CN115098623A - Physical training data feature extraction method based on BERT - Google Patents
Physical training data feature extraction method based on BERT Download PDFInfo
- Publication number
- CN115098623A CN115098623A CN202210628842.6A CN202210628842A CN115098623A CN 115098623 A CN115098623 A CN 115098623A CN 202210628842 A CN202210628842 A CN 202210628842A CN 115098623 A CN115098623 A CN 115098623A
- Authority
- CN
- China
- Prior art keywords
- training
- bert
- population
- physical
- physical training
- 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.)
- Pending
Links
- 238000012549 training Methods 0.000 title claims abstract description 149
- 238000000605 extraction Methods 0.000 title claims abstract description 23
- 230000035790 physiological processes and functions Effects 0.000 claims abstract description 15
- 230000004927 fusion Effects 0.000 claims abstract description 11
- 238000012795 verification Methods 0.000 claims abstract description 7
- 238000012360 testing method Methods 0.000 claims abstract description 5
- 238000000034 method Methods 0.000 claims description 24
- 230000003993 interaction Effects 0.000 claims description 4
- 238000002790 cross-validation Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 description 6
- 238000003058 natural language processing Methods 0.000 description 6
- 238000007405 data analysis Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
Images
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
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- 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
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the invention provides a physical training data feature extraction method based on BERT, which comprises the steps of taking 80% of daily training data of people who perform physical training as a training set of a model, taking 10% as a verification set and taking 10% as a test set, performing K-fold cross verification, and clustering the training by using imbedding of a word2vec pre-training physical training item in a BERT part; improving MLM, changing the form of the whole training information including attributes of the original direct mask into only one part of information for mask training; the fusion part fuses the age, the sex, the physiological state and the psychological state distribution of the population developing physical training, which meet the preset conditions, into a network; the output part is used for classifying and predicting the age, the sex, the physiological state and the psychological state distribution of the trainee and outputting the user class probability for realizing effective training by adopting different trainings; and training the model until the model is converged.
Description
Technical Field
The invention relates to the technical field of data processing, in particular to a physical training data feature extraction method based on BERT.
Background
Along with the continuous development of economic level and national quality, the crowd who carries out physical training is huge day by day, and the data of all kinds of physical training is also abundant constantly. When a large amount of physical training data are classified and further analyzed, only the physical training data are subjected to feature extraction, and high-dimensional original data are mapped to low-dimensional manifold, so that the physical training data can become separable.
Journal articles 'application of data mining technology in physical training of fire-fighting troops [ J ] fire-fighting technology and product information, 2017(08): 74-77' use a fuzzy classification method to realize system classification, and provide basis for classification training. And (4) calculating by decomposition by combining the AHP method with training scores and standards of different items of the trainees.
However, in the above-mentioned paper technical solutions, the fuzzy classification method and the AHP method adopted by the method use subjective factors of experts as the basis for determining key features, so that the subjective factors have an excessive influence, it is difficult to implement implicit knowledge mining, and the scientific and effective feature extraction cannot be achieved, which will affect the data analysis effect.
Disclosure of Invention
The embodiment of the invention provides a physical training data feature extraction method based on BERT, which is characterized in that corresponding key features are extracted according to the difference of time sequence physical training data through feature engineering, the feature dimension of the training data is reduced, the data analysis difficulty of physical training is further reduced, the data analysis efficiency and effectiveness are improved, and the prediction of the training target effect is finally realized.
The embodiment of the invention provides a physical training data feature extraction method based on BERT, which takes 80% of daily training data of people who perform physical training as a training set of a model, 10% as a verification set and 10% as a test set, and performs K-fold cross verification, and comprises the following steps:
a BERT part, which is embodied by using word2vec to pre-train the embedding of the physical training item, and clustering the training; improving MLM, changing the form of the whole training information including attributes of the original direct mask into only one part of information for mask training;
the fusion part is used for fusing the distribution of the age, the sex, the physiological state and the psychological state of the population who carries out physical training and meets the preset conditions into a network;
the output part is used for classifying and predicting the age, the sex, the physiological state and the psychological state distribution of the trained personnel and outputting the user category probability which can realize effective training by adopting different training;
and training the model until the model is converged, and finishing the training, wherein the BERT part of the model is the training feature extractor.
