Production Method, Apparatus and Electronic Device

Abstract

The present invention belongs to the field of online education technology and provides a method, device and electronic device for producing questions. The method includes: obtaining question information items used by each user to search for questions within a set time period, and matching the question information items in a question library to obtain matching results; calculating the matching degree between the question information items and the corresponding matching results, and filtering out question information items with matching degrees lower than a first preset threshold from each question information item based on the matching degree as candidate question information items; clustering the candidate question information items to classify similar question information items into the same information item cluster, and filtering out information item clusters containing a number of question information items higher than a second preset threshold; processing questions based on the question information in the filtered information item cluster to generate standardized questions. The present invention can produce questions with high user demand, improve question production efficiency, and at the same time meet user needs, improving user learning efficiency and search experience.

Description

Technical Field

[0001] The present invention belongs to the field of educational technology and is particularly applicable to online education. More specifically, it relates to a question generation method, device, electronic device and computer-readable medium. Background Art

[0002] With the development of modern information technology and the needs of the education market, online education, as an emerging educational concept, is being continuously promoted and popularized. Online teaching has brought great convenience to teachers and students. Through online education, teachers can teach at home, and students can also take classes and exams at home simultaneously. In the existing technology, when users encounter questions they don’t know how to answer or need to verify whether the answers to the questions they have done are correct, they can search for the corresponding questions and answers through educational apps. However, there may be situations where the same or even similar questions cannot be found, resulting in users being unable to obtain the correct answers in the first place. In order for other users to search for the same question later, the question needs to be generated in the first place. Therefore, how to screen and generate popular questions searched by users in real time becomes a technical problem to be solved. Summary of the Invention

[0003] (1) Technical issues to be resolved

[0004] The present invention aims to solve the technical problem of how to filter out hot topics searched by users in real time to meet the needs of users.

[0005] (2) Technical solution

[0006] In order to solve the above technical problems, one aspect of the present invention provides a method for producing a topic, comprising:

[0007] Obtaining the topic information items used by each user to search for topics within a set time period, and matching results obtained by matching the topic information items in a topic database;

[0008] Calculating a matching degree between the topic information item and the corresponding matching result, and screening, from each topic information item, topic information items whose matching degree is lower than a first preset threshold value as candidate topic information items according to the matching degree;

[0009] Clustering the candidate topic information items to classify similar topic information items into the same information item cluster, and screening out, from the multiple information item clusters formed by the clustering, an information item cluster containing a number of topic information items exceeding a second preset threshold;

[0010] The questions are processed according to the question information in the screened information item cluster to generate standardized questions that meet the storage requirements.

[0011] According to a preferred embodiment of the present invention, the calculating of the matching degree between the topic information items within a time period and the corresponding real-time matching results further includes:

[0012] A machine learning model is used to calculate the similarity between the title information item and the corresponding matching result, and the similarity is used as the matching degree.

[0013] According to a preferred embodiment of the present invention, the topic information item and the matching result both include text information;

[0014] The calculation of the similarity includes: converting the title information item and the text information of the matching result into feature vectors respectively, and inputting the feature vectors into a machine learning model to calculate the similarity.

[0015] According to a preferred embodiment of the present invention, before clustering the candidate topic information items, the topic generation method further includes:

[0016] The topics whose matching degree is lower than a first preset threshold due to the unqualified topic information items uploaded by the user are removed from the candidate topic information items.

[0017] According to a preferred embodiment of the present invention, clustering the candidate topic information items includes:

[0018] Pre-classification: converting the text information in the candidate topic information items into feature vectors, and grouping the candidate topic information items corresponding to similar feature vectors into a set based on the similarity of the feature vectors, thereby obtaining multiple sets containing similar candidate topic information items;

[0019] Clustering: Cluster the candidate topic information items in each set obtained by pre-classification.

[0020] According to a preferred embodiment of the present invention, clustering candidate topic information items includes:

[0021] Taking each candidate topic information item as a vertex and the similarity between two candidate topic information items as an edge, when the similarity between two candidate topic information items is higher than a preset first threshold, connecting the vertices corresponding to the candidate topic information items to form a connected graph;

[0022] A connectivity graph between the candidate topic information items is calculated according to a preset rule, and each connectivity graph is regarded as an information item cluster.

[0023] According to a preferred embodiment of the present invention, a set-finding algorithm is used to calculate a connectivity graph of candidate topic information items;

[0024] Optionally, the second preset threshold is set according to the length of the set time period.

[0025] According to a preferred embodiment of the present invention, processing the topic information in the filtered information item cluster to generate standardized topics further includes:

[0026] Screening out the topic information items suitable for topic production from the information item cluster according to a predetermined rule, and processing the topic based on the screened topic information items to generate a standardized topic;

[0027] Optionally, the predetermined rule includes a text information completeness indicator.

