Internet-based language teaching automatic test paper composition and evaluation method, medium and equipment

Abstract

The present invention relates to an automatic test paper composition and evaluation method, medium and device for language teaching based on the Internet, which realizes the automatic generation of articles in reading questions by going from a first text theme to a first article context and then to the form of a first basic text, performs knowledge point recognition on the first basic text, and then selects a first to-be-modified sentence segment from a first basic sentence segment through a first modification information, allocates a plurality of first to-be-modified sentence segments based on a first difficulty value, and replaces the first to-be-modified sentence segment with the selected first candidate sentence segment until the plurality of first to-be-modified sentence segments are replaced in sequence, obtains a first modified text, provides answers for a terminal, and records a first answer result and a first examination score. The first modified text obtained at the end is made to be more in line with the requirements of the first modification information and the first difficulty configuration information, realizes the controllable adjustment of the text of the reading article, and is more in line with the requirements of the examination.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an Internet-based language teaching automatic test paper composition and evaluation method, medium and device. Background Art

[0002] Cloze is a form of exercise for English language learning and testing that requires students to choose the correct word or phrase to fill in the blank given according to the context and grammar rules. Reading practice is a form of exercise that helps students improve their reading comprehension and language skills. By reading different types of texts, students can improve their reading speed, understand the main idea and details of the text, and develop critical thinking skills.

[0003] Existing cloze and reading articles are generally completed through article selection, independent writing, and textbook text selection. This method has many problems, such as the generated articles cannot fully cover the knowledge points tested, the writing logic becomes deliberate in order to use multiple knowledge points tested, and the reading difficulty is uncontrollable, resulting in poor adjustability of the articles. Summary of the invention

[0004] In view of the above problems, the present invention provides an Internet-based language teaching automatic test paper composition and evaluation method, medium and device, which solves the problem of poor adjustability of articles used in existing reading test questions.

[0005] To achieve the above-mentioned object, in a first aspect, the present invention provides an Internet-based language teaching automatic test paper composition and evaluation method, the method comprising:

[0006] Obtaining a first text topic, inputting the first text topic and a vocabulary library into a first training model, and obtaining a first vocabulary set associated with the first text topic, wherein the first vocabulary set contains a plurality of first words;

[0007] Generate a first article context according to the first text theme, and obtain a first basic text according to the first article context;

[0008] Performing segmentation on the first basic text to obtain a first basic segment set, wherein the first basic segment set includes a plurality of first basic segments;

[0009] Performing knowledge point recognition on each first basic sentence segment to obtain a plurality of first examination knowledge points and a first vocabulary associated with each first examination knowledge point, mapping and storing the first examination knowledge point and the corresponding first vocabulary to obtain a first examination knowledge point set;

[0010] Acquire first modification information, the first modification information including a first set of required knowledge points, the first set of required knowledge points including at least part of the first examination knowledge points;

[0011] According to the first modification information, a plurality of first basic segments to be modified are selected from the first basic segment set and recorded as first segments to be modified;

[0012] Inputting the first to-be-modified segments into the second training model one by one to obtain a plurality of first candidate segments corresponding to each first to-be-modified segment;

[0013] Acquire first difficulty configuration information, and generate a first difficulty value for each first sentence segment to be modified according to the first difficulty configuration information;

[0014] Perform the following steps for each first segment to be modified:

[0015] Selecting a first candidate sentence segment within a first difficulty threshold value range from among the plurality of first candidate sentence segments, recording the first candidate sentence segment as a first replacement sentence segment, and replacing the first to-be-modified sentence segment with the first replacement sentence segment to obtain a first modified text;

[0016] Obtain a first inspection text according to the first modified text, and display the first inspection text on the terminal;

[0017] The first answer result of the collection terminal is output, the first examination score is output, and the first answer result and the first examination text are mapped and stored.

[0018] Further, the first text topic is obtained according to the following method:

[0019] Constructing a first basic document set, the first basic document set including at least one of hot articles within a first preset time period, test question documents within a second preset time period, and subject documents within a third preset time period;

[0020] Inputting the first basic document set into the third training model, obtaining a plurality of first keywords and a semantic association degree between each first keyword and other first keywords, recorded as a first semantic association degree;

[0021] Based on the first semantic association degree, a vocabulary relationship network corresponding to each first keyword is constructed, which is recorded as the first relationship network;

[0022] Obtaining a plurality of first association sets according to the plurality of first relationship networks, each first association set including a plurality of first keywords with a close association degree;

[0023] At least one first text topic is extracted from the plurality of first association sets according to the syntactic dependency relationship.

[0024] Furthermore, generating a first article context according to the first text theme, and obtaining a first basic text according to the first article context also includes:

[0025] Acquire a first article selection associated with a first text topic in an article database, wherein the article database stores a plurality of first articles, each of which includes a plurality of first paragraphs, and the first article selection is a plurality of first paragraphs in a first article;

[0026] Marking the first article selection with words to obtain a plurality of second keywords, and dividing the first article selection according to the second keywords to obtain the sentence segments where the second keywords are located, which are recorded as second key sentence segments;

[0027] Arranging the paragraph dependencies of the plurality of second key sentences according to the logical order of the preceding and following contexts to obtain the first article context corresponding to the first article selection, wherein the paragraph dependencies include at least one of a parallel relationship, a progressive relationship, and a general-specific relationship;

[0028] The first article context is input into the fourth training model to obtain a first basic text.

[0029] Furthermore, the first article excerpt associated with the first text topic is obtained according to the following method:

[0030] Filtering at least one first article from an article database according to a first text topic;

[0031] Performing text theme analysis on the first paragraphs of the first articles selected one by one to obtain theme information of each first paragraph, which is recorded as the theme of the first paragraph;

[0032] Constructing a second semantic relevance degree set for each first paragraph, wherein the second semantic relevance degree set includes the second semantic relevance degrees of the remaining first paragraphs and the current first paragraph;

[0033] Matching the first paragraph topics corresponding to the first paragraphs with the first text topics one by one, obtaining a number of first paragraphs with the highest degree of relevance to the first text topic, which are recorded as first related paragraphs;

[0034] Obtaining a second semantic relevance set of the first associated paragraphs, and selecting the first paragraphs that meet a first preset condition as second associated paragraphs, wherein the first preset condition is included in the second semantic relevance set corresponding to the first associated paragraphs, and the second semantic relevance degree with the first associated paragraphs is within a preset relevance degree threshold range;

[0035] Integrate the first related paragraph and the second related paragraph according to the order of writing to obtain the first article excerpt.

[0036] Further, obtaining the first modification information includes:

[0037] Obtain user information, including basic information, historical answer information, and learning level information corresponding to the current user;

[0038] Generate a first set of knowledge points to be tested and a second set of knowledge points to be tested according to the user information, wherein the first set of knowledge points to be tested includes a plurality of first knowledge points to be tested, which are knowledge points marked as difficult to understand in the historical answer information, and the second set of knowledge points to be tested includes a plurality of second knowledge points to be tested, which are knowledge points that the user is required to master in the teaching plan;

[0039] Obtain a first test weight value for each first knowledge point to be tested according to the historical answer information, and obtain a second test weight value for each second knowledge point to be tested according to the teaching plan;

[0040] Obtaining a first preset number of first knowledge points to be checked according to the learning level information, and obtaining a second preset number of second knowledge points to be checked according to the learning level information;

[0041] Arrange the first knowledge points to be tested in descending order of the first test weight values, and extract a first preset number of first knowledge points to be tested that are arranged in front of the first knowledge point set to be tested, and record them as first basic examination knowledge points;

[0042] Arrange the second knowledge points to be tested in descending order of the second test weight values, and extract a second preset number of second knowledge points to be tested that are arranged in front of the second knowledge point set to be tested, and record them as second basic examination knowledge points;

[0043] The first basic examination knowledge points and the second basic examination knowledge points are integrated to obtain a first set of compulsory examination knowledge points, and first modification information is generated according to the first set of compulsory examination knowledge points.

[0044] Furthermore, obtaining the first test weight value of each first knowledge point to be tested according to the historical answer information also includes:

[0045] A corresponding answer record table is set for each first knowledge point to be tested, wherein the answer record table records the answer time, and the number of answers and answer results of the questions of the current first knowledge point to be tested arranged in the order of answer time;

[0046] In the answer record table, select the number of wrong answers to the first knowledge point to be tested and record it as the number of wrong answers;

[0047] Calculate the time interval between the answering times corresponding to two consecutive wrong answer results, and record it as the wrong question interval period;

[0048] The number of wrong answers and the wrong answer interval period are mapped and stored to obtain a wrong answer record set;

[0049] A correct rate prediction curve is obtained according to the set of wrong question records, wherein the independent variable of the correct rate prediction curve is the time interval, and the dependent variable of the correct rate prediction curve is the predicted answer correct rate of the current first knowledge point to be tested;

[0050] Calculate the time interval between the current timestamp and the last answer time to obtain a first time interval to be tested, and obtain a first predicted answer accuracy of the first knowledge point to be tested according to the first time interval to be tested and the accuracy prediction curve;

[0051] A first test weight value is generated according to the first predicted answer accuracy rate.

