CN112434746A - Pre-labeling method based on hierarchical transfer learning and related equipment thereof - Google Patents
Pre-labeling method based on hierarchical transfer learning and related equipment thereof Download PDFInfo
- Publication number
- CN112434746A CN112434746A CN202011364408.9A CN202011364408A CN112434746A CN 112434746 A CN112434746 A CN 112434746A CN 202011364408 A CN202011364408 A CN 202011364408A CN 112434746 A CN112434746 A CN 112434746A
- Authority
- CN
- China
- Prior art keywords
- scene
- scenes
- data
- model
- labeling
- 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
- 238000002372 labelling Methods 0.000 title claims abstract description 136
- 238000013526 transfer learning Methods 0.000 title claims abstract description 26
- 238000013508 migration Methods 0.000 claims abstract description 119
- 230000005012 migration Effects 0.000 claims abstract description 119
- 238000000034 method Methods 0.000 claims abstract description 27
- 238000012546 transfer Methods 0.000 claims abstract description 10
- 239000013598 vector Substances 0.000 claims description 55
- 238000012549 training Methods 0.000 claims description 39
- 238000012937 correction Methods 0.000 claims description 11
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 230000008569 process Effects 0.000 description 9
- 230000000694 effects Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 238000004891 communication Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 230000006835 compression Effects 0.000 description 2
- 238000007906 compression Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000003058 natural language processing Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Medical Informatics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Image Analysis (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the application belongs to the field of artificial intelligence and relates to a pre-labeling method based on hierarchical transfer learning and related equipment thereof, wherein the method comprises the steps of clustering a plurality of different pre-received scenes based on a preset clustering algorithm to obtain a clustering result; determining a first class scene and a second class scene according to the clustering result, wherein the first class scene comprises the first scene, and the data volume of the marking data in the first scene is larger than that of any scene in the second class scene; performing migration learning on the first type of scenes based on a preset identification model to obtain pre-labeled data and a migration model of each scene in the first type of scenes; and performing transfer learning on the second type of scenes based on a transfer model to obtain the pre-labeled data of each scene in the second type of scenes. The pre-labeled data of each scene can be stored in a block chain. The method and the device can rapidly obtain better pre-marked data in different scenes.
Description
Technical Field
The application relates to the technical field of artificial intelligence, in particular to a pre-labeling method based on hierarchical transfer learning and related equipment thereof.
Background
The development of random science and technology and the intelligent identification technology are widely applied, many data can be directly pre-labeled through the model without being labeled by personnel, and the personnel only need to correct the result of the pre-labeling, so that the labor cost and the labeling time are effectively reduced.
However, at present, before the model is actually put into use and pre-labeled, a large amount of labeled data is needed to train the model, and the effect of model training directly affects the result of pre-labeling. However, when training models of different scenes, personnel are still required to label a large amount of data in each scene, which causes too high labor cost and too slow labeling speed, and a large amount of better labeled data in different scenes cannot be obtained quickly in a short time.
The current mode is to directly transfer the trained model in the old scene to the new scene for pre-labeling, and then the personnel correct the pre-labeling result, but the different scenes may have larger difference, so that the pre-labeling result of the model is poorer, and even the model cannot be used at all.
Disclosure of Invention
The embodiment of the application aims to provide a pre-labeling method based on hierarchical transfer learning and related equipment thereof, so that better pre-labeling data in different scenes can be quickly obtained.
In order to solve the above technical problem, an embodiment of the present application provides a pre-labeling method based on hierarchical migration learning, which adopts the following technical solutions:
a pre-labeling method based on hierarchical transfer learning comprises the following steps:
clustering a plurality of different pre-received scenes based on a preset clustering algorithm to obtain a clustering result;
determining a first class of scenes and a second class of scenes according to the clustering result, wherein the first class of scenes comprises a first scene, and the data volume of the marking data in the first scene is larger than that of any scene in the second class of scenes;
performing migration learning on the first type of scenes based on a preset identification model to obtain pre-labeled data and a migration model of each scene in the first type of scenes;
and performing transfer learning on the second type of scenes based on the transfer model to obtain the pre-labeled data of each scene in the second type of scenes.
Further, the clustering of a plurality of different pre-received scenes based on a preset clustering algorithm to obtain a clustering result includes:
receiving a plurality of different scene texts, wherein the scene texts correspond to the scenes one to one;
respectively inputting the different scene texts into pre-trained vector models to respectively obtain scene vectors;
and clustering the scene vectors based on an unsupervised K-means algorithm to obtain a clustering result.
Further, the step of inputting the plurality of different scene texts into pre-trained vector models respectively to obtain scene vectors respectively includes:
sequentially inputting each scene text into a classification memory network and a bag-of-words network of a pre-trained vector model respectively to obtain a first text characteristic vector and a second text characteristic vector respectively;
and assembling the first text feature vector and the second text feature vector to obtain the scene vector.
Further, the first class of scenes at least includes a first scene and a second scene, and the step of performing migration learning on the first class of scenes based on a preset identification model to obtain the pre-labeled data and the migration model of each scene in the first class of scenes includes:
training a preset recognition model based on the labeling data contained in the first scene to obtain a first recognition model, wherein the first scene comprises corresponding labeling data;
adjusting the first recognition model based on labeling data contained in the second scene to obtain a second recognition model, wherein the second scene comprises corresponding labeling data and non-labeling data;
inputting unmarked data contained in the second scene into the second recognition model to obtain pre-marked data corresponding to the second scene, and correcting the pre-marked data to obtain a first marking result;
training the second recognition model through the first labeling result to obtain a third recognition model corresponding to a second scene;
judging whether the number of scenes which are learned by the recognition model is equal to the number of scenes contained in the first class of scenes or not;
if the first identification model and the second identification model are equal, the third identification model is used as the migration model, and pre-labeled data of each scene in the migration model and the first class of scenes are obtained;
if not, learning the next scene through the third identification model until the number of the learned scenes of the identification model is equal to the number of the scenes contained in the first class of scenes, and obtaining the pre-labeled data of each scene in the migration model and the first class of scenes.