[0028] According to a preferred embodiment of the present invention, a topic is generated based on a plurality of topic information items of the same information item cluster.

[0029] A second aspect of the present invention provides a title production device, comprising:

[0030] A topic acquisition module is used to obtain topic information items used by each user to search for topics within a set period of time, and to obtain matching results by matching the topic information items in a topic database;

[0031] a topic screening module, configured to calculate a matching degree between the topic information item and the corresponding matching result, and to screen, from each topic information item, topic information items whose matching degree is lower than a first preset threshold value as candidate topic information items based on the matching degree;

[0032] a topic classification module, configured to cluster the candidate topic information items to classify similar topic information items into the same information item cluster, and to select, from the plurality of information item clusters formed by the clustering, an information item cluster containing a number of topic information items exceeding a second preset threshold;

[0033] The topic production module is used to process the topics according to the topic information in the screened information item cluster to generate standardized topics that meet the storage requirements.

[0034] A third aspect of the present invention provides an electronic device, comprising a processor and a memory, wherein the memory is used to store a computer executable program, and when the computer program is executed by the processor, the processor executes the method described.

[0035] A fourth aspect of the present invention further proposes a computer-readable medium storing a computer-executable program, which implements the method described above when the computer-executable program is executed.

[0036] (3) Beneficial effects

[0037] The present invention calculates the similarity between the questions searched by the user and the questions matched in the question bank, determines whether the questions matched in the question bank are accurate, and clusters all inaccurately matched questions within a period of time, from which questions with a large number of user searches are screened out for production and stored in the question bank. When other users search for the same question, the correct questions can be matched from the question bank and fed back to the user. The present invention can accurately produce the questions required by most users under limited resources, thereby improving the efficiency of question production, meeting user needs, and improving user learning efficiency and search experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a schematic flow chart of a method for producing a title according to one embodiment of the present invention;

[0039] Figure 2a This is a schematic diagram of a user search topic according to an embodiment of the present invention;

[0040] Figure 2b This is a schematic diagram of question bank matching results after user search according to an embodiment of the present invention;

[0041] Figure 2c This is a schematic diagram of a user search topic according to another embodiment of the present invention;

[0042] Figure 3 is a schematic diagram of a title production device according to an embodiment of the present invention;

[0043] Figure 4 is a schematic structural diagram of an electronic device according to an embodiment of the present invention;

[0044] Figure 5 is a schematic diagram of a computer-readable recording medium according to an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In the introduction of specific embodiments, the detailed description of the structure, performance, effect or other features is intended to enable those skilled in the art to fully understand the embodiments. However, this does not preclude those skilled in the art from implementing the present invention with a technical solution that does not include the aforementioned structure, performance, effect or other features under specific circumstances.

[0046] The flowcharts in the accompanying drawings are merely illustrative of the process flow and do not necessarily include all of the content, operations, and steps in the flowcharts, nor do they necessarily imply that all of the steps in the flowcharts must be executed in the order shown. For example, some of the steps in the flowcharts may be separated, some may be combined or partially combined, and so on. The execution order shown in the flowcharts may be changed according to actual circumstances without departing from the spirit of the present invention.

[0047] Frames in the accompanying drawings Figure 1 The term "functional entity" generally refers to a functional entity and does not necessarily correspond to a physically independent entity. That is, these functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different networks and / or processing unit devices and / or microcontroller devices.

[0048] The same reference numerals in the accompanying drawings represent the same or similar elements, components or parts, and thus repeated descriptions of the same or similar elements, components or parts may be omitted below. It should also be understood that although the first, second, third and other numbered adjectives may be used herein to describe various devices, elements, components or parts, these devices, elements, components or parts should not be limited by these adjectives. In other words, these adjectives are only used to distinguish one from another. For example, the first device may also be called the second device, but this does not deviate from the essential technical solution of the present invention. In addition, the terms "and / or" and "and / or" refer to all combinations including any one or more of the listed items.

[0049] To solve the above technical problems, the present invention proposes a method for generating questions. When students encounter questions they don't know how to answer or want to verify whether their answers are correct, they upload the questions to the APP client. They can take a photo of the question with an electronic device and upload the picture, or they can manually enter the text information of the question and upload it. Since uploading a picture after taking a photo is more convenient and quicker than manually entering the text of the question, most users upload pictures after taking a photo. Therefore, the embodiment of the present invention uses the photo form as an example for explanation. However, it will be understood by those skilled in the art that the specific form of user search does not affect the implementation of the technical solution of the present invention.