[0052] Furthermore, obtaining the first investigation text according to the first modified text also includes:

[0053] Obtaining test question type information, which includes reading questions and / or cloze questions;

[0054] Generate an assessment outline based on the assessment question type information, which includes the question type and assessment knowledge points;

[0055] Generate multiple final examination questions based on the examination outline;

[0056] generating a first examination text according to a plurality of final examination questions and a first revised text;

[0057] Generating multiple final examination questions based on the examination outline also includes:

[0058] Determine one by one whether the test questions in the test outline belong to the category of vocabulary examination. If so, obtain the test vocabulary corresponding to the test question, and generate a similar vocabulary set based on the test vocabulary. The similar vocabulary set includes test vocabulary and confusing vocabulary. The confusing vocabulary includes vocabulary with similar meaning or similar word form to the test vocabulary.

[0059] Generate multiple initial test questions corresponding to the test words of the current test question according to the similar word set;

[0060] Acquire second difficulty configuration information, and generate a second difficulty value corresponding to each initial assessment question according to the second difficulty configuration information;

[0061] An initial examination question corresponding to a second difficulty level value closest to the learning level information of the user information is selected from the plurality of initial examination questions, and is recorded as a final examination question.

[0062] Furthermore, the method also includes:

[0063] Determine whether the first examination result is qualified, and if not, obtain the first answer result and the first examination text;

[0064] Filter out the second test knowledge points based on the first answer result and the first test text;

[0065] Generating a second text theme according to the second inspection knowledge point, inputting the second text theme and the vocabulary library into the first training model to obtain a second vocabulary set associated with the second text theme, wherein the second vocabulary set contains a plurality of second words;

[0066] Generate a second article context according to the second text theme, and obtain a second basic text according to the second article context;

[0067] Performing segmentation on the second basic text to obtain a second basic segment set, wherein the second basic segment set includes a plurality of second basic segments;

[0068] Performing knowledge point recognition on each second basic sentence segment to obtain a plurality of second examination knowledge points and second vocabulary associated with each second examination knowledge point, mapping and storing the second examination knowledge points and the second vocabulary to obtain a second examination knowledge point set;

[0069] Acquire second modification information, the second modification information including a second set of compulsory knowledge points, the second set of compulsory knowledge points including at least part of the second examination knowledge points;

[0070] According to the second modification information, a plurality of second basic segments to be modified are selected from the second basic segment set and recorded as second segments to be modified;

[0071] Inputting the second to-be-modified segments into the second training model one by one to obtain a plurality of second candidate segments corresponding to each second to-be-modified segment;

[0072] Acquire third difficulty configuration information, and generate a third difficulty value for each second sentence segment to be modified according to the third difficulty configuration information;

[0073] Perform the following steps for each second segment to be modified:

[0074] Selecting a second candidate sentence segment within a third difficulty threshold value range from among the plurality of second candidate sentence segments, recording the second candidate sentence segment as a second replacement sentence segment, and replacing the second sentence segment to be modified to obtain a second modified text;

[0075] A second inspection text is obtained according to the second modified text, and the second inspection text is displayed on the terminal.

[0076] In a second aspect, the present invention further provides a computer-readable storage medium storing computer program instructions, which implement the method described in the first aspect when executed by a processor.

[0077] In a third aspect, the present invention further provides an electronic device, comprising a memory and a processor, wherein the memory is used to store one or more computer program instructions, wherein the one or more computer program instructions are executed by the processor to implement the method described in the first aspect.

[0078] Different from the prior art, in the above technical solution, a first text theme is obtained, the first text theme and the vocabulary library are input into a first training model to obtain a first vocabulary set associated with the first text theme, a first article context is generated according to the first text theme, and a first basic text is obtained according to the first article context; the first basic text is segmented to obtain a first basic sentence segment set, knowledge points are identified for each first basic sentence segment to obtain multiple first examination knowledge points and first vocabulary associated with each first examination knowledge point, the first examination knowledge point is mapped and stored with the corresponding first vocabulary to obtain a first examination knowledge point set; first modification information is obtained, and a plurality of first knowledge points that need to be modified are screened out from the first basic sentence segment set according to the first modification information. A basic sentence segment is recorded as the first sentence segment to be modified; the first sentence segments to be modified are input into the second training model one by one to obtain a number of first candidate sentence segments corresponding to each first sentence segment to be modified; the first difficulty configuration information is obtained, and a first difficulty value of each first sentence segment to be modified is generated according to the first difficulty configuration information; a first candidate sentence segment that is within the first difficulty value threshold range is selected from the several first candidate sentence segments, recorded as a first replacement sentence segment, and the first replacement sentence segment is used to replace the first sentence segment to be modified to obtain a first modified text; a first test text is obtained according to the first modified text, and the first test text is displayed on the terminal; the first answer result of the terminal is collected, the first test score is output, and the first answer result is mapped and stored with the first test text. This technical solution realizes the automatic generation of articles in reading questions from the first text theme to the first article context, and then to the form of the first basic text. On this basis, the first basic text is identified with knowledge points to obtain a first test knowledge point set, and then the first to-be-modified sentence segment is screened out from the first basic sentence segment through the first modification information, and a plurality of first candidate sentence segments are generated for the first to-be-modified sentence segment, and the plurality of first candidate sentence segments are allocated based on the first difficulty value, and the plurality of first candidate sentence segments corresponding to each first to-be-modified sentence segment are screened according to the first difficulty value threshold range, and the screened first candidate sentence segments replace the first to-be-modified sentence segment until the plurality of first to-be-modified sentence segments are replaced in sequence to obtain a first modified text, and then the first test text is obtained based on the first modified text and displayed for the terminal to answer, and the first answer result and the first test score are recorded. This method allows the reading articles, that is, the first test text, to partially or completely modify, adjust and replace sentences in the first basic text according to the first difficulty configuration information and in combination with the first set of required knowledge points, so that the final first modified text is more in line with the requirements of the first modification information and the first difficulty configuration information, thereby achieving controllable adjustment of the text of the reading articles and better meeting the examination requirements.

[0079] The above-mentioned records related to the content of the invention are only an overview of the technical solution of the present invention. In order to enable ordinary technicians in the field to more clearly understand the technical solution of the present invention, and then implement it according to the text of the specification and the contents recorded in the drawings, and in order to make the above-mentioned purpose and other purposes, features and advantages of the present invention easier to understand, the following is a description in combination with the specific implementation mode of the present invention and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] The drawings are only used to illustrate the principles, implementations, applications, features and effects of the specific embodiments of the present invention and other related contents, and shall not be considered as limiting the present invention.

[0081] In the drawings of the specification:

[0082] Figure 1 This is a flowchart of steps S101 to S111 of the automatic test paper composition and evaluation method according to a specific embodiment of the present invention;

[0083] Figure 2 This is a flowchart of steps S201 to S205 of the automatic test paper composition and evaluation method according to a specific embodiment of the present invention;

[0084] Figure 3 This is a flowchart of steps S301 to S307 of the automatic test paper composition and evaluation method according to a specific embodiment of the present invention;

[0085] Figure 4 This is a flowchart of steps S401 to S407 of the automatic test paper composition and evaluation method according to a specific embodiment of the present invention;

[0086] Figure 5 This is a diagram of an electronic device for the automatic test paper composition and evaluation method described in a specific implementation mode of the present invention.

[0087] The reference numerals in the above drawings are described as follows:

[0088] 1. Electronic equipment;

[0089] 11. Memory

[0090] 12. Processor. DETAILED DESCRIPTION

[0091] In order to explain in detail the possible application scenarios, technical principles, specific schemes that can be implemented, and the purposes and effects of the present invention, the following is a detailed description in combination with the specific embodiments listed and the accompanying drawings. The embodiments described herein are only used to more clearly illustrate the technical solutions of the present invention, and are therefore only used as examples, and cannot be used to limit the scope of protection of the present invention.

[0092] Reference to "embodiment" herein means that the specific features, structures or characteristics described in conjunction with the embodiment may be included in at least one embodiment of the present invention. The word "embodiment" appearing in various places in the specification does not necessarily refer to the same embodiment, nor does it particularly limit its independence or association with other embodiments. In principle, in the present invention, as long as there is no technical contradiction or conflict, the various technical features mentioned in each embodiment can be combined in any way to form a corresponding implementable technical solution.

[0093] Unless otherwise defined, the technical terms used herein have the same meanings as those generally understood by those skilled in the art to which the present invention belongs; the use of relevant terms herein is only for describing specific embodiments and is not intended to limit the present invention.

[0094] In the description of the present invention, the term "and / or" is an expression used to describe the logical relationship between objects, indicating that three relationships may exist, for example, A and / or B, which means: A exists, B exists, and A and B exist at the same time. In addition, the character " / " in this article generally indicates that the objects before and after are in a logical relationship of "or".

[0095] In the present invention, terms such as "first" and "second" are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship of quantity, priority or sequence between these entities or operations.

