CN110334204A - A kind of exercise similarity calculation recommended method based on user record - Google Patents
A kind of exercise similarity calculation recommended method based on user record Download PDFInfo
- Publication number
- CN110334204A CN110334204A CN201910444120.3A CN201910444120A CN110334204A CN 110334204 A CN110334204 A CN 110334204A CN 201910444120 A CN201910444120 A CN 201910444120A CN 110334204 A CN110334204 A CN 110334204A
- Authority
- CN
- China
- Prior art keywords
- exercise
- vector
- word
- user
- target
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/335—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Machine Translation (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The exercise similarity calculation recommended method based on user record that the invention discloses a kind of, the present invention is by utilizing item2vec thought and convolutional neural networks respectively advantage, the two is effectively combined, it solves in current exercise recommendation since topic includes that a large amount of formal notation content structures are complicated, the problem of semantic complications are to be difficult to match similar topic type, and can segment exercise in the angle of natural language processing, learn the specific grammer meaning of exercise, similar topic type is matched on the meaning of a word.It finally allows exercise recommender system preferably to recommend more matched similar topic type, promotes exercise and recommend matter.
Description
Technical field:
The invention belongs to software fields, and in particular to a kind of exercise similarity calculation recommended method based on user record.
Background technique:
The most common mode based on machine learning is the algorithm TIIDF of text similarity detection, and LSA, the methods of LDA exist
Data format, clean do it is fairly proper in the case where can obtain certain accuracy, but this is only in the weaker meaning of one's words
On it is similar, therefore effect is not very good in actual be recommended to use, and the topic of recommendation is substantially closely similar (belong to
In the topic that one is looked like), algorithm how is improved here in the understanding of meaning of one's words level, really obtains the correlation of exercise semantically
It is very important, is used here based on the algorithm of deep learning by many scenes, be based on LSTM, the model of CNN can be certain
The meaning of one's words of sentence is learnt and indicated in degree, therefore is existed using the algorithm of deep learning in progress text similarity matching
It is more much better than traditional machine learning method in effect, but since exercise and urtext sentence have very big difference,
Sentence looks like more tortuous change, and a variety of texts mix (mathematic sign and formula etc.), and these be substantially on model by
It filters out, the matching accuracy of sentence will be greatly reduced in this way.The existing mode based on deep learning is also difficult to obtain
Satisfactory result.
Explanation of nouns:
Word2vec: being the word incorporation model that Google 2013 proposes, actually a kind of neural network model of shallow-layer,
There are two types of network structures, respectively CBOW and Skip-gram.This patent mainly uses the network model of word2vec to be
Skip-gram。
Item2vec: mainly the method for word2vec is used in recommender system, using item of merchandise as in word2vec
The word in face, using the item of merchandise set of user's single purchase as the sentence inside word2vec.
Skip-gram network model: being by input layer, mapping layer, the neural network of output layer composition, is by target list
Word speculates context, that is, inputs a target word, obtain context words.
Softmax: to normalize exponential function, the output of multiple neurons is mapped in (0,1) section, can be considered as
Probability.
Cross entropy: as the loss function of model, the main difference distance for calculating reality output and desired output.Intersect
Entropy is smaller, and reality output and desired output are with regard to closer.
Primary practice: the setting quantity once done with the set or client of the exercise done in client's certain time period
Exercise.
Same type exercise: the i.e. exercise that two exercises are a subject or the form of the same race under a knowledge point.
Summary of the invention:
The exercise similarity calculation recommended method based on user record that the invention discloses a kind of；The present invention can be answered preferably
Cause similarity calculation is improper exercise is made to recommend the problem for inaccuracy occur exercise content structure complexity, effective improve is practised
Topic recommends accuracy.
In order to reach the invention purpose, the invention adopts the following technical scheme:
A kind of exercise similarity calculation recommended method based on user record, comprising the following steps:
Step 1: carry out word segmentation processing using each exercise as sentence, the word for obtaining segmenting in exercise be embedded in
Amount, then the word insertion vector of all words of each exercise is connected as a matrix according to the successive word order that word in exercise occurs,
The exercise matrix for obtaining representing exercise information is handled exercise matrix using convolutional neural networks model: convolutional Neural net
Network model carries out convolution using various sizes of filter, obtains multiple output features, and the result for exporting feature is carried out pond
Processing, is spliced into a vector v ector1；
Step 2: as a whole by exercise, compute exercises between similarity: regard exercise itself as a word,
The set for the exercise that each user was once done is as a sentence；Two exercises are calculated in the set of same exercise simultaneously
Similarity of the probability of appearance as two exercises；Finally obtain the insertion vector of each exercise itself, i.e. vector v ector2；
Step 3: vector1 and vector2 are spliced to obtain final vector v ector, carried out by vector v ector
Training obtains trained model；
Step 4: the nearest exercise done to trained mode input user, output result is to own in exam pool
It is same category of probability is to recommend probability that exercise, which corresponds to the exercise that user is done, to the probability of exercise all in result into
Row sequence processing chooses the exercise for recommending a user of maximum probability not do also and user is showed to complete recommendation task；A is
The exercise of setting recommends number.
