CN112364167A

CN112364167A - Deep learning-based intention recognition method, system, device and storage medium

Info

Publication number: CN112364167A
Application number: CN202011311131.3A
Authority: CN
Inventors: 江小林; 罗超; 胡泓; 李巍; 邹宇
Original assignee: Ctrip Computer Technology Shanghai Co Ltd
Current assignee: Ctrip Computer Technology Shanghai Co Ltd
Priority date: 2020-11-20
Filing date: 2020-11-20
Publication date: 2021-02-12

Abstract

The invention provides an intention identification method, system, equipment and storage medium based on deep learning, wherein the method comprises the following steps: obtaining statement information to be analyzed and preprocessing the statement information; using a word segmentation tool to segment words of the sentence information; weighting all the participles by using a pre-trained word vector model; obtaining a tone classification corresponding to statement information by using a pre-trained first neural network; using a pre-trained second neural network to obtain sentence pattern classification corresponding to the sentence information; obtaining an expression classification corresponding to the statement information by using a pre-trained third neural network; and predicting the intention category of the statement information according to the mood classification, the sentence pattern classification and the expression classification of the statement information by using a pre-trained fourth neural network. The method and the device can accurately determine the type of the intention expressed by the sentence from the sentence text information by analyzing the word vector of each participle in the sentence text information without extracting the characteristic information of the sentence text information in the corresponding field.

Description

Deep learning-based intention recognition method, system, device and storage medium

Technical Field

The present invention relates to the field of language identification, and in particular, to an intention identification method, system, device and storage medium based on deep learning.

Background

The dialog System (dialog System) is a natural language-based human-computer interaction System. Through a dialog system, a person may use natural language and a computer to perform multiple rounds of interaction to accomplish specific tasks, such as information query, service acquisition, and the like. The dialogue system provides a more natural and convenient man-machine interaction mode, and is widely applied to scenes such as vehicle-mounted scenes, home furnishing scenes, customer service scenes and the like.

Among them, Natural Language Understanding (Natural Language Understanding) is a core module in a dialog system. The goal of natural language understanding is to convert the textual information of natural language into a Semantic Representation (Semantic Representation) that can be processed by a computer, i.e., a structured data to represent the meaning expressed by a sentence. That is, the natural language understanding aims to determine an intention that a user wants to express and a condition satisfying the user intention from text information to be parsed.

In the related art, one of the most common technical solutions is a rule-based natural language understanding. The rule-based natural language understanding manner performs intention recognition by defining several rules and the intention elements acquire keywords of the intention elements required to express the corresponding intention from the text information. For example, when the input text conforms to the rule "order/X to Y/flight", the system outputs the intent to order an air ticket, X and Y being departure and arrival cities, respectively.

However, the rule-based natural language understanding method has at least the following technical problems:

(1) the setting of the rules generally depends on domain knowledge, and requires a professional to set the rules, which is labor-intensive.

(2) Because of the diversity of natural language expressions, a large number of rules need to be set to cover different expressions, but even then, all the rules cannot be exhausted.

(3) The rule-based method cannot perform self-learning and evolves continuously according to actual data, and can only solve problems by continuously setting new rules.

(4) If there are more and more rules, conflicts between the rules may even occur.

In addition, natural language understanding can also be performed by a conventional statistical machine learning method, and the natural language understanding in this way requires input features of a manually designed model (e.g., n-gram, tf-idf (term-update-accumulation-frequency, a commonly used weighting technique for information retrieval data mining)), which brings about two problems. First, the quality of the features affects the effectiveness of the model, and the design of the features depends on domain knowledge, thus requiring a professional to exert a great deal of effort on the design of the features. Second, because most features are domain knowledge-related, the same feature is effective in solving a problem in one domain and does not mean that the feature is effective in solving a problem in another domain, which makes the method less capable of migrating across domains.

Accordingly, the present invention provides a method, system, device and storage medium for deep learning based intention recognition.

Disclosure of Invention

Aiming at the problems in the prior art, the invention aims to provide an intention identification method, system, equipment and storage medium based on deep learning, which overcome the difficulties in the prior art, can accurately determine the type of an intention to be expressed by a sentence from sentence text information by analyzing a word vector of each participle in the sentence text information without extracting the feature information of the sentence text information in the corresponding field.