In some embodiments of the present invention, the merging the distribution of the age, sex, physiological status and psychological status of the population who performs physical training and meets the preset conditions into the network includes:
dividing the population who meets the preset conditions and carries out physical training into a plurality of parts;
if the trainee performs multiple training and comes from the first part of the population, the rest of the population is used as seed population, the age, sex, physiological state and psychological state distribution of each training are calculated, and 12-layer transformers are adopted for fusion.
In some embodiments of the present invention, the population for developing physical training who meets the preset condition is divided into a plurality of parts, specifically:
the population who meets the preset conditions for developing physical training is divided into 64 parts.
In some embodiments of the present invention, the fusion performed by using 12 layers of transformers specifically includes:
and (3) coding by using a BERT model containing 12 layers of transformers as a coder, wherein each layer of Transformer takes the hidden variable of the previous layer as input and outputs the hidden variable with higher information interaction degree.
In some embodiments of the present invention, the predetermined condition is that a predetermined training time period is satisfied and training is performed at a predetermined frequency.
In some embodiments of the present invention, the method for extracting characteristics of physical training data based on BERT further includes:
when the improved MLM prediction training is adopted, the latest specific project name ID and the attributes thereof can be used for prediction, so that the embedding of different attributes are aligned in a semantic space.
The physical training data feature extraction method based on BERT provided by the embodiment of the invention has the following advantages: after physical training data input is received, physical training feature extraction is carried out through an improved BERT method, each training is assumed to be a word, a training sequence of a trained user is regarded as a sentence, namely, an original problem is converted into a text classification problem in the NLP (natural language processing) field, and with the help of a mature embedding method in the field, a training target effect prediction problem is converted into a classification problem according to the type of a stable training person, so that physical training data feature extraction and training effect prediction are achieved. A physical training feature extraction method based on BERT is established, and effective extraction of physical training data features is realized by means of mature and effective language feature extraction capability of the BERT.
Drawings
FIG. 1 is a flowchart of a method for extracting the characteristics of the BERT-based physical training data according to an embodiment of the present invention;
fig. 2 is a block diagram of a method for extracting physical training data features based on BERT according to an embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, those skilled in the art will further describe the present invention with reference to the accompanying drawings and the detailed description.
The phrases "in one embodiment," "in another embodiment," "in yet another embodiment," "in an embodiment," "in some embodiments," or "in other embodiments" may be used in this specification to refer to one or more of the same or different embodiments in accordance with the invention.
Specific embodiments of the present invention are described hereinafter with reference to the accompanying drawings; however, it is to be understood that the disclosed embodiments are merely exemplary of the invention, which can be embodied in various forms. Well-known and/or repeated functions and structures have not been described in detail so as to not unnecessarily obscure the present invention, but may be learned by the operation of the user's history and actual intent. Therefore, specific structural and functional details disclosed herein are not to be interpreted as limiting, but merely as a basis for the claims and as a representative basis for teaching one skilled in the art to variously employ the present invention in virtually any appropriately detailed structure.
The embodiment of the invention provides a physical training data feature extraction method based on BERT, as shown in fig. 1 and fig. 2, the method takes 80% of daily training data of a population performing physical training as a training set of a model, 10% as a verification set and 10% as a test set, and performs K-fold cross-validation, and the method comprises the following steps:
a BERT (bidirectional Encoder reproduction From Transformer) part, which is to pre-train the embedding of the physical training items by using word2vec, and cluster the training; improving MLM, changing the form of the whole training information including attributes of the original direct mask into only one part of information for mask training;
the fusion part is used for fusing the distribution of the age, the sex, the physiological state and the psychological state of the population who carries out physical training and meets the preset conditions into a network;
the output part is used for classifying and predicting the age, the sex, the physiological state and the psychological state distribution of the trained personnel and outputting the user category probability which can realize effective training by adopting different training;
and training the model until the model is converged, and finishing the training, wherein the BERT part of the model is the training feature extractor.
