CN110619042A - Neural network-based teaching question and answer system and method - Google Patents
Neural network-based teaching question and answer system and method Download PDFInfo
- Publication number
- CN110619042A CN110619042A CN201911066052.8A CN201911066052A CN110619042A CN 110619042 A CN110619042 A CN 110619042A CN 201911066052 A CN201911066052 A CN 201911066052A CN 110619042 A CN110619042 A CN 110619042A
- Authority
- CN
- China
- Prior art keywords
- answer
- learner
- question
- module
- subsystem
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Abstract
The invention designs a neural network-based teaching question and answer system, which comprises a front end and a rear end, wherein the front end comprises a registration module, a login module, a question-asking module, an answer module, an evaluation module and a cognitive evaluation tree module; the back end comprises three subsystems, namely a cognitive evaluation subsystem, an answer reuse subsystem and an answer generation subsystem; the invention retrieves the answers of the existing questions in the community by extracting the information of the question sentence; if there are no similar questions, an answer is generated in the offline document using the trained model, and the learner is guided to expand the reading in the returned answer.
Description
Technical Field
The invention relates to a neural network-based teaching question and answer system and method, and belongs to the field of neural networks and automatic question and answer.
Background
Research in the fields of learning science and computer-aided learning provides evidence that learner's dialogue is an important influencing factor in the construction of knowledge networks. In the online teaching, the classroom question-answering is a very important component in the traditional online teaching mode, so that the online question-answering technology of the online learning community restores the online teaching scene, provides an online environment interacting with learning partners for learners and teachers, and promotes knowledge flow and experience sharing between teachers and learners and between learners and learners in the community. However, online questioning and answering lacks real-time, for example, in the questioning and answering platform of MOOC platform "courrera" created by stanford university, the median time for a question posed by a learner to be answered is up to 22 minutes. Therefore, the learner cannot solve the doubt in time, and the teacher cannot know the knowledge point mastery condition of the learner in time and cannot achieve the mode of answering questions in time in a face-to-face mode in the off-line teaching.
Aiming at the defect of real-time online question answering, the introduction of an automatic question answering technology allows a learner to narrate through a natural language, and a system combines a natural language processing technology to 'understand' the problem of the learner and retrieve and obtain a corresponding answer to return to the learner in real time. However, due to the increasing number of learners in the online learning community, the questions asked by the learners in the community cannot be answered by other users timely and accurately. Meanwhile, since the descriptions of the questions by different learners are different, the same questions with answers already existed cannot be fed back to the learners in time. However, merely feeding back the answer to the question to the learner can make the learner passively receive the answer, and cannot make the learner actively extend reading, nor reflect the cognitive changes of the learner in participating in the question discussion. Therefore, a question-answering system, namely a question-answering system, which can not only solve the confusion of the learner accurately in time, but also guide the learner to read in an expanding way and finally reflect the cognitive change of the learner in time is needed to solve the problems.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the system and the method have the advantages that the problem of instantaneity and accuracy of on-line question answering is overcome, the problem of confusion of learners can be solved timely and accurately, the learners can be guided to read in an expanding mode, and finally the change of cognition of the learners is reflected timely.