The embodiment of the invention provides an intention identification method based on deep learning, which comprises the following steps:

s110, obtaining statement information to be analyzed and preprocessing the statement information;

s120, performing word segmentation on the sentence information by using a word segmentation tool;

s130, weighting all the participles by using a pre-trained word vector model;

s140, acquiring a tone classification corresponding to the statement information by using a pre-trained first neural network;

s150, obtaining sentence pattern classification corresponding to the sentence information by using a pre-trained second neural network;

s160, obtaining an expression classification corresponding to the statement information by using a pre-trained third neural network;

and S170, predicting the intention category of the statement information according to the mood classification, sentence pattern classification and expression classification of the statement information by using a pre-trained fourth neural network.

Preferably, in step S140, the statement information is vector-encoded using a pre-trained first neural network, where the first neural network is a mood recognition model trained using a BI-LSTM algorithm, captures semantic information of the comment text, and obtains a mood classification corresponding to the statement information through an output layer.

Preferably, in step S150, the sentence information is vector-coded using a pre-trained second neural network, where the second neural network is a sentence pattern recognition model trained using a BI-LSTM algorithm, captures semantic information of the comment text, and obtains a sentence pattern classification corresponding to the sentence information through the output layer.

Preferably, in step S160, a pre-trained third neural network is used to perform expression classification on the statement information, where the third neural network is an expression recognition model, and obtains an expression classification corresponding to the statement information, where the expression classification includes positive expression and negative expression.

Preferably, in step S120, a HanLP participle tool is used to perform participle on the sentence information.

Preferably, the word vector model pre-trained by word2vec and GloVe in step S130 weights all the participles.

Preferably, the first neural network, the second neural network, the third neural network and the fourth neural network are respectively one of a CNN neural network, an LTSM neural network and a transform model.

The embodiment of the present invention further provides an intention recognition system based on deep learning, which is used for implementing the above intention recognition method based on deep learning, and the intention recognition system based on deep learning includes:

the preprocessing module acquires statement information to be analyzed for preprocessing;

the sentence segmentation module is used for segmenting words of the sentence information by using a segmentation tool;

the word segmentation weighting module is used for weighting all the word segmentations by using a pre-trained word vector model;

the tone classification module is used for acquiring tone classification corresponding to the statement information by using a pre-trained first neural network;

the sentence pattern classification module is used for acquiring sentence pattern classification corresponding to the sentence information by using a pre-trained second neural network;

the expression classification module is used for acquiring expression classification corresponding to the statement information by using a pre-trained third neural network;

and the intention category module predicts the intention category of the statement information according to the mood classification, sentence pattern classification and expression classification of the statement information by using a pre-trained fourth neural network.

An embodiment of the present invention further provides an intention recognition apparatus based on deep learning, including:

a processor;

a memory having stored therein executable instructions of the processor;

wherein the processor is configured to perform the steps of the above deep learning based intent recognition method via execution of the executable instructions.

Embodiments of the present invention also provide a computer-readable storage medium storing a program that, when executed, implements the steps of the above-described deep learning-based intention recognition method.

The invention aims to provide an intention identification method, an intention identification system, intention identification equipment and a storage medium based on deep learning, which can accurately determine the type of an intention expressed by a sentence from the sentence text information by analyzing a word vector of each participle in the sentence text information without extracting the characteristic information of the sentence text information in the corresponding field.

Drawings

Other features, objects and advantages of the present invention will become more apparent upon reading of the following detailed description of non-limiting embodiments thereof, with reference to the accompanying drawings.

FIG. 1 is a flow chart of the deep learning based intent recognition method of the present invention.

FIG. 2 is a block schematic diagram of the deep learning based intent recognition system of the present invention.

Fig. 3 is a schematic structural diagram of the deep learning-based intention recognition apparatus of the present invention.

Fig. 4 is a schematic structural diagram of a computer-readable storage medium according to an embodiment of the present invention.

Detailed Description

Example embodiments will now be described more fully with reference to the accompanying drawings. Example embodiments may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of example embodiments to those skilled in the art. The same reference numerals in the drawings denote the same or similar structures, and thus their repetitive description will be omitted.

FIG. 1 is a flow chart of the deep learning based intent recognition method of the present invention. As shown in fig. 1, an embodiment of the present invention provides an intention recognition method based on deep learning, including the following steps:

and S110, obtaining statement information to be analyzed and preprocessing the statement information.