Further to improve, the step 1 includes the following steps:
Step 1: being segmented using third party library jieba Chinese word segmentation component to each exercise, obtained participle is made
It is trained with the skip-gram network model of word2vec, each word in exercise is mapped to a d dimension term vector, it will
The term vector of all participles is attached according to the semantic sequence in exercise in each exercise, is obtained one and is represented exercise square
Battle array；The maximum exercise of word number is taken, word number is n；Each exercise is handled as n*d matrix, the exercise that word number is less than n carries out mending 0 behaviour
Make, so that the data of input keep dimension consistent；Learn exercise matrix using convolution model, 2*d, tri- rulers of 3*d, 5*d are set
Very little and each size uses three filters to carry out convolution operation respectively, and output feature is carried out maximum pondization operation；It will processing
Nine output features results be spliced into one include exercise semantic information vector v ector1.
Further to improve, the step 2 includes the following steps:
The insertion vector of each exercise is obtained with skip-gram network model: done in first once practicing user
Exercise as one gather, if user once practice done in exercise quantity be S, exercise is respectively W1, W2, W3 ... WS；
Current goal exercise Wi is chosen, is amounted to using the output of skip-gram network model and current goal exercise Wi in a workbook
Other existing exercise, that is, positive samples；The training of model will make the target exercise Wi in all exercise set once practice with user
In remaining exercise co-occurrence two-by-two conditional probability it is maximum, i.e.,It is maximum；
Wherein,
Wherein uiIt is the vector of target exercise Wi, vjBe with target exercise Wi and meanwhile appear in set in exercise
Vector；I represents the exam pool comprising all exercises；K indicates the exercise of input exam pool；Wj indicate with
The exercise different from target exercise Wi in the exercise that family is once practiced；
With the method for negative sampling, that is, multiple and target exercise Wi is randomly selected not in the exercise of identity set i.e. negative sample
Come the optimization exported, the training calculation amount of model is reduced；Finally obtain the insertion vector expression of exercise: vector itself
vector2。
Further to improve, the step 3 includes the following steps:
Vector v ector1 and vector v ector2 are spliced to obtain final vector v ector, vector v ector is input to
In one neural network connected entirely, learning training is then carried out by final vector v ector: with same type exercise work
For a training set, several training set cooperations are a training set, input target exercise, the exercise of desired output is and target
Exercise belongs to other exercises of same type exercise, so that it is the general of other same type of exercises that output, which is with preceding target exercise,
Rate is maximum, is not the probability minimum of other same type of exercises calculated with current goal exercise, obtains trained
Model.
It is further to improve, acceleration training is carried out using the method for negative sampling in trained process, i.e., for a target
The input of exercise, randomly selects e and non-sum target exercise appears in the exercise i.e. negative sample of identity set to parameter
Renewal process optimizes, and reduces calculation amount, accelerates the training speed of network.
Beneficial effects of the present invention: the present invention, which can preferably cope with exercise content structure complexity, causes similarity calculation improper
So that exercise recommends the problem for inaccuracy occur, the effective exercise that improves recommends accuracy.
Detailed description of the invention
Attached drawing is used to provide further understanding of the present invention, and constitutes part of specification, with tool of the invention
Body embodiment is used to explain the present invention together, is not construed as limiting the invention.In the accompanying drawings:
Fig. 1 is the flow diagram of step 1.
Fig. 2 is final flowsheet schematic diagram of the invention.
Specific embodiment:
Specific embodiments of the present invention are described in further detail with reference to the accompanying drawing, it should be understood that described herein
Embodiment for instruction and explanation of the present invention, be not intended to limit the present invention.