[0050] After receiving the question image uploaded by the user, the client uses OCR technology to identify the text information and question category information in the image in real time. If the question is a graphic question, it can also identify the graphic information in the graphic question, and identify the question information item from the text information in combination with semantics. The question information item refers to the characteristics of the question or its expression, which can be used to detect questions in the question bank or the Internet. For example, it can include keywords in the text information, such as feature fields, etc., and perform real-time matching in the question bank based on the question information item to obtain the most similar question in the current question bank. Due to the timeliness of the questions, the questions searched by the user are not necessarily the same as the questions matched in the final question bank. The questions fed back to the user may not be what the user wants, and therefore cannot meet the user's requirements. At this time, it is necessary to produce questions that are highly searched by users as soon as possible so that the user's search needs can be met later.

[0051] The number of people needed to produce questions is limited, so it is generally necessary to first screen out questions with high user search frequency before producing them. The question production method provided by the present invention first obtains the questions searched by each user over a period of time, compares the similarity between each searched question and the corresponding matching result in the question bank, and screens out those with low similarity to the matching results and that do not meet the user's requirements from the user's searched questions. These questions are then clustered. The clustering result will classify the same or highly similar questions among the questions searched by users over this period into the same cluster. From these clusters, several clusters containing a large number of questions are selected, and the questions in the clusters are generated and stored in the question bank.

[0052] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings.

[0053] Figure 1 It is a flowchart of a method for producing a title according to an embodiment of the present invention.

[0054] like Figure 1 As shown, the method includes:

[0055] S101: Acquire topic information items used by each user to search for topics within a set period of time, and obtain matching results by matching the topic information items in a topic database.

[0056] This step obtains the title information of all user photo searches (hereinafter referred to as photo searches) within a specific time period, so as to select the most popular search topics. This step also obtains the matching results corresponding to the user's photo search topics, so as to identify / filter out those topics with low matching scores, that is, to find those topics that cannot be found in the resource library after photo searches.

[0057] For example, the user takes a photo of the question and uploads the photo to the client. After the client receives the photo containing the question, it transmits the photo to the server. The server uses OCR technology to identify the question information items in the photo, including the text information and graphic information of the question. OCR technology can be used to form multiple row structures with the text information in the photo as a unit. The server splices adjacent row structures, predicts which row structures belong to the same question based on the context, and extracts the question information items of the identified complete question. If a photo contains multiple questions, each question can also be identified separately and the question information items of each question can be extracted.

[0058] After extracting the title information item of the title in the image, a real-time match is performed in the title library based on the title information item, and the title information item is input into the search model. The search model filters out the title with the highest similarity to the title information item searched by the user from the title library (question bank or resource library), and feeds the title information item of the title as the real-time matching result of the retrieval back to the user through the client. At this time, it is possible that there is no title in the title library (question bank or resource library) that is the same as or the same as the title the user wants to search for, which is manifested as the returned real-time matching result having a low degree of match with the title the user wants to search for, for example, below a first threshold. Therefore, based on the degree of match between the title the user wants to search for and the returned matching result, the user can filter out the topics that the user wants to search for but cannot find based on the matching degree between the topic the user wants to search for and the returned matching result.

[0059] In addition, when searching with photos, if it is recognized that the picture taken by the user contains multiple questions, the client will first display the multiple questions to the user. The user selects and confirms the question to be searched. After confirmation and submission, the server matches the question confirmed by the user in the question library and feedback to the user.

[0060] Since the process from the user searching for a question to the server feeding back the question to the user is completed online in real time, and the retrieval time is extremely short, the search model algorithm used for similarity matching is preferably real-time.

[0061] S102: Calculate the matching degree between the topic information item and the corresponding matching result, and select, from among the topic information items, topic information items whose matching degree is lower than a first preset threshold value as candidate topic information items based on the matching degree.

[0062] This step filters out the topics that the user wants to search for but cannot find based on the matching degree. When the matching degree between the topic information item and the corresponding matching result is lower than the first preset threshold, it means that no identical or similar topics have been retrieved. The similarity here means that the retrieved topics can solve the user's needs for obtaining answers, analysis, etc., and can satisfy the user. There may be many topics to be produced within the set time period, but the service provider has limited productivity. Therefore, these topics that cannot meet user needs and need to be produced or modified are regarded as candidates / alternatives, and topics that are hot searches for users are selected from them, and topics with high search frequency are given priority.

[0063] This solution first requires setting a time period, such as one or two days. For example, if the time period is set to one day, the server retrieves all user search questions during the day, extracts the question information item for each question, and the corresponding real-time matching result from the server. Each question information item and the corresponding real-time matching result are input into the trained matching model, which then outputs the matching degree between the question and the real-time matching result.