And S120, segmenting the sentence information by using a segmentation tool.

S130, weighting all the participles by using the pre-trained word vector model.

And S140, carrying out vector coding on the statement information by using a pre-trained first neural network, wherein the first neural network is a tone recognition model and acquires tone classification corresponding to the statement information.

And S150, carrying out vector coding on the statement information by using a pre-trained second neural network, wherein the second neural network is a sentence pattern recognition model and obtains sentence pattern classification corresponding to the statement information.

And S160, carrying out expression classification on the statement information by using a pre-trained third neural network, wherein the third neural network is an expression recognition model and obtains expression classification corresponding to the statement information, and the expression classification comprises positive expression and negative expression.

And S170, predicting the intention category of the statement information according to the mood classification, the sentence pattern classification and the expression classification of the statement information by using a pre-trained fourth neural network.

In a preferred embodiment, in step S140, the sentence information is vector-encoded using a pre-trained first neural network, which is a mood recognition model trained using a BI-LSTM algorithm, the semantic information of the comment text is captured, and the mood classification corresponding to the sentence information is obtained through the output layer.

In a preferred embodiment, in step S150, the sentence information is vector-coded using a pre-trained second neural network, where the second neural network is a sentence pattern recognition model trained using BI-LSTM algorithm, captures semantic information of the comment text, and obtains a sentence pattern classification corresponding to the sentence information through the output layer.

In a preferred embodiment, in step S160, the statement information is subjected to expression classification using a pre-trained third neural network, where the third neural network is an expression recognition model, and the expression classification corresponding to the statement information is obtained, and includes positive expression and negative expression.

In a preferred embodiment, the sentence information is participled using the HanLP participle tool in step S120. The HanLP is a toolkit composed of a series of model pre-algorithms, combines distributed natural language processing of a deep neural network, has the characteristics of complete functions, high performance, clear architecture, novel linguistic data, customization and the like, provides the functions of lexical analysis, syntactic analysis, text analysis, emotion analysis and the like, and is a natural language processing technology which is most popular with GitHub, has the largest user quantity (more than 13000 star) and has the highest community liveness. HanLP is a fully open source, including dictionaries. Independent of other jar, the bottom layer adopts a series of high-speed data structures, such as a double-array Trie tree, a DAWG, an AhoCorasickDoubleArrayTrie and the like, and the basic elements are open sources.

In a preferred embodiment, the word vector model pre-trained by word2vec and GloVe in step S130 weights all the participles.

Word2vec, a group of correlation models used to generate Word vectors. These models are shallow, two-layer neural networks that are trained to reconstruct linguistic word text. The network is represented by words and the input words in adjacent positions are guessed, and the order of the words is unimportant under the assumption of the bag-of-words model in word2 vec. After training is completed, the word2vec model can be used to map each word to a vector, which can be used to represent word-to-word relationships, and the vector is a hidden layer of the neural network.

GloVe's full name Global Vectors for Word Representation is a Word Representation tool based on Global Word frequency statistics (count-based & overall statistics) that can represent a Word as a vector consisting of real numbers that capture some semantic properties between words, such as similarity (similarity), analogy (analogy), etc. We can compute the semantic similarity between two words by operations on the vectors, such as euclidean distance or cosine similarity.

In a preferred embodiment, the first neural network, the second neural network, the third neural network and the fourth neural network are respectively one of a CNN neural network, an LTSM neural network and a transform model.

Convolutional Neural Networks (CNN) are a class of feed forward Neural Networks (fed forward Neural Networks) that contain convolution computations and have a deep structure, and are one of the representative algorithms for deep learning (deep learning) [1-2 ]. Convolutional Neural Networks have a feature learning (representation learning) capability, and can perform Shift-Invariant classification (Shift-Invariant classification) on input information according to a hierarchical structure thereof, and are also called Shift-Invariant Artificial Neural Networks (SIANN).

The Long Short-Term Memory network (LSTM) is a time-cycle neural network, which is specially designed to solve the Long-Term dependence problem of the general RNN (cyclic neural network), and all RNNs have a chain form of repeated neural network modules. In the standard RNN, this repeated structure block has only a very simple structure, e.g. one tanh layer.

