content does not cover all the answer points corresponding to the second target question, the answer points not covered by the second response content will trigger at least one round of interaction, so that the second response content, as well as the input response content received during at least one round of interaction, covers all the answer points corresponding to the second target question.

[0071] All key points for answering the second objective question are determined based on the reading content.

[0072] The second response may cover all the key points of the response to the second target question, or it may cover some of the key points of the response to the second target question, or it may not cover any of the key points of the response to the second target question.

[0073] The processing flow after receiving the second reply from the user can be found in the aforementioned processing flow after receiving the first reply from the user, and will not be repeated here.

[0074] In an optional embodiment, the method for assisting reading provided by the embodiments of the present application can further comprise:

[0075] After the first reply content and the reply content received in the at least one round of interaction process cover all the answer points corresponding to the first target question, a second target question corresponding to the reading content is output.

[0076] After guiding the user to input the information covering all the answer points corresponding to the first target question, if there is a second target question in the reading content that has not been output, the second target question is output to guide the user to input the information covering all the answer points corresponding to the second target question. The interaction process after the second target question is output can refer to the foregoing embodiments, which will not be described here.

[0077] If all the target questions corresponding to the reading content have been output after the first reply content and the reply content received in the at least one round of interaction process cover all the answer points corresponding to the first target question, it indicates that the learning guidance for the reading content has ended. Then, after the first reply content and the reply content received in the at least one round of interaction process cover all the answer points corresponding to the first target question, the assisting reading for the reading content ends, that is, the intelligent guidance process for the reading content ends.

[0078] In an optional embodiment, the at least one round of interaction process triggered by the answer points not covered by the first reply content comprises:

[0079] In any round of the at least one round of interaction process, a historical interaction record starting from the reading content is obtained, that is, the reading content and the interaction record after the reading content is output are obtained as the historical interaction record. That is to say, the historical interaction record is the historical interaction record (including the latest reply content input by the user) from the output of the reading content to the historical interaction record before the round of interaction.

[0080] The historical interaction record and all the answer points corresponding to the first target question are processed by a natural language processing model to generate a to-be-replied question. If the historical reply content in the historical interaction record does not cover all the answer points corresponding to the first target question, the answer corresponding to the to-be-replied question covers at least part of the answer points in the answer points not covered by the first reply content.

[0081] That is, if the historical reply content in the historical interaction record does not cover all the reply points corresponding to the first target question, the question to be replied to is used to guide the user to input information covering part of the reply points in the reply points not covered by the first reply content, or the question to be replied to is used to guide the user to input information covering all the reply points in the reply points not covered by the first reply content.

[0082] If the historical reply content in the historical interaction record covers all the reply points corresponding to the first target question, the second target question will be included in the question to be replied to.

[0083] The natural language processing model of the present application can be a large language model, a multi-modal large language model, a multi-modal super large language model, etc. As an example, the generative model can include but is not limited to: a model of the Transformer architecture, such as GPT (Generative Pre-Training)-3, GPT-4, etc. The generative model can also be other generative models, such as PaLM (Pathways Language Model), T5 (Text-to-Text Transfer Transformer), etc.

[0084] As an example, the above historical interaction record and all the reply points corresponding to the first target question can be added to the preset first guiding instruction instruction template to obtain a first guiding instruction instruction prompt, and the first guiding instruction instruction prompt is input into the natural language processing model to obtain the question to be replied to (i.e. the question to be answered by the user) generated by the natural language processing model.

[0085] Output the question to be replied to.

[0086] Receive the reply content input for the question to be replied to.

[0087] After receiving the reply content input for the question to be replied to, return to execute the step of obtaining the historical interaction record starting from the reading content and the subsequent steps, at this time, the historical interaction record is increased with the above question to be replied to and the reply content input for the question to be replied to.

[0088] The capability of the natural language processing model to generate questions to be replied can be trained by a first training dataset. Each training sample (denoted as a first type of training sample for convenience and distinction) in the first training dataset includes a reading material sample, at least one question corresponding to the reading material sample, at least one reply gist corresponding to each question, and at least one round of dialogue. The at least one round of dialogue is a process in which a teacher asks questions to a student and guides the student to learn the reply gist corresponding to each question in the first type of training sample. Different first type of training samples include different reading materials.