The technical scheme of the invention is as follows:
a neural network-based lead learning question-answering system comprises a front end and a rear end, wherein the front end comprises a registration module, a login module, a question-asking module, an answer module, an evaluation module and a cognitive evaluation tree module; the back end comprises three subsystems, namely a cognitive evaluation subsystem, an answer reuse subsystem and an answer generation subsystem;
the registration module provides a function of learner registration, and the learner can log in only after the learner successfully registers;
the login module provides a login function for the learner, and the learner can use the functions of the questioning module, the answering module and the evaluation module only after successfully logging in;
the questioning module provides a function of questioning a learner, and the learner writes a question title, a question description and a question classification in a questioning interface;
the answer module provides the function of answering the questions of other learners by the learner, and the learner writes answer contents on the answer interface;
the evaluation module provides the learner with the answer function of evaluating other learners, and the praise and the deprecate of the learner are marked as standard answers or the learner writes evaluation contents in the evaluation interface;
the evaluation tree module is used for providing a function of the posting cognition degree in the course browsing of the teacher;
the cognitive evaluation subsystem is used for evaluating the cognitive behavior posted by the learner, and finally displaying a cognitive evaluation tree for each concept and knowledge point so as to provide a basis for a teacher to intervene in the conversation of the learner in the online learning community; the cognitive evaluation subsystem comprises a cognitive feature selection module and a cognitive evaluation module; the cognitive characteristic selection module obtains a characteristic vector of the cognitive state of the student according to the posting content of the student; the cognitive evaluation module obtains a cognitive evaluation tree according to the cognitive state of the student;
the answer reusing subsystem is used for repeatedly utilizing the questions containing high-quality answers existing in the online learning community, reducing the redundancy of the answers and quickly corresponding to the questions of the learner; the answer reuse subsystem includes: the question vector generation module obtains vector representation of the question according to the text content of the question; the question similarity calculation module obtains the similarity degree of the question similarity according to the vector representation of the questions;
the answer generation subsystem is used for generating answers to the questions of the learner according to the offline document data when similar questions cannot be found; the answer generation subsystem comprises a document screening module, a feature selection module and an answer selection module; the document screening module obtains a similar document set according to the similarity degree among the documents; the feature selection module obtains the digital vector representation of the post according to the text content of the post; the answer selection module obtains segments which can be used as answers according to the output of the neural network;
and the modules in the front-end interface and the modules in the back-end system interact through an HTTP protocol.
The invention discloses a neural network-based lead learning question and answer method, which comprises the following steps:
(1) for the question of a learner, firstly, using a Boolean expression to classify concepts and knowledge points, and for the classified questions, parallelly entering an answer reuse subsystem and a cognitive evaluation subsystem;
(2) in the answer reusing subsystem, the questions which are similar and have answers and are approved by other learners or marked as answers are screened out according to the knowledge points of the questions, similarity calculation is carried out on the questions and the questions of the learners, the most similar questions are screened out, and the answers with the most approved number or marked as standard answers under the questions are returned;
(3) if the answer of the similar question does not exist, entering an answer generation subsystem, generating the answer of the question according to the offline document, if the score of the answer is greater than a given threshold value, considering that the answer is high in quality, and returning the answer;
(4) if the question is not answered by the learner, or the answer is not approved or marked as the answer, the automatic answer is abandoned, and other learners are waited to answer the question;
(5) through the cognitive evaluation tree generated by the cognitive evaluation subsystem, a teacher intervenes the online learning community conversation so as to guide a learner to learn; the answer returned by the answer reusing subsystem and the answer generating subsystem is added with the answer source to guide the learner to expand reading.
Compared with the prior art, the invention has the advantages that:
(1) the invention is subdivided into two modules of answer reuse and answer generation, search for the most similar question set through improving TF-IDF search algorithm, search for the most probable answer in the off-line file through the neural network method at the same time, combine the results of the two to provide the accurate answer in time for the question of the learner;
(2) the cognitive classification framework can subdivide different cognition into specific conversation behaviors, and accurately define each cognitive behavior by combining the characteristics of community conversation contents, so that a classification model can accurately perform cognitive classification on the conversation in an online learning community;
(3) the invention uses a neural network model to carry out semantic analysis on the question of a learner, retrieves related contents in an offline document according to semantics, extracts word vectors and text features of the document and the question, finally obtains deep hidden layer vector features through a multilayer bidirectional cyclic neural network, and finally finds the position of the answer appearing in the document through a bilinear classifier and returns the position to the learner. We verified the improvement in the effect of the model compared to two reference models and two improved models through experiments modeling answer generation.
Drawings
FIG. 1 is a block diagram of the components of the system of the present invention;
FIG. 2 is a diagram of a cognitive assessment subsystem implementation process of the present invention;
FIG. 3 is a diagram of an answer reuse subsystem implementation of the present invention;
FIG. 4 is a diagram of an answer generation subsystem implementation of the present invention.
Detailed Description
The following detailed description is made with reference to the accompanying drawings and implementation procedures.