The transducer model of google was first used for machine translation tasks, and the SOTA effect was achieved at that time. The Transformer improves the defect of slow training of the RNN most suffered from the problem, and realizes quick parallelism by using a self-attention mechanism. And the Transformer can be increased to a very deep depth, so that the characteristics of the DNN model are fully explored, and the model accuracy is improved.

The intention identification method based on deep learning can accurately determine the type of the intention expressed by the sentence from the sentence text information by analyzing the word vector of each participle in the sentence text information without extracting the characteristic information of the sentence text information in the corresponding field.

The invention is mainly realized by the following method, which mainly comprises the following steps:

first, training phase

Tone recognition:

identifying the query and non-query: stating tone, imperative tone, exclamatory tone

The query language may contain different types:

whether to ask a sentence: whether or not to

In particular to question: asking for a specific content

Selecting a question: selecting a certain condition

Question of positive and negative: the positive and negative forms of the predicate are concatenated and then asked

Question back: positive or negative meanings are expressed in terms of interrogative sentences. "positive form" expresses negative meaning, while negative form expresses positive meaning

Such as:

"who can't receive order" expression can receive order

"who says can order" expression can not order

Expression of "not having been grafted" is grafted

1.1 collecting and annotating data

1.2 text preprocessing

Filtering pure numbers and standardizing: such as full angle to half angle, traditional Chinese to simplified Chinese, conversion from capital to small form

1.3 dialogue user statement text vectorization

And for the text obtained in the previous step, performing word segmentation by using a hand word segmentation tool, adding some professional words in the scene in the word segmentation process to ensure that the word segmentation accuracy is higher, and weighting by using word vector models pre-trained by word2vec and glove to obtain word vectors of all words to be used as model input.

1.4 vector coding of sentences using neural networks

The method comprises the steps of training texts marked with various tone types by using a BI-LSTM algorithm, capturing semantic information of the comment texts, and obtaining the tone types of the texts through a softmax layer.

1.5, training neural network parameters to obtain a tone recognition model

Obtaining a deep learning model after multiple iterations

Alternatively, the common neural network classification model can be selected from models based on cnn, lstm, transormer and the like.

Secondly, sentence pattern recognition: relevant words with or without explicit display

Two or more items are stated in a plurality of main and subordinate structures, the matters stated in the main sentence are emphasized, and the appearances of the clauses are used for more comprehensively explaining the meaning of the main sentence.

Common complex sentence patterns include:

2.1, assume sentence pattern

The case of expression of a clause under an assumed condition or a specific condition does not necessarily occur. Commonly used associated words: if so, as long as. . . Then it is completed. . . Only, then. Just before

If he does not wish to add money, he cannot cancel it free of charge

The intention is to focus on the following "cannot be cancelled"

2.2 conditional sentence pattern

The presentation of a particular condition produces a particular result, or produces a result regardless of any condition

Regardless of the amount of money deducted, the single can not be cancelled

The intention is to focus on the following 'Cancel'

2.3 turning sentence pattern

Preceding and following clauses are intended to be completely or partially opposite

While we have policies that can be cancelled, the emphasis of his timeout being too much to cancel the intent is on the following "cannot be cancelled" rather than the preceding "can be cancelled"

2.4 causal sentence pattern

The preceding and following sentences explain the cause and effect respectively

Because I do not start the computer, I wait for you to see the list again

The emphasis of the intention is on the rear "to see"

2.5 parallel sentence pattern

A plurality of simple sentences are formed in parallel in a primary and secondary non-independent relationship, the front and rear sentences are objective fact orders with tail numbers of 1234, 5678 cannot be confirmed

The emphasis on "1234 may", "5678 does not" is on multiple intents, both of which are important.

2.6 selecting sentence pattern

Enumerating multiple cases and selecting one case from them

Can continue to live and can also deduct money to cancel

Both intentions are important

2.7, progressive sentence pattern

The relation between the preceding sentence and the following sentence is one level in meaning, and the preceding sentence and the following sentence are both objective facts

We have free cancellation policy, and he or vip can cancel the free

The emphasis of the intention is that "can cancel free" to obtain one

And (3) sentence pattern model training:

1.1 collecting and annotating data

1.2 text preprocessing

1.3 dialogue user statement text vectorization

1.4 vector coding of sentences using neural networks

And training the text marked with various sentence types by using a BI-LSTM algorithm, capturing semantic information of the comment text, and acquiring the sentence type to which the text belongs through a softmax layer.

