CN113094481A

CN113094481A - Intention recognition method and device, electronic equipment and computer readable storage medium

Info

Publication number: CN113094481A
Application number: CN202110232614.2A
Authority: CN
Inventors: 徐正虹; 吴科; 吴立楠; 徐懿
Original assignee: Beijing Zhichi Bochuang Technology Co ltd
Current assignee: Beijing Zhichi Bochuang Technology Co ltd
Priority date: 2021-03-03
Filing date: 2021-03-03
Publication date: 2021-07-09

Abstract

The application provides an intention identification method, an intention identification device, electronic equipment and a computer readable storage medium, wherein the intention identification method comprises the following steps: acquiring current voice sent by a client in an outbound scene, and acquiring a corresponding target text based on the current voice; and determining the intention type contained in the target text, and acquiring an intention recognition result corresponding to the target text based on the intention type. The scheme converts the voice of the client in the outbound scene into the corresponding target text, then determines the intention type in the target text, adopts different intention identification models for different intention types, and identifies the intention of the client.

Description

Intention recognition method and device, electronic equipment and computer readable storage medium

Technical Field

The present application relates to the field of artificial intelligence technologies, and in particular, to an intention recognition method, an intention recognition device, an electronic device, and a computer-readable storage medium.

Background

With the rapid development of economy, many industries have precipitated massive data while providing users with richer products. The utilization of these data for valuable outbound services is a popular product marketing approach. By Outbound (Outbound) is meant: the telephone automatically dials out the user's telephone through the computer, and plays the recorded voice to the user through the computer, which is an indispensable component of the modern customer service center system integrating the computer and the telephone. Outbound calls are generally divided into two phases: the method comprises the steps of obtaining outbound data and initiating an outbound action.

In the intelligent outbound service, a corresponding dialect template is set in advance according to different marketing products, and a marketing robot (or an outbound platform) carries out product marketing according to a node flow in the dialect template. Therefore, in the outbound marketing dialogue, the quick judgment of the acceptance of the marketing product by the client is very important, so that not only can the time and the cost be saved, but also the marketing dialogue can be replaced to continuously recommend the product with uncertain acceptance of the client, and the sales conversion is facilitated. This requires that the robot be able to quickly and accurately identify the user's acceptance of the marketed product.

At present, the recognition data of the intelligent outbound is converted from voice to spoken language, and when the spoken language has a plurality of language words and repeated words and a sentence has a plurality of intentions, the intentions in the spoken language are difficult to be accurately recognized, so that the existing intention recognition method needs to be improved.

Disclosure of Invention

The purpose of this application is to solve at least one of the above technical defects, and the technical solution provided by this application embodiment is as follows:

in a first aspect, an embodiment of the present application provides an intention identification method, which includes:

acquiring current voice sent by a client in an outbound scene, and acquiring a corresponding target text based on the current voice;

and determining the intention type contained in the target text, and acquiring an intention recognition result corresponding to the target text based on the intention type.

In an optional embodiment of the present application, determining an intention type corresponding to the target text includes:

and inputting the target text into a pre-trained binary classification model, and outputting an intention type corresponding to the target text.

In an optional embodiment of the present application, obtaining an intention recognition result corresponding to a target text based on an intention type includes:

acquiring a corresponding pre-trained intention recognition model based on the intention type;

and acquiring an intention recognition result corresponding to the target text by using the pre-trained intention recognition model.

In an optional embodiment of the present application, if the intention type is a general intention type indicating a degree of reception of the outbound marketing object by the client, the intention recognition model corresponding to the general intention type is a pre-trained general intention recognition model;

acquiring an intention recognition result corresponding to a target text by using a pre-trained intention recognition model, wherein the method comprises the following steps:

and acquiring an intention recognition result corresponding to the target text by using the pre-trained general intention recognition model.

In an optional embodiment of the present application, obtaining an intention recognition result corresponding to a target text by using a pre-trained general intention recognition model includes:

if the number of characters of the target text is not less than a preset numerical value, splitting the target text into one or more target text segments;

respectively inputting each target text segment into a pre-trained general purpose intention recognition model, and outputting a corresponding general purpose intention;

and acquiring an intention recognition result corresponding to the target text based on the general intention.

In an optional embodiment of the present application, obtaining an intention recognition result corresponding to a target text based on one or more common intention recognition results includes:

acquiring the priority level of each universal intention recognition result based on a preset universal intention classification table;

and taking the general purpose intention with the highest priority in all the general purpose intention recognition results as the intention recognition result corresponding to the target text.