As shown in figure 1 of the drawings, in which,
the model and the framework of the system are written by Pycharm IDE under a MacOS 10.12.6 platform and developed by using Python language and Scala language. Training of deep learning models and the like is completed in a GTX 1080+ Cuda 8.0 environment under a Ubuntu 16.04 operating system. In the implementation of the neural network model, the current popular PyTorch deep learning framework is used.
Collecting the questions of learners, classifying concepts and knowledge points by using a Boolean expression, and entering the classified questions into an answer reuse subsystem and a cognitive evaluation subsystem in parallel; in the answer reusing subsystem, the similar questions with high quality answers are screened out according to the knowledge points of the questions, the similarity calculation is carried out on the questions and the questions of the learner, the most similar questions are screened out, and the answers with the most praise numbers or marked as standard answers under the questions are returned; if the answer of the similar question does not exist, entering an answer generation subsystem, generating the answer of the question according to the offline document, if the score of the answer is greater than a given threshold value, considering that the answer is high in quality, and returning the answer; if the answer quality is low, giving up the automatic answer and waiting for other learners to answer the question; finally, through a cognitive evaluation tree generated by the cognitive evaluation subsystem, a teacher can intervene in community conversation so as to guide learners to learn; the answer returned by the answer reusing subsystem and the answer generating subsystem is added with the answer source to guide the learner to expand reading.
The back end is mainly divided into three subsystems, namely a cognitive evaluation subsystem, an answer reuse subsystem and an answer generation subsystem, wherein each subsystem has the functions as follows:
(1) cognitive evaluation subsystem
The cognitive evaluation subsystem mainly comprises a cognitive characteristic selection module and a cognitive evaluation module, and is mainly used for evaluating the cognitive behavior posted by the learner, finally displaying a cognitive evaluation tree for each concept and knowledge point, and providing a basis for a teacher to intervene the conversation of a learner in the online learning community;
(2) answer reuse subsystem
The answer reusing subsystem mainly comprises a question vector generating module and a question similarity calculating module, and aims to recycle the existing questions containing high-quality answers in the online learning community, reduce the redundancy of the answers and quickly correspond to the questions of the learner;
(3) answer generation subsystem
The answer generating subsystem mainly comprises a document screening module, a feature selecting module and an answer selecting module, and aims to generate an answer to a question of a learner according to offline document data when the answer reusing subsystem cannot find similar questions.
The modules in the front-end interface and the modules in the back-end system interact with each other through an HTTP protocol, and specific interaction requests are set forth in the design and implementation of each subsystem in the following.
As shown in fig. 2, the answer reuse process can be described as follows:
(1) preprocessing the question presented by the learner on the text, such as taking out punctuation marks, stop words and the like;
(2) classifying the preprocessed questions by a Boolean rule, and finding out a target question set which is classified in the same way and contains the questions with high-quality answers in the question set according to the classification;
(3) performing semantic expansion on words in the preprocessed question to enrich the number of key words of the question;
(4) weighting calculation of key words is carried out on the questions in the target question set and the expanded question through TF-IDF to obtain a feature vector;
(5) carrying out similarity calculation on the problem vectors in the target problem set and the problem vector of the learner; and returning the answer with the most praise number or marked as the standard answer under the question with the highest similarity to the learner, and simultaneously returning the URL link of the question to guide the learner to read the similar question.
As shown in fig. 3, the answer generation process can be described as:
(1) preprocessing the question proposed by the learner, wherein part of punctuations are required to be reserved as features;
(2) performing n-gram expansion on the preprocessed problem and hashing each word into one bit in a binary system;
(3) searching the binary codes in the document set, and screening out the most relevant document set;
(4) extracting features of the problems and the documents respectively to obtain vectors input into the neural network;
(5) respectively inputting the problem vector and the document vector into two multilayer bidirectional RNN networks to obtain hidden layer feature vectors of the problem and the document;
(6) inputting the question and the hidden layer feature vector of the document into a bilinear classifier to predict the positions of the beginning and the end of the answer, returning the generated answer to the user, and attaching a link of a document source to guide the learner into link expansion reading.