1.5, training neural network parameters to obtain a tone recognition model

And obtaining a deep learning model after multiple iterations. Alternatively, the common neural network classification model can be selected from models based on cnn, lstm, transormer and the like. Meanwhile, common sentence pattern associated words are collected to serve as model supplementary knowledge, and sentence pattern recognition is carried out on the basis of associated word dictionaries and models.

Three, positive/negative identification

Conventional positive negation

Double negation

By negative word recognition:

negative words: no, none, no,

double negative words: not, not possible, not … not …, not … not …, not … not …, not … … not …, not necessarily not.

Multi-task learning of intention identification and tone and sentence pattern

Optionally, the intention recognition classification model and the mood and sentence pattern detection model are used as a joint task, so that the overall generalization of the model is improved. As shown in the figure, different task losses are calculated by the aid of three different tasks in a parameter sharing mode, the losses of the three tasks are added and calculated, and combined training is carried out to obtain a model.

FIG. 2 is a block schematic diagram of the deep learning based intent recognition system of the present invention. As shown in fig. 2, the deep learning based intention recognition system 5 of the present invention includes:

the preprocessing module 51 obtains the statement information to be analyzed for preprocessing.

The sentence segmentation module 52 performs segmentation on the sentence information by using a segmentation tool.

The segmentation weighting module 53 weights all the segmentations using the pre-trained word vector model.

The mood classification module 54 obtains a mood classification corresponding to the sentence information by using the pre-trained first neural network.

And the sentence pattern classification module 55 acquires sentence pattern classification corresponding to the sentence information by using a pre-trained second neural network.

And the expression classification module 56 acquires the expression classification corresponding to the statement information by using the pre-trained third neural network.

The intention category module 57 predicts an intention category of the sentence information from the mood classification, sentence pattern classification, and expression classification of the sentence information using a pre-trained fourth neural network.

The intention recognition system based on deep learning can accurately determine the type of the intention expressed by the sentence from the sentence text information by analyzing the word vector of each participle in the sentence text information without extracting the characteristic information of the sentence text information in the corresponding field.

The embodiment of the invention also provides the intention recognition equipment based on the deep learning, which comprises a processor. A memory having stored therein executable instructions of the processor. Wherein the processor is configured to perform the steps of the deep learning based intent recognition method via execution of executable instructions.

As shown above, the deep learning-based intention recognition system of the embodiment of the present invention can accurately determine the type of intention to be expressed by a sentence from the sentence text information by analyzing the word vector of each participle in the sentence text information without extracting feature information of the sentence text information in the corresponding field.

As will be appreciated by one skilled in the art, aspects of the present invention may be embodied as a system, method or program product. Thus, various aspects of the invention may be embodied in the form of: an entirely hardware embodiment, an entirely software embodiment (including firmware, microcode, etc.) or an embodiment combining hardware and software aspects that may all generally be referred to herein as a "circuit," module "or" platform.

Fig. 3 is a schematic structural diagram of the deep learning-based intention recognition apparatus of the present invention. An electronic device 600 according to this embodiment of the invention is described below with reference to fig. 3. The electronic device 600 shown in fig. 3 is only an example and should not bring any limitation to the functions and the scope of use of the embodiments of the present invention.

As shown in fig. 3, the electronic device 600 is embodied in the form of a general purpose computing device. The components of the electronic device 600 may include, but are not limited to: at least one processing unit 610, at least one memory unit 620, a bus 630 connecting the different platform components (including the memory unit 620 and the processing unit 610), a display unit 640, etc.

Wherein the storage unit stores program code executable by the processing unit 610 to cause the processing unit 610 to perform steps according to various exemplary embodiments of the present invention described in the above-mentioned electronic prescription flow processing method section of the present specification. For example, processing unit 610 may perform the steps as shown in fig. 1.

The storage unit 620 may include readable media in the form of volatile memory units, such as a random access memory unit (RAM)6201 and/or a cache memory unit 6202, and may further include a read-only memory unit (ROM) 6203.

The memory unit 620 may also include a program/utility 6204 having a set (at least one) of program modules 6205, such program modules 6205 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may comprise an implementation of a network environment.