[0096] Without further restrictions, in the present invention, the words "include", "comprises", "has" or other similar open-ended expressions used in sentences are intended to cover non-exclusive inclusion. These expressions do not exclude the presence of additional elements in the process, method or product including the elements, so that the process, method or product including a series of elements may include not only those limited elements, but also other elements not explicitly listed, or also include elements inherent to such process, method or product.

[0097] Similar to the understanding in the Examination Guidelines, in the present invention, expressions such as "greater than", "less than", "exceed" and the like are understood to exclude the number itself; expressions such as "above", "below", "within" and the like are understood to include the number itself. In addition, in the description of the embodiments of the present invention, "multiple" means more than two (including two), and similar expressions related to "multiple" are also understood in this way, such as "multiple groups", "multiple times", etc., unless otherwise clearly and specifically limited.

[0098] See also Figure 1 In a first aspect, this embodiment provides an Internet-based language teaching automatic test paper composition and evaluation method, the method comprising:

[0099] S101, obtaining a first text topic, inputting the first text topic and a vocabulary library into a first training model, obtaining a first vocabulary set associated with the first text topic, wherein the first vocabulary set contains a plurality of first words;

[0100] S102, generating a first article context according to the first text theme, and obtaining a first basic text according to the first article context;

[0101] S103, segmenting the first basic text to obtain a first basic segment set, where the first basic segment set includes a plurality of first basic segments;

[0102] S104, performing knowledge point recognition on each first basic sentence segment to obtain a plurality of first examination knowledge points and a first vocabulary associated with each first examination knowledge point, mapping and storing the first examination knowledge point and the corresponding first vocabulary to obtain a first examination knowledge point set;

[0103] S105, obtaining first modification information, wherein the first modification information includes a first set of compulsory knowledge points, and the first set of compulsory knowledge points includes at least part of the first examination knowledge points;

[0104] S106, selecting a plurality of first basic segments to be modified from the first basic segment set according to the first modification information, and recording them as first segments to be modified;

[0105] S107, inputting the first to-be-modified segments into the second training model one by one, and obtaining a plurality of first candidate segments corresponding to each first to-be-modified segment;

[0106] S108, obtaining first difficulty configuration information, and generating a first difficulty value for each first sentence segment to be modified according to the first difficulty configuration information;

[0107] Perform the following steps for each first segment to be modified:

[0108] S109, selecting a first candidate sentence segment within a first difficulty level threshold range from among the plurality of first candidate sentence segments, recording the first candidate sentence segment as a first replacement sentence segment, and replacing the first sentence segment to be modified with the first replacement sentence segment to obtain a first modified text;

[0109] S110, obtaining a first inspection text according to the first modified text, and displaying the first inspection text on a terminal;

[0110] S111, collecting the first answer result of the terminal, outputting the first examination score, and mapping and storing the first answer result and the first examination text.

[0111] In step S101, the first text theme can be selected according to actual needs. The first text theme shown in this embodiment can be pre-generated and stored in the text theme list, or it can be generated in real time based on big data. A method for generating the first text theme is shown below. Optionally, in some other embodiments, the first text theme can be obtained by setting a multi-level label and arranging and combining each level of labels. For example, according to the subject, predicate and object form of the text theme, it is divided into a first-level label, a second-level label and a third-level label. There are multiple subjects in the first-level label, multiple predicates in the second-level label, and multiple objects in the third-level label. Through the arrangement and combination of the first-level label, the second-level label and the third-level label, multiple first text themes can be obtained, and then selected according to actual needs; in some embodiments, the first text theme can also be implemented in a manually generated form, such as manually inputting the text theme, or manually selecting words in the multi-level labels for combination. It should be noted that the first text theme shown in this embodiment can be understood as the central idea of ​​the text summarized in one sentence. For ease of understanding, the articles corresponding to the cloze test and the articles corresponding to the reading questions are all classified as reading articles. The first text theme is also a one-sentence summary of the central idea of ​​the reading articles.

[0112] In this embodiment, the vocabulary library can be a vocabulary library obtained after being divided by grade or subject, or it can be a complete vocabulary library, and the vocabulary library contains all the words or phrases and word groups that need to be mastered in the teaching process. The first text theme and the vocabulary library are input into the first training model to obtain a first vocabulary set, and the first vocabulary set is associated with the first text theme, that is, the first vocabulary contained in the first vocabulary set is the vocabulary in the vocabulary library that the first text theme may need to use in the process of forming an article. In this embodiment, the first training model can be a neural network model obtained by training a convolutional neural network. The first training model is repeatedly trained using a sample text theme and a sample vocabulary library, and the output result of the first training model is a sample vocabulary library associated with the sample text theme. This step can associate the vocabulary in the vocabulary library with the first text theme to form a set centered on the first text theme and the first vocabulary as data.

[0113] In step S102, the first article context is generated according to the first text theme, and the first basic text is obtained based on the first article context. It should be noted that the first vocabulary set is also added on the basis of the first article context, and the first vocabulary and the first article context are used to form the first basic text, that is, the first basic text includes sentences composed of the first vocabulary. The detailed generation process is described below. Furthermore, the first text theme is the central idea expressed in a sentence, and the first article context can be understood as the article outline obtained based on the first text theme. On this basis, the first basic text is developed from the article outline.

[0114] In step S103, the first basic text is segmented. Specifically, punctuation marks contained in the first basic text may be identified, and the period is used as a segmentation feature to obtain a plurality of first basic sentence segments, which are then integrated to form a first basic sentence segment set.

[0115] In step S104, knowledge points are identified for each first basic sentence segment. Specifically, the knowledge points contained in the first basic sentence segment include the use of vocabulary and phrases. Each first basic sentence segment is split according to words as units, and the grammatical relationship between words is identified, so that the division of words and phrases can be realized. For the convenience of distinction, the usage of the vocabulary or phrase that needs to be examined is recorded as the first examination knowledge point. For example, a certain first basic sentence segment contains auxiliary verbs, non-finite verbs and prepositions to form a phrase A that needs to be examined. The use of phrase A is that a connecting verb is required behind phrase A to complete the expression of the general future tense of the entire sentence, then phrase A and the application form of phrase A are recorded as the first examination knowledge point, and the vocabulary associated with the first examination knowledge point is also the first vocabulary, and phrase A is also the first vocabulary. In other words, the first vocabulary can be a single word or a phrase with a specific usage. Based on this idea, knowledge points are identified for each first basic sentence segment one by one to obtain a first examination knowledge point set. It should be noted that not every first basic sentence segment has a first examination knowledge point. For example, there are 20 words in the first vocabulary set, and the first basic article has 40 complete first basic sentence segments. The first words can be selectively distributed in several of the first basic sentence segments. Even a complete first basic sentence segment can have two or more first examination knowledge points. This embodiment does not limit this.

[0116] In step S105, the first modification information is obtained. The first modification information may be pre-set or obtained as shown in the following steps. The first modification information contains a first set of required knowledge points. It should be noted that in this step, the division of the first required knowledge points can be understood as required knowledge points in the current classroom teaching, or as required knowledge points in the teaching objectives of this semester. The difference between the first required knowledge point set and the first examination knowledge point set is that the first required knowledge point set contains at least part of the first examination knowledge points, and the first required knowledge point set may also contain all of the first examination knowledge points. This embodiment does not limit this, that is, the first required knowledge point set is a subset of the first examination knowledge point set.

[0117] In step S106, several first basic sentence segments that need to be modified are screened out from the first basic sentence segment set according to the first modification information, that is, the first basic sentence segments containing the first examination knowledge point in the first compulsory knowledge point set are extracted from the first basic sentence segment set to obtain several first basic sentence segments that need to be modified, and are recorded as the first segments to be modified.

[0118] Further, in step S107, the first modified segment is input into the second training model to obtain a plurality of first candidate segments corresponding to each first segment to be modified. A first segment to be modified has a plurality of first candidate segments, and both the first segment to be modified and the plurality of first candidate segments contain the first examination knowledge point, and the difference is that the expression forms of the first segment to be modified and the first candidate segments are different. Specifically, in addition to the first vocabulary corresponding to the first examination knowledge point in the first segment to be modified, in order to express the semantic information consistent with the context of the first article, the first segment to be modified also contains a subject, an object, an adjective, etc. The requirements for nouns and adjectives of different grades are different, and other nouns except the first vocabulary corresponding to the first examination knowledge point can be selectively replaced. The generation of the first candidate segment can be: select semantically similar vocabulary to perform partial vocabulary replacement on the first segment to be modified. Optionally, this embodiment uses the second training model to modify the first segment to be modified to obtain a plurality of first candidate segments. Specifically, the second training model can be a SIM similarity algorithm model, and the second training model is trained with samples in an unsupervised form. Furthermore, sample difficulty configuration information can be introduced in the training process of the second training model. That is, the multiple first candidate sentences generated by the first sentence to be modified have different degrees of difficulty. It should be noted that this degree of difficulty lies in the change and adjustment of the expression method and nouns. For example, the vocabulary of the college level 6 English vocabulary library is used to replace the nouns of the college level 4 English vocabulary in the current first sentence to be modified, and the semantics remain unchanged. Then the target users of the college level 4 have a certain degree of difficulty in understanding this first candidate sentence; for example, the sentence composition in the first sentence to be modified can be modified, such as changing the current general future tense to the general present tense, modifying from the perspective of tense appropriation, and increasing the difficulty of users confusing the tense used in the first test knowledge point.