As shown in FIG. 4, the present invention divides the learner's question asking system to return an answer into the following steps:
(1) the question-answering system firstly classifies concepts and knowledge points by using a Boolean expression for the question of a learner, and the classified questions enter an answer reuse subsystem and a cognitive evaluation subsystem in parallel;
(2) in the answer reusing subsystem, the similar questions with high quality answers are screened out according to the knowledge points of the questions, the similarity calculation is carried out on the questions and the questions of the learner, the most similar questions are screened out, and the answers with the most praise numbers or marked as standard answers under the questions are returned;
(3) if the answer of the similar question does not exist, entering an answer generation subsystem, generating the answer of the question according to the offline document, if the score of the answer is greater than a given threshold value, considering that the answer is high in quality, and returning the answer;
(4) if the answer quality is low, giving up the automatic answer and waiting for other learners to answer the question; through the cognitive evaluation tree generated by the cognitive evaluation subsystem, a teacher can intervene in community conversation, so that a learner is guided to learn; the answer returned by the answer reusing subsystem and the answer generating subsystem is added with the answer source to guide the learner to expand reading.
Table 1 shows the predicted results of the answer generation system and comparison system on the validation set constructed by the present invention using default model parameters and features.
TABLE 1 comparison of answer generation model predictions
As can be seen from the table, other systems have improved F1 to a different degree than the benchmark model because these models improve the model structure and introduce new text semantic expression features. The value of F1 is the harmonic mean value of the precision rate and the recall rate, i.e. F1 is 2PR/P + R, P is the precision rate, and is the ratio of the number of positive samples predicted to be correct to the number of all positive samples, in this experiment, the number of questions answered correctly by the system is the ratio of the number of questions answered by the system; r is the recall rate, is the ratio of the number of positive samples which are predicted to be correct and all samples which are judged to be positive, and in the experiment, is the ratio of the number of questions which are answered correctly by the system to all the questions. The F1 value is one of the important indicators for evaluating the predicted effect of the system.
Compared with the BiDAF system and the A3Net system, the introduction of the characteristics adds the attribute characteristics of the phrases on the characteristics, and the system can not only represent the phrases semantically, but also label the phrases on the part of speech and the entity information. Then, in the aspect of attention mechanism, self-alignment attention characteristics of the problems are added, so that the model can not only add weights to the words in the documents according to the problems, but also update the weights in the training process according to the importance degree of the words in the problems. Finally, on the output implicit layer vector mapping, two linear classifiers are used for respectively predicting the starting position and the ending position of the answer instead of one softmax classifier, so that the predictions are not interfered, and the prediction accuracy is improved. In summary, the lead quiz system is improved in F1 value compared to the BiDAF system and the A3Net system, indicating that the system has improved accuracy in answering the learner's question.
Claims (2)
1. A neural network-based lead learning question-answering system is characterized in that: the system comprises a front end and a rear end, wherein the front end comprises a registration module, a login module, a question-asking module, an answer module, an evaluation module and a cognitive evaluation tree module; the back end comprises three subsystems, namely a cognitive evaluation subsystem, an answer reuse subsystem and an answer generation subsystem;
the registration module provides a function of learner registration, and the learner can log in only after the learner successfully registers;
the login module provides a login function for the learner, and the learner can use the functions of the questioning module, the answering module and the evaluation module only after successfully logging in;
the questioning module provides a function of questioning a learner, and the learner writes a question title, a question description and a question classification in a questioning interface;
the answer module provides the function of answering the questions of other learners by the learner, and the learner writes answer contents on the answer interface;
the evaluation module provides the learner with the answer function of evaluating other learners, and the praise and the deprecate of the learner are marked as standard answers or the learner writes evaluation contents in the evaluation interface;
the evaluation tree module is used for providing a function of the posting cognition degree in the course browsing of the teacher;
the cognitive evaluation subsystem is used for evaluating the cognitive behavior posted by the learner, and finally displaying a cognitive evaluation tree for each concept and knowledge point so as to provide a basis for a teacher to intervene in the conversation of the learner in the online learning community; the cognitive evaluation subsystem comprises a cognitive feature selection module and a cognitive evaluation module; the cognitive characteristic selection module obtains a characteristic vector of the cognitive state of the student according to the posting content of the student; the cognitive evaluation module obtains a cognitive evaluation tree according to the cognitive state of the student;
the answer reusing subsystem is used for repeatedly utilizing the questions containing high-quality answers existing in the online learning community, reducing the redundancy of the answers and quickly corresponding to the questions of the learner; the answer reuse subsystem includes: the question vector generation module obtains vector representation of the question according to the text content of the question; the question similarity calculation module obtains the similarity degree of the question similarity according to the vector representation of the questions;
the answer generation subsystem is used for generating answers to the questions of the learner according to the offline document data when similar questions cannot be found; the answer generation subsystem comprises a document screening module, a feature selection module and an answer selection module; the document screening module obtains a similar document set according to the similarity degree among the documents; the feature selection module obtains the digital vector representation of the post according to the text content of the post; the answer selection module obtains segments which can be used as answers according to the output of the neural network;
and the modules in the front-end interface and the modules in the back-end system interact through an HTTP protocol.