In an optional embodiment of the application, if the intention type is a knowledge intention type indicating that the client asks for relevant information of the outbound marketing object, an intention identification model corresponding to the knowledge intention type is a pre-trained knowledge intention identification model;

and acquiring an intention recognition result corresponding to the target text by using the pre-trained knowledge intention recognition model.

In an optional embodiment of the present application, obtaining an intention recognition result corresponding to a target text by using a pre-trained knowledge intention recognition model includes:

acquiring a text to be matched with an entity label and/or a keyword label from a preset knowledge meaning library;

respectively inputting the key position vector of the target text and the key position vector of each text to be matched into a pre-trained knowledge intention recognition model, and outputting the matching probability of the target text and each text to be matched;

and acquiring an intention recognition result corresponding to the target text based on each matching probability.

In an optional embodiment of the present application, obtaining an intention recognition result corresponding to a target text based on each matching probability includes:

and taking the knowledge intention corresponding to the text to be matched with the target text with the maximum matching probability as an intention identification result corresponding to the target text.

In an optional embodiment of the present application, the intention type includes a general intention type and a knowledge intention type, and the intention recognition model corresponding to the general intention type includes a pre-trained general intention recognition model and a pre-trained knowledge intention recognition model;

respectively utilizing a pre-trained general intention recognition model and a pre-trained knowledge intention recognition model to obtain a general intention and a knowledge intention corresponding to a target text;

and acquiring an intention recognition result corresponding to the target text based on the general intention and the knowledge intention.

In an optional embodiment of the present application, acquiring a corresponding target text based on a current speech includes:

acquiring a previous voice of an outbound platform corresponding to the current voice, and acquiring a corresponding previous text based on the previous voice;

if the content of the previous text is the answer content to the client, determining the text corresponding to the current voice as the target text;

and if the content of the previous text is the content of a question to the client, splitting the text corresponding to the current voice to obtain one or more target texts.

In an optional embodiment of the present application, the method further comprises:

acquiring corresponding reply content based on the intention recognition result;

and sending the reply content to the client through the outbound platform.

In a second aspect, the present application provides an intent recognition apparatus comprising:

the target text acquisition module is used for acquiring current voice sent by a client in an outbound scene and acquiring a corresponding target text based on the current voice;

and the intention recognition result acquisition module is used for determining the intention type contained in the target text and acquiring the intention recognition result corresponding to the target text based on the intention type.

In an optional embodiment of the present application, the intention identification result obtaining module is specifically configured to:

In an optional embodiment of the present application, if the intention type is a general intention type indicating a degree of reception of the outbound marketing object by the client, the intention recognition model corresponding to the general intention type is a pre-trained general intention recognition model; the intention recognition result acquisition module is specifically configured to:

In an optional embodiment of the present application, the intention-recognition-result obtaining module is further configured to:

In an optional embodiment of the application, if the intention type is a knowledge intention type indicating that the client asks for relevant information of the outbound marketing object, an intention identification model corresponding to the knowledge intention type is a pre-trained knowledge intention identification model; the intention recognition result acquisition module is specifically configured to:

In an optional embodiment of the present application, the intention type includes a general intention type and a knowledge intention type, and the intention recognition model corresponding to the general intention type includes a pre-trained general intention recognition model and a pre-trained knowledge intention recognition model; the intention recognition result acquisition module is specifically configured to:

In an optional embodiment of the present application, the target text acquiring module is specifically configured to:

In an optional embodiment of the present application, the apparatus further comprises a reply module for:

and sending the reply content to the client through the outbound platform.

In a third aspect, an embodiment of the present application provides an electronic device, including a memory and a processor;

the memory has a computer program stored therein;

a processor configured to execute a computer program to implement the method provided in the embodiment of the first aspect or any optional embodiment of the first aspect.

In a fourth aspect, this application provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the method provided in the embodiments of the first aspect or any optional embodiment of the first aspect.

The beneficial effect that technical scheme that this application provided brought is:

the method comprises the steps of converting the voice of a client in an outbound scene into a corresponding target text, then determining the intention type in the target text, and identifying the intention of the client by adopting different intention identification models for different intention types.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings used in the description of the embodiments of the present application will be briefly described below.