[0119] In step S108, first difficulty configuration information is obtained. The first difficulty configuration information can be pre-configured on the teacher side. A first difficulty value of each first sentence to be modified is generated according to the first difficulty configuration information. The difficulty level can be divided according to actual needs, such as 0%-100% as the manifestation of difficulty level. Furthermore, the difficulty level can be evaluated by the specific mastery of the vocabulary mastered by students in the current grade. For example, vocabulary that can be fully mastered by a small number of students is recorded as more difficult vocabulary, and vocabulary that can be fully mastered by most students is recorded as simple vocabulary. The vocabulary is then expressed in the form of percentage to achieve the quantitative characteristics of the difficulty level. The first difficulty configuration information can be the difficulty configuration requirement for the current first basic text. For example, the difficulty of the current first basic text is set to 60%, and the difficulty value of each first sentence to be modified is set according to the difficulty of the first basic text, which is recorded as the first difficulty value. For example, there are 5 first sentences to be modified with a first difficulty value of 50%, 6 first sentences to be modified with a first difficulty value of 80%, 3 first sentences to be modified with a difficulty of 30%, and the difficulty of the remaining first basic sentences is 20%. The overall difficulty of the composed first basic text is 60%, but the first difficulty values ​​of different first basic sentences are different, so that users have different understanding difficulties in the process of interpreting an article, thereby realizing the growth of users' reading ability in reading articles and the understanding and application of examination knowledge points.

[0120] In step S109, the first candidate sentence segments are screened and replaced for each first sentence segment to be modified. Specifically, each first difficulty value corresponds to a first difficulty value threshold range. For example, if the first difficulty value of the first sentence segment f to be modified is set to 60%, then the first difficulty value threshold range corresponding to it is 50%~70%. From the multiple first candidate sentence segments of the first sentence segment f to be modified, the difficulty value of each first candidate sentence segment is obtained, and the first candidate sentence segment in the range of 50%~70% is screened as the first replacement sentence segment, and the first replacement sentence segment is used to replace the first sentence segment to be modified, which is considered to complete the modification of the first sentence segment to be modified. Repeat the above screening and replacement steps until all the first sentence segments to be modified are replaced, and the first modified text is obtained.

[0121] In step S110, a first test text is obtained according to the first modified text and displayed. It should be noted that the first modified text is the first basic text with multiple first sentences to be modified replaced, that is, the first modified text belongs to a reading article after the difficulty is adjusted, and the first test text is a reading test question based on the first modified text, that is, the first test text includes the first modified text and the test question.

[0122] In step S111, after the first examination text is displayed on the terminal, the terminal answers the first examination text, collects the first answer result of the terminal, and outputs the first examination score. The first answer result and the first examination text are mapped and stored, realizing the logic of automatic generation and automatic grading of reading test questions.

[0123] From the perspective of article generation, this embodiment expands the first text theme into the first article context, and introduces the first vocabulary set to obtain the first basic text. On this basis, the first modification information and the first difficulty configuration information are obtained, and the first sentence to be modified in the first basic text is selectively replaced to achieve sentence adjustment and modification of the first basic text to obtain a first modified text. Based on the first modified text, a first examination text is obtained for examination, so that the first modified text finally obtained is more in line with the requirements of the first modification information and the first difficulty configuration information, and the text of the reading article can be controlled and adjusted to better meet the examination requirements. In addition, the first answer result and the first examination score can be automatically collected, and the automatic composition and grading of reading texts are realized, thereby improving the efficiency of classroom teaching.

[0124] See also Figure 2 Further, in some embodiments, the first text topic is obtained according to the following method:

[0125] S201, constructing a first basic document set, the first basic document set including at least one of hot articles within a first preset time period, test question documents within a second preset time period, and subject documents within a third preset time period;

[0126] S202, inputting the first basic document set into the third training model to obtain a plurality of first keywords and a semantic association degree between each first keyword and the other first keywords, recorded as a first semantic association degree;

[0127] S203, constructing a vocabulary relationship network corresponding to each first keyword based on the first semantic association degree, recorded as a first relationship network;

[0128] S204, obtaining a plurality of first association sets according to the plurality of first relationship networks, each first association set including a plurality of closely associated first keywords;

[0129] S205: Extract at least one first text topic from multiple first association sets according to the syntactic dependency relationship.

[0130] In step S201, a first basic document set is constructed. Specifically, the first basic document set includes at least one of hot articles, test documents and subject documents. Hot articles can be understood as Chinese and English articles, essays, etc. reported in current newspapers, news, and media. Test documents can be understood as reading articles in the past test questions of this subject in the current grade. Subject documents can be understood as articles in tutorial books such as textbooks and teaching plans. Further, this embodiment also adds time segment restrictions to hot articles, test documents and subject documents, that is, hot articles within the first preset time period, test documents within the second preset time period, and subject documents within the third preset time period. The first preset time period can be determined by dividing and determining from the perspectives of the duration of different hot spots, the promotion level, etc. For example, if it is national competitive sports news, the first preset time period can be divided into one year, and if it is a certain explosive news, the first preset time period can be divided into one week or one month. Specifically, different first preset time periods can be set according to different selected hot spot types, and then multiple hot spots can be selectively combined. The second preset time period can be pre-set or manually input. For example, the teacher can set the selected real questions to be the real questions of the past ten years up to the present according to the current teaching situation, or can select the real questions within the past five years up to the present, and so on. The setting of the third preset time period can also be understood with reference to the second preset time period.

[0131] In step S202, the first basic document set is input into the third training model. Specifically, the third training model can be a community discovery model. Specifically, the use process of the third training model for the first basic document set can be: pre-processing the hot articles, test documents and subject documents in the first basic document set, removing irrelevant characters, punctuation marks and special symbols, segmenting each article, and removing modal particles or auxiliary words that do not have semantics; the pre-processed first basic document set is recorded as the first document set to be processed, and the first document set to be processed contains multiple first documents to be processed, and each first document to be processed is an article with word segmentation. The words in the multiple first documents to be processed are evaluated for word association. Specifically, the Word2Vec word vector algorithm model or the TF-IDF model can be used to evaluate the degree of association of each word, and the final multiple first keywords and the degree of semantic association between each first keyword and the remaining first keywords are obtained, which is recorded as the first semantic association degree.

[0132] Furthermore, in step S203, a lexical relationship network is established for each first keyword, recorded as a first relationship network. Specifically, the first relationship network is centered on the first keyword, and the remaining first keywords are clusters radiating with the corresponding first semantic association degree as the connection length. The closer the distance to the first keyword, the higher the first semantic association degree.

[0133] In step S204, a number of first association sets are obtained based on the multiple first relationship networks. Specifically, a community discovery algorithm can be used to evaluate the multiple first relationship networks to obtain multiple first association sets. For example, each first relationship network is regarded as a separate community, and the internal modularity index and the external NMI index can be used to evaluate the connection relationship between each first relationship network and other first relationship networks, and finally multiple first association sets are obtained, each of which includes multiple first keywords that are closely associated and can be connected in series under syntactic dependencies to form a clear theme.

[0134] In step S205, among the multiple first association sets, topics are extracted for each first association set according to syntactic dependencies. It should be noted that multiple first association sets can generate multiple first text topics, and this embodiment does not limit the number of first text topics.

[0135] It should be noted that when using the first text theme, the first text theme required to generate the first examination text can be screened from multiple first text themes through manual selection by the teacher, or one or several first text themes can be randomly selected from the multiple generated first text themes using a random model. This embodiment does not limit this.

[0136] The method steps shown in this embodiment extract multiple first keywords and the first correlation degree corresponding to each first keyword from hot articles, test documents and subject documents, and construct a first relationship network for each first keyword. Multiple first correlation sets are obtained through the first relationship network, and then a first text theme is generated in the first correlation set. It can take into account existing hot current affairs and fully and reasonably combine the contents of history test questions and subject textbooks. Multiple first text themes are obtained on the basis of a clear document scope, which increases the diversity and thinking height of the first text theme, so that the first examination text obtained based on the first text theme has distinct characteristics, is not easy to be similar to existing articles and documents, and increases the confidentiality of the examination.

[0137] Furthermore, in some embodiments, generating a first article context according to the first text theme, and obtaining a first basic text according to the first article context further includes:

[0138] Acquire a first article selection associated with a first text topic in an article database, wherein the article database stores a plurality of first articles, each of which includes a plurality of first paragraphs, and the first article selection is a plurality of first paragraphs in a first article;

[0139] Marking the first article selection with words to obtain a plurality of second keywords, and dividing the first article selection according to the second keywords to obtain the sentence segments where the second keywords are located, which are recorded as second key sentence segments;

[0140] Arranging the paragraph dependencies of the plurality of second key sentences according to the logical order of the preceding and following contexts to obtain the first article context corresponding to the first article selection, wherein the paragraph dependencies include at least one of a parallel relationship, a progressive relationship, and a general-specific relationship;

[0141] The first article context is input into the fourth training model to obtain a first basic text.