[0064] In an embodiment of the present invention, the similarity between the title information item of the user's searched title and the real-time matching result is calculated, and the calculated similarity is used as the matching degree. A first preset threshold is pre-set. When the matching degree between the title information item of the user's searched title and the real-time matching result is lower than the first preset threshold, it can be determined that the title fed back to the user by the server is inaccurate. For example, the first preset threshold is set to 99.5%. That is, when the similarity between the title information item of the user's searched title and the real-time matching result is lower than 99.5%, it is determined that the real-time matching result fed back by the server does not meet the user's needs. This method is used to filter out the titles that do not meet the user's needs from all the titles searched by users in a day. The first preset threshold can be determined by those skilled in the art based on actual conditions such as experience and / or experimental results, and is not limited in this embodiment.

[0065] The similarity matching performed by the above matching model can be completed offline by the server in the background. Compared with the search model, it has higher accuracy and uses a more complex algorithm.

[0066] Figure 2a This is a schematic diagram of a topic searched by a user according to an embodiment of the present invention. Figure 2b FIG. 1 is a schematic diagram of the matching results returned by the question bank after a user searches in one embodiment of the present invention. Figure 2a As shown in the figure, the user uploads a picture containing a question by taking a photo. The server extracts the question information item of the question after identifying the picture, and then searches for similar questions in the question library through the search model. The searched questions are as follows: Figure 2b As shown. The subsequent matching model can extract Figure 2a The sentence containing parameters in the title information item, such as 20% off, saving 200 yuan, actual amount of money spent, etc., is matched. Figure 2b The sentence containing parameters in the question information item includes 20% off, saving 16 yuan, how much yuan is actually spent, etc. Combining the semantics of the question information item to compare the similarity between the two, the similarity is less than 99.5%, and it is judged that Figure 2b The title in does not meet the user's search requirements, Figure 2a The topic information item corresponding to the topic in is taken as the candidate topic information item.

[0067] The matching model in the embodiment of the present invention may use a binary gradient boosting decision tree (GBDT) model to calculate the similarity between two questions.

[0068] In some embodiments, after filtering out the topics that do not meet the user's needs, these topics are screened again to filter out topics that meet the production requirements from these topics (i.e., remove topics that cannot be searched because the photos taken are unqualified). Optionally, the topic information items of these topics can be input into the neural network model to determine whether the topic information items of each topic are complete. Since the user may take pictures at an incorrect angle or focus, resulting in unclear pictures or failure to capture the complete topic, the server may not be able to find the same topic when searching in the topic library. Therefore, it is necessary to use a neural network model to determine whether the topic is complete. For example, multiple-choice questions have no options, and application questions only have stems without questions. These are all incomplete topics. Among the topics that do not meet the user's needs, these topics that do not meet the production requirements are removed (this type of topic may already have answers, but they have not been retrieved). After filtering out the topics that meet the production requirements through the neural network model in this step, the topic information items of these topics are used as candidate topic information items.

[0069] That is, preferably, the topics whose matching degree is lower than the first preset threshold due to unqualified topic information items uploaded by users can also be removed from the candidate topic information items to ensure the integrity of the topics to be clustered subsequently.

[0070] S103: Clustering the candidate topic information items to classify similar topic information items into the same information item cluster, and screening out, from the multiple information item clusters formed by the clustering, an information item cluster containing a number of topic information items exceeding a second preset threshold.

[0071] For example, we can cluster the candidate topic information items screened within a day, calculate the similarity between these candidate topic information items, and then divide the candidate topic information items into different information item clusters based on the similarity. Information item clusters containing a large number of candidate topic information items indicate that these topics are searched frequently and are hot topics.

[0072] Due to the huge amount of searches by users every day, tens of millions of candidate topic information items are generated in a day. If the similarity between any two candidate topic information items is calculated, the calculation scale is huge and it will consume a lot of manpower and material resources. Therefore, preferably, before clustering the candidate topic information items, the candidate topic information items are first pre-classified.

[0073] Pre-classification: convert the text information in the candidate topic information items into feature vectors, and according to the similarity of the feature vectors, group the candidate topic information items corresponding to similar feature vectors into a set, thereby obtaining multiple sets containing similar candidate topic information items. In subsequent clustering, cluster the candidate topic information items in each set.