In some embodiments of the present invention, the integrating the distribution of the age, sex, physiological status and psychological status of the population performing physical training meeting the preset conditions into the network includes:
dividing the population for developing physical training meeting the preset conditions into a plurality of parts;
if the trainee performs multiple training and comes from the first part of the population, the rest of the population is used as seed population, the age, sex, physiological state and psychological state distribution of each training are calculated, and 12-layer transformers are adopted for fusion.
In some embodiments of the present invention, the population for developing physical training who meets the preset condition is divided into a plurality of parts, specifically:
the population who meets the preset conditions for developing physical training is divided into 64 parts.
In some embodiments of the present invention, the fusion performed by using 12 layers of transformers specifically includes:
and (3) coding by using a BERT model containing 12 layers of transformers as a coder, wherein each layer of Transformer takes the hidden variable of the previous layer as input and outputs the hidden variable with higher information interaction degree.
In some embodiments of the present invention, the predetermined condition is that a predetermined training duration is satisfied and training is performed at a predetermined frequency.
In some embodiments of the present invention, the method for extracting characteristics of physical training data based on BERT further includes:
when the improved MLM prediction training is adopted, the latest specific project name ID and the attribute thereof can be used for prediction, so that embedding of different attributes are aligned in a semantic space.
According to the technical scheme, the method for extracting the physical training data features based on the BERT provided by the embodiment of the invention has the following advantages: after physical training data input is received, physical training characteristic extraction is carried out through an improved BERT method, each training is assumed to be a word, a training sequence of a trained user is regarded as a sentence, namely, an original problem is converted into a text classification problem in the NLP (natural language processing) field, and with the help of a mature embedding method in the field, a training target effect prediction problem is converted into a classification problem according to a stable training person type, so that physical training data characteristic extraction and training effect prediction are achieved. A physical training feature extraction method based on BERT is established, and effective extraction of physical training data features is realized by means of mature and effective language feature extraction capability of BERT.
In order to facilitate detailed description of the above technical solution, a specific example is described in detail below, and 80% of daily training data of a population who often performs physical training is selected as a training set of a model, 10% is selected as a verification set, and 10% is selected as a test set, and K-fold cross-validation is performed. FIG. 1 is a model introduction diagram of the technical solution of the present invention. As shown in fig. 1, the physical ability training feature extraction method of the present invention includes the following steps:
(1) BERT part: assuming that each training is a word, the training sequence of the trained user is regarded as a sentence, and the original question can be converted into a text classification question in the field of NLP (Natural Language Processing). And (3) coding by using a BERT model containing 12 layers of transformers as a coder, wherein each layer of Transformer takes the hidden variable of the previous layer as input, and finally outputs the hidden variable with higher information interaction degree. The BERT is trained by adopting a pre-training method of the following two stages: firstly, using word2vec to pre-train embedding of physical training items, and clustering similar training; and secondly, improving the MLM, and changing the form of the whole training information including the attributes of the original direct mask into the form of only one part of information (training items or attributes) of the mask training. When the improved mlm (masked language model) is used for prediction training, not only the latest specific item name ID but also the attribute of the latest specific item name ID can be used for prediction, so that embedding of different attributes can be aligned in a semantic space.
(2) And (3) fusion part: the method is characterized in that the age, the sex, the physiological state and the psychological state distribution of the population capable of stably developing physical training for a long time are integrated into a network, the population is divided into 64 parts, and if the trainee performs n training and comes from the first population, the rest population can be used as seed population, the age, the sex, the physiological state and the psychological state distribution of each training are calculated, and 12 layers of transformers are used for integration.
(3) An output section: the ages, the sexes, the physiological states and the psychological state distribution of the trained personnel are classified and predicted, and the user category probability of effective training can be output by adopting different training.
And (5) training the model until the model is converged, and finishing the training, wherein the BERT part of the model is the training feature extractor.