[0142] In this embodiment, the article database can be constructed in real time or in advance. For example, before using the method shown in this embodiment, an article database can be pre-integrated in the system. Specifically, the article database can be a shared database of other document platforms or an article database collected on the Internet. Multiple articles are recorded in the article database. For ease of expression, an article associated with the first text theme obtained from the article database is recorded as the first article, a paragraph in the first article is recorded as the first paragraph, and several selected first paragraphs in the same first article are recorded as the first article excerpt.

[0143] Further, in some embodiments, the first article selection associated with the first text topic is obtained according to the following method:

[0144] Filtering at least one first article from an article database according to a first text topic;

[0145] Performing text theme analysis on the first paragraphs of the first articles selected one by one to obtain theme information of each first paragraph, which is recorded as the theme of the first paragraph;

[0146] Constructing a second semantic relevance degree set for each first paragraph, wherein the second semantic relevance degree set includes the second semantic relevance degrees of the remaining first paragraphs and the current first paragraph;

[0147] Matching the first paragraph topics corresponding to the first paragraphs with the first text topics one by one, obtaining a number of first paragraphs with the highest degree of relevance to the first text topic, which are recorded as first related paragraphs;

[0148] Obtaining a second semantic relevance set of the first associated paragraphs, and selecting the first paragraphs that meet a first preset condition as second associated paragraphs, wherein the first preset condition is included in the second semantic relevance set corresponding to the first associated paragraphs, and the second semantic relevance degree with the first associated paragraphs is within a preset relevance degree threshold range;

[0149] Integrate the first related paragraph and the second related paragraph according to the order of writing to obtain the first article excerpt.

[0150] In this embodiment, the first article in the article database is pre-analyzed for text topic analysis. This step can be obtained through a language algorithm model to obtain multiple first article topics. The first text topic and the multiple first article topics are input into a cosine similarity algorithm model to obtain the similarity between different first article topics and the first text topics. Finally, the first article corresponding to the first article topic within the similarity threshold range is screened out from the article database.

[0151] Furthermore, after selecting at least one first article, a text topic analysis is performed on the first paragraph in each first article to obtain multiple first paragraph topics. A second semantic association degree set is constructed for multiple first paragraph topics in the same first article. The specific association degree evaluation process can refer to the first semantic association degree evaluation process to obtain the second semantic association degree between each first paragraph and other first paragraphs.

[0152] Furthermore, firstly, the multiple first paragraph topics in the same first article are matched with the first text topic to obtain several first paragraphs with the highest degree of association with the first text topic, which are recorded as first associated paragraphs; then the second semantic association degree set corresponding to each first associated paragraph is obtained, and the first paragraph that meets the first preset condition is screened out, which is recorded as the second associated paragraph. It should be noted that the first preset condition is to screen multiple first paragraphs in the second semantic association degree set in the first associated paragraph, and obtain the first paragraph whose association degree with the current first associated paragraph is within the preset association degree threshold range, which is recorded as the second associated paragraph. The first associated paragraph and the corresponding multiple second associated paragraphs are integrated in the order of writing to obtain the first article excerpt; when the number of first associated paragraphs is multiple, the multiple first associated paragraphs can be integrated with the corresponding second associated paragraphs, and finally the first article excerpt of the same first article is obtained; further, when the number of first articles is multiple, each first article can be processed in the above steps to obtain multiple first article excerpts. In some embodiments, the integration process of the text sequence may be: obtaining the paragraph sequence of the second related paragraph and the first related paragraph in the same first article, and arranging the second related paragraph and the first related paragraph in the paragraph sequence to obtain the first article excerpt.

[0153] In this embodiment, after obtaining the first article selection, the content in the first article selection is annotated with words, that is, the vocabulary in the first article selection is segmented in units of words or characters, and after removing the noise data, multiple second keywords are obtained. It should be noted that the noise data shown in this embodiment refers to the elimination of redundant expression words in the first article selection, such as: words that do not have clear expression semantics, words that are not closely related to the theme of each first paragraph in the entire first article selection, and words that are obviously modified, etc., and finally multiple second keywords are obtained, and the second keywords are strongly related to the first paragraph theme of the first paragraph in the first article selection where they are located. Further, the method of achieving noise data removal can be: input the first article selection into the language algorithm model for word recognition, and then use the community discovery algorithm to eliminate redundant expression words to obtain the second keyword. Afterwards, the first article selection is divided according to the second keyword, and the sentence segment where the second keyword is located is retained and recorded as the second key sentence segment.

[0154] Furthermore, the paragraph dependencies of the multiple second key sentences are sorted according to the logical order of the context to obtain the first article context corresponding to the first article selection. The paragraph dependency relationship includes one or more of a parallel relationship, a progressive relationship, and a total-to-specific relationship. In some optional embodiments, the first article context can be presented in the form of a tree diagram, so that the teacher can adjust the individual vocabulary or paragraph dependencies in the first article context according to actual needs.

[0155] Finally, the first article context and the first vocabulary set are input into the fourth training model to obtain the first basic text. Optionally, the fourth training model can be a neural network model trained based on a convolutional neural network model, the input samples of the fourth training model are sample article contexts and sample vocabulary sets, and the output result of the fourth training model is a sample basic text. This method can achieve the expansion and filling of the first article context in text using the first vocabulary set, and finally form a complete first basic text, which is convenient for performing subsequent modification operations.

[0156] The method steps shown in this embodiment realize the specific generation process from the first text theme to the first article context, and then from the first article context to the first basic text, which reduces the operations on the teacher side and improves the efficiency of test paper compilation.

[0157] See also Figure 3 Further, in some embodiments, obtaining the first modification information includes:

[0158] S301, obtaining user information, the user information including basic information, historical answer information and learning level information corresponding to the current user;

[0159] S302, generating a first set of knowledge points to be checked and a second set of knowledge points to be checked according to the user information, wherein the first set of knowledge points to be checked includes a plurality of first knowledge points to be checked, which are knowledge points marked as difficult to understand in the historical answer information, and the second set of knowledge points to be checked includes a plurality of second knowledge points to be checked, which are knowledge points that the user is required to master in the teaching plan;

[0160] S303, obtaining a first test weight value for each first knowledge point to be tested according to the historical answer information, and obtaining a second test weight value for each second knowledge point to be tested according to the teaching plan;

[0161] S304, obtaining a first preset number of first knowledge points to be checked according to the learning level information, and obtaining a second preset number of second knowledge points to be checked according to the learning level information;

[0162] S305, arranging the first knowledge points to be checked in descending order of the first check weight values, and extracting a first preset number of first knowledge points to be checked that are arranged in front of the first knowledge point set to be checked, and recording them as first basic examination knowledge points;

[0163] S306, arranging the second knowledge points to be tested in descending order of the second test weight values, and extracting a second preset number of second knowledge points to be tested that are arranged at the front in the set of second knowledge points to be tested, and recording them as second basic examination knowledge points;

[0164] S307: Integrate the first basic examination knowledge points and the second basic examination knowledge points to obtain a first set of compulsory examination knowledge points, and generate first modification information according to the first set of compulsory examination knowledge points.

[0165] In step S301, user information is obtained. It should be noted that the user shown in this article is a student and the terminal is a student terminal. The user information includes basic information, historical answer information and learning level information corresponding to the current user, wherein the basic information includes subject information, student name, student ID information, grade information, etc. The historical answer information is the first answer result and the first examination score in the historical record formed after the current user answered the previous first examination text in the previous classroom teaching, and can also include a variety of information extracted from the first answer result, such as the knowledge points with difficulty in understanding described later, etc. The learning level information can be converted for the comprehensive ability level of this user in the current subject. Specifically, the comprehensive ability level includes the ranking information of the current user in the grade exam of this subject, the proportion of the number of knowledge points clearly mastered in each classroom teaching assignment to the number of knowledge points planned to be mastered, etc., and the comprehensive ability level is obtained by weighted calculation.

[0166] In step S302, a first set of knowledge points to be tested and a second set of knowledge points to be tested are generated based on user information, wherein the first set of knowledge points to be tested are knowledge points marked as difficult to understand in historical answer information, and the second set of knowledge points to be tested are knowledge points that the user is required to master in the teaching plan. Optionally, the second set of knowledge points to be tested are knowledge points that the user is required to master in the teaching plan but has not yet mastered.

[0167] In step S303, a first test weight value of each first knowledge point to be tested is obtained according to the historical answer information, and a second test weight value of each second knowledge point to be tested is obtained according to the teaching plan. Specifically, the first test weight value can be obtained by referring to the steps described below, and the second test weight value can be obtained according to the user's learning level information.

[0168] In step S304, the first knowledge points to be tested are arranged in order of the first test weight values, and a first preset number of first knowledge points to be tested that are arranged in front of the first knowledge point set to be tested are extracted and recorded as the first basic examination knowledge points. The first preset number can be set according to actual needs.