[0074] The specific method may be: vectorizing the candidate topic information items, converting the text information in the candidate topic information items into feature vectors by using the Word2vec algorithm model, each candidate topic information item corresponds to a feature vector, and all feature vectors are stored in a vector library. When pre-classifying one of the candidate topic information items, a feature search engine may be used to search the vector library for all feature vectors similar to the feature vector corresponding to the candidate topic information item. Here, the similarity may be judged by calculating the inner product of the feature vectors. In this way, the feature search engine will obtain all feature vectors with high similarity to the searched feature vector within 10 milliseconds, and the accuracy of the similarity is about 80%. That is to say, if the similarity between two feature vectors is higher than 80%, the feature search engine will classify the two feature vectors into one category. In this way, all candidate topic information items within a day are pre-classified and divided into multiple sets.

[0075] After pre-classifying the candidate topic information items, multiple sets containing the candidate topic information items are obtained. The candidate topic information items in each set are clustered, and the matching model in the above embodiment can be used to calculate the similarity between two candidate topic information items.

[0076] After calculating the similarity between every two candidate topic information items in all sets, a union-find algorithm is used to calculate a connectivity graph of the candidate topic information items based on the similarity. Each candidate topic information item is used as a vertex, and the similarity between two candidate topic information items is used as an edge. If the similarity between two candidate topic information items exceeds a preset first threshold, the two vertices are connected to form a connectivity graph. Information item clusters are formed based on the connectivity graph, where each connectivity graph corresponds to an information item cluster, and the candidate topic information items within the same information item cluster are identical.

[0077] Because a large number of information item clusters are formed, and some information item clusters may contain only one candidate topic information item, to improve resource utilization, it is preferable to first detect the number of candidate topic information items contained in each information item cluster, select the information item clusters whose number exceeds a second preset threshold, and select them as the information item clusters to be produced. The topics corresponding to the candidate topic information items in the information item clusters to be produced are then produced first. Alternatively, all information item clusters can be sorted by the number of candidate topic information items they contain, with the topics in the first several information item clusters containing the largest number of candidate topic information items being produced first.

[0078] The size of the second preset threshold is determined according to the length of the set time period. For example, if the time period is set to one day, the second preset threshold can be set to 1000. That is to say, if the number of searches for the same question in one day is greater than 1000 times, the question can be produced first.

[0079] The questions selected by this method are those that are searched more frequently by users. Giving priority to these questions improves the efficiency of question production, meets user needs, and improves user learning efficiency and search experience.

[0080] S104: Process the questions according to the question information in the screened information item cluster to generate standardized questions that meet the storage requirements.

[0081] Because each of the selected information items to be produced contains multiple candidate question information items, rules need to be set to determine which candidate question information item to base the production on. For example, the rule could be: the server determines the completeness of the text information in each cluster of information items to be produced. Text completeness includes whether the question stem and question are complete, and whether there is any useless redundant information. Based on this rule, a text information completeness model is established. The text information of each candidate question information item in the cluster of information items to be produced is input into the model to obtain a text information completeness score for each candidate question information item.

[0082] In addition, a comprehensive scoring judgment can be made in combination with other factors that affect the production of questions, such as whether the photos corresponding to the candidate question information items are clear.

[0083] Figure 2c This is a schematic diagram of a user search topic according to another embodiment of the present invention, combined with Figure 2a , the questions in the two figures are from the same cluster of information items to be produced, but Figure 2c contains a lot of redundant information (handwritten answers), so Figure 2c The candidate question information item score will be lower than Figure 2a The candidate question information item score is Figure 2a The candidate question information items are used as the basis for question production, and the candidate question information items are processed into standardized questions and corresponding analyses and answers.

[0084] Questions can be generated based on multiple topic information items from the same information item cluster. For example, some topic information items may be incomplete but the topic photo may be clear, while others may have complete information items and good text recognition but incomplete graphics. These topics can be combined to generate questions more efficiently.

[0085] Preferably, in order to ensure the highest integrity of the candidate topic information items serving as the basic topic, after scoring the candidate topic information items in a cluster of information items to be produced, the candidate topic information item with the highest score can be used as the main item, and the candidate topic information items with the top scores can be used as auxiliary items. After integration, complete text information is formed, and the text information is used as the basic topic for production.

[0086] In summary, the selected information clusters contain highly searched questions and are suitable as a basis for producing standardized questions. Standardized questions generally need to meet the requirements for inclusion in the database, including at least the question and answer, to satisfy basic user needs.

[0087] Those skilled in the art will appreciate that all or part of the steps for implementing the above embodiments are implemented as a program (computer program) executed by a computer data processing device. When the computer program is executed, the above method provided by the present invention can be implemented. Moreover, the computer program can be stored in a computer-readable storage medium, which can be a readable storage medium such as a disk, an optical disk, a ROM, a RAM, or a storage array composed of multiple storage media, such as a disk or tape storage array. The storage medium is not limited to centralized storage, and can also be distributed storage, such as cloud storage based on cloud computing.