Bus 630 may be one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

The electronic device 600 may also communicate with one or more external devices 700 (e.g., keyboard, pointing device, bluetooth device, etc.), with one or more devices that enable a user to interact with the electronic device 600, and/or with any devices (e.g., router, modem, etc.) that enable the electronic device 600 to communicate with one or more other computing devices. Such communication may occur via an input/output (I/O) interface 650. Also, the electronic device 600 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network such as the Internet) via the network adapter 660. The network adapter 660 may communicate with other modules of the electronic device 600 via the bus 630. It should be appreciated that although not shown in the figures, other hardware and/or software modules may be used in conjunction with the electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms, to name a few.

The embodiment of the invention also provides a computer readable storage medium for storing a program, and the steps of the intention identification method based on deep learning are realized when the program is executed. In some possible embodiments, the aspects of the present invention may also be implemented in the form of a program product comprising program code for causing a terminal device to perform the steps according to various exemplary embodiments of the present invention described in the above-mentioned electronic prescription flow processing method section of this specification, when the program product is run on the terminal device.

Fig. 4 is a schematic structural diagram of a computer-readable storage medium of the present invention. Referring to fig. 4, a program product 800 for implementing the above method according to an embodiment of the present invention is described, which may employ a portable compact disc read only memory (CD-ROM) and include program code, and may be run on a terminal device, such as a personal computer. However, the program product of the present invention is not limited in this regard and, in the present document, a readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

A computer readable storage medium may include a propagated data signal with readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A readable storage medium may also be any readable medium that is not a readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a readable storage medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computing device (e.g., through the internet using an internet service provider).

In summary, the present invention is directed to an intention recognition method, an intention recognition system, an intention recognition device, and a storage medium based on deep learning, where the intention recognition system based on deep learning can accurately determine the type of an intention to be expressed by a sentence from sentence text information by analyzing a word vector of each word segment in the sentence text information without extracting feature information of the sentence text information in a corresponding field.

The foregoing is a more detailed description of the invention in connection with specific preferred embodiments and it is not intended that the invention be limited to these specific details. For those skilled in the art to which the invention pertains, several simple deductions or substitutions can be made without departing from the spirit of the invention, and all shall be considered as belonging to the protection scope of the invention.

Claims

1. An intention recognition method based on deep learning is characterized by comprising the following steps:

s130, weighting all the participles by using a pre-trained word vector model;

2. The deep learning-based intention recognition method of claim 1, wherein in step S140, the sentence information is vector-encoded using a pre-trained first neural network, the first neural network is a mood recognition model trained using BI-LSTM algorithm, semantic information of the comment text is captured, and a mood classification corresponding to the sentence information is obtained through the output layer.

3. The deep learning-based intention recognition method of claim 1, wherein in step S150, the sentence information is vector-coded using a pre-trained second neural network, the second neural network is a sentence pattern recognition model trained by using a BI-LSTM algorithm, semantic information of the comment text is captured, and the sentence pattern classification corresponding to the sentence information is obtained through the output layer.

4. The deep learning-based intention recognition method of claim 1, wherein in step S160, a pre-trained third neural network is used for performing expression classification on the sentence information, wherein the third neural network is an expression recognition model, and the expression classification corresponding to the sentence information is obtained, and the expression classification comprises positive expression and negative expression.

5. The deep learning-based intention recognition method of claim 1, wherein in the step S120, a HanLP participle tool is used to participle the sentence information.

6. The deep learning-based intention recognition method of claim 1, wherein the word vector model pre-trained by word2vec and GloVe in the step S130 weights all the participles.

7. The deep learning based intention recognition method of claim 1, wherein the first, second, third and fourth neural networks are each one of CNN, LTSM, and Transformer models.

8. A deep learning based intention recognition system for implementing the deep learning based intention recognition method of claim 1, characterized by comprising:

9. An intent recognition device based on deep learning, comprising:

a processor;

a memory having stored therein executable instructions of the processor;

wherein the processor is configured to perform the steps of the deep learning based intention recognition method of any one of claims 1 to 7 via execution of the executable instructions.

10. A computer-readable storage medium storing a program, wherein the program is configured to implement the steps of the deep learning based intention recognition method of any one of claims 1 to 7 when executed.