[0169] Similarly, in step S305, the second knowledge points to be tested are arranged in order of the second test weight values, and a second preset number of second knowledge points to be tested that are arranged at the front in the second knowledge point set to be tested are extracted and recorded as second basic examination knowledge points; the second preset number can be set according to actual needs.

[0170] In step S306, the first basic examination knowledge points and the second basic examination knowledge points are integrated to generate a first set of required knowledge points. As mentioned above, the first set of required knowledge points contains at least part of the first examination knowledge points, that is, the first examination knowledge points in the first required knowledge points are the first basic examination knowledge points and / or the second basic examination knowledge points. The first modification information is generated according to the first set of required knowledge points, so that the first basic text can be modified through the first modification information, the operation steps on the teacher side can be reduced, and the difficulty of modifying the first basic text can be reduced.

[0171] See also Figure 4 Further, in some embodiments, obtaining the first inspection weight value of each first knowledge point to be inspected according to the historical answer information also includes:

[0172] S401, setting a corresponding answer record table for each first knowledge point to be tested, wherein the answer record table records the answer time, and the number of answers and answer results of the current first knowledge point to be tested arranged in the order of answer time;

[0173] S402, filtering out the number of wrong answers to the first knowledge point to be tested in the answer record table, and recording it as the number of wrong answers;

[0174] S403, respectively calculating the time interval between the answering times corresponding to two adjacent wrong answer results, and recording it as the wrong answer interval period;

[0175] S404, mapping and storing the number of wrong answers and the wrong answer interval period to obtain a wrong answer record set;

[0176] S405, obtaining a correct rate prediction curve according to the set of wrong question records, wherein the independent variable of the correct rate prediction curve is the time interval, and the dependent variable of the correct rate prediction curve is the predicted answer correct rate of the current first knowledge point to be tested;

[0177] S406, calculating the time interval between the current timestamp and the last answer time to obtain a first time interval to be tested, and obtaining a first predicted answer accuracy of the first knowledge point to be tested according to the first time interval to be tested and the accuracy prediction curve;

[0178] S407. Generate a first test weight value according to the first predicted answer accuracy rate.

[0179] In step S401, an answer record table is set for each first knowledge point to be tested, and the answer record table records the number of answers to the questions of the current first knowledge point to be tested, arranged in order of answer time, the answer results, and the answer interval period between two adjacent answer times, wherein the answer results are correct or incorrect, and each answer number corresponds to a timestamp. The answer interval period is obtained by converting the timestamps between two adjacent answer times.

[0180] Furthermore, in step S402, based on the answer results, the number of answers to the questions that are answered incorrectly is screened out in the answer record table, and recorded as the number of wrong questions; in step S403, the time interval corresponding to the number of wrong questions with wrong answer results between two adjacent questions is calculated, and recorded as the wrong question interval period. It should be noted that step S403 is only executed when the answer result is wrong. In step S404, the number of wrong questions with wrong answer results in the historical answer information is mapped and stored with the wrong question interval period to obtain a wrong question record set.

[0181] In step S405, a correctness prediction curve is obtained based on the set of wrong question records. Specifically, a hyperbola fitting model or other fitting models can be used to generate a correctness prediction curve for the current first knowledge point to be tested by utilizing the set of wrong question records in the historical answer information. The independent variable of the correctness prediction curve is the time interval, and the dependent variable of the correctness prediction curve is the predicted answer accuracy of the current first knowledge point to be tested.

[0182] In step S406, the time interval between the current timestamp and the last answer time is calculated to obtain the first time interval to be tested, and the first predicted answer accuracy of the first knowledge point to be tested is obtained based on the first time interval to be tested and the accuracy prediction curve; it should be noted that the current timestamp is the timestamp up to now. Assuming that the current first knowledge point to be tested is the first examination knowledge point, the current timestamp is used as the timestamp corresponding to the number of answers given by the user in this first knowledge point to be tested, the first time interval to be tested is calculated, and the first predicted answer accuracy is obtained based on this.

[0183] In step S407, a first test weight value is generated according to the first predicted answer accuracy. Specifically, it is determined whether the first predicted answer accuracy is within the preset accuracy threshold range. If so, the first test weight value is increased. If not, the first test weight value is kept unchanged, or the first test weight value is reduced. For example, if the preset accuracy threshold range is 30%, when the first predicted answer accuracy is lower than 30%, the first test weight value is increased, thereby increasing the probability of the first knowledge point to be tested being selected.

[0184] To facilitate understanding, further examples are given as follows: in the historical answer information of the first knowledge point A to be tested, the intervals between wrong questions in the wrong question record set are: 10 days, 3 days, 1 day, 20 days, 35 days, and 40 days, respectively. This means that the longest memory period of the user's impression of wrong questions in the first knowledge point A to be tested is 40 days. Assuming that the current first test knowledge point contains the number of answers and the answer results of the first knowledge point to be tested, the answer interval period can be obtained by calculating the number of answers and the number of answers of the previous first knowledge point A to be tested. The answer interval period is 15 days. This 15-day answer interval period is compared with the accuracy prediction curve of the first knowledge point A to be tested, and the predicted answer accuracy of the current first knowledge point A to be tested is 62.5%. This means that the current accuracy of the first knowledge point A to be tested is higher than 30%, and the user has a good grasp of the first knowledge point A to be tested. The first test weight value of the first knowledge point A to be tested is lowered, and the first knowledge point A to be tested does not have a high test significance in the current first test text.

[0185] Furthermore, in some embodiments, obtaining the first inspection text according to the first modified text further includes:

[0186] Obtaining test question type information, which includes reading questions and / or cloze questions;

[0187] Generate an assessment outline based on the assessment question type information, which includes the question type and assessment knowledge points;

[0188] Generate multiple final examination questions based on the examination outline;

[0189] generating a first examination text according to a plurality of final examination questions and a first revised text;

[0190] Generating multiple final examination questions based on the examination outline also includes:

[0191] Determine one by one whether the test questions in the test outline belong to the category of vocabulary examination. If so, obtain the test vocabulary corresponding to the test question, and generate a similar vocabulary set based on the test vocabulary. The similar vocabulary set includes test vocabulary and confusing vocabulary. The confusing vocabulary includes vocabulary with similar meaning or similar word form to the test vocabulary.

[0192] Generate multiple initial test questions corresponding to the test words of the current test question according to the similar word set;

[0193] Acquire second difficulty configuration information, and generate a second difficulty value corresponding to each initial assessment question according to the second difficulty configuration information;

[0194] An initial examination question corresponding to a second difficulty level value closest to the learning level information of the user information is selected from the plurality of initial examination questions, and is recorded as a final examination question.

[0195] In this embodiment, the assessment question type information is pre-input by the teacher, and the assessment question type information includes reading questions and / or cloze questions. The specific form of the reading questions can refer to the existing English reading questions, such as: a reading article C and 5 multiple-choice questions corresponding to the reading article C; similarly, the cloze questions can be: a cloze reading article D and 20 cloze reading cloze questions matching the cloze reading article D, and leave corresponding spaces in the cloze reading article D.

[0196] An assessment outline is generated according to the assessment question type information, and the assessment outline includes the question type of the assessment question and the assessment knowledge points, which are the knowledge points that each assessment question needs to examine. It should be noted that in this embodiment, the assessment outline is generated from the assessment question type information, which can be obtained through a preset template. For example, the teacher can input the current reading question in advance, and the system can automatically select the question combination logic corresponding to the reading question to obtain the assessment outline, which is the assessment outline for the first modified text.

[0197] Furthermore, multiple final examination questions are generated based on the examination outline. The final examination questions are complete questions containing confusing options. The examination questions in the examination outline are just an allocation of examination knowledge points and a setting of question types. It is impossible to answer directly from the examination questions in the examination outline.

[0198] Further, it is determined one by one whether the test questions in the test outline belong to the category of vocabulary examination. The category of vocabulary examination can be understood as test questions specifically for test knowledge points with vocabulary as the test point. As mentioned above, the test vocabulary corresponding to the test question is obtained, and a set of similar vocabulary is generated based on the test vocabulary. The specific generation method can be: input the test vocabulary into the similarity algorithm model to obtain multiple confusing vocabulary, and the confusing vocabulary is a vocabulary with similar meaning and similar part of speech as the test vocabulary. It should be noted that the test vocabulary includes words and phrases, and the phrases can involve grammatical application knowledge, that is, the test vocabulary includes words and grammatical application assessments. Similarly, confusing vocabulary can also include phrases with similar grammatical application methods to the tested phrases, and so on.

[0199] Multiple initial test questions corresponding to the test vocabulary of the current test question are generated based on a set of similar words; it should be noted that different initial test questions have different confusing words selected, and different selections of confusing words can adjust the difficulty of different initial test questions.

[0200] The second difficulty configuration information is obtained, and a second difficulty value corresponding to each initial assessment question is generated according to the second difficulty configuration information; the second difficulty configuration information is similar to the first difficulty configuration information and can be obtained by pre-inputting the teacher terminal. Further, the second difficulty value of each initial assessment question is obtained according to the difficulty standard provided by the second difficulty configuration information.