Fig. 1 is a schematic flowchart of an intention identification method according to an embodiment of the present application;

FIG. 2 is a flowchart illustrating an exemplary intent recognition process according to an embodiment of the present application;

fig. 3 is a block diagram illustrating an intention identifying apparatus according to an embodiment of the present disclosure;

fig. 4 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

Reference will now be made in detail to embodiments of the present application, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are exemplary only for the purpose of explaining the present application and are not to be construed as limiting the present invention.

As used herein, the singular forms "a", "an", "the" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. It will be understood that when an element is referred to as being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element or intervening elements may also be present. Further, "connected" or "coupled" as used herein may include wirelessly connected or wirelessly coupled. As used herein, the term "and/or" includes all or any element and all combinations of one or more of the associated listed items.

To make the objects, technical solutions and advantages of the present application more clear, embodiments of the present application will be described in further detail below with reference to the accompanying drawings.

Fig. 1 is a schematic flowchart of an intention identification method provided in an embodiment of the present application, and as shown in fig. 1, the method may include the following steps:

step S101, obtaining current voice sent by a client in an outbound scene, and obtaining a corresponding target text based on the current voice.

The current voice is the voice which needs to be analyzed and obtained at present, and the current voice is converted into the corresponding text, so that the target text for subsequent intention analysis is obtained.

Specifically, after the current Speech uttered by the client is acquired, the current Speech may be converted into a corresponding text through an ASR (Automatic Speech Recognition) algorithm, and further, a target text is acquired according to the text, and then, the target text is analyzed to identify an intention contained therein.

Step S102, determining the intention type contained in the target text, and acquiring an intention identification result corresponding to the target text based on the intention type.

Specifically, since multiple types of intentions may be contained in the target text, in order to accurately recognize the intentions of different intention types, the intention types contained in the target text may be determined first, and then, for different intention types, different intention recognition models are used for recognition, so as to obtain corresponding intentions.

According to the scheme provided by the application, the voice of the client in the outbound scene is converted into the corresponding target text, the intention type in the target text is determined, different intention recognition models are adopted for different intention types, the intention of the client is recognized, and due to the fact that different intention recognition models are adopted for different intention types, the recognition accuracy rate of the voice containing multiple intentions can be improved.

And performing classification judgment on the target text, and judging whether the target text contains general purpose intentions or knowledge intentions or both. The two-class model uses a two-class model common to NLP (Natural Language Processing), and the optional models include but are not limited to Bert, TextCNN, FastText, SVM, logistic regression, random forest, etc.

Specifically, after the intention types contained in the target text are acquired, the corresponding intention recognition models are acquired to perform intention recognition on the target text, and corresponding intention recognition results are output. Specifically, if the target text is determined to contain only the general purpose type, the target text is recognized by adopting a pre-trained general purpose recognition model, if the target text is determined to contain only the knowledge purpose type, the target text is recognized by adopting the pre-trained knowledge purpose recognition model, and if the target text is determined to contain both the general purpose type and the knowledge purpose type, the target text is recognized by respectively adopting the pre-trained general purpose recognition model and the pre-trained knowledge purpose recognition model. The above three emotions will be described in detail below.

In an optional embodiment of the present application, obtaining an intention recognition result corresponding to a target text based on a general intention recognition result includes:

Specifically, after determining that the intention type included in the target text is a general intention type, in general intention recognition, if a sentence exceeds a preset numerical value (for example, 16 characters) and a sentence mark, a question mark, a comma, an exclamation mark and other separators exist, a sentence break model is used for carrying out sentence break on the target text, all target text segments after the sentence break are respectively subjected to general intention recognition, if the target text segments after the sentence break belong to various general intentions, the target text segments are sorted according to the priority of the general intentions (namely the priority of the general intention recognition results), and finally one general intention is output as the final user intention of the target text.

For example, the general intentions in the outbound scenario may be divided into 5 large classes and 8 middle classes, where the 8 middle classes are: (1) the method comprises the following steps of obtaining user expression, (2) user acceptance, (3) continuing conversation, (4) repeating conversation, (5) changing marketing conversation, (6) ending conversation (turning to sending message), (7) mood soothing, and (8) complaining. The categories are further subdivided into outbound general intent subdivisions, as shown in table 1 (general intent classification table).

TABLE 1

When the outbound call system (or called outbound call platform) is configured, different conversation templates are configured according to the subdivided general purpose intention for conversation. The robot in the outbound system dialogues with the user according to the dialogues template.