Embodiment 1
Shown in specific steps Fig. 1 and Fig. 2 of the present invention:
1) each exercise is segmented using third party library jieba Chinese word segmentation first, obtained participle is used
The skip-gram network model of word2vec is trained, so that each word is mapped to a d dimension term vector.By exercise
The term vector of all participles is attached, and obtains the matrix for representing exercise.The maximum exercise of word number is taken, word number is n.It will be every
A exercise processing is n*d matrix, and exercise of the word number less than n carries out mending 0 dimension for maintaining input.Finally all exercises are indicated
For n*d matrix.Next learn exercise matrix with convolutional neural networks, 2*d is set, tri- sizes of 3*d, 5*d distinguish three mistakes
Filter carries out convolution operation, and will the maximum pondization operation output maximum value of output feature.The knot of nine output features will be handled
Fruit be spliced into one include exercise semantic information vector v ector1.
2) as a whole by exercise, attempt to obtain by the thought of item2vec skip-gram network model every
The insertion vector of a exercise, the exercise that user is once done is gathered as one, if this exercise quantity done of user is
S, exercise W1, W2, W3 ... WS.We choose current goal exercise Wi, then needing skip-gram network output and current
Target exercise gathers other exercise, that is, positive samples of co-occurrence at one, and the exercise not occurred in a set is negative sample.
The training of model will make the conditional probability of the exercise of co-occurrence two-by-two in an exercise set maximum.The corresponding objective function of model
It is as follows:
Wherein p (Wj | Wi) is a softmax function:
Wherein uiIt is the vector of target exercise Wi, vjBe with target exercise Wi and meanwhile appear in set in exercise vector；
I represents the exam pool comprising all exercises；K indicates the exercise of input exam pool；
With the method for negative sampling, that is, randomly selects and multiple do not exported in the exercise of current collection i.e. negative sample
Optimization only needs to update a small amount of parameter every time to finally obtain the insertion vector expression of exercise: vector itself to accelerate to train
vector2。
3) vector v ector1 and vector v ector2 are spliced to obtain final vector v ector, vector v ector is inputted
In the neural network connected entirely to one, then carry out learning training: input target exercise, the exercise of desired output are and target
Exercise belongs to other exercises of a classification, a target exercise vector v ector is specially inputted, by multilayer neural network
Obtain the probability for the exercise that each exercise and current exercise in exam pool are the same type later with softmax function normalization,
The fit object of model be make be with current goal exercise other same type of exercises the maximum probability calculated, make and
Current goal exercise is not that the probability of other same type of exercises calculated is minimum.Model being capable of root after training is completed
Other exercises in exam pool are calculated according to target exercise vector and it be same type of probability is the probability for recommending these exercises.
The quantity of exercise is larger, if may require that a large amount of calculating and time using normal sample training method, adopts here
The optimization exported with the thought of negative sampling, concrete measure are to randomly select several negative samples for a target exercise
(general setting quantity is 3-7), is trained in the form of cross entropy, so that the training of model is completed, which comparison
The training of total amount exercise more saves training and calculates cost and training time.
4) by the trained model of step 3, the done exercise sample of user is inputted, output result is that all exercises correspond to the habit
The probability recommended is inscribed, this probability represents and the exercise belongs to same category of probability, to exercise all in result
Probability is ranked up processing, chooses the exercise that maximum a user did not did also and user is showed to complete recommendation task.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.
Claims (5)
1. a kind of exercise similarity calculation recommended method based on user record, which comprises the following steps:
Step 1: carrying out word segmentation processing for each exercise as a sentence, the word segmented in exercise insertion vector is obtained, then
The word insertion vector of all words of each exercise is connected as a matrix according to the successive word order that word in exercise occurs, obtains generation
The exercise matrix of table exercise information is handled exercise matrix using convolutional neural networks model: convolutional neural networks model
Convolution is carried out using various sizes of filter, obtains multiple output features, the result for exporting feature is subjected to pond processing, is spelled
It is connected in a vector v ector1；
Step 2: as a whole by exercise, compute exercises between similarity: regard exercise itself as a word, will be each
The set for the exercise that a user once did is as a sentence；Two exercises are calculated in the set of same exercise while being occurred
Similarity of the probability as two exercises；Finally obtain the insertion vector of each exercise itself, i.e. vector v ector2；
Step 3: vector1 and vector2 are spliced to obtain final vector v ector, it is trained by vector v ector
Obtain trained model；
Step 4: the nearest exercise done to trained mode input user, output result is all exercises in exam pool
The exercise that corresponding user is done is that same category of probability is to recommend probability, is arranged the probability of exercise all in result
Sequence processing chooses the exercise for recommending a user of maximum probability not do also and user is showed to complete recommendation task；A is setting
Exercise recommend number.