[0089] An example of the first type of training sample provided by an embodiment of the present application is as follows:

[0090] "In the beautiful countryside... \nThere is a reading material above, now you are the reading teacher Xiao Fei, I am the student. Please ask me questions based on the above material, your discussion should focus on these questions: \n0. What hardships did the ugly duckling suffer in childhood? Reference answer: The ugly duckling was beaten, excluded, laughed at everywhere in the duck and chicken groups, the ugly duckling's brothers and sisters and mother also did not like him; met the hunting dog and gun in the marsh; in the farmhouse, the ugly duckling was questioned and mocked by the cat and the hen; in the shrubbery, the ugly duckling was frozen and fainted; in the farmer's house, the ugly duckling was mocked. 1. Although the ugly duckling suffered a lot of hardships in childhood, but he always wanted to go to a wider world. What kind of ugly duckling do you see from this? Reference answer: From this, I see an ugly duckling who is not afraid of difficulties and yearns for beauty. \nNow please ask the first question:

[0091] Teacher: Little friend, can you tell me what hardships did the ugly duckling suffer in the story when he was young?

[0092] Student: The rejection of duck mother and brothers and sisters, the suppression and bullying of other animals.

[0093] Teacher: You answered very well! So in the story, when the ugly duckling saw the beautiful white swan, what did he decide to do to approach these noble birds?

[0094] Student: The ugly duckling flew towards those white swans at the risk of his life.

[0095] Teacher: Great, little friend! So in the story, what qualities do we see from this persistent ugly duckling?

[0096] Student: Always strive to approach what you yearn for, be able to stick to your own ideas and always stick to them without giving up.

[0097] Teacher: That's wonderful! Your understanding makes me very happy. We should cherish our uniqueness and strive to explore and showcase our potential. I hope this story will have a positive impact on your future growth. Thank you for discussing this beautiful story with me! &The End.

[0098] In the example of the first type of training sample above, the phrase "in the beautiful countryside..." should be the main text of the story "The Ugly Duckling". This is an example provided here due to space limitations.

[0099] The process of training a natural language processing model to generate response questions using any first-class training samples can include:

[0100] For any question in any first-class training sample, following the dialogue flow of at least one round of dialogue in that first-class training sample, each time the reading material sample, the student's one response, the interaction history before that response, and the key points of each answer corresponding to that question are input into the natural language processing model, the output of the natural language processing model is obtained as one response content; with the goal of the response content output by the natural language processing model approaching the teacher's evaluation and / or question following the student's response in any first-class training sample, the parameters of the natural language processing model are updated.

[0101] Specifically, when inputting the reading material sample, a student's response content, the interaction history before the response content, and the key points of each answer corresponding to any question into the natural language processing model, the reading material sample, the student's response content, the interaction history before the response content, and the key points of each answer corresponding to any question can be added to a preset first guidance instruction template to obtain a first guidance instruction prompt, and the first guidance instruction prompt can be input into the natural language processing model.

[0102] In the above embodiments, it is not necessary to identify whether the user's input reply covers all the key points of the answer to the first target question. At least one round of interaction is triggered directly even if the user's reply does not cover all the key points of the answer to the first target question.

[0103] In an optional embodiment, it is also possible to first identify whether the user's response content covers all the key points of the answer corresponding to the first target question, and then generate a question to be answered based on the identification result, historical interaction records, and all the key points of the answer corresponding to the first target question. Specifically,

[0104] The degree of coincidence of the reply content in the last round of interaction process with each reply gist corresponding to the first target question can be obtained. For any target question and any reply gist corresponding to the target question, the degree of coincidence of the reply content of the user with the reply gist includes the following cases: complete coincidence, partial coincidence or no coincidence. Only when the degree of coincidence of the reply content of the user with each reply gist corresponding to the target question is complete coincidence, the reply content of the user covers all the reply gists corresponding to the target question, otherwise, the reply content of the user does not cover all the reply gists corresponding to the target question.

[0105] The historical interaction record starting from the reading content is obtained, that is, the reading content and the interaction record after outputting the reading content are obtained as the historical interaction record.