Further, the limited level of each general purpose is preset, and the priority of the general purpose in table 1 is as follows: 81>71>63>62>61>52>51>42>41>33>32>31>21>11, wherein the number is the "user intention number" in the general intention classification table, the previous user intention number indicating a higher priority. And if the number of the characters of the target text is not less than the preset numerical value, carrying out priority sequencing according to the general intention recognition results of the multiple target text segments after sentence break, and taking the general intention recognition result with the highest priority as the final general intention of the target text. For example, after the target text is split, if the model identification includes a plurality of general intents 81, 71, 63, etc., the final general intention is 81. Specifically, the common intention of the target text is ranked according to the common intention recognition results of the target text segments obtained after sentence break. Because many sentences tend to contain multiple intentions according to past experience, for example, the robot says, "we have a lesson of 19.9, do you not consider? "if the user replies to" kay ", then the general intent is 21, i.e." accept & ok ". If the user replies to "kay, consider" which includes two general intentions, wherein "kay" indicates general intent 21, i.e., "accept & affirm" and "consider" indicates general intent "consider" belonging to callout general intent sequence number 52. If the user replies "kay, consider under, or don't care", then the post-sentence identification will contain three general intentions, "accept & affirm" and "consider under" and "don't care", respectively, belonging to general intent 61, i.e. "don't need". Such three general intents are prioritized with 61 being preferred over 52 over 21, so the final general intent is 61, i.e. "don't need". Therefore, the recognition accuracy rate of one target text containing a plurality of general intents is improved. The determination of the priority can also be adjusted according to different service requirements.

The general purpose recognition model is a plurality of multi-classification models, such as a Bert classification, FastText, TextCNN, RNN _ Attention, Idcnn, and the like. For example, each target text is converted into a word vector through a Bert pre-training model by using a Bert multi-classification model, and besides the word vectors (Token entries), Position vectors (Position entries), and sentence vectors (Segment entries) already included in the Bert model, a word segmentation vector is added to the model as an additional feature, because in general purpose, words such as "unused", "hung", "needed", and the like in the target text are important for the purpose determination of the target text, and the purpose determination accuracy can be improved by adding the word segmentation feature.

Specifically, when the intention type contained in the target text is determined to be the knowledge intention type, the target text is identified through a knowledge intention identification model, and a proper answer is selected from a preset knowledge intention library of knowledge intention. The knowledge intention recognition model used in the embodiment of the application is a classification model, the target text is paired with all intentions to be matched in a preset knowledge intention library one by one, whether the pairing is correct or not is judged by using the classification model, and the knowledge intention which is matched correctly with the maximum probability of the classification model is selected as the final knowledge intention. For example, a predetermined knowledge-meaning library is shown in table 2.

TABLE 2

The knowledge intention recognition model is a Bert classification model with keywords. And inputting the text to be matched with the keywords and the target text into the Bert for classification, wherein the model classification result is to judge whether the input target text and the text to be matched belong to the same knowledge intention. The input of the traditional Bert model is a continuous sentence, a target text and a text to be matched in a preset knowledge and intention library are respectively paired one by one in a classification model of the embodiment of the application, if more than one text to be matched exists under each knowledge intention, the group of texts to be matched with the same knowledge intention in the preset knowledge and intention library are merged, the merged texts are paired with the target text, the matched texts are input into a knowledge and intention identification model after word segmentation and word stop, labeling entities, keywords and the like are carried out, and finally, the knowledge intention of which the classification model is successfully paired at the maximum probability (namely, the maximum matching probability) is selected as the final knowledge intention, so that an intention identification result corresponding to the target text is obtained.

The specific operations of labeling entities, keywords and the like comprise: labeling a text to be matched in a preset knowledge and meaning library according to a first level (entity) and a second level (keyword) in the preset knowledge and meaning library, namely after the text to be matched is divided into words, if the words are entities or keywords, the position of the characters in the words is marked as 1, if the words are not entities or keywords, the position of the characters in the words after word division is marked as 0, and thus obtaining a key position vector consisting of 0 and 1. Similarly, the target text is processed in the same way according to the domain entity dictionary and the domain keyword dictionary, the position of the entity or the keyword is marked as 1 after the word is segmented and the word is stopped, and the positions of other words are marked as 0. And inputting the finally matched preset knowledge intention library and the target texts into a classification model (namely a knowledge intention recognition model) together with the key position vectors of the target texts.