2. the exercise similarity calculation recommended method based on user record as described in claim 1, which is characterized in that the step
Rapid one includes the following steps:
Step 1: being segmented using third party library jieba Chinese word segmentation component to each exercise, obtained participle is used
The skip-gram network model of word2vec is trained, and each word in exercise is mapped to a d dimension term vector, will be every
The term vector of all participles is attached according to the semantic sequence in exercise in a exercise, is obtained one and is represented exercise matrix；
The maximum exercise of word number is taken, word number is n；Each exercise is handled as n*d matrix, the exercise that word number is less than n carries out mending 0 operation,
So that the data of input keep dimension consistent；Using convolution model learn exercise matrix, be arranged 2*d, tri- sizes of 3*d, 5*d and
Each size uses three filters to carry out convolution operation respectively, and output feature is carried out maximum pondization operation；By the nine of processing
The result of a output feature is spliced into the vector v ector1 comprising exercise semantic information.
3. the exercise similarity calculation recommended method based on user record as claimed in claim 2, which is characterized in that the step
Rapid two include the following steps:
The insertion vector of each exercise is obtained with skip-gram network model: exercise done in first once practicing user
As one gather, if user once practice done in exercise quantity be S, exercise is respectively W1, W2, W3 ... WS；It chooses
Current goal exercise Wi is amounted to using the output of skip-gram network model and current goal exercise Wi in a workbook existing
Other exercise, that is, positive samples；The training of model to make in all exercise set target exercise Wi and user once practice in its
The conditional probability of remaining exercise co-occurrence two-by-two is maximum, i.e.,It is maximum；
Wherein,
Wherein uiIt is the vector of target exercise Wi, vjBe with target exercise Wi and meanwhile appear in set in exercise vector；I generation
Table includes the exam pool of all exercises；K indicates the exercise of input exam pool；Wj indicates to practise in the exercise that user once practices with target
Inscribe the different exercise of Wi；
With the method for negative sampling, that is, randomly select it is multiple with target exercise Wi not the exercise of identity set i.e. negative sample come into
The optimization of row output, reduces the training calculation amount of model；Finally obtain the insertion vector expression of exercise: vector v ector2 itself.
4. the exercise similarity calculation recommended method based on user record as claimed in claim 3, which is characterized in that the step
Rapid three include the following steps:
Vector v ector1 and vector v ector2 are spliced to obtain final vector v ector, vector v ector is input to one
In the neural network connected entirely, then pass through final vector v ector and carry out learning training: using same type exercise as one
A trained set, several training set cooperations are a training set, input target exercise, and the exercise of desired output is and target exercise
Belong to other exercises of same type exercise so that output for preceding target exercise be the probability of other same type of exercises most
Greatly, it is not the probability minimum of other same type of exercises calculated with current goal exercise, obtains trained model.
5. the exercise similarity calculation recommended method based on user record as claimed in claim 4, which is characterized in that training
Acceleration training, the i.e. input for a target exercise are carried out using the method for negative sampling in the process, randomly select out e not
The exercise i.e. negative sample of identity set is appeared in target exercise to optimize to the renewal process of parameter, reduces calculation amount,
Accelerate the training speed of network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910444120.3A CN110334204B (en) | 2019-05-27 | 2019-05-27 | Exercise similarity calculation recommendation method based on user records |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910444120.3A CN110334204B (en) | 2019-05-27 | 2019-05-27 | Exercise similarity calculation recommendation method based on user records |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110334204A true CN110334204A (en) | 2019-10-15 |
| CN110334204B CN110334204B (en) | 2022-10-18 |
Family
ID=68140298
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910444120.3A Active CN110334204B (en) | 2019-05-27 | 2019-05-27 | Exercise similarity calculation recommendation method based on user records |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110334204B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111143604A (en) * | 2019-12-25 | 2020-05-12 | 腾讯音乐娱乐科技（深圳）有限公司 | Audio similarity matching method and device and storage medium |
| CN117688248A (en) * | 2024-02-01 | 2024-03-12 | 安徽教育网络出版有限公司 | Online course recommendation method and system based on convolutional neural network |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107832453A (en) * | 2017-11-24 | 2018-03-23 | 重庆科技学院 | Virtual test paper recommendation method oriented to personalized learning scheme |