[0106] The historical interaction record, all reply gists corresponding to the first target question and the degree of coincidence of each historical reply in the historical interaction record with each reply gist corresponding to the first target question are processed by a natural language processing model to generate a question to be replied; wherein if the degree of coincidence of each historical reply content in the historical interaction record with each reply gist corresponding to the first target question represents that each historical reply content in the historical interaction record does not cover all the reply gists corresponding to the first target question, the answer corresponding to the question to be replied covers at least part of the reply gists that are not covered by the first reply content among all the reply gists corresponding to the first target question.

[0107] That is, if the degree of coincidence of each historical reply content in the historical interaction record with each reply gist corresponding to the first target question represents that each historical reply content in the historical interaction record does not cover all the reply gists corresponding to the first target question, the question to be replied is used to guide the user to input information covering at least part of the reply gists that are not covered by the first reply content.

[0108] If the degree of coincidence of each historical reply content in the historical interaction record with each reply gist corresponding to the first target question represents that each historical reply content in the historical interaction record covers all the reply gists corresponding to the first target question, the second target question will be included in the question to be replied.

[0109] As an example, the above historical interaction record, all reply points corresponding to the first target question, and the degree of coincidence of each historical reply in the historical interaction record with each reply point corresponding to the first target question (of course, some historical replies do not calculate the degree of coincidence with each reply point corresponding to the first target question. At this time, the degree of coincidence of the historical reply with each reply point corresponding to the first target question is empty) can be added to the preset second guidance instruction template to obtain a second guidance instruction prompt. The second guidance instruction prompt is input into the natural language processing model to obtain the question to be replied (i.e., the question to be answered by the user) generated by the natural language processing model.

[0110] By obtaining the degree of coincidence of each historical reply in the historical interaction record with each reply point corresponding to the first target question, the natural language processing model can more clearly know whether the historical reply content in the historical interaction record covers all the reply points corresponding to the first target question, thereby further improving the accuracy of the natural language processing model.

[0111] Output the question to be replied.

[0112] Receive the reply content input for the question to be replied.

[0113] The capability of the natural language processing model to generate the question to be replied can be trained by a second training data set. Each training sample (referred to as a second type of training sample for ease of description and differentiation) in the second training data set includes a reading material sample, at least one question corresponding to the reading material sample, at least one reply point corresponding to each question, at least one round of dialogue, and the degree of coincidence of the reply content of the student in each round of dialogue with each reply point. The at least one round of dialogue is the process of the teacher asking the student questions and guiding the student to learn the reply points corresponding to each question in the second type of training sample. Different second type of training samples include different reading materials.

[0114] The process of training the capability of the natural language processing model to generate the question to be replied by using any second type of training sample can include:

[0115] For any question in any second type of training sample, according to the dialogue flow of at least one round of dialogue in the any second type of training sample, the reading material sample, the reply content of the student, the respective reply key points corresponding to the any question, the degree of coincidence of the reply content with the respective reply key points corresponding to the any question, and the interactive history record before the reply content are input into the natural language processing model to obtain the reply content to be replied by the natural language processing model.

[0116] When the reading material sample, the reply content of the student, the respective reply key points corresponding to the any question, the degree of coincidence of the reply content with the respective reply key points corresponding to the any question, and the interactive history record before the reply content are input into the natural language processing model, the reading material sample, the reply content of the student, the respective reply key points corresponding to the any question, the degree of coincidence of the reply content with the respective reply key points corresponding to the any question, and the interactive history record before the reply content can be added to a preset second guidance instruction template to obtain a second guidance instruction prompt, and the second guidance instruction prompt is input into the natural language processing model.

[0117] In an optional embodiment, the above-mentioned implementation manner of obtaining the degree of coincidence of the reply content in the previous round of interaction with the respective reply key points corresponding to the first target question can not include:

[0118] The reply content in the previous round of interaction, the first target question, and the respective reply key points corresponding to the first target question are processed by the key point correction model to obtain the degree of coincidence of the reply content in the previous round of interaction with the respective reply key points corresponding to the first target question.