Further, the step of adjusting the first recognition model based on the annotation data included in the second scene to obtain a second recognition model includes:
dividing the first recognition model into a fixed layer and a layer to be adjusted according to a preset dividing position;
training the layer to be adjusted based on the labeling data contained in the second scene to obtain an adjusting layer;
and combining the fixed layer and the adjusting layer to obtain the second recognition model.
Further, the step of correcting the pre-labeled data to obtain a labeling result includes:
displaying the pre-marked data corresponding to the second scene in a preset front-end page;
sending a correction notice to related personnel to inform the related personnel to correct the pre-marked data in the front-end page;
and after receiving a correction completion notification returned by the related personnel, acquiring the pre-marked data corrected by the related personnel as a marking result.
Further, the second class of scenes at least includes a third scene, and the step of performing migration learning on the second class of scenes based on the migration model to obtain the pre-labeled data of each scene in the second class of scenes includes:
adjusting the migration model based on the labeling data contained in the third scene to obtain a first migration model, wherein the third scene comprises corresponding labeling data and non-labeling data;
pre-labeling unmarked data contained in the third scene through the first migration model to obtain pre-labeled data corresponding to the third scene, and correcting the pre-labeled data to obtain a second labeling result;
training the first migration model through the second labeling result to obtain a second migration model corresponding to a third scene;
judging whether the number of scenes which are learned by the migration model is equal to the number of scenes contained in the second class of scenes or not;
if the scene data are equal to the preset data, determining to finish the pre-labeling of each scene, and obtaining the pre-labeling data of each scene in the second class of scenes;
and if not, learning the next scene through the second migration model until the number of the scenes which are learned by the migration model is equal to the number of the scenes contained in the second class of scenes, and obtaining the pre-labeled data of each scene in the second class of scenes.
In order to solve the above technical problem, an embodiment of the present application further provides a pre-labeling device based on hierarchical migration learning, which adopts the following technical scheme:
a pre-labeling apparatus based on hierarchical migration learning, comprising:
the scene clustering module is used for clustering a plurality of different pre-received scenes based on a preset clustering algorithm to obtain a clustering result;
the category determination module is used for determining a first category of scenes and a second category of scenes according to the clustering result, wherein the first category of scenes comprises a first scene, and the data volume of the marking data in the first scene is larger than that of any scene in the second category of scenes;
the first labeling module is used for performing transfer learning on the first type of scenes based on a preset identification model to obtain pre-labeling data and a transfer model of each scene in the first type of scenes;
and the second labeling module is used for performing transfer learning on the second type of scenes based on the transfer model to obtain the pre-labeling data of each scene in the second type of scenes.
In order to solve the above technical problem, an embodiment of the present application further provides a computer device, which adopts the following technical solutions:
a computer device includes a memory and a processor, the memory stores computer readable instructions, and the processor implements the steps of the pre-labeling method based on hierarchical migration learning when executing the computer readable instructions.
In order to solve the above technical problem, an embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solutions:
a computer readable storage medium having computer readable instructions stored thereon, which when executed by a processor, implement the steps of the aforementioned pre-labeling method based on hierarchical migration learning.
Compared with the prior art, the embodiment of the application mainly has the following beneficial effects:
according to the method, under the same task type, a plurality of different scenes are clustered to obtain clustering results, the scenes are divided according to the clustering results, the preset recognition model is firstly subjected to transfer learning in the same scene, and due to the fact that the scenes are similar, the recognition model can capture fine-grained knowledge in the similar scene, and therefore better pre-marked data which can be output by the recognition model in the different scenes in the same scene are recognized. After the migration learning of similar scenes of the same class is completed, the obtained migration model is input into scenes of another class, and the coarse-grained knowledge in dissimilar scenes under the same task type can still be captured, so that the problem that high-quality labeled data in the early stage of model training is insufficient can be solved, the cost for designing and developing models under different scenes of the same task type is reduced, the effect of pre-labeling is improved, and the effective labeled data can be efficiently output by a computer.
Drawings
In order to more clearly illustrate the solution of the present application, the drawings needed for describing the embodiments of the present application will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present application, and that other drawings can be obtained by those skilled in the art without inventive effort.
FIG. 1 is an exemplary system architecture diagram in which the present application may be applied;
FIG. 2 is a flow diagram of one embodiment of a pre-labeling method based on hierarchical migration learning according to the present application;
FIG. 3 is a schematic structural diagram of an embodiment of a pre-labeling apparatus based on hierarchical migration learning according to the present application;
FIG. 4 is a schematic block diagram of one embodiment of a computer device according to the present application.
Reference numerals: 200. a computer device; 201. a memory; 202. a processor; 203. a network interface; 300. a pre-labeling device based on hierarchical transfer learning; 301. a scene clustering module; 302. a category determination module; 303. a first labeling module; 304. and a second labeling module.
Detailed Description
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "including" and "having," and any variations thereof, in the description and claims of this application and the description of the above figures are intended to cover non-exclusive inclusions. The terms "first," "second," and the like in the description and claims of this application or in the above-described drawings are used for distinguishing between different objects and not for describing a particular order.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.
In order to make the technical solutions better understood by those skilled in the art, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.
As shown in fig. 1, the system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 serves as a medium for providing communication links between the terminal devices 101, 102, 103 and the server 105. Network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, to name a few.
The user may use the terminal devices 101, 102, 103 to interact with the server 105 via the network 104 to receive or send messages or the like. The terminal devices 101, 102, 103 may have various communication client applications installed thereon, such as a web browser application, a shopping application, a search application, an instant messaging tool, a mailbox client, social platform software, and the like.
The terminal devices 101, 102, 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III, mpeg compression standard Audio Layer 3), MP4 players (Moving Picture Experts Group Audio Layer IV, mpeg compression standard Audio Layer 4), laptop portable computers, desktop computers, and the like.
The server 105 may be a server providing various services, such as a background server providing support for pages displayed on the terminal devices 101, 102, 103.
It should be noted that the pre-labeling method based on hierarchical migration learning provided in the embodiments of the present application is generally executed by a server/terminal device, and accordingly, the pre-labeling apparatus based on hierarchical migration learning is generally disposed in the server/terminal device.
It should be understood that the number of terminal devices, networks, and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.
With continued reference to FIG. 2, a flow diagram of one embodiment of a pre-annotation methodology based on hierarchical migration learning in accordance with the present application is illustrated. The pre-labeling method based on hierarchical transfer learning comprises the following steps:
s1: clustering a plurality of different pre-received scenes based on a preset clustering algorithm to obtain a clustering result, wherein the different scenes belong to the same task.
In this embodiment, a plurality of different scenes are clustered, so that the scenes are classified, similar scenes are determined, and subsequent transfer learning on the similar scenes is facilitated, so that fine-grained knowledge under the similar scenes can be captured, wherein the same task refers to the fact that the types of tasks of different scenes are the same by the model, for example, if the current task is intention identification, the tasks of different scenes are intention identification by the model.
In this embodiment, an electronic device (for example, the server/terminal device shown in fig. 1) on which the hierarchical migration learning-based pre-labeling method operates may receive a plurality of different scenes in advance through a wired connection manner or a wireless connection manner. It should be noted that the wireless connection means may include, but is not limited to, a 3G/4G connection, a WiFi connection, a bluetooth connection, a WiMAX connection, a Zigbee connection, a uwb (ultra wideband) connection, and other wireless connection means now known or developed in the future.
Specifically, the clustering of a plurality of different pre-received scenes based on a preset clustering algorithm includes:
receiving a plurality of different scene texts, wherein the scene texts correspond to the scenes one to one;
respectively inputting the different scene texts into pre-trained vector models to respectively obtain scene vectors;
and clustering the scene vectors based on an unsupervised K-means algorithm to obtain a clustering result.
In this embodiment, the scene text specifically refers to text data in the scene, for example, in a scene in which the external robot has a conversation with the client, the content of the conversation between the external robot and the client is the scene text of the scene. In the actual operation process, the scene text may be adaptively selected and adjusted according to the actual needs and the model performance, for example, a description text of the scene is selected as the scene text. The vector model is Doc2vec, and Doc2vec is an unsupervised algorithm, and can obtain the vector expression of sentences/paragraphs/texts. The scene text is converted into the scene vector through the vector model, so that the scene vector is conveniently clustered through an algorithm, and further the scene clustering is realized.
The step of respectively inputting the different scene texts into pre-trained vector models and respectively obtaining scene vectors comprises:
sequentially inputting each scene text into a classification memory network and a bag-of-words network in a vector model trained in advance respectively to obtain a first text characteristic vector and a second text characteristic vector respectively;
and assembling the first text feature vector and the second text feature vector to obtain the scene vector.
In this embodiment, the vector representation output by Doc2vec is more accurate in terms of the subject matter relative to other vector models. The Doc2vec includes two networks, which are respectively: a classified Memory network (DM) and a Bag of Words network (DBOW). And obtaining a scene vector by assembling the first text feature vector and the second text feature vector. For example, the first text feature vector output by the classification memory network is (p1, p2, p3, p4), the second text feature vector output by the bag-of-words network is (q1, q2, q3, q4), and the scene vector is (p1, p2, p3, p4, q1, q2, q3, q 4).
S2: and determining a first class of scenes and a second class of scenes according to the clustering result, wherein the first class of scenes comprises a first scene, and the data volume of the marking data in the first scene is larger than that of any scene in the second class of scenes.
In this embodiment, a plurality of different scenes are divided into different categories by clustering. And determining a first type of scene and a second type of scene according to the types of the scenes. In the present application, the first-class scene and the second-class scene do not refer to a certain class of scenes, but refer to any class, and in the actual application process, the class of the selected scene may be determined according to actual needs, for example, a class with more corresponding scenes is selected as the first-class scene, or a class with a pre-identified scene with a high importance level is selected as the first-class scene. And determining that the importance level is high if the importance level exceeds a preset level threshold. And after the first class of scenes and the second class of scenes are determined according to the clustering result of the scenes, the subsequent transfer learning of the same class of scenes is facilitated. However, it should be noted that the data amount of the annotation data of at least one scene (referred to as a first scene in this application) always exists in the first class of scenes, which is larger than the data amount of the annotation data of any scene in the second class of scenes, so as to ensure that the recognition model initially trained through the first scene does not change greatly in the subsequent training process, and ensure that enough annotation data exists in the first scene for training the recognition model. In other words, because the data volume of the label data in the first scene is large, the label data is suitable for being used as the scene for training the recognition model, and when the recognition model trained through the first scene is transferred to other scenes initially, the recognition model is fine-tuned only by a small amount of label data in the other scenes, so that a good recognition model can be obtained.
S3: and performing transfer learning on the first type of scenes based on a preset identification model to obtain the pre-labeled data and the transfer model of each scene in the first type of scenes.
In this embodiment, in the same type of scene, the scenes all belong to similar scenes, and the first type of scene is subjected to transfer learning through a preset identification model, so that the model performs transfer learning on the similar scenes first, and it can be ensured that the identification model can capture fine-grained knowledge in the similar scenes and output better pre-labeled data.
Specifically, the first-class scenes at least include a first scene and a second scene, and the step of performing migration learning on the first-class scenes based on a preset identification model to obtain the pre-labeled data and the migration model of each scene in the first-class scenes includes:
training a preset recognition model based on the labeling data contained in the first scene to obtain a first recognition model, wherein the first scene comprises corresponding labeling data;
adjusting the first recognition model based on labeling data contained in the second scene to obtain a second recognition model, wherein the second scene comprises corresponding labeling data and non-labeling data;
inputting unmarked data contained in the second scene into the second recognition model to obtain pre-marked data corresponding to the second scene, and correcting the pre-marked data to obtain a first marking result;
training the second recognition model through the first labeling result to obtain a third recognition model corresponding to a second scene;
judging whether the number of scenes which are learned by the recognition model is equal to the number of scenes contained in the first class of scenes or not;
if the first identification model and the second identification model are equal, the third identification model is used as the migration model, and pre-labeled data of each scene in the migration model and the first class of scenes are obtained;
if not, learning the next scene through the third identification model until the number of the learned scenes of the identification model is equal to the number of the scenes contained in the first class of scenes, and obtaining the pre-labeled data of each scene in the migration model and the first class of scenes.