2. A neural network-based lead learning question-answering method is characterized by comprising the following steps:
(1) for the question of a learner, firstly, using a Boolean expression to classify concepts and knowledge points, and for the classified questions, parallelly entering an answer reuse subsystem and a cognitive evaluation subsystem;
(2) in the answer reusing subsystem, the questions which are similar and have answers and are approved by other learners or marked as answers are screened out according to the knowledge points of the questions, similarity calculation is carried out on the questions and the questions of the learners, the most similar questions are screened out, and the answers with the most approved number or marked as standard answers under the questions are returned;
(3) if the answer of the similar question does not exist, entering an answer generation subsystem, generating the answer of the question according to the offline document, if the score of the answer is greater than a given threshold value, considering that the answer is high in quality, and returning the answer;
(4) if the question is not answered by the learner, or the answer is not approved or marked as the answer, the automatic answer is abandoned, and other learners are waited to answer the question;
(5) through the cognitive evaluation tree generated by the cognitive evaluation subsystem, a teacher intervenes the online learning community conversation so as to guide a learner to learn; the answer returned by the answer reusing subsystem and the answer generating subsystem is added with the answer source to guide the learner to expand reading.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910187119.7A CN109871441A (en) | 2019-03-13 | 2019-03-13 | One kind knowledge neural network based of leading answers system and method |
CN2019101871197 | 2019-03-13 |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110619042A true CN110619042A (en) | 2019-12-27 |
CN110619042B CN110619042B (en) | 2022-08-16 |
Family
ID=66920369
Family Applications (2)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910187119.7A Pending CN109871441A (en) | 2019-03-13 | 2019-03-13 | One kind knowledge neural network based of leading answers system and method |
CN201911066052.8A Active CN110619042B (en) | 2019-03-13 | 2019-11-04 | Neural network-based teaching question and answer system and method |
Family Applications Before (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910187119.7A Pending CN109871441A (en) | 2019-03-13 | 2019-03-13 | One kind knowledge neural network based of leading answers system and method |
Country Status (1)
Country | Link |
---|---|
CN (2) | CN109871441A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111445153A (en) * | 2020-03-31 | 2020-07-24 | 华中师范大学 | Objective test question attribute mode estimation and correction method and system for education measurement |
TWI750665B (en) * | 2020-05-19 | 2021-12-21 | 南一書局企業股份有限公司 | System for marking e-book exercises for questioning |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
FR3087308B1 (en) * | 2018-10-10 | 2021-09-10 | Happy Ledger | MULTIPLE IDENTIFICATION SYSTEM AND METHOD BY INTELLIGENT BLOCK CHAIN CONTRACTS |
CN111353032B (en) * | 2020-02-27 | 2022-06-14 | 福州大学 | Community question and answer oriented question classification method and system |
CN113127734B (en) * | 2021-03-30 | 2022-11-25 | 江苏乐易智慧科技有限公司 | Personalized test paper pushing method based on genetic algorithm |
Citations (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1686513A2 (en) * | 2005-01-28 | 2006-08-02 | Expert-24 Ltd. | Improvements relating to expert systems |
CN101000600A (en) * | 2006-12-30 | 2007-07-18 | 南京凌越教育科技服务有限公司 | Study management system and method |