[0119] The key point correction model can be a model specially used for identifying the degree of coincidence of the reply content in the previous round of interaction with the respective reply key points corresponding to the first target question, or can be the aforementioned natural language processing model.

[0120] As an example, the reply content of the student, the any question in the any second type of training sample, and the respective reply key points corresponding to the any question can be added to a preset correction instruction template to obtain a correction instruction prompt, and the correction instruction prompt is input into the natural language processing model to obtain the degree of coincidence of the reply content of the student generated by the natural language processing model with the respective reply key points corresponding to the any question.

[0121] The ability of a natural language processing model to identify how well a user's response matches the key points of the question can be obtained by training on the second training dataset mentioned above.

[0122] The process of training a natural language processing model to identify the degree of conformity of each answer to a question using any second-type training sample may include:

[0123] According to the dialogue flow of at least one round of dialogue in any of the second type of training samples, the student's one-time response, any question in any of the second type of training samples and its corresponding key points of each response are input into the natural language processing model. The output of the natural language processing model is the degree of conformity between the student's one-time response and the key points of each response to any question. The parameters of the natural language processing model are updated with the goal of making the degree of conformity between the student's one-time response and the key points of each response to any question in any of the second type of training samples approach the level of conformity between the student's one-time response and the key points of each response to any question.

[0124] When a student's response, any question in any second-type training sample, and its corresponding key points of each response are input into the natural language processing model, the student's response, any question in any second-type training sample, and its corresponding key points of each response can be added to a preset grading instruction template to obtain a grading instruction prompt. The grading instruction prompt is then input into the natural language processing model.

[0125] In an optional embodiment, before outputting the first target question corresponding to the reading content, the assisted reading method of this application may further include:

[0126] The reading content is processed to generate multiple target questions corresponding to the content, as well as key points for answering each target question. The first and second target questions are among these multiple target questions.

[0127] As an example, the above natural language processing model can be used to process the reading content to generate multiple target questions corresponding to the reading content, as well as the key points of the answers to each target question.

[0128] As an example, the above reading content can be added to a preset question prompt template to obtain a question prompt. The question prompt can then be input into a natural language processing model to obtain multiple target questions corresponding to the reading content generated by the natural language processing model, as well as the answer points for each target question.

[0129] The capability of the natural language processing model to generate a target question and a corresponding answer point based on reading content can be trained by a third training data set. Each training sample in the third training data set (referred to as a third type of training sample for the sake of description and differentiation) includes a reading material sample, at least one question corresponding to the reading material, and an answer point corresponding to each question.

[0130] The process of utilizing any third type of training sample to train the capability of the natural language processing model to generate a target question and a corresponding answer point based on reading content can include:

[0131] The reading material sample in the third type of training sample is input into the natural language processing model to obtain at least one question generated by the natural language model and an answer point corresponding to each question. The parameters of the natural language processing model are updated with the goal of the at least one question generated by the natural language model and the answer point corresponding to each question approaching the at least one question and the answer point corresponding to each question in the third type of training sample.

[0132] When the reading material sample in the third type of training sample is input into the natural language processing model, the reading material sample can be added to a preset question indication instruction template to obtain a question indication instruction prompt, and the question indication instruction prompt is input into the natural language processing model.

[0133] In the above embodiments, the training process of various capabilities of the natural language processing model is described. In an optional embodiment, before training the natural language processing model for various capabilities, the natural language processing model can be pre-trained in an unsupervised manner using general domain corpus (including but not limited to news articles, books, web page texts, videos, audios, etc.). The goal is to enable the natural language processing model to understand and master basic language rules and knowledge in a broad language environment. In the unsupervised training stage, for any text domain corpus, the domain corpus is input into the natural language processing model in sequence by character, and the next character of the input string of the natural language processing model is predicted by the natural language processing model until the entire domain corpus is input into the natural language processing model. The parameters of the natural language processing model are updated with the goal of the output string of the natural language processing model approaching the string after the first character in the domain corpus. For any audio or video domain corpus, the domain corpus is input into the natural language processing model in sequence by audio frame or video frame, and the next audio frame or video frame of the input audio frame or video frame of the natural language processing model is predicted by the natural language processing model until the entire domain corpus is input into the natural language processing model. The parameters of the natural language processing model are updated with the goal of the output audio frame or video frame of the natural language processing model approaching the audio frame or video frame after the first audio frame or video frame in the domain corpus.