[0088] The following describes an apparatus embodiment of the present invention, which can be used to perform the method embodiment of the present invention. Details described in the apparatus embodiment of the present invention should be considered supplementary to the above-described method embodiment; details not disclosed in the apparatus embodiment of the present invention can be implemented with reference to the above-described method embodiment.

[0089] Figure 3 The present invention is a title production device of an embodiment, such as Figure 3 As shown, the device 200 includes:

[0090] The topic acquisition module 201 is used to obtain topic information items used by each user to search for topics within a set period of time, and to obtain matching results by matching the topic information items in the topic database;

[0091] a topic screening module 202 for calculating a matching degree between the topic information item and the corresponding matching result, and screening, based on the matching degree, topic information items with a matching degree lower than a first preset threshold from among the topic information items as candidate topic information items;

[0092] a topic classification module 203 for clustering the candidate topic information items to classify similar topic information items into the same information item cluster, and screening out, from the plurality of information item clusters formed by the clustering, an information item cluster containing a number of topic information items exceeding a second preset threshold;

[0093] The topic production module 204 is used to process topics according to the topic information in the screened information item cluster to generate standardized topics that meet the storage requirements.

[0094] The topic screening module 202 is further configured to use a machine learning model to calculate the similarity between the topic information item and the corresponding matching result, using the similarity as the matching degree. Optionally, both the topic information item and the matching result include text information. Calculating the similarity includes converting the text information of the topic information item and the matching result into feature vectors, respectively, and inputting the feature vectors into the machine learning model to calculate the similarity. Before clustering the candidate topic information items, the topic screening module 202 is further configured to remove topics from the candidate topic information items whose matching degree is lower than a first preset threshold due to unqualified topic information items uploaded by users.

[0095] The topic classification module 203 also includes: a pre-classification module for converting the text information in the candidate topic information items into feature vectors, and according to the similarity of the feature vectors, grouping the candidate topic information items corresponding to similar feature vectors into a set, thereby obtaining multiple sets containing similar candidate topic information items; a clustering module for clustering the candidate topic information items in each set obtained by pre-classification. Each candidate topic information item can be used as a vertex, and the similarity between two candidate topic information items can be used as an edge. When the similarity between two candidate topic information items is higher than a preset first threshold, the vertices corresponding to the candidate topic information items are connected to form a connected graph; the connected graph between the candidate topic information items is calculated according to preset rules, and each connected graph is regarded as an information item cluster; optionally, a union-find algorithm is used to calculate the connected graph of the candidate topic information items; optionally, the second preset threshold is set according to the set time period length.

[0096] The topic production module 204 is further configured to select the topic information items suitable for topic production from the information item cluster according to predetermined rules, and process the topics based on the selected topic information items to generate standardized topics; optionally, the predetermined rules include a text information completeness index.

[0097] The topic generating module 204 generates a topic based on multiple topic information items of the same information item cluster.

[0098] Figure 4 1 is a schematic structural diagram of an electronic device according to an embodiment of the present invention. The electronic device includes a processor and a memory. The memory is used to store a computer executable program. When the computer program is executed by the processor, the processor executes a question generation method.

[0099] like Figure 4As shown, the electronic device is implemented as a general-purpose computing device. The processor may be one or multiple processors working in concert. The present invention also does not exclude distributed processing, meaning that the processors may be dispersed across different physical devices. The electronic device of the present invention is not limited to a single entity but may also be the sum of multiple physical devices.

[0100] The memory stores a computer executable program, typically a machine-readable code, which can be executed by the processor to enable the electronic device to perform the method of the present invention, or at least some of the steps in the method.

[0101] The memory includes a volatile memory, such as a random access memory unit (RAM) and / or a cache memory unit, and may also be a non-volatile memory, such as a read-only memory unit (ROM).

[0102] Optionally, in this embodiment, the electronic device further includes an I / O interface for exchanging data with an external device. The I / O interface may represent one or more of several types of bus structures, including a storage unit bus or storage unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus structures.

[0103] It should be understood that Figure 4 The electronic device shown is merely an example of the present invention. The electronic device of the present invention may also include elements or components not shown in the above examples. For example, some electronic devices also include display units such as screens, and some electronic devices also include human-computer interaction elements such as buttons and keyboards. As long as the electronic device can execute a computer-readable program stored in its memory to implement the method of the present invention or at least some of the steps of the method, it is considered an electronic device covered by the present invention.