US20090186330A1 (en) * | 2008-01-19 | 2009-07-23 | International Business Machines Corporation | Reusable ad hoc self-posed questions and answers for social network profiles |
CN105760417A (en) * | 2015-01-02 | 2016-07-13 | 国际商业机器公司 | Cognitive Interactive Searching Method And System Based On Personalized User Model And Context |
CN106295792A (en) * | 2016-08-05 | 2017-01-04 | 北京光年无限科技有限公司 | Dialogue data interaction processing method based on multi-model output and device |
CN106960245A (en) * | 2017-02-24 | 2017-07-18 | 中国科学院计算技术研究所 | A kind of individualized medicine evaluation method and system based on cognitive process chain |
CN107908803A (en) * | 2017-12-26 | 2018-04-13 | 上海智臻智能网络科技股份有限公司 | The response method and device, storage medium, terminal of question and answer interaction |
CN108021703A (en) * | 2017-12-26 | 2018-05-11 | 广西师范大学 | A kind of talk formula intelligent tutoring system |
CN108288468A (en) * | 2017-06-29 | 2018-07-17 | 腾讯科技(深圳)有限公司 | Audio recognition method and device |
CN108345585A (en) * | 2018-01-11 | 2018-07-31 | 浙江大学 | A kind of automatic question-answering method based on deep learning |
CN108763284A (en) * | 2018-04-13 | 2018-11-06 | 华南理工大学 | A kind of question answering system implementation method based on deep learning and topic model |
CN108959556A (en) * | 2018-06-29 | 2018-12-07 | 北京百度网讯科技有限公司 | Entity answering method, device and terminal neural network based |
CN108984778A (en) * | 2018-07-25 | 2018-12-11 | 南京瓦尔基里网络科技有限公司 | A kind of intelligent interaction automatically request-answering system and self-teaching method |
CN109062958A (en) * | 2018-06-26 | 2018-12-21 | 华中师范大学 | It is a kind of based on the primary school of TextRank and convolutional neural networks write a composition automatic classification method |
-
2019
- 2019-03-13 CN CN201910187119.7A patent/CN109871441A/en active Pending
- 2019-11-04 CN CN201911066052.8A patent/CN110619042B/en active Active
Patent Citations (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1686513A2 (en) * | 2005-01-28 | 2006-08-02 | Expert-24 Ltd. | Improvements relating to expert systems |
CN101000600A (en) * | 2006-12-30 | 2007-07-18 | 南京凌越教育科技服务有限公司 | Study management system and method |
US20090186330A1 (en) * | 2008-01-19 | 2009-07-23 | International Business Machines Corporation | Reusable ad hoc self-posed questions and answers for social network profiles |
CN105760417A (en) * | 2015-01-02 | 2016-07-13 | 国际商业机器公司 | Cognitive Interactive Searching Method And System Based On Personalized User Model And Context |
CN106295792A (en) * | 2016-08-05 | 2017-01-04 | 北京光年无限科技有限公司 | Dialogue data interaction processing method based on multi-model output and device |
CN106960245A (en) * | 2017-02-24 | 2017-07-18 | 中国科学院计算技术研究所 | A kind of individualized medicine evaluation method and system based on cognitive process chain |
CN108288468A (en) * | 2017-06-29 | 2018-07-17 | 腾讯科技(深圳)有限公司 | Audio recognition method and device |
CN107908803A (en) * | 2017-12-26 | 2018-04-13 | 上海智臻智能网络科技股份有限公司 | The response method and device, storage medium, terminal of question and answer interaction |
CN108021703A (en) * | 2017-12-26 | 2018-05-11 | 广西师范大学 | A kind of talk formula intelligent tutoring system |
CN108345585A (en) * | 2018-01-11 | 2018-07-31 | 浙江大学 | A kind of automatic question-answering method based on deep learning |
CN108763284A (en) * | 2018-04-13 | 2018-11-06 | 华南理工大学 | A kind of question answering system implementation method based on deep learning and topic model |
CN109062958A (en) * | 2018-06-26 | 2018-12-21 | 华中师范大学 | It is a kind of based on the primary school of TextRank and convolutional neural networks write a composition automatic classification method |
CN108959556A (en) * | 2018-06-29 | 2018-12-07 | 北京百度网讯科技有限公司 | Entity answering method, device and terminal neural network based |