[0134] After unsupervised pre-training is completed, before various field capability training is performed, special corpus in education, literature, dialogue and the like can be collected to perform supervised fine-tuning (SFT) training on the model, so that the natural language processing model can better understand and generate text in a specific field. In the SFT stage, specific instruction instructions and standard answers are designed by experts for input data in the special corpus of the specific field, and when training, the input data is input into the natural language processing model after specific instruction instructions are constructed, so that the answer output by the natural language processing model approaches the standard answer.

[0135] As shown in FIG. 1, an example of an auxiliary reading interface provided by an embodiment of the present application is provided, in which the auxiliary reading device plays a video for a user (a child), when the video plays to Long Socks Pippi, the auxiliary reading device guides the user to answer the question "Can you describe the appearance of Long Socks Pippi in combination with the picture?". Figure 2 As shown in FIG. 2, the auxiliary reading device pauses the video when the video plays to the interface as shown in FIG. 2, and displays the question "Can you describe the appearance of Long Socks Pippi in combination with the picture?" in the video interface, and displays the question "Can you tell me the appearance characteristics of Long Socks Pippi?" in the interactive interface (the same question as the video interface, but the description is different), so that the child describes the appearance characteristics of Long Socks Pippi, the child says that his feet are very large, the auxiliary reading device evaluates the child's answer, and guides the child to describe the features of Pippi's hair, nose and clothes, the child adds that his hair is yellow, his nose is large, and his clothes are blue, and then the auxiliary reading device supplements the child's description, and guides the child to observe the color of Pippi's socks, the child says that the description of Pippi's socks is a striped color, the auxiliary reading device affirms the child's answer, and guides the child to describe Pippi's appearance using the features of Pippi's hair, nose, clothes and socks, then the child describes Pippi's appearance in detail according to the information obtained in the previous interaction process, and finally the auxiliary reading device evaluates and summarizes the child's description. Figure 2 In the above example, the auxiliary reading device can output the interaction content in the form of voice when interacting with the child, and the child can input the interaction content in the form of voice.

[0136] Corresponding to the method embodiment, the present application also provides an auxiliary reading device, and a structure diagram of an auxiliary reading device provided by an embodiment of the present application is shown in FIG. 3, which can include:

[0137] Figure 3

[0138] ​​The output module 301, the receiving module 302 and the processing module 303; wherein

[0139] The output module 301 is configured to output a first target question corresponding to the reading content.

[0140] The receiving module 302 is configured to receive input first reply content.

[0141] The processing module 303 is configured to, if the first reply content does not cover all reply points corresponding to the first target question, trigger at least one round of interaction process corresponding to the reply points not covered by the first reply content, so that the first reply content and the input reply content received in the at least one round of interaction process cover all reply points corresponding to the first target question.

[0142] All reply points corresponding to the first target question are determined based on the reading content.

[0143] The auxiliary reading device provided by the embodiments of the present application outputs a target question corresponding to the reading content, and after obtaining the input first reply content, if the first reply content does not cover all reply points corresponding to the first target question, guides the user to speak all content related to the target question in the reading content through at least one round of interaction, thereby providing a scheme of guiding and assisting the user in reading and learning through human-computer interaction, and improving the intelligence of the auxiliary reading.

[0144] In an optional embodiment, the output module 301 is further configured to:

[0145] If the first reply content covers all reply points corresponding to the first target question, output a second target question corresponding to the reading content.

[0146] Alternatively,

[0147] After the first reply content and the input reply content received in the at least one round of interaction process cover all reply points corresponding to the first target question, output a second target question corresponding to the reading content.

[0148] In an optional embodiment, when the processing module 303 performs the at least one round of interaction process, it is configured to:

[0149] In each round of interaction process, a historical interaction record starting from the reading content is obtained.