[0104] Figure 5 FIG is a schematic diagram of a computer readable recording medium according to an embodiment of the present invention. Figure 5As shown, a computer-readable recording medium stores a computer-executable program, and when the computer-executable program is executed, the above-mentioned method for producing the title of the present invention is implemented. The computer-readable storage medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The readable storage medium may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in combination with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the above.

[0105] The computer-readable medium carries one or more programs. When the one or more programs are executed by a device, the computer-readable medium implements the following functions: obtaining the topic information items used by each user to search for topics within a set time period, and matching results obtained by matching the topic information items in a topic library; calculating the matching degree between the topic information items and the corresponding matching results, and filtering out topic information items with matching degrees lower than a first preset threshold from each topic information item based on the matching degree as candidate topic information items; clustering the candidate topic information items to classify similar topic information items into the same information item cluster, and filtering out, from the multiple information item clusters formed by the clustering, information item clusters containing a number of topic information items higher than a second preset threshold; processing topics based on the topic information in the filtered information item clusters to generate standardized topics that meet the storage requirements.

[0106] The program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages such as Java, C++, and the like, as well as conventional procedural programming languages such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a stand-alone software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving a remote computing device, the remote computing device may be connected to the user computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0107] Through the above description of the implementation mode, it is easy for those skilled in the art to understand that the present invention can be implemented by hardware capable of executing a specific computer program, such as the system of the present invention, and the electronic processing unit, server, client, mobile phone, control unit, processor, etc. contained in the system. The present invention can also be implemented by computer software that executes the method of the present invention. However, it should be noted that the computer software that executes the method of the present invention is not limited to being executed by one or a specific hardware entity, and it can also be implemented in a distributed manner by unspecified hardware. For example, some method steps executed by the computer program can be executed on a mobile client, and another part can be executed in a smart meter, smart recognition pen, etc. For computer software, the software product can be stored in a computer-readable storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), or it can be distributed and stored on a network, as long as it enables the electronic device to execute the method according to the present invention.

[0108] In summary, the present invention can be implemented in hardware, or in a software module running on one or more processors, or in a combination thereof. It will be appreciated by those skilled in the art that general data processing equipment such as a microprocessor or a digital signal processor (DSP) can be used in practice to implement some or all of the functions of some or all of the components in the embodiments of the present invention. The present invention can also be implemented as a device or apparatus program (e.g., a computer program and a computer program product) for executing a part or all of the methods described herein. Such a program for implementing the present invention can be stored on a computer-readable medium, or can have the form of one or more signals. Such a signal can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0109] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the present invention is not inherently related to any specific computer, virtual device, or electronic device, and various general-purpose devices can also implement the present invention. The above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included in the scope of protection of the present invention.

Claims

1. A method for producing standardized questions based on question information, characterized in that: include: Obtaining all topic information items used by users to search for topics within a set period of time, so as to select topics with high search popularity, and performing real-time matching in a topic database based on the topic information items to obtain matching results; the topic information items include characteristics of the topic or the description of the topic; Calculating the matching degree between the topic information item and the corresponding real-time matching result, and filtering out topic information items whose matching degree is lower than a first preset threshold from each topic information item based on the matching degree, and determining whether the filtered topic information items are complete through a neural network model, and using the filtered complete topic information items as candidate topic information items, so as to find the same or similar topics that are needed by all users but cannot be searched within the set time period as topics that meet the production requirements; Clustering the candidate topic information items includes: converting the text information in the candidate topic information items into feature vectors, so that each candidate topic information item corresponds to a feature vector; pre-classification, using a feature search engine to search from a vector library for all feature vectors similar to the feature vector corresponding to a candidate topic information item, wherein the similarity is judged by calculating the inner product of the feature vectors, and the candidate topic information items corresponding to similar feature vectors are grouped into one set, thereby obtaining a plurality of sets containing similar candidate topic information items; and clustering, clustering the candidate topic information items in each set obtained by pre-classification: calculating each similarity between two candidate topic information items in the set, calculating a connectivity graph of the candidate topic information items based on the similarity, forming information clusters based on the connectivity graph, and classifying similar topic information items into the same information item cluster; screening, from the multiple information item clusters formed, information item clusters containing a number of candidate topic information items exceeding a second preset threshold as information item clusters to be produced, and giving priority to producing corresponding topics; or sorting all information item clusters according to the number of candidate topic information items they contain, and giving priority to producing topics corresponding to the first several information item clusters containing a large number of candidate topic information items, wherein the second preset threshold is set according to the length of the set time period; Topic processing is performed based on the topic information in the screened information item cluster to generate standardized topics that meet the storage requirements, including: screening the topic information items suitable for topic production from the information item cluster according to predetermined rules, performing topic processing based on the screened topic information items, and producing topics based on multiple topic information items in the same information item cluster to generate standardized topics; the predetermined rules include a text information completeness index, inputting the text information of each candidate topic information item in the information item cluster to be produced into a text information completeness model, obtaining a text information completeness score for the candidate topic information item, using the candidate topic information item with the highest score as the main item, and using the candidate topic information items with the top scores as auxiliary items, and forming complete text information after integration, and using the text information as the basic topic for production.