For example, the matching target text and the text to be matched are "when the invoice can be issued" and "can my invoice be issued", respectively, since the invoice, or the like belong to an entity or a keyword, the input after the position identification is: "11000001" and "001100101". And if the knowledge intentions of the paired target text and the text to be matched, which are expressed during the recognition of the knowledge intention recognition model, are the same or different, and if the knowledge intentions are the same, the obtained final knowledge intention of the target text is the knowledge intention corresponding to the paired text to be matched. The key position vector is obtained by marking the entity and the key word position, entity and key word information can be introduced into the model, and the entity and the key word are important characteristics in the knowledge and intention identification process, and are particularly helpful for improving the accuracy of intention identification under spoken language scenes such as overlong target text length, more language words, more repeated words and the like. If the pairing calculation is carried out purely according to the text similarity, the characters with high similarity are classified into the same intention, but the intentions may be different.

Specifically, if it is determined that the target text contains the two intention types, two corresponding intention recognition models are respectively used for intention recognition, and the specific recognition process refers to the foregoing description and is not repeated herein.

In particular, when a robot asks a question in a conversational template setting, the answer to the client will be a punctuation first, since the client's answer following the robot question is likely to both answer the robot's question and ask its own new question in a sentence. The method comprises the steps of carrying out sentence breaking on answers of a client to obtain a plurality of target texts, and then carrying out binary model judgment on the target texts after sentence breaking, wherein the input knowledge intention identification model belongs to knowledge intentions, and the input general intention identification model belongs to general intentions. In this way, in the next sentence, when the robot answers the customer question, the robot can continue to walk the appropriate sentence according to the feedback of the customer. When the robot answers the question of the user instead of the robot question set in the dialect template, the return of the client is directly used as the target text, whether the target text belongs to the general purpose intention or the knowledge intention is judged firstly, and the knowledge intention identification model is directly input if the target text belongs to the knowledge intention without sentence break. If the general intentions belong to the general intentions, the sentences are firstly broken, then the general intentions are input into the general intention recognition model for recognition, and finally the most suitable general intentions are selected as the final general intentions according to the priority.

In summary, as shown in fig. 2, which is a schematic flow chart of an implementation process of the embodiment of the present application, as shown in fig. 2, the method may include the following steps:

firstly, acquiring a previous voice of an outbound platform corresponding to a current voice, and acquiring a corresponding previous text based on the previous voice; if the content of the previous text is the answer content (namely the robot answer) to the client, determining the text corresponding to the current voice as the target text; if the content of the previous text is the question content of the client (namely, the robot question), splitting the text corresponding to the current voice to obtain one or more target texts.

And then, inputting the target text into a pre-trained binary classification model, and outputting an intention type corresponding to the target text. And if the intention type is a general intention type indicating the receiving degree of the client to the outbound marketing object, acquiring an intention recognition result corresponding to the target text by using a pre-trained general intention recognition model. And if the intention type is a knowledge intention type indicating the question of the client about the relevant information of the outbound marketing object, acquiring an intention recognition result corresponding to the target text by using a pre-trained knowledge intention recognition model.

Finally, acquiring corresponding reply content based on the intention recognition result; sending the reply content to the client through the outbound call platform (i.e. giving knowledge base answers, or absolute robotics trends)

Fig. 3 is a block diagram of an intention recognition apparatus according to an embodiment of the present application, and as shown in fig. 3, the apparatus 300 may include a target text obtaining module 301 and an intention recognition result obtaining module 302, where:

the target text acquisition module 301 is configured to acquire a current voice sent by a client in an outbound scene, and acquire a corresponding target text based on the current voice;

the intention recognition result obtaining module 302 is configured to determine an intention type included in the target text, and obtain an intention recognition result corresponding to the target text based on the intention type.

and sending the reply content to the client through the outbound platform.

Referring now to fig. 4, shown is a schematic diagram of an electronic device (e.g., a terminal device or a server that performs the method shown in fig. 1) 400 suitable for implementing embodiments of the present application. The electronic device in the embodiments of the present application may include, but is not limited to, a mobile terminal such as a mobile phone, a notebook computer, a digital broadcast receiver, a PDA (personal digital assistant), a PAD (tablet), a PMP (portable multimedia player), a vehicle terminal (e.g., a car navigation terminal), a wearable device, and the like, and a stationary terminal such as a digital TV, a desktop computer, and the like. The electronic device shown in fig. 4 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present application.