[0150] generate a question to be replied, by processing the historical interaction record, all the answer points corresponding to the first target question, and the degree of coincidence of each historical reply in the historical interaction record with each answer point corresponding to the first target question, by using a natural language processing model, wherein if the historical reply content in the historical interaction record does not cover all the answer points corresponding to the first target question, the answer corresponding to the question to be replied covers at least part of the answer points in the answer points corresponding to the first target question which are not covered by the first reply content;

[0151] output the question to be replied;

[0152] receive the reply content input for the question to be replied.

[0153] In an optional embodiment, the processing module 303 is configured to, when performing at least one round of interactive process:

[0154] In each round of interactive process, obtain the degree of coincidence of the reply content in the previous round of interactive process with each answer point corresponding to the first target question;

[0155] obtain the historical interaction record starting from the reading content;

[0156] generate a question to be replied, by processing the historical interaction record, all the answer points corresponding to the first target question, and the degree of coincidence of each historical reply in the historical interaction record with each answer point corresponding to the first target question, by using a natural language processing model, wherein if the historical reply content in the historical interaction record does not cover all the answer points corresponding to the first target question, the answer corresponding to the question to be replied covers at least part of the answer points in the answer points corresponding to the first target question which are not covered by the first reply content;

[0157] output the question to be replied;

[0158] receive the reply content input for the question to be replied.

[0159] In an optional embodiment, the processing module 303 obtains the degree of coincidence of the reply content in the previous round of interactive process with each answer point corresponding to the first target question, by:

[0160] obtain the degree of coincidence of the reply content in the previous round of interactive process with each answer point corresponding to the first target question, by processing the reply content in the previous round of interactive process, the first target question, and each answer point corresponding to the first target question, by using a point grading model.

[0161] In an optional embodiment, if the historical reply content in the historical interaction record covers all the reply points corresponding to the first target question, the question to be replied to includes the second target question corresponding to the reading content.

[0162] In an optional embodiment, the auxiliary reading device further includes a generating module configured to:

[0163] Before outputting the first target question corresponding to the reading content, the reading content is processed to generate a plurality of target questions corresponding to the reading content and reply points corresponding to each target question;

[0164] The first target question belongs to the plurality of target questions.

[0165] The auxiliary reading device provided by the embodiments of the present application can be applied to an auxiliary reading device, such as a PC terminal, a mobile terminal, a cloud platform, a learning machine, a server, a server cluster, and the like. Optionally, Figure 4 A hardware structure block diagram of the auxiliary reading device is shown, and refer to Figure 4 The hardware structure of the auxiliary reading device can include at least one processor 1, at least one communication interface 2, at least one memory 3, and at least one communication bus 4.

[0166] In the embodiments of the present application, the number of the processor 1, the communication interface 2, the memory 3, and the communication bus 4 is at least one, and the processor 1, the communication interface 2, and the memory 3 complete the communication among each other through the communication bus 4.

[0167] The processor 1 can be a central processing unit CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application, etc.

[0168] The memory 3 can include a high-speed RAM memory, and can also include a non-volatile memory, such as at least one disk memory.

[0169] The memory stores a program, and the processor can call the program stored in the memory, and the program is configured to:

[0170] Output the first target question corresponding to the reading content.

[0171] Receive the input first reply content.

[0172] if the first reply content does not cover all the reply points corresponding to the first target question, triggering at least one round of interactive process corresponding to the reply points not covered by the first reply content, so that the first reply content and the inputted reply content received in the at least one round of interactive process cover all the reply points corresponding to the first target question.

[0173] All the reply points corresponding to the first target question are determined based on the reading content.

[0174] Optionally, the refinement function and the extension function of the program can refer to the above description.

[0175] The embodiment of the application further provides a storage medium which can store a program suitable for processor execution, and the program is used for:

[0176] outputting a first target question corresponding to reading content;

[0177] receiving inputted first reply content;

[0178] if the first reply content does not cover all the reply points corresponding to the first target question, triggering at least one round of interactive process corresponding to the reply points not covered by the first reply content, so that the first reply content and the inputted reply content received in the at least one round of interactive process cover all the reply points corresponding to the first target question;

[0179] All the reply points corresponding to the first target question are determined based on the reading content.

[0180] Optionally, the refinement function and the extension function of the program can refer to the above description.