2. The method for producing a topic according to claim 1, wherein: The calculating the matching degree between the topic information item and the corresponding real-time matching result includes: A machine learning model is used to calculate the similarity between the title information item and the corresponding matching result, and the similarity is used as the matching degree.

3. The method for producing a topic according to claim 2, wherein: Also includes: The topic information item and the matching result both include text information; The calculation of the similarity includes: converting the title information item and the text information of the matching result into feature vectors respectively, and inputting the feature vectors into a machine learning model to calculate the similarity.

4. The method for producing a topic according to claim 1, wherein: Before clustering the candidate topic information items, the topic generation method further includes: The topics whose matching degree is lower than a first preset threshold due to the unqualified topic information items uploaded by the user are removed from the candidate topic information items.

5. The method for producing a topic according to any one of claims 1 to 4, characterized in that: Calculating a connectivity graph of candidate topic information items based on similarity, and forming information clusters based on the connectivity graph further includes: Taking each candidate topic information item as a vertex and the similarity between two candidate topic information items as an edge, when the similarity between two candidate topic information items is higher than a preset first threshold, connecting the vertices corresponding to the candidate topic information items to form a connected graph; A connectivity graph between the candidate topic information items is calculated according to a preset rule, and each connectivity graph is regarded as an information item cluster.

6. The method for producing a topic according to claim 1, wherein: Also includes: The union-find algorithm is used to calculate the connectivity graph of candidate question information items.

7. A question generation device, characterized in that: include: The topic acquisition module is used to obtain topic information items used by all users to search for topics within a set period of time, so as to select topics with high search popularity, and to obtain matching results by performing real-time matching in the topic database based on the topic information items; the topic information items include the characteristics of the topic or the description of the topic; a topic screening module, configured to calculate a degree of matching between the topic information items and the corresponding real-time matching results, and based on the degree of matching, screen out topic information items with a degree of matching lower than a first preset threshold from each topic information item, and determine whether the screened topic information items are complete through a neural network model, and use the screened complete topic information items as candidate topic information items to find the same or similar topics that all users need but cannot search for within the set time period as topics that meet production requirements; The topic classification module is used to cluster the candidate topic information items, including: converting the text information in the candidate topic information items into feature vectors, so that each candidate topic information item corresponds to a feature vector; pre-classification, using a feature search engine to search from a vector library for all feature vectors similar to the feature vector corresponding to a candidate topic information item, wherein the similarity is determined by calculating the inner product of the feature vectors, and the candidate topic information items corresponding to similar feature vectors are grouped into one set, thereby obtaining multiple sets containing similar candidate topic information items; and clustering, clustering the candidate topic information items in each set obtained by the pre-classification: Calculating the similarity between two candidate topic information items in each set, calculating a connectivity graph of the candidate topic information items based on the similarity, and forming information clusters based on the connectivity graph to classify similar topic information items into the same information item cluster; screening, from the multiple information item clusters formed, information item clusters containing a number of candidate topic information items exceeding a second preset threshold as information item clusters to be produced, and giving priority to producing corresponding topics; or, sorting all information item clusters according to the number of candidate topic information items they contain, and giving priority to producing topics corresponding to the first several information item clusters containing a large number of candidate topic information items, wherein the second preset threshold is set according to the length of the set time period; The topic production module is used to process topics based on the topic information in the screened information item cluster to generate standardized topics that meet the storage requirements, including: screening the topic information items suitable for topic production from the information item cluster according to predetermined rules, processing topics based on the screened topic information items, and producing topics based on multiple topic information items in the same information item cluster to generate standardized topics; the predetermined rules include a text information completeness index, inputting the text information of each candidate topic information item in the information item cluster to be produced into a text information completeness model, obtaining a text information completeness score for the candidate topic information item, taking the candidate topic information item with the highest score as the main item, and taking the candidate topic information items with the top scores as auxiliary items, and forming complete text information after integration, and using the text information as the basic topic for production.

8. An electronic device comprising a processor and a memory, wherein the memory is used to store a computer executable program, wherein: When the computer-executable program is executed by the processor, the processor performs the method according to any one of claims 1 to 6.

9. A computer-readable medium storing a computer-executable program, characterized in that: When the computer executable program is executed, the method according to any one of claims 1 to 6 is implemented.

Images