The electronic device includes: a memory for storing a program for executing the method of the above-mentioned method embodiments and a processor; the processor is configured to execute programs stored in the memory. The processor may be referred to as a processing device 401 described below, and the memory may include at least one of a Read Only Memory (ROM)402, a Random Access Memory (RAM)403, and a storage device 408, which are described below:

as shown in fig. 4, electronic device 400 may include a processing device (e.g., central processing unit, graphics processor, etc.) 401 that may perform various appropriate actions and processes in accordance with a program stored in a Read Only Memory (ROM)402 or a program loaded from a storage device 408 into a Random Access Memory (RAM) 403. In the RAM403, various programs and data necessary for the operation of the electronic apparatus 400 are also stored. The processing device 401, the ROM 402, and the RAM403 are connected to each other via a bus 404. An input/output (I/O) interface 405 is also connected to bus 404.

Generally, the following devices may be connected to the I/O interface 405: input devices 406 including, for example, a touch screen, touch pad, keyboard, mouse, camera, microphone, accelerometer, gyroscope, etc.; an output device 407 including, for example, a Liquid Crystal Display (LCD), a speaker, a vibrator, and the like; storage 408 including, for example, tape, hard disk, etc.; and a communication device 409. The communication means 409 may allow the electronic device 400 to communicate wirelessly or by wire with other devices to exchange data. While fig. 4 illustrates an electronic device having various means, it is to be understood that not all illustrated means are required to be implemented or provided. More or fewer devices may alternatively be implemented or provided.

In particular, according to embodiments of the application, the processes described above with reference to the flow diagrams may be implemented as computer software programs. For example, embodiments of the present application include a computer program product comprising a computer program carried on a non-transitory computer readable medium, the computer program containing program code for performing the method illustrated by the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network via the communication device 409, or from the storage device 408, or from the ROM 402. The computer program, when executed by the processing device 401, performs the above-described functions defined in the methods of the embodiments of the present application.

It should be noted that the computer readable storage medium mentioned above in the present application may be a computer readable signal medium or a computer readable storage medium or any combination of the two. A computer 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 of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, 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. In the present application, a computer 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. In this application, however, a computer readable signal medium may include a propagated data signal with computer 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 computer readable signal medium may also be any computer readable medium that is not a computer 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 computer readable medium may be transmitted using any appropriate medium, including but not limited to: electrical wires, optical cables, RF (radio frequency), etc., or any suitable combination of the foregoing.

In some embodiments, the clients, servers may communicate using any currently known or future developed network Protocol, such as HTTP (HyperText Transfer Protocol), and may interconnect with any form or medium of digital data communication (e.g., a communications network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), the Internet (e.g., the Internet), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future developed network.

The computer readable medium may be embodied in the electronic device; or may exist separately without being assembled into the electronic device.

The computer readable medium carries one or more programs which, when executed by the electronic device, cause the electronic device to:

acquiring current voice sent by a client in an outbound scene, and acquiring a corresponding target text based on the current voice; and determining the intention type contained in the target text, and acquiring an intention recognition result corresponding to the target text based on the intention type.

Computer program code for carrying out operations for aspects of the present application may be written in any combination of one or more programming languages, including but not limited to an object oriented programming language such as Java, Smalltalk, C + +, 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 computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).

The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

The modules or units described in the embodiments of the present application may be implemented by software or hardware. The name of a module or a unit does not in some cases constitute a limitation of the unit itself, and for example, the target text acquiring module may also be described as a "module acquiring a target text".

The functions described herein above may be performed, at least in part, by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field Programmable Gate Arrays (FPGAs), Application Specific Integrated Circuits (ASICs), Application Specific Standard Products (ASSPs), systems on a chip (SOCs), Complex Programmable Logic Devices (CPLDs), and the like.

In the context of this application, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, 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.

The apparatus provided in the embodiment of the present application may implement at least one of the modules through an AI model. The functions associated with the AI may be performed by the non-volatile memory, the volatile memory, and the processor.

The processor may include one or more processors. At this time, the one or more processors may be general-purpose processors, such as a Central Processing Unit (CPU), an Application Processor (AP), or the like, or pure graphics processing units, such as a Graphics Processing Unit (GPU), a Vision Processing Unit (VPU), and/or AI-specific processors, such as a Neural Processing Unit (NPU).