[0181] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the application.

[0182] In several embodiments provided in the application, it should be understood that the disclosed system, device and method can be realized in other ways. In addition, the coupling or direct coupling or communication connection between the displayed or discussed elements can be indirect coupling or communication connection through some interfaces, devices or units, and can be electrical, mechanical or other forms.

[0183] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, i.e. may be located in one place, or may be distributed on multiple network units. Part or all of the units may be selected to achieve the purpose of the embodiment scheme according to actual needs.

[0184] In addition, the functional units in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present alone, or two or more units can be integrated in one unit.

[0185] If the functions are realized in the form of software functional units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the parts of the technical solutions that essentially contribute to the prior art can be embodied in the form of software products, which are stored in a storage medium and include a number of instructions for making a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in the embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0186] Finally, it should be noted that in this paper, relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the processes, methods, articles or devices including a series of elements not only include those elements, but also include other elements not explicitly listed or inherent to such processes, methods, articles or devices. Without more limitations, the element defined by the statement "including a…" does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0187] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other.

[0188] The foregoing description of the disclosed embodiments enables a person skilled in the art to make or use the application. Modifications of these embodiments will occur to persons of skill in the art, and that the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Therefore, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for assisting reading, characterized in that, include: Output the primary objective question corresponding to the reading content; Receive the first response content input; If the first response content does not cover all the key points of the answer corresponding to the first target question, at least one round of interaction is triggered for the key points of the answer not covered by the first response content, so that the first response content and the input response content received in the at least one round of interaction cover all the key points of the answer corresponding to the first target question; All key points for the response to the first target question are determined based on the reading content.

2. The method according to claim 1, characterized in that, Also includes: If the first response covers all the key points of the answer to the first target question, output the second target question corresponding to the reading content; or, After the first response content and the input response content received during the at least one round of interaction cover all the key points of the answer corresponding to the first target question, the second target question corresponding to the reading content is output.

3. The method according to claim 1, characterized in that, The at least one round of interaction process includes: In each round of interaction, a historical interaction record starting from the reading content is obtained; The historical interaction records and all key points of the responses corresponding to the first target question are processed using a natural language processing model to generate a question to be answered. Output the question to be answered; Receive the response content input for the question to be answered.

4. The method according to claim 1, characterized in that, The at least one round of interaction process includes: In each round of interaction, the degree to which the response content in the previous round of interaction conforms to each key point of the response to the first target question is obtained; Obtain the historical interaction record starting from the read content; The natural language processing model is used to process the historical interaction records, all the key points of the responses to the first target question, and the degree of conformity of each historical response in the historical interaction records to the key points of each response to the first target question, to generate a question to be answered. Output the question to be answered; Receive the response content input for the question to be answered.

5. The method according to claim 4, characterized in that, The degree to which the responses from the previous round of interaction conform to the key points of each answer to the first target question includes: The key point correction model processes the response content from the previous round of interaction, the first target question, and the key points of each response to the first target question to obtain the degree of conformity between the response content from the previous round of interaction and the key points of each response to the first target question.

6. The method according to any one of claims 3-5, characterized in that, If the historical response content in the historical interaction record covers all the key points of the response to the first target question, the question to be answered includes the second target question corresponding to the reading content.

7. The method according to claim 1, characterized in that, Before outputting the first target question corresponding to the reading content, it also includes: The reading content is processed to generate multiple target questions corresponding to the reading content, as well as key points for answering each target question; The first objective problem belongs to the plurality of objective problems.

8. An auxiliary reading device, characterized in that, The device includes: The output module is used to output the first target question corresponding to the reading content; The receiving module is used to receive the first response content input; The processing module is configured to trigger at least one round of interaction for the answer points not covered by the first response content if the first response content does not cover all the answer points corresponding to the first target question, so that the first response content and the input response content received during the at least one round of interaction cover all the answer points corresponding to the first target question. All key points for the response to the first target question are determined based on the reading content.

9. An auxiliary reading device, characterized in that, Including memory and processor; The memory is used to store programs; The processor is configured to execute the program to implement the various steps of the assisted reading method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the assisted reading method as described in any one of claims 1-7.

Images