CN108984778A (en) * | 2018-07-25 | 2018-12-11 | 南京瓦尔基里网络科技有限公司 | A kind of intelligent interaction automatically request-answering system and self-teaching method |
Non-Patent Citations (4)
Title |
---|
KAVYA KISHORE等: "Document Summarization in Malayalam with sentence framing", 《2016 INTERNATIONAL CONFERENCE ON INFORMATION SCIENCE (ICIS)》 * |
吴文峻: "面向智慧教育的学习大数据分析技术", 《电化教育研究》 * |
曾帅等: "面向知识自动化的自动问答研究进展", 《自动化学报》 * |
高翔: "基于神经网络的问答系统关键技术研究", 《中国优秀博硕士学位论文全文数据库(硕士)信息科技辑》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111445153A (en) * | 2020-03-31 | 2020-07-24 | 华中师范大学 | Objective test question attribute mode estimation and correction method and system for education measurement |
CN111445153B (en) * | 2020-03-31 | 2023-04-14 | 华中师范大学 | Method and system for estimating and correcting objective test question attribute mode for education measurement |
TWI750665B (en) * | 2020-05-19 | 2021-12-21 | 南一書局企業股份有限公司 | System for marking e-book exercises for questioning |
Also Published As
Publication number | Publication date |
---|---|
CN110619042B (en) | 2022-08-16 |
CN109871441A (en) | 2019-06-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110619042B (en) | Neural network-based teaching question and answer system and method | |
CN110196901B (en) | Method and device for constructing dialog system, computer equipment and storage medium | |
Palasundram et al. | Sequence to sequence model performance for education chatbot | |
US10861344B2 (en) | Personalized learning system and method for the automated generation of structured learning assets based on user data | |
Ondáš et al. | How chatbots can be involved in the education process | |
CN111081220B (en) | Vehicle-mounted voice interaction method, full-duplex dialogue system, server and storage medium | |
CN110825867B (en) | Similar text recommendation method and device, electronic equipment and storage medium | |
Van Nguyen et al. | Enhancing lexical-based approach with external knowledge for Vietnamese multiple-choice machine reading comprehension | |
US20220139248A1 (en) | Knowledge-grounded dialogue system and method for language learning | |
KR102644992B1 (en) | English speaking teaching method using interactive artificial intelligence avatar based on the topic of educational content, device and system therefor | |
CN112818106A (en) | Evaluation method of generating type question and answer | |
CN110991195A (en) | Machine translation model training method, device and storage medium | |
CN114218379A (en) | Intelligent question-answering system-oriented method for attributing questions which cannot be answered | |
Liu et al. | Cross-domain slot filling as machine reading comprehension: A new perspective | |
CN110675292A (en) | Child language ability evaluation method based on artificial intelligence | |
CN114330483A (en) | Data processing method, model training method, device, equipment and storage medium | |
CN117252259A (en) | Deep learning-based natural language understanding method and AI teaching aid system | |
Ye et al. | A sentiment based non-factoid question-answering framework | |
Zhu et al. | YUN111@ Dravidian-CodeMix-FIRE2020: Sentiment Analysis of Dravidian Code Mixed Text. | |
US20220309936A1 (en) | Video education content providing method and apparatus based on artificial intelligence natural language processing using characters | |
Krichen et al. | Online analysis of children handwritten words in dictation context | |
Liang et al. | Intelligent chat robot in digital campus based on deep learning | |
La et al. | Improving reading comprehension step by step using Online-Boost text readability classification system | |
JP2018031828A (en) | Program, device, and method automatically grading according to learner's speech sound | |
Young et al. | Evaluation of the Potential Usage of ChatGPT for Providing Easier Reading Materials for ESL Students |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |