CN111581978A - Intention identification method through domain migration and antagonistic learning - Google Patents
Intention identification method through domain migration and antagonistic learning Download PDFInfo
- Publication number
- CN111581978A CN111581978A CN202010348401.1A CN202010348401A CN111581978A CN 111581978 A CN111581978 A CN 111581978A CN 202010348401 A CN202010348401 A CN 202010348401A CN 111581978 A CN111581978 A CN 111581978A
- Authority
- CN
- China
- Prior art keywords
- data
- mapping layer
- source
- target
- steps
- 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.)
- Withdrawn
Links
Images
Classifications
-
- 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
- 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
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- 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
- G06F16/355—Class or cluster creation or modification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Databases & Information Systems (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Machine Translation (AREA)
Abstract
An intention identification method through domain migration and antagonistic learning relates to the technical field of data processing analysis and comprises the following steps: step 1): performing feature extraction on a large amount of labeled data and unlabeled data by adopting a pre-training model extractor; step 2): extracting input sentences input into each dialogue, and expressing the input sentences into a semantic vector through an extractor; step 3): inputting the semantic vector into a mapping layer, wherein source data corresponds to the source mapping layer, and target data is mapped to the target mapping layer; step 4): the target mapping layer distinguishes whether the output target data is real or not through a classifier. The invention inhibits the efficiency of the classifier in the field of target data, constructs real labels and false labels aiming at the target data so as to achieve the effect of resisting learning, and can train the source data, the mapping layer and the classifier thereof in advance, thereby improving the accuracy and reducing the cost consumption of manpower and material resources.
Description
Technical Field
The invention relates to the technical field of data processing and analysis, in particular to an intention identification method for migration and counterstudy through fields.
Background
The intention recognition is used as a component of a modularized dialogue system, plays roles in recognizing user intention and judging dialogue fields, the current intention recognition task is usually used as a classification model in machine learning for modeling, the used learning mode is mostly supervised learning, and the supervised learning is to give a section of data representation and a corresponding label to train the model discrimination capability. However, supervised classification learning is successfully built on an assumption that the data distribution of training data is consistent or similar to test data, but in the real world, this assumption often does not exist; secondly, a large amount of labeled data is needed for model training in supervised learning, which brings cost consumption of manpower and material resources.
Disclosure of Invention
Technical problem to be solved
Aiming at the defects of the prior art, the invention provides an intention identification method through field migration and antagonistic learning, which solves the problem that supervised classification learning is successfully established on the assumption that the data distribution of training data is consistent or similar to test data, but in the real world, the assumption does not exist; secondly, a large amount of labeled data is needed for model training in supervised learning, which brings the problem of cost consumption of manpower and material resources.
(II) technical scheme
In order to achieve the purpose, the invention is realized by the following technical scheme: an intention recognition method through domain migration and antagonistic learning, comprising the steps of:
step 1): performing feature extraction on a large amount of labeled data and unlabeled data by adopting a pre-training model extractor;
step 2): extracting input sentences input into each dialogue, and expressing the input sentences into a semantic vector through an extractor;
step 3): inputting the semantic vector into a mapping layer, wherein source data corresponds to the source mapping layer, and target data is mapped to the target mapping layer;
step 4): the target mapping layer distinguishes whether the output target data is real or not through a classifier.
Preferably, the pre-training model extractor adopts a Chinese ERNIE model.
Preferably, the labeled data is source data, and the unlabeled data is target data.
Preferably, the mapping layer is implemented by a recurrent neural network.
Preferably, the source mapping layer outputs to a source classifier for classifying the marked source data.
Preferably, for the multi-label classification output, a plurality of sigmoid functions are adopted to perform a plurality of binary classification outputs.
(III) advantageous effects
The invention provides an intention identification method for migration and counterstudy through a field. Has the following beneficial effects:
(1) the method for identifying the intention of the user to move and resist learning through the field comprises the steps of extracting features of a large amount of labeled data and unlabeled data by adopting a pre-training model extractor; extracting input sentences input into each dialogue, and expressing the input sentences into a semantic vector through an extractor; inputting the semantic vector into a mapping layer, wherein source data corresponds to the source mapping layer, and target data is mapped to the target mapping layer; the target mapping layer distinguishes whether the output real target data exists through a classifier, the effect of resisting learning is achieved, the source data, the mapping layer and the classifier are trained in advance, the accuracy is improved, and the cost consumption of manpower and material resources is reduced.
Drawings
FIG. 1 is a flow chart of an intent recognition method of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, the present invention provides a technical solution: an intention recognition method through domain migration and antagonistic learning, comprising the steps of:
step 1): performing feature extraction on a large amount of labeled data and unlabeled data by adopting a pre-training model extractor, wherein the pre-training model extractor adopts a Chinese ERNIE model, labeled data are source data, and unlabeled data are target data;
step 2): extracting input sentences input into each dialogue, and expressing the input sentences into a semantic vector through an extractor;
step 3): inputting the semantic vector into a mapping layer, wherein the mapping layer is realized by a recurrent neural network, source data corresponds to a source mapping layer, the source mapping layer outputs the source data to a source classifier, the labeled source data is classified, and target data is mapped to a target mapping layer;
step 4): the target mapping layer distinguishes whether the output target data is real or not through a classifier, and multiple S-shaped functions are adopted to carry out multiple two-classification output aiming at the multi-label classification output.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation. The use of the phrase "comprising one of the elements does not exclude the presence of other like elements in the process, method, article, or apparatus that comprises the element.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (6)
1. An intention recognition method through domain migration and antagonistic learning, characterized in that: the method comprises the following steps:
step 1): performing feature extraction on a large amount of labeled data and unlabeled data by adopting a pre-training model extractor;
step 2): extracting input sentences input into each dialogue, and expressing the input sentences into a semantic vector through an extractor;
step 3): inputting the semantic vector into a mapping layer, wherein source data corresponds to the source mapping layer, and target data is mapped to the target mapping layer;
step 4): the target mapping layer distinguishes whether the output target data is real or not through a classifier.
2. The method of claim 1, wherein the method comprises the following steps: the pre-training model extractor adopts a Chinese ERNIE model.
3. The method of claim 1, wherein the method comprises the following steps: the marked data are source data, and the unmarked data are target data.
4. The method of claim 1, wherein the method comprises the following steps: the mapping layer is implemented by a recurrent neural network.
5. The method of claim 1, wherein the method comprises the following steps: the source mapping layer is output to a source classifier for classifying the marked source data.
6. The method of claim 1, wherein the method comprises the following steps: and aiming at the multi-label classified output, a plurality of S-shaped functions are adopted to carry out a plurality of two-classification outputs.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010348401.1A CN111581978A (en) | 2020-04-28 | 2020-04-28 | Intention identification method through domain migration and antagonistic learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010348401.1A CN111581978A (en) | 2020-04-28 | 2020-04-28 | Intention identification method through domain migration and antagonistic learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111581978A true CN111581978A (en) | 2020-08-25 |
Family
ID=72111785
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010348401.1A Withdrawn CN111581978A (en) | 2020-04-28 | 2020-04-28 | Intention identification method through domain migration and antagonistic learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111581978A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113139063A (en) * | 2021-06-21 | 2021-07-20 | 平安科技（深圳）有限公司 | Intention recognition method, device, equipment and storage medium |
-
2020
- 2020-04-28 CN CN202010348401.1A patent/CN111581978A/en not_active Withdrawn
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113139063A (en) * | 2021-06-21 | 2021-07-20 | 平安科技（深圳）有限公司 | Intention recognition method, device, equipment and storage medium |
| CN113139063B (en) * | 2021-06-21 | 2021-09-14 | 平安科技（深圳）有限公司 | Intention recognition method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108874782B (en) | A kind of more wheel dialogue management methods of level attention LSTM and knowledge mapping | |
| CN104572958B (en) | A kind of sensitive information monitoring method based on event extraction | |
| CN109002834B (en) | Fine-grained image classification method based on multi-modal representation | |
| CN109919106B (en) | Progressive target fine recognition and description method | |
| CN108960073A (en) | Cross-module state image steganalysis method towards Biomedical literature | |
| CN109685065B (en) | Layout analysis method and system for automatically classifying test paper contents | |
| CN111506729B (en) | Information processing method, device and computer readable storage medium | |
| CN113505200A (en) | Sentence-level Chinese event detection method combining document key information | |
| CN106934055B (en) | Semi-supervised webpage automatic classification method based on insufficient modal information | |
| CN113434688B (en) | Data processing method and device for public opinion classification model training | |
| Dey et al. | A two-stage CNN-based hand-drawn electrical and electronic circuit component recognition system | |
| CN110633475A (en) | Natural language understanding method, device and system based on computer scene and storage medium | |
| CN111639185B (en) | Relation information extraction method, device, electronic equipment and readable storage medium | |
| CN116910571A (en) | Open-domain adaptation method and system based on prototype comparison learning | |
| Li et al. | Printed/handwritten texts and graphics separation in complex documents using conditional random fields | |
| CN112632963A (en) | Method and device for constructing Chinese metaphor information knowledge base based on government work report | |
| Alon et al. | Deep-hand: a deep inference vision approach of recognizing a hand sign language using american alphabet | |
| CN117454987B (en) | Mine event knowledge graph construction method and device based on event automatic extraction | |
| CN111581978A (en) | Intention identification method through domain migration and antagonistic learning | |
| CN112579730A (en) | High-expansibility multi-label text classification method and device | |
| CN110413856A (en) | Classification annotation method, apparatus, readable storage medium storing program for executing and equipment | |
| CN109886289A (en) | A kind of deep learning method, equipment and computer readable storage medium | |
| CN111898528B (en) | Data processing method, device, computer readable medium and electronic equipment | |
| de las Heras et al. | Notation-invariant patch-based wall detector in architectural floor plans | |
| Barveen et al. | Meme Expressive Classification in Multimodal State with Feature Extraction in Deep Learning |
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 | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20200825 |
|
| WW01 | Invention patent application withdrawn after publication |