The one or more processors control the processing of the input data according to predefined operating rules or Artificial Intelligence (AI) models stored in the non-volatile memory and the volatile memory. Predefined operating rules or artificial intelligence models are provided through training or learning.

Here, the provision by learning means that a predefined operation rule or an AI model having a desired characteristic is obtained by applying a learning algorithm to a plurality of learning data. This learning may be performed in the device itself in which the AI according to the embodiment is performed, and/or may be implemented by a separate server/system.

The AI model may include a plurality of neural network layers. Each layer has a plurality of weight values, and the calculation of one layer is performed by the calculation result of the previous layer and the plurality of weights of the current layer. Examples of neural networks include, but are not limited to, Convolutional Neural Networks (CNNs), Deep Neural Networks (DNNs), Recurrent Neural Networks (RNNs), Restricted Boltzmann Machines (RBMs), Deep Belief Networks (DBNs), Bidirectional Recurrent Deep Neural Networks (BRDNNs), generative confrontation networks (GANs), and deep Q networks.

A learning algorithm is a method of training a predetermined target device (e.g., a robot) using a plurality of learning data to make, allow, or control the target device to make a determination or prediction. Examples of the learning algorithm include, but are not limited to, supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning.

It is clear to those skilled in the art that, for convenience and brevity of description, the specific method implemented by the computer-readable medium described above when executed by the electronic device may refer to the corresponding process in the foregoing method embodiments, and will not be described herein again.

Claims

1. An intent recognition method, comprising:

2. The method of claim 1, wherein the determining the type of intent corresponding to the target text comprises:

3. The method according to claim 1, wherein the obtaining of the intention recognition result corresponding to the target text based on the intention type includes:

4. The method of claim 3, wherein if the intent type is a general intent type indicating a degree of receipt of an outbound marketing object by the customer, the intent recognition model corresponding to the general intent type is a pre-trained general intent recognition model;

the obtaining of the intention recognition result corresponding to the target text by using the pre-trained intention recognition model includes:

5. The method according to claim 4, wherein the obtaining the intention recognition result corresponding to the target text by using the pre-trained general intention recognition model comprises:

respectively inputting each target text segment into the pre-trained general purpose intention recognition model, and outputting corresponding general purpose intents;

6. The method according to claim 5, wherein the obtaining the intention recognition result corresponding to the target text based on the general intention recognition result comprises:

and taking the general intention with the highest priority in all the general intention recognition results as the intention recognition result corresponding to the target text.

7. The method of claim 3, wherein if the intention type is a knowledge intention type indicating the customer asking about information related to an outbound marketing object, the intention recognition model corresponding to the knowledge intention type is a pre-trained knowledge intention recognition model;

8. The method according to claim 7, wherein the obtaining the intention recognition result corresponding to the target text by using the pre-trained knowledge intention recognition model comprises:

inputting the key position vector of the target text and the key position vector of each text to be matched into the pre-trained knowledge intention recognition model, and outputting the matching probability of the target text and each text to be matched;

9. The method according to claim 8, wherein the obtaining the intention recognition result corresponding to the target text based on each matching probability comprises:

and taking the knowledge intention corresponding to the text to be matched with the target text with the maximum matching probability as an intention recognition result corresponding to the target text.

10. The method of claim 3, wherein the intent types include a general intent type and a knowledge intent type, and wherein the intent recognition models corresponding to the general intent type include a pre-trained general intent recognition model and a pre-trained knowledge intent recognition model;

respectively utilizing the pre-trained general intention recognition model and the pre-trained knowledge intention recognition model to obtain a general intention and a knowledge intention corresponding to the target text;

11. The method according to any one of claims 1-10, wherein obtaining corresponding target text based on the current speech comprises:

acquiring a previous voice of the outbound platform corresponding to the current voice, and acquiring a corresponding previous text based on the previous voice;

and if the content of the previous text is the content of the questions of the client, splitting the text corresponding to the current voice to obtain one or more target texts.

12. The method of claim 1, further comprising:

and sending the reply content to the client through the outbound platform.

13. An intention recognition apparatus, comprising:

14. An electronic device comprising a memory and a processor;

the memory has stored therein a computer program;

the processor for executing the computer program to implement the method of any one of claims 1 to 12.

15. A computer-readable storage medium, characterized in that a computer program is stored on the computer-readable storage medium, which computer program, when being executed by a processor, carries out the method of any one of claims 1 to 12.