While the present invention has been described in detail with reference to the preferred embodiments, it should be understood that the above description should not be taken as limiting the invention. Various modifications and alterations to this invention will become apparent to those skilled in the art upon reading the foregoing description. Accordingly, the scope of the invention should be determined from the following claims.
Claims (6)
1. A method for extracting physical training data features based on BERT is characterized in that 80% of daily training data of a population subjected to physical training is used as a training set of a model, 10% is used as a verification set, and 10% is used as a test set, and K-fold cross validation is performed, wherein the method comprises the following steps:
a BERT part, which is embodied by using word2vec to pre-train the embedding of the physical training item, and clustering the training; improving MLM, changing the form of the whole training information including attributes of the original direct mask into only one part of information for mask training;
the fusion part is used for fusing the distribution of the age, the sex, the physiological state and the psychological state of the population who carries out physical training and meets the preset conditions into a network;
the output part is used for classifying and predicting the age, the sex, the physiological state and the psychological state distribution of the trained personnel and outputting the user category probability which can realize effective training by adopting different training;
and training the model until the model is converged, and finishing the training, wherein the BERT part of the model is the training feature extractor.
2. The method for extracting characteristics of physical training data based on BERT according to claim 1, wherein the step of integrating the distribution of ages, sexes, physiological states and psychological states of the population who carries out physical training and meets preset conditions into a network comprises the following steps:
dividing the population for developing physical training meeting the preset conditions into a plurality of parts;
if the trainee performs multiple training and comes from the first part of the population, the rest of the population is used as seed population, the age, sex, physiological state and psychological state distribution of each training are calculated, and 12-layer transformers are adopted for fusion.
3. The BERT-based physical fitness training data feature extraction method according to claim 2, wherein the population for developing physical fitness training that meets preset conditions is divided into a plurality of parts, specifically:
the population who performs physical training and meets the preset conditions is divided into 64 parts.
4. The method for extracting physical training data features based on BERT according to claim 3, wherein the 12-layer Transformer is adopted for fusion, specifically:
and (3) coding by using a BERT model containing 12 layers of transformers as a coder, wherein each layer of Transformer takes the hidden variable of the previous layer as input and outputs the hidden variable with higher information interaction degree.
5. The method of feature extraction for BERT-based physical training data according to claim 4,
and the preset condition is that the preset training time is met and the training is carried out at the preset frequency.
6. The method of claim 5, further comprising:
when the improved MLM prediction training is adopted, the latest specific project name ID and the attributes thereof can be used for prediction, so that the embedding of different attributes are aligned in a semantic space.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210628842.6A CN115098623A (en) | 2022-06-06 | 2022-06-06 | Physical training data feature extraction method based on BERT |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210628842.6A CN115098623A (en) | 2022-06-06 | 2022-06-06 | Physical training data feature extraction method based on BERT |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115098623A true CN115098623A (en) | 2022-09-23 |
Family
ID=83289393
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210628842.6A Pending CN115098623A (en) | 2022-06-06 | 2022-06-06 | Physical training data feature extraction method based on BERT |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115098623A (en) |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110647612A (en) * | 2019-09-18 | 2020-01-03 | 合肥工业大学 | Visual conversation generation method based on double-visual attention network |
US20210089724A1 (en) * | 2019-09-25 | 2021-03-25 | Google Llc | Contrastive Pre-Training for Language Tasks |