[0201] The initial test questions corresponding to the second difficulty level value closest to the learning level information of the user information are selected from multiple initial test questions, and recorded as the final test questions. This method can realize the selective screening of the initial test questions, so that the final first test text is more consistent with the user's learning level information, and the final test questions of different users are different, which improves the confidentiality and security of the test.

[0202] Specifically, in a cloze test, if the test vocabulary is test phrase H, then the test questions are multiple-choice questions centered around phrase H; and the final test questions include confusing phrase E, confusing phrase T, and confusing phrase P, which together with test phrase H form four options for users to choose from, and so on.

[0203] Furthermore, the first test text is generated based on multiple final test questions and the first modified text. This step can be understood as adjusting the first modified text accordingly according to the final test questions, and then organizing it into a complete reading test question, that is, the first test text. For example, if the reading questions do not need to modify the first modified text, the final test questions can be directly attached to the back of the first modified text to form the first test text; for another example, the cloze questions need to correspond to the location of the test vocabulary in the blanked article, and then the final test questions can be attached to the back of the first modified text to form the first test text.

[0204] Furthermore, in some embodiments, the method further comprises:

[0205] Determine whether the first examination result is qualified, and if not, obtain the first answer result and the first examination text;

[0206] Filter out the second test knowledge points based on the first answer result and the first test text;

[0207] Generating a second text theme according to the second inspection knowledge point, inputting the second text theme and the vocabulary library into the first training model to obtain a second vocabulary set associated with the second text theme, wherein the second vocabulary set contains a plurality of second words;

[0208] Generate a second article context according to the second text theme, and obtain a second basic text according to the second article context;

[0209] Performing segmentation on the second basic text to obtain a second basic segment set, wherein the second basic segment set includes a plurality of second basic segments;

[0210] Performing knowledge point recognition on each second basic sentence segment to obtain a plurality of second examination knowledge points and second vocabulary associated with each second examination knowledge point, mapping and storing the second examination knowledge points and the second vocabulary to obtain a second examination knowledge point set;

[0211] Acquire second modification information, the second modification information including a second set of compulsory knowledge points, the second set of compulsory knowledge points including at least part of the second examination knowledge points;

[0212] According to the second modification information, a plurality of second basic segments to be modified are selected from the second basic segment set and recorded as second segments to be modified;

[0213] Inputting the second to-be-modified segments into the second training model one by one to obtain a plurality of second candidate segments corresponding to each second to-be-modified segment;

[0214] Acquire third difficulty configuration information, and generate a third difficulty value for each second sentence segment to be modified according to the third difficulty configuration information;

[0215] Perform the following steps for each second segment to be modified:

[0216] Selecting a second candidate sentence segment within a third difficulty threshold value range from among the plurality of second candidate sentence segments, recording the second candidate sentence segment as a second replacement sentence segment, and replacing the second sentence segment to be modified to obtain a second modified text;

[0217] A second inspection text is obtained according to the second modified text, and the second inspection text is displayed on the terminal.

[0218] In this embodiment, if the first answer result is unqualified, it means that the user's ability level in this first examination text is relatively weak. For the first answer result and the first examination text, the second examination knowledge point is screened. Specifically, the wrong test questions can be screened out in the first answer result, and the first examination knowledge point in the first examination text can be extracted from the test questions, and the first examination knowledge point is the second examination knowledge point. The second text theme is generated according to the second examination knowledge point. This generation method can be obtained by setting a neural network training model. It should be noted that the second text theme generates a second article context, and the second basic text is obtained based on the second article context, and then the second modification information is obtained. According to the second modification information and the third difficulty configuration information, the second basic text is modified and replaced to obtain the second modified text, and then the specific process of obtaining the second examination text according to the second modified text can refer to the generation process of the first modified text and the first examination text described above, and this embodiment will not be redundantly described here.

[0219] This method can achieve secondary integration of multiple first examination knowledge points that unqualified users have a weak grasp of, obtain second examination knowledge points, and obtain a second examination text based on the second examination knowledge points, so as to achieve targeted training for users and improve teaching efficiency.

[0220] In a second aspect, this embodiment further provides a computer-readable storage medium on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described in the first aspect is implemented.

[0221] See also Figure 5 In a third aspect, this embodiment further provides an electronic device 1, comprising a memory 11 and a processor 12, wherein the memory 11 is used to store one or more computer program instructions, wherein the one or more computer program instructions are executed by the processor 12 to implement the method described in the first aspect.

[0222] The storage medium / memory 11 includes but is not limited to: RAM, ROM, disk, tape, CD, flash memory, USB flash drive, mobile hard disk, memory card, memory stick, network server storage, network cloud storage, etc. The processor 12 includes but is not limited to CPU (central processing unit), GPU (graphics processing unit), MCU (microprocessor), etc.

[0223] In the above technical scheme, a first text theme is obtained, the first text theme and the vocabulary library are input into a first training model to obtain a first vocabulary set associated with the first text theme, a first article context is generated according to the first text theme, and a first basic text is obtained according to the first article context; the first basic text is segmented to obtain a first basic sentence segment set, knowledge points are identified for each first basic sentence segment to obtain multiple first examination knowledge points and first vocabulary associated with each first examination knowledge point, the first examination knowledge point is mapped and stored with the corresponding first vocabulary to obtain a first examination knowledge point set; first modification information is obtained, and a plurality of first basic sentence segments that need to be modified are screened out from the first basic sentence segment set according to the first modification information , recorded as the first sentence segment to be modified; input the first sentence segment to be modified into the second training model one by one to obtain a number of first candidate sentence segments corresponding to each first sentence segment to be modified; obtain the first difficulty configuration information, and generate a first difficulty value for each first sentence segment to be modified according to the first difficulty configuration information; select the first candidate sentence segment that is within the first difficulty value threshold range from the several first candidate sentence segments, record it as the first replacement sentence segment, and use the first replacement sentence segment to replace the first sentence segment to be modified to obtain a first modified text; obtain a first examination text according to the first modified text, and display the first examination text on the terminal; collect the first answer result of the terminal, output the first examination score, and map the first answer result with the first examination text for storage. This technical solution realizes the automatic generation of articles in reading questions from the first text theme to the first article context, and then to the form of the first basic text. On this basis, the first basic text is identified with knowledge points to obtain a first test knowledge point set, and then the first to-be-modified sentence segment is screened out from the first basic sentence segment through the first modification information, and a plurality of first candidate sentence segments are generated for the first to-be-modified sentence segment, and the plurality of first candidate sentence segments are allocated based on the first difficulty value, and the plurality of first candidate sentence segments corresponding to each first to-be-modified sentence segment are screened according to the first difficulty value threshold range, and the screened first candidate sentence segments replace the first to-be-modified sentence segment until the plurality of first to-be-modified sentence segments are replaced in sequence to obtain a first modified text, and then the first test text is obtained based on the first modified text and displayed for the terminal to answer, and the first answer result and the first test score are recorded. This method allows the reading articles, that is, the first test text, to partially or completely modify, adjust and replace sentences in the first basic text according to the first difficulty configuration information and in combination with the first set of required knowledge points, so that the final first modified text is more in line with the requirements of the first modification information and the first difficulty configuration information, thereby achieving controllable adjustment of the text of the reading articles and better meeting the examination requirements.

[0224] Finally, it should be noted that although the above embodiments have been described in the specification and drawings of the present invention, the patent protection scope of the present invention cannot be limited thereby. All technical solutions generated by replacing or modifying equivalent structures or equivalent processes based on the essential concept of the present invention and using the contents recorded in the specification and drawings of the present invention, as well as directly or indirectly implementing the technical solutions of the above embodiments in other related technical fields, are included in the patent protection scope of the present invention.