In this embodiment, the annotation data in the first scene, which is high-quality annotated data picked up by the relevant person, is used as an input to the recognition model. The recognition model of the present application is an NLP (Natural Language Processing) model. Training a recognition model through high-quality labeling data in a first scene, in an intention recognition task, a Cross Entropy Loss function can be adopted as a Loss function of the recognition model, the Cross Entropy Loss function (Cross Entropy Loss) is used for representing a difference value between a real sample label and a prediction probability, and when the accuracy of the recognition model reaches a preset accuracy threshold, the training is determined to be completed to obtain a first recognition model. In addition, since the annotation data in the first scene is used for training the recognition model and the annotation data in the second scene is used for adjusting the recognition model, the data amount of the annotation data in the first scene needs to be larger than that of the annotation data in the second scene, so as to ensure that the model does not change excessively in the second scene. And determining whether the recognition model has performed transfer learning on all scenes in the first class of scenes and whether the pre-labeling of each scene is completed by judging whether the number of the scenes which have been learned by the recognition model is equal to the number of the scenes contained in the first class of scenes. If the unlabeled data exists in the first scene, the unlabeled data in the first scene can be directly labeled after the recognition model is trained in the first scene. And the output pre-labeled data has higher quality due to the similar scenes and the adjustment of the model, thereby effectively and quickly obtaining a large amount of high-quality pre-labeled data of different scenes.
Further, the step of adjusting the first recognition model based on the annotation data included in the second scene to obtain a second recognition model includes:
dividing the first recognition model into a fixed layer and a layer to be adjusted according to a preset dividing position;
training the layer to be adjusted based on the labeling data contained in the second scene to obtain an adjusting layer;
and combining the fixed layer and the adjusting layer to obtain the second recognition model.
In this embodiment, the fixed layer is a front n layers of the first recognition model, the layer to be adjusted is a rear m layer of the first recognition model, and n and m are positive integers. The fixed layer and the layer to be adjusted constitute a first recognition model. And training the layer to be adjusted by using the labeled data in the second scene, namely representing the parameters of the first layers of the fixed first model, and only training the parameters of the second layers of the first recognition model by using the labeled data in the second scene. The second recognition model of the features under the second scene is obtained through fine tuning. By the method, the problem that the labeled data in the second scene are insufficient is solved, a large amount of labeled data in the second scene are not needed, the second recognition model with better performance can be trained, high-quality labeling can be performed on the unlabeled data in the second scene through the second recognition model, and the efficiency of data labeling is effectively improved.
It should be noted that: the training process of training the second recognition model according to the first labeling result to obtain a third recognition model corresponding to a second scene may be to train and iterate all parameters of the second recognition model, or to adjust the second recognition model, that is, only train a layer to be adjusted of the second recognition model.
In addition, the step of correcting the pre-labeled data to obtain a labeling result includes:
displaying the pre-marked data corresponding to the second scene in a preset front-end page;
sending a correction notice to related personnel to inform the related personnel to correct the pre-marked data in the front-end page;
and after receiving a correction completion notification returned by the related personnel, acquiring the pre-marked data corrected by the related personnel as a marking result.
In this embodiment, the pre-labeled data is displayed on the front-end page, and the relevant personnel is notified to correct the pre-labeled data, and the corrected data is the high-quality labeled data in the second scene. Due to the fact that the pre-labeling quality is good, the speed of correcting the pre-labeling data by related personnel is effectively improved, and high-quality labeling data can be quickly obtained.
S4: and performing transfer learning on the second type of scenes based on the transfer model to obtain the pre-labeled data of each scene in the second type of scenes.
In this embodiment, after the recognition model completes the migration learning in the first type of scene, a migration model is obtained. Migration learning of second class of scenes based on migration model
Specifically, the second class of scenes at least includes a third scene, and the step of performing migration learning on the second class of scenes based on the migration model to obtain the pre-labeled data of each scene in the second class of scenes includes:
adjusting the migration model based on the labeling data contained in the third scene to obtain a first migration model, wherein the third scene comprises corresponding labeling data and non-labeling data;
pre-labeling unmarked data contained in the third scene through the first migration model to obtain pre-labeled data corresponding to the third scene, and correcting the pre-labeled data to obtain a second labeling result;
training the first migration model through the second labeling result to obtain a second migration model corresponding to a third scene;
judging whether the number of scenes which are learned by the migration model is equal to the number of scenes contained in the second class of scenes or not;
if the scene data are equal to the preset data, determining to finish the pre-labeling of each scene, and obtaining the pre-labeling data of each scene in the second class of scenes;
and if not, learning the next scene through the second migration model until the number of the scenes which are learned by the migration model is equal to the number of the scenes contained in the second class of scenes, and obtaining the pre-labeled data of each scene in the second class of scenes.
In this embodiment, the migration model is adjusted by the annotation data in the third scenario. The step of adjusting the migration model based on the annotation data included in the third scene to obtain the first migration model is consistent with the step of adjusting the first recognition model based on the annotation data included in the second scene to obtain the second recognition model, and details are not repeated here. It should be noted that: in the first recognition model, the fixed layer is the front n layers of the first recognition model, and the layer to be adjusted is the rear m layer of the first recognition model. In the migration model, the fixed layer is a front k layer of the migration model, and the layer to be adjusted is a rear h layer of the migration model. Where n ≠ k and m ≠ h, or n ≠ k and m ≠ h. The model which completes the migration learning in the similar scene is input into another scene, and the input migration model is adjusted through the labeled data in another scene, so that the migration model can learn the characteristics of the scene, the coarse-grained knowledge in the scene can be captured, the quality of the output pre-labeled data is better than that of the output pre-labeled data of the model which is not adjusted, the labeling efficiency is effectively improved, and the model design and development cost in different scenes in the same task type is reduced.
It should be noted that: the process of training the first migration model according to the second labeling result to obtain the second migration model corresponding to the third scenario may be to train and iterate all parameters of the first migration model, or to adjust the first migration model, that is, only the layer to be adjusted of the first migration model needs to be trained.