CN112732874A (en) * | 2021-01-15 | 2021-04-30 | 北京明略昭辉科技有限公司 | Training method and device for user identification call-through model |
CN113434699A (en) * | 2021-06-30 | 2021-09-24 | 平安科技(深圳)有限公司 | Pre-training method of BERT model, computer device and storage medium |
CN113723106A (en) * | 2021-07-29 | 2021-11-30 | 北京工业大学 | Zero sample text classification method based on label extension |
CN114218921A (en) * | 2021-11-30 | 2022-03-22 | 中国医学科学院医学信息研究所 | Problem semantic matching method for optimizing BERT |
US20220139384A1 (en) * | 2020-11-03 | 2022-05-05 | Salesforce.Com, Inc. | System and methods for training task-oriented dialogue (tod) language models |
CN114580418A (en) * | 2022-03-04 | 2022-06-03 | 河北师范大学 | Knowledge map system for police physical training |
-
2022
- 2022-06-06 CN CN202210628842.6A patent/CN115098623A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110647612A (en) * | 2019-09-18 | 2020-01-03 | 合肥工业大学 | Visual conversation generation method based on double-visual attention network |
US20210089724A1 (en) * | 2019-09-25 | 2021-03-25 | Google Llc | Contrastive Pre-Training for Language Tasks |
US20220139384A1 (en) * | 2020-11-03 | 2022-05-05 | Salesforce.Com, Inc. | System and methods for training task-oriented dialogue (tod) language models |
CN112732874A (en) * | 2021-01-15 | 2021-04-30 | 北京明略昭辉科技有限公司 | Training method and device for user identification call-through model |
CN113434699A (en) * | 2021-06-30 | 2021-09-24 | 平安科技(深圳)有限公司 | Pre-training method of BERT model, computer device and storage medium |
CN113723106A (en) * | 2021-07-29 | 2021-11-30 | 北京工业大学 | Zero sample text classification method based on label extension |
CN114218921A (en) * | 2021-11-30 | 2022-03-22 | 中国医学科学院医学信息研究所 | Problem semantic matching method for optimizing BERT |
CN114580418A (en) * | 2022-03-04 | 2022-06-03 | 河北师范大学 | Knowledge map system for police physical training |
Non-Patent Citations (2)
Title |
---|
张媛媛;: "数据挖掘技术在消防部队体能训练中的应用", 消防技术与产品信息, no. 08, 15 August 2017 (2017-08-15) * |
徐菲菲;冯东升;: "文本词向量与预训练语言模型研究", 上海电力大学学报, no. 04, 15 August 2020 (2020-08-15) * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Ahmed et al. | Tweeting India’s Nirbhaya protest: A study of emotional dynamics in an online social movement | |
CN109885670A (en) | A kind of interaction attention coding sentiment analysis method towards topic text | |
CN112749341B (en) | Important public opinion recommendation method, readable storage medium and data processing device | |
CN105701210A (en) | Microblog theme emotion analysis method based on mixed characteristic calculation | |
CN107688870B (en) | Text stream input-based hierarchical factor visualization analysis method and device for deep neural network | |
CN109558492A (en) | A kind of listed company's knowledge mapping construction method and device suitable for event attribution | |
US20110264443A1 (en) | Information processing device, information processing method, and program | |
Liu et al. | Learning to predict population-level label distributions | |
Jorge-Botana et al. | Bridging the theoretical gap between semantic representation models without the pressure of a ranking: some lessons learnt from LSA | |
Mirza | Extracting temporal and causal relations between events | |
Jimenez et al. | Word2set: WordNet-based word representation rivaling neural word embedding for lexical similarity and sentiment analysis | |
CN116383430A (en) | Knowledge graph construction method, device, equipment and storage medium | |
CN117909507A (en) | AI-based data classification system | |
Lu et al. | An emotion analysis method using multi-channel convolution neural network in social networks | |
CN112836043A (en) | Long text clustering method and device based on pre-training language model | |
CN115600602B (en) | Method, system and terminal device for extracting key elements of long text | |
Feng et al. | Incident threading for news passages | |
Lipizzi et al. | A computational model implementing subjectivity with the'Room Theory'. The case of detecting Emotion from Text | |
Ireland et al. | Sentimental analysis for AIML-based e-health conversational agents | |
CN115098623A (en) | Physical training data feature extraction method based on BERT | |
Bai et al. | Gated character-aware convolutional neural network for effective automated essay scoring | |
CN111625623B (en) | Text theme extraction method, text theme extraction device, computer equipment, medium and program product | |
Westbury | Prenominal adjective order is such a fat big deal because adjectives are ordered by likely need | |
CN111815426B (en) | Data processing method and terminal related to financial investment and research | |
Ji et al. | Improving non-negative positive-unlabeled learning for news headline classification |
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 |