| US20180218429A1 (en) * | 2017-01-31 | 2018-08-02 | Wal-Mart Stores, Inc. | Systems and methods for utilizing a convolutional neural network architecture for visual product recommendations |
| CN109271401A (en) * | 2018-09-26 | 2019-01-25 | 杭州大拿科技股份有限公司 | Method, apparatus, electronic equipment and storage medium are corrected in a kind of search of topic |
| CN109299380A (en) * | 2018-10-30 | 2019-02-01 | 浙江工商大学 | Exercise personalized recommendation method in online education platform based on multidimensional characteristic |
| CN109635100A (en) * | 2018-12-24 | 2019-04-16 | 上海仁静信息技术有限公司 | A kind of recommended method, device, electronic equipment and the storage medium of similar topic |
-
2019
- 2019-05-27 CN CN201910444120.3A patent/CN110334204B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180218429A1 (en) * | 2017-01-31 | 2018-08-02 | Wal-Mart Stores, Inc. | Systems and methods for utilizing a convolutional neural network architecture for visual product recommendations |
| CN107832453A (en) * | 2017-11-24 | 2018-03-23 | 重庆科技学院 | Virtual test paper recommendation method oriented to personalized learning scheme |
| CN109271401A (en) * | 2018-09-26 | 2019-01-25 | 杭州大拿科技股份有限公司 | Method, apparatus, electronic equipment and storage medium are corrected in a kind of search of topic |
| CN109299380A (en) * | 2018-10-30 | 2019-02-01 | 浙江工商大学 | Exercise personalized recommendation method in online education platform based on multidimensional characteristic |
| CN109635100A (en) * | 2018-12-24 | 2019-04-16 | 上海仁静信息技术有限公司 | A kind of recommended method, device, electronic equipment and the storage medium of similar topic |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111143604A (en) * | 2019-12-25 | 2020-05-12 | 腾讯音乐娱乐科技（深圳）有限公司 | Audio similarity matching method and device and storage medium |
| CN111143604B (en) * | 2019-12-25 | 2024-02-02 | 腾讯音乐娱乐科技（深圳）有限公司 | Similarity matching method and device for audio frequency and storage medium |
| CN117688248A (en) * | 2024-02-01 | 2024-03-12 | 安徽教育网络出版有限公司 | Online course recommendation method and system based on convolutional neural network |
| CN117688248B (en) * | 2024-02-01 | 2024-04-26 | 安徽教育网络出版有限公司 | Online course recommendation method and system based on convolutional neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110334204B (en) | 2022-10-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111415740B (en) | Method and device for processing inquiry information, storage medium and computer equipment | |
| Miech et al. | Howto100m: Learning a text-video embedding by watching hundred million narrated video clips | |
| CN106503055B (en) | A kind of generation method from structured text to iamge description | |
| Trischler et al. | Natural language comprehension with the epireader | |
| Wang et al. | Multi-perspective context matching for machine comprehension | |
| CN107230174B (en) | Online interactive learning system and method based on network | |
| CN108984530A (en) | A kind of detection method and detection system of network sensitive content | |
| Xu et al. | How images inspire poems: Generating classical Chinese poetry from images with memory networks | |
| CN110516245A (en) | Fine granularity sentiment analysis method, apparatus, computer equipment and storage medium | |
| CN112084299B (en) | Reading comprehension automatic question-answering method based on BERT semantic representation | |
| CN106502985A (en) | A kind of neural network modeling approach and device for generating title | |
| CN108549658A (en) | A kind of deep learning video answering method and system based on the upper attention mechanism of syntactic analysis tree | |
| CN109710744A (en) | A kind of data matching method, device, equipment and storage medium | |
| CN111737427B (en) | Method for recommending lesson forum posts by combining forum interaction behaviors and user reading preference | |
| CN105446955A (en) | Adaptive word segmentation method | |
| CN109165040A (en) | A method of the code copy suspicion detection based on Random Forest model | |
| CN107688583A (en) | The method and apparatus for creating the training data for natural language processing device | |
| CN110334204A (en) | A kind of exercise similarity calculation recommended method based on user record | |
| US20170193333A1 (en) | Analyzing and exploring images posted on social media | |
| CN113065757A (en) | Method and device for evaluating on-line course teaching quality | |
| Yu et al. | UGAN: unified generative adversarial networks for multidirectional text style transfer | |
| Shi et al. | Prompt space optimizing few-shot reasoning success with large language models | |
| CN117473041A (en) | Programming knowledge tracking method based on cognitive strategy | |
| Schmaltz | On the utility of lay summaries and AI safety disclosures: Toward robust, open research oversight | |
| Giabbanelli et al. | Generative AI for Systems Thinking: Can a GPT Question-Answering System Turn Text into the Causal Maps Produced by Human Readers? |
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 |