In addition, it should be noted that the present application can be applied to a variety of different pre-labeling scenarios. Taking the intention recognition task as an example, the task type may be a gift giving conversation scene, an article selling conversation scene, an after-sale conversation scene, and the like. In the clustering process, the gift giving conversation scene and the article sale conversation scene are classified into the first kind of scene, and the after-sale conversation scene is classified into the second kind of scene. The gift giving conversation scene contains a large amount of high-quality intention labeling data, and the article selling conversation scene and the after-sale conversation scene only contain a small amount of high-quality intention labeling data.
According to the method, under the same task type, a plurality of different scenes are clustered to obtain clustering results, the scenes are divided according to the clustering results, the preset recognition model is firstly subjected to transfer learning in the same scene, and due to the fact that the scenes are similar, the recognition model can capture fine-grained knowledge in the similar scene, and therefore better pre-marked data which can be output by the recognition model in the different scenes in the same scene are recognized. After the migration learning of similar scenes of the same class is completed, the obtained migration model is input into scenes of another class, and the coarse-grained knowledge in dissimilar scenes under the same task type can still be captured, so that the problem that high-quality labeled data in the early stage of model training is insufficient can be solved, the cost for designing and developing models under different scenes of the same task type is reduced, the effect of pre-labeling is improved, and the effective labeled data can be efficiently output by a computer.
It should be emphasized that, in order to further ensure the privacy and security of the pre-labeled data of each scene, the pre-labeled data of each scene may also be stored in a node of a block chain.
The block chain referred by the application is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, a consensus mechanism, an encryption algorithm and the like. A block chain (Blockchain), which is essentially a decentralized database, is a series of data blocks associated by using a cryptographic method, and each data block contains information of a batch of network transactions, so as to verify the validity (anti-counterfeiting) of the information and generate a next block. The blockchain may include a blockchain underlying platform, a platform product service layer, an application service layer, and the like.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware associated with computer readable instructions, which can be stored in a computer readable storage medium, and when executed, can include processes of the embodiments of the methods described above. The storage medium may be a non-volatile storage medium such as a magnetic disk, an optical disk, a Read-Only Memory (ROM), or a Random Access Memory (RAM).
It should be understood that, although the steps in the flowcharts of the figures are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and may be performed in other orders unless explicitly stated herein. Moreover, at least a portion of the steps in the flow chart of the figure may include multiple sub-steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, which are not necessarily performed in sequence, but may be performed alternately or alternately with other steps or at least a portion of the sub-steps or stages of other steps.
With further reference to fig. 3, as an implementation of the method shown in fig. 2, the present application provides an embodiment of a pre-labeling apparatus based on hierarchical migration learning, where the embodiment of the apparatus corresponds to the embodiment of the method shown in fig. 2, and the apparatus may be specifically applied to various electronic devices.
As shown in fig. 3, the pre-labeling apparatus 300 based on hierarchical migration learning according to this embodiment includes: a scene clustering module 301, a category determination module 302, a first labeling module 303, and a second labeling module 304. Wherein: the scene clustering module 301 is configured to cluster a plurality of different pre-received scenes based on a preset clustering algorithm to obtain a clustering result; a category determining module 302, configured to determine a first category of scenes and a second category of scenes according to the clustering result, where the first category of scenes includes a first scene, and a data amount of label data in the first scene is greater than a data amount of label data of any scene in the second category of scenes; the first labeling module 303 is configured to perform migration learning on the first type of scenes based on a preset identification model, and obtain pre-labeling data and a migration model of each scene in the first type of scenes; a second labeling module 304, configured to perform migration learning on the second class of scenes based on the migration model, and obtain pre-labeling data of each scene in the second class of scenes.
In this embodiment, the method and the device cluster a plurality of different scenes under the same task type to obtain a clustering result, and realize scene division according to the clustering result, and the preset recognition model firstly performs migration learning in the same class of scenes. After the migration learning of similar scenes of the same class is completed, the obtained migration model is input into scenes of another class, and the coarse-grained knowledge in dissimilar scenes under the same task type can still be captured, so that the problem that high-quality labeled data in the early stage of model training is insufficient can be solved, the cost for designing and developing models under different scenes of the same task type is reduced, the effect of pre-labeling is improved, and the effective labeled data can be efficiently output by a computer.
The scene clustering module 301 includes a receiving sub-module, an input sub-module, and a clustering sub-module. The receiving submodule is used for receiving a plurality of different scene texts, wherein the scene texts correspond to the scenes one to one; the input submodule is used for respectively inputting the different scene texts into a vector model trained in advance to respectively obtain scene vectors; and the clustering submodule is used for clustering the scene vectors based on an unsupervised K-means algorithm to obtain a clustering result.
The input submodule comprises an input unit and an assembly unit. The input unit is used for sequentially inputting each scene text into a classification memory network and a bag-of-words network in a pre-trained vector model respectively to obtain a first text feature vector and a second text feature vector respectively; the assembling unit is used for assembling the first text feature vector and the second text feature vector to obtain the scene vector.
The first class of scenes at least comprises a first scene and a second scene, and the first labeling module 303 comprises a first training submodule, a first adjusting submodule, a first correcting submodule, a first obtaining submodule, a first judging submodule, a first equaling submodule and a first terminating submodule. The first training submodule is used for training a preset recognition model based on the marking data contained in the first scene to obtain a first recognition model, wherein the first scene comprises corresponding marking data; the first adjusting submodule is used for adjusting the first recognition model based on the labeling data contained in the second scene to obtain a second recognition model, wherein the second scene comprises corresponding labeling data and non-labeling data; the first correction submodule is used for inputting unmarked data contained in the second scene into the second recognition model, obtaining pre-marked data corresponding to the second scene, correcting the pre-marked data and obtaining a first marking result; the first obtaining submodule is used for training the second recognition model through the first labeling result to obtain a third recognition model corresponding to a second scene; the first judgment submodule is used for judging whether the number of scenes which are learned by the identification model is equal to the number of scenes contained in the first class of scenes or not; the first equality submodule is used for taking the third identification model as the migration model when the number of scenes which are learned by the identification model is equal to the number of scenes contained in the first class of scenes, and obtaining the pre-labeled data of each scene in the migration model and the first class of scenes; the first termination sub-module is used for learning the next scene through the third recognition model when the number of the learned scenes of the recognition model is not equal to the number of the scenes contained in the first class of scenes, until the number of the learned scenes of the recognition model is equal to the number of the scenes contained in the first class of scenes, and obtaining the pre-labeled data of each scene in the migration model and the first class of scenes.
The first adjusting submodule comprises a dividing unit, a training unit and a combining unit. The dividing unit is used for dividing the first recognition model into a fixed layer and a layer to be adjusted according to a preset dividing position; the training unit is used for training the layer to be adjusted based on the labeling data contained in the second scene to obtain an adjusting layer; the combination unit is used for combining the fixed layer and the adjusting layer to obtain the second identification model.
The first correction submodule comprises a display unit, a notification unit and a marking unit. The display unit is used for displaying the pre-marked data corresponding to the second scene in a preset front-end page; the notification unit is used for sending a correction notification to related personnel to notify the related personnel to correct the pre-marked data in the front-end page; and the marking unit is used for acquiring the pre-marked data corrected by the relevant personnel as a marking result after receiving the correction completion notification returned by the relevant personnel.
The second type of scene at least includes a third scene, and the second labeling module 304 includes a second adjusting sub-module, a second correcting sub-module, a second obtaining sub-module, a second judging sub-module, a second equaling sub-module, and a second terminating sub-module. The second adjusting submodule is used for adjusting the migration model based on the annotation data contained in the third scene to obtain a first migration model, wherein the third scene comprises corresponding annotation data and unmarked data; the second correction submodule is used for pre-labeling the unmarked data contained in the third scene through the first migration model, obtaining pre-labeled data corresponding to the third scene, correcting the pre-labeled data and obtaining a second labeling result; the second obtaining submodule is used for training the first migration model according to the second labeling result to obtain a second migration model corresponding to a third scene; the second judgment submodule is used for judging whether the number of the learned scenes of the migration model is equal to the number of the scenes contained in the second class of scenes or not; the second equal submodule is used for determining that the pre-labeling of each scene is finished when the number of the scenes which are learned by the migration model is not equal to the number of the scenes contained in the second class of scenes, and obtaining the pre-labeling data of each scene in the second class of scenes; and the second termination sub-module is used for learning the next scene through the second migration model when the number of the scenes which have been learned by the migration model is not equal to the number of the scenes contained in the second class of scenes until the number of the scenes which have been learned by the migration model is equal to the number of the scenes contained in the second class of scenes, and obtaining the pre-labeled data of each scene in the second class of scenes.
According to the method, under the same task type, a plurality of different scenes are clustered to obtain clustering results, the scenes are divided according to the clustering results, the preset recognition model is firstly subjected to transfer learning in the same scene, and due to the fact that the scenes are similar, the recognition model can capture fine-grained knowledge in the similar scene, and therefore better pre-marked data which can be output by the recognition model in the different scenes in the same scene are recognized. After the migration learning of similar scenes of the same class is completed, the obtained migration model is input into scenes of another class, and the coarse-grained knowledge in dissimilar scenes under the same task type can still be captured, so that the problem that high-quality labeled data in the early stage of model training is insufficient can be solved, the cost for designing and developing models under different scenes of the same task type is reduced, the effect of pre-labeling is improved, and the effective labeled data can be efficiently output by a computer.
In order to solve the technical problem, an embodiment of the present application further provides a computer device. Referring to fig. 4, fig. 4 is a block diagram of a basic structure of a computer device according to the present embodiment.
The computer device 200 comprises a memory 201, a processor 202, a network interface 203 communicatively connected to each other via a system bus. It is noted that only computer device 200 having components 201 and 203 is shown, but it is understood that not all of the illustrated components are required and that more or fewer components may alternatively be implemented. As will be understood by those skilled in the art, the computer device is a device capable of automatically performing numerical calculation and/or information processing according to a preset or stored instruction, and the hardware includes, but is not limited to, a microprocessor, an Application Specific Integrated Circuit (ASIC), a Programmable Gate Array (FPGA), a Digital Signal Processor (DSP), an embedded device, and the like.
The computer device can be a desktop computer, a notebook, a palm computer, a cloud server and other computing devices. The computer equipment can carry out man-machine interaction with a user through a keyboard, a mouse, a remote controller, a touch panel or voice control equipment and the like.
The memory 201 includes at least one type of readable storage medium including a flash memory, a hard disk, a multimedia card, a card type memory (e.g., SD or DX memory, etc.), a Random Access Memory (RAM), a Static Random Access Memory (SRAM), a Read Only Memory (ROM), an Electrically Erasable Programmable Read Only Memory (EEPROM), a Programmable Read Only Memory (PROM), a magnetic memory, a magnetic disk, an optical disk, etc. In some embodiments, the storage 201 may be an internal storage unit of the computer device 200, such as a hard disk or a memory of the computer device 200. In other embodiments, the memory 201 may also be an external storage device of the computer device 200, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), or the like, provided on the computer device 200. Of course, the memory 201 may also include both internal and external storage devices of the computer device 200. In this embodiment, the memory 201 is generally used for storing an operating system and various application software installed on the computer device 200, such as computer readable instructions of a pre-labeling method based on hierarchical migration learning. Further, the memory 201 may also be used to temporarily store various types of data that have been output or are to be output.
The processor 202 may be a Central Processing Unit (CPU), controller, microcontroller, microprocessor, or other data Processing chip in some embodiments. The processor 202 is generally operative to control overall operation of the computer device 200. In this embodiment, the processor 202 is configured to execute computer readable instructions stored in the memory 201 or process data, for example, execute computer readable instructions of the pre-labeling method based on hierarchical migration learning.
The network interface 203 may comprise a wireless network interface or a wired network interface, and the network interface 203 is generally used for establishing communication connection between the computer device 200 and other electronic devices.
In the embodiment, the problem that the high-quality labeled data in the early stage of model training is insufficient is solved, the effect of pre-labeling is improved, and the effective labeled data can be efficiently output by a computer.
The present application further provides another embodiment, which is to provide a computer-readable storage medium storing computer-readable instructions executable by at least one processor to cause the at least one processor to perform the steps of the pre-labeling method based on hierarchical migration learning as described above.
In the embodiment, the problem that the high-quality labeled data in the early stage of model training is insufficient is solved, the effect of pre-labeling is improved, and the effective labeled data can be efficiently output by a computer.
Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solutions of the present application may be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal device (such as a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the method according to the embodiments of the present application.
It is to be understood that the above-described embodiments are merely illustrative of some, but not restrictive, of the broad invention, and that the appended drawings illustrate preferred embodiments of the invention and do not limit the scope of the invention. This application is capable of embodiments in many different forms and is provided for the purpose of enabling a thorough understanding of the disclosure of the application. Although the present application has been described in detail with reference to the foregoing embodiments, it will be apparent to one skilled in the art that the present application may be practiced without modification or with equivalents of some of the features described in the foregoing embodiments. All equivalent structures made by using the contents of the specification and the drawings of the present application are directly or indirectly applied to other related technical fields and are within the protection scope of the present application.
Claims (10)
1. A pre-labeling method based on hierarchical transfer learning is characterized by comprising the following steps:
clustering a plurality of different pre-received scenes based on a preset clustering algorithm to obtain a clustering result, wherein the different scenes belong to the same task;
determining a first class of scenes and a second class of scenes according to the clustering result, wherein the first class of scenes comprises a first scene, and the data volume of the marking data in the first scene is larger than that of any scene in the second class of scenes;
performing migration learning on the first type of scenes based on a preset identification model to obtain pre-labeled data and a migration model of each scene in the first type of scenes;
and performing transfer learning on the second type of scenes based on the transfer model to obtain the pre-labeled data of each scene in the second type of scenes.
2. The pre-labeling method based on hierarchical migration learning according to claim 1, wherein the pre-clustering algorithm clusters a plurality of different pre-received scenes to obtain a clustering result, and the step of obtaining the clustering result comprises:
receiving a plurality of different scene texts, wherein the scene texts correspond to the scenes one to one;
respectively inputting the different scene texts into pre-trained vector models to respectively obtain scene vectors;
and clustering the scene vectors based on an unsupervised K-means algorithm to obtain a clustering result.
3. The method according to claim 2, wherein the step of inputting the different scene texts into pre-trained vector models respectively to obtain scene vectors respectively comprises:
sequentially inputting each scene text into a classification memory network and a bag-of-words network in a vector model trained in advance respectively to obtain a first text characteristic vector and a second text characteristic vector respectively;
and assembling the first text feature vector and the second text feature vector to obtain the scene vector.
4. The prelabeling method based on hierarchical migration learning according to claim 1, wherein the first class of scenes at least includes a first scene and a second scene, the step of performing migration learning on the first class of scenes based on a preset recognition model to obtain the prelabeling data and the migration model of each scene in the first class of scenes comprises:
training a preset recognition model based on the labeling data contained in the first scene to obtain a first recognition model, wherein the first scene comprises corresponding labeling data;
adjusting the first recognition model based on labeling data contained in the second scene to obtain a second recognition model, wherein the second scene comprises corresponding labeling data and non-labeling data;
inputting unmarked data contained in the second scene into the second recognition model to obtain pre-marked data corresponding to the second scene, and correcting the pre-marked data to obtain a first marking result;
training the second recognition model through the first labeling result to obtain a third recognition model corresponding to a second scene;
judging whether the number of scenes which are learned by the recognition model is equal to the number of scenes contained in the first class of scenes or not;
if the first identification model and the second identification model are equal, taking the third identification model as the migration model to obtain the pre-labeled data of each scene in the migration model and the first class of scenes;
if not, learning the next scene through the third identification model until the number of the learned scenes of the identification model is equal to the number of the scenes contained in the first class of scenes, and obtaining the pre-labeled data of each scene in the migration model and the first class of scenes.
5. The pre-labeling method based on hierarchical migration learning according to claim 4, wherein the step of adjusting the first recognition model based on the labeling data included in the second scenario to obtain a second recognition model comprises:
dividing the first recognition model into a fixed layer and a layer to be adjusted according to a preset dividing position;
training the layer to be adjusted based on the labeling data contained in the second scene to obtain an adjusting layer;
and combining the fixed layer and the adjusting layer to obtain the second recognition model.
6. The pre-labeling method based on hierarchical migration learning according to claim 4, wherein the step of correcting the pre-labeled data to obtain a labeling result comprises:
displaying the pre-marked data corresponding to the second scene in a preset front-end page;
sending a correction notice to related personnel to inform the related personnel to correct the pre-marked data in the front-end page;
and after receiving a correction completion notification returned by the related personnel, acquiring the pre-marked data corrected by the related personnel as a marking result.
7. The pre-labeling method based on hierarchical migration learning according to claim 1, wherein the second class of scenes at least includes a third scene, and the step of performing migration learning on the second class of scenes based on the migration model to obtain pre-labeling data of each scene in the second class of scenes comprises:
adjusting the migration model based on the labeling data contained in the third scene to obtain a first migration model, wherein the third scene comprises corresponding labeling data and non-labeling data;
pre-labeling unmarked data contained in the third scene through the first migration model to obtain pre-labeled data corresponding to the third scene, and correcting the pre-labeled data to obtain a second labeling result;
training the first migration model through the second labeling result to obtain a second migration model corresponding to a third scene;
judging whether the number of scenes which are learned by the migration model is equal to the number of scenes contained in the second class of scenes or not;
if so, determining to finish the pre-labeling of each scene, and obtaining the pre-labeling data of each scene in the second class of scenes;
and if not, learning the next scene through the second migration model until the number of the scenes which are learned by the migration model is equal to the number of the scenes contained in the second class of scenes, and obtaining the pre-labeled data of each scene in the second class of scenes.
8. A pre-labeling apparatus based on hierarchical migration learning, comprising:
the scene clustering module is used for clustering a plurality of different pre-received scenes based on a preset clustering algorithm to obtain a clustering result;
the category determination module is used for determining a first category of scenes and a second category of scenes according to the clustering result, wherein the first category of scenes comprises a first scene, and the data volume of the marking data in the first scene is larger than that of any scene in the second category of scenes;
the first labeling module is used for performing transfer learning on the first type of scenes based on a preset identification model to obtain pre-labeling data and a transfer model of each scene in the first type of scenes;
and the second labeling module is used for performing transfer learning on the second type of scenes based on the transfer model to obtain the pre-labeling data of each scene in the second type of scenes.