Claims

1. An Internet-based language teaching automatic test paper composition and evaluation method, characterized in that: The method comprises: Acquire a first text topic, input the first text topic and a vocabulary library into a first training model, and obtain a first vocabulary set associated with the first text topic, wherein the first vocabulary set contains a plurality of first words; Generate a first article context according to the first text theme, and obtain a first basic text according to the first article context; Performing segmentation on the first basic text to obtain a first basic segment set, wherein the first basic segment set includes a plurality of first basic segments; Performing knowledge point recognition on each of the first basic sentence segments to obtain a plurality of first examination knowledge points and a first vocabulary associated with each of the first examination knowledge points, mapping and storing the first examination knowledge points and the corresponding first vocabulary to obtain a first examination knowledge point set; Acquire first modification information, wherein the first modification information includes a first set of compulsory knowledge points, and the first set of compulsory knowledge points includes at least part of the first examination knowledge points; According to the first modification information, a plurality of first basic segments to be modified are selected from the first basic segment set and recorded as first segments to be modified; Inputting the first to-be-modified segments into the second training model one by one to obtain a plurality of first candidate segments corresponding to each of the first to-be-modified segments; Acquire first difficulty configuration information, and generate a first difficulty value for each first sentence segment to be modified according to the first difficulty configuration information; The following steps are performed for each of the first to-be-modified segments: Selecting a first candidate sentence segment within a first difficulty level threshold from among the plurality of first candidate sentence segments, recording the first candidate sentence segment as a first replacement sentence segment, and replacing the first to-be-modified sentence segment with the first replacement sentence segment to obtain a first modified text; Obtaining a first inspection text according to the first modified text, and displaying the first inspection text on a terminal; Collecting the first answer result of the terminal, outputting a first examination score, and mapping and storing the first answer result and the first examination text; The first text topic is obtained according to the following method: Constructing a first basic document set, wherein the first basic document set includes at least one of hot articles within a first preset time period, test question documents within a second preset time period, and subject documents within a third preset time period; Inputting the first basic document set into a third training model to obtain a plurality of first keywords and a semantic association degree between each of the first keywords and the other first keywords, recorded as a first semantic association degree; constructing a vocabulary relationship network corresponding to each of the first keywords based on the first semantic association degree, recorded as a first relationship network; Obtaining a plurality of first association sets according to the plurality of the first relationship networks, each of the first association sets including a plurality of first keywords with a close degree of association; Extracting at least one of the first text topics from the plurality of the first association sets according to the syntactic dependency relationship; Generating a first article context according to the first text theme, and obtaining a first basic text according to the first article context also includes: Acquire a first article selection associated with the first text topic in an article database, wherein the article database stores a plurality of first articles, each of the first articles includes a plurality of first paragraphs, and the first article selection is a plurality of first paragraphs in one of the first articles; Performing word tagging on the first article selection to obtain a plurality of second keywords, and dividing the first article selection according to the second keywords to obtain the sentence segments where the second keywords are located, which are recorded as second key sentence segments; Arranging the paragraph dependency relationships of the plurality of the second key sentences according to the logical order of the preceding and following contexts to obtain the first article context corresponding to the first article selection, wherein the paragraph dependency relationship includes at least one of a parallel relationship, a progressive relationship, and a general-specific relationship; The first article context is input into a fourth training model to obtain the first basic text.

2. The method for automatically composing and evaluating language teaching papers based on the Internet according to claim 1, characterized in that: The first article excerpt associated with the first text topic is obtained according to the following method: Filtering out at least one first article in the article database according to the first text topic; Performing text theme analysis on the first paragraphs in the selected first articles one by one to obtain theme information of each of the first paragraphs, which is recorded as the first paragraph theme; constructing a second semantic association degree set for each of the first paragraphs, wherein the second semantic association degree set includes second semantic association degrees between the remaining first paragraphs and the current first paragraph; Matching the first paragraph topics corresponding to the first paragraphs with the first text topic one by one, obtaining a number of first paragraphs with the highest degree of relevance to the first text topic, which are recorded as first related paragraphs; Obtaining a second semantic relevance set of the first associated paragraphs, and screening out first paragraphs that meet a first preset condition as second associated paragraphs, wherein the first preset condition is included in the second semantic relevance set corresponding to the first associated paragraphs, and the second semantic relevance degree with the first associated paragraph is within a preset relevance degree threshold range; The first related paragraph and the second related paragraph are integrated according to the order of writing to obtain the first article excerpt.

3. The method for automatic composition and evaluation of language teaching based on the Internet according to claim 1, characterized in that: Obtaining the first modification information includes: Obtaining user information, including basic information, historical answer information, and learning level information corresponding to the current user; Generate a first set of knowledge points to be tested and a second set of knowledge points to be tested according to the user information, wherein the first set of knowledge points to be tested includes a plurality of first knowledge points to be tested, the first knowledge points to be tested are knowledge points marked as difficult to understand in the historical answer information, and the second set of knowledge points to be tested includes a plurality of second knowledge points to be tested, the second knowledge points to be tested are knowledge points that the user is required to master in the teaching plan; Obtain a first inspection weight value for each of the first knowledge points to be inspected according to the historical answer information, and obtain a second inspection weight value for each of the second knowledge points to be inspected according to the teaching plan; Obtaining a first preset number of the first knowledge points to be checked according to the learning level information, and obtaining a second preset number of the second knowledge points to be checked according to the learning level information; Arrange the first to-be-tested knowledge points in descending order of first test weight values, and extract a first preset number of first to-be-tested knowledge points arranged at the front in the first to-be-tested knowledge point set, and record them as first basic examination knowledge points; Arrange the second knowledge points to be tested in descending order of second test weight values, and extract a second preset number of second knowledge points to be tested that are arranged at the front in the set of the second knowledge points to be tested, and record them as second basic examination knowledge points; The first basic examination knowledge points and the second basic examination knowledge points are integrated to obtain the first set of compulsory examination knowledge points, and the first modification information is generated according to the first set of compulsory examination knowledge points.

4. The method for automatically composing and evaluating language teaching papers based on the Internet according to claim 3 is characterized in that: Obtaining the first inspection weight value of each of the first knowledge points to be inspected according to the historical answer information further includes: Setting a corresponding answer record table for each of the first knowledge points to be tested, wherein the answer record table records the answer time, and the number of answers and answer results of the current question of the first knowledge point to be tested arranged in order of answer time; Filter out the number of wrong answers to the first knowledge point to be tested in the answer record table, and record it as the number of wrong answers; Calculate the time interval between the answering times corresponding to two consecutive wrong answer results, and record it as the wrong question interval period; The number of wrong answers and the wrong answer interval period are mapped and stored to obtain a wrong answer record set; Obtaining a correctness prediction curve according to the set of wrong question records, wherein the independent variable of the correctness prediction curve is the time interval, and the dependent variable of the correctness prediction curve is the predicted answer correctness of the current first knowledge point to be tested; Calculate the time interval between the current timestamp and the last answer time to obtain a first time interval to be tested, and obtain a first predicted answer accuracy of the first knowledge point to be tested according to the first time interval to be tested and the accuracy prediction curve; The first test weight value is generated according to the first predicted answer accuracy rate.

5. The method for automatic composition and evaluation of language teaching based on the Internet according to claim 3, characterized in that: The first inspection text obtained according to the first modified text also includes: Acquiring examination question type information, wherein the examination question type information includes reading questions and / or cloze questions; Generate an examination outline according to the examination question type information, wherein the examination outline includes the examination question type and the examination knowledge points; Generate multiple final examination questions according to the examination outline; Generate the first examination text according to the plurality of the final examination questions and the first modified text; Generating a plurality of final examination questions according to the examination outline also includes: Determine one by one whether the test questions in the test outline belong to the category of vocabulary examination. If so, obtain the test vocabulary corresponding to the test question, and generate a similar vocabulary set based on the test vocabulary, wherein the similar vocabulary set includes the test vocabulary and confusing vocabulary, and the confusing vocabulary includes vocabulary with similar meaning or similar word form to the test vocabulary; Generating a plurality of initial examination questions corresponding to the examination words of the current examination question according to the similar word set; Acquire second difficulty configuration information, and generate a second difficulty value corresponding to each of the initial assessment questions according to the second difficulty configuration information; An initial examination question corresponding to a second difficulty level value closest to the learning level information of the user information is selected from the plurality of initial examination questions, and is recorded as a final examination question.

6. The method for automatic composition and evaluation of language teaching based on the Internet according to claim 1, characterized in that: The method further comprises: Determine whether the first examination result is qualified, and if not, obtain the first answer result and the first examination text; Filtering out second examination knowledge points according to the first answer result and the first examination text; Generating a second text topic according to the second inspection knowledge point, inputting the second text topic and the vocabulary library into the first training model to obtain a second vocabulary set associated with the second text topic, wherein the second vocabulary set contains a plurality of second words; generating a second article context according to the second text theme, and obtaining a second basic text according to the second article context; Performing segmentation on the second basic text to obtain a second basic segment set, wherein the second basic segment set includes a plurality of second basic segments; Performing knowledge point recognition on each of the second basic sentence segments to obtain a plurality of second examination knowledge points and a second vocabulary associated with each of the second examination knowledge points, mapping and storing the second examination knowledge points and the second vocabulary to obtain a second examination knowledge point set; Acquire second modification information, wherein the second modification information includes a second set of compulsory knowledge points, and the second set of compulsory knowledge points includes at least part of the second examination knowledge points; According to the second modification information, a plurality of second basic segments to be modified are selected from the second basic segment set and recorded as second segments to be modified; Inputting the second to-be-modified segments into the second training model one by one to obtain a plurality of second candidate segments corresponding to each of the second to-be-modified segments; Acquire third difficulty configuration information, and generate a third difficulty value for each second sentence segment to be modified according to the third difficulty configuration information; For each of the second segments to be modified, perform the following steps: Selecting a second candidate sentence segment within the third difficulty threshold value range from the plurality of second candidate sentence segments, recording the second candidate sentence segment as a second replacement sentence segment, and replacing the second sentence segment to be modified to obtain a second modified text; A second inspection text is obtained according to the second modified text, and the second inspection text is displayed on the terminal.

7. A computer-readable storage medium storing computer program instructions, characterized in that: The computer program instructions implement the method according to any one of claims 1 to 6 when executed by a processor.

8. An electronic device, comprising a memory and a processor, characterized in that: The memory is used to store one or more computer program instructions, wherein the one or more computer program instructions are executed by the processor to implement the method according to any one of claims 1 to 6.

Images