9. A computer device comprising a memory having computer readable instructions stored therein and a processor that when executed implements the steps of the method of pre-labeling based on hierarchical migration learning according to any of claims 1 to 7.
10. A computer-readable storage medium, having computer-readable instructions stored thereon, which, when executed by a processor, implement the steps of the pre-labeling method based on hierarchical migration learning according to any one of claims 1 to 7.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011364408.9A CN112434746B (en) | 2020-11-27 | 2020-11-27 | Pre-labeling method based on hierarchical migration learning and related equipment thereof |
| PCT/CN2021/082942 WO2022001233A1 (en) | 2020-11-27 | 2021-03-25 | Pre-labeling method based on hierarchical transfer learning and related device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011364408.9A CN112434746B (en) | 2020-11-27 | 2020-11-27 | Pre-labeling method based on hierarchical migration learning and related equipment thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112434746A true CN112434746A (en) | 2021-03-02 |
| CN112434746B CN112434746B (en) | 2023-10-27 |
Family
ID=74697880
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011364408.9A Active CN112434746B (en) | 2020-11-27 | 2020-11-27 | Pre-labeling method based on hierarchical migration learning and related equipment thereof |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN112434746B (en) |
| WO (1) | WO2022001233A1 (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022001233A1 (en) * | 2020-11-27 | 2022-01-06 | 平安科技（深圳）有限公司 | Pre-labeling method based on hierarchical transfer learning and related device |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116340335B (en) * | 2023-05-30 | 2023-09-05 | 广东石油化工学院 | Industrial equipment data storage system based on alliance chain and data processing method thereof |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180314975A1 (en) * | 2017-04-27 | 2018-11-01 | Futurewei Technologies, Inc. | Ensemble transfer learning |
| CN109657552A (en) * | 2018-11-16 | 2019-04-19 | 北京邮电大学 | The vehicle type recognition device being cold-started across scene and method are realized based on transfer learning |
| CN110009038A (en) * | 2019-04-04 | 2019-07-12 | 北京百度网讯科技有限公司 | Training method, device and the storage medium of screening model |
| CN111860674A (en) * | 2020-07-28 | 2020-10-30 | 平安科技（深圳）有限公司 | Sample class identification method and device, computer equipment and storage medium |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110728294A (en) * | 2019-08-30 | 2020-01-24 | 北京影谱科技股份有限公司 | Cross-domain image classification model construction method and device based on transfer learning |
| CN111259812B (en) * | 2020-01-17 | 2023-04-18 | 上海交通大学 | Inland ship re-identification method and equipment based on transfer learning and storage medium |
| CN112434746B (en) * | 2020-11-27 | 2023-10-27 | 平安科技（深圳）有限公司 | Pre-labeling method based on hierarchical migration learning and related equipment thereof |
-
2020
- 2020-11-27 CN CN202011364408.9A patent/CN112434746B/en active Active
-
2021
- 2021-03-25 WO PCT/CN2021/082942 patent/WO2022001233A1/en active Application Filing
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180314975A1 (en) * | 2017-04-27 | 2018-11-01 | Futurewei Technologies, Inc. | Ensemble transfer learning |
| CN109657552A (en) * | 2018-11-16 | 2019-04-19 | 北京邮电大学 | The vehicle type recognition device being cold-started across scene and method are realized based on transfer learning |
| CN110009038A (en) * | 2019-04-04 | 2019-07-12 | 北京百度网讯科技有限公司 | Training method, device and the storage medium of screening model |
| CN111860674A (en) * | 2020-07-28 | 2020-10-30 | 平安科技（深圳）有限公司 | Sample class identification method and device, computer equipment and storage medium |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022001233A1 (en) * | 2020-11-27 | 2022-01-06 | 平安科技（深圳）有限公司 | Pre-labeling method based on hierarchical transfer learning and related device |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2022001233A1 (en) | 2022-01-06 |
| CN112434746B (en) | 2023-10-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2022142014A1 (en) | Multi-modal information fusion-based text classification method, and related device thereof | |
| CN112101437B (en) | Fine granularity classification model processing method based on image detection and related equipment thereof | |
| CN112528029A (en) | Text classification model processing method and device, computer equipment and storage medium | |
| WO2021174877A1 (en) | Processing method for smart decision-based target detection model, and related device | |
| CN111126347B (en) | Human eye state identification method, device, terminal and readable storage medium | |
| CN113220734A (en) | Course recommendation method and device, computer equipment and storage medium | |
| CN112995414B (en) | Behavior quality inspection method, device, equipment and storage medium based on voice call | |
| WO2022001233A1 (en) | Pre-labeling method based on hierarchical transfer learning and related device | |
| CN112686243A (en) | Method and device for intelligently identifying picture characters, computer equipment and storage medium | |
| CN112084752A (en) | Statement marking method, device, equipment and storage medium based on natural language | |
| CN112417121A (en) | Client intention recognition method and device, computer equipment and storage medium | |
| CN112395390A (en) | Training corpus generation method of intention recognition model and related equipment thereof | |
| CN115510186A (en) | Instant question and answer method, device, equipment and storage medium based on intention recognition | |
| CN113283222B (en) | Automatic report generation method and device, computer equipment and storage medium | |
| CN113240071A (en) | Graph neural network processing method and device, computer equipment and storage medium | |
| CN110532448B (en) | Document classification method, device, equipment and storage medium based on neural network | |
| CN112507141A (en) | Investigation task generation method and device, computer equipment and storage medium | |
| CN111143568A (en) | Method, device and equipment for buffering during paper classification and storage medium | |
| CN114241411B (en) | Counting model processing method and device based on target detection and computer equipment | |
| CN112966150A (en) | Video content extraction method and device, computer equipment and storage medium | |
| CN112395450A (en) | Picture character detection method and device, computer equipment and storage medium | |
| CN112396111A (en) | Text intention classification method and device, computer equipment and storage medium | |
| CN114359582B (en) | Small sample feature extraction method based on neural network and related equipment | |
| CN113139490B (en) | Image feature matching method and device, computer equipment and storage medium | |
| CN117688193B (en) | Picture and text unified coding method, device, computer equipment and medium |
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 |