CN112231474A - Intention recognition method, system, electronic device and storage medium - Google Patents
Intention recognition method, system, electronic device and storage medium Download PDFInfo
- Publication number
- CN112231474A CN112231474A CN202011091700.8A CN202011091700A CN112231474A CN 112231474 A CN112231474 A CN 112231474A CN 202011091700 A CN202011091700 A CN 202011091700A CN 112231474 A CN112231474 A CN 112231474A
- Authority
- CN
- China
- Prior art keywords
- domain
- classification
- interactive text
- result
- recognition
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 62
- 230000002452 interceptive effect Effects 0.000 claims abstract description 105
- 230000009467 reduction Effects 0.000 claims abstract description 31
- 238000012545 processing Methods 0.000 claims abstract description 19
- 238000013145 classification model Methods 0.000 claims description 55
- 238000012549 training Methods 0.000 claims description 30
- 238000007781 pre-processing Methods 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 5
- 238000003058 natural language processing Methods 0.000 abstract description 5
- 230000008569 process Effects 0.000 description 17
- 238000004422 calculation algorithm Methods 0.000 description 9
- 238000013475 authorization Methods 0.000 description 5
- 238000013461 design Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 230000001755 vocal effect Effects 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the invention relates to natural language processing and discloses an intention identification method, an intention identification system, electronic equipment and a storage medium. In the invention, interactive texts to be identified are obtained; performing domain classification on the interactive text to obtain a classification result of the domain to which the interactive text belongs; performing dimension reduction processing on the first domain set comprising all authorized domains according to the classification result to obtain a second domain set related to the classification result; and performing intention recognition on the interactive text according to the second domain set to obtain a recognition result. According to the embodiment of the invention, the interactive text is subjected to field classification, the first field set is subjected to dimensionality reduction according to the classification result to obtain the second field set, the second field set is subjected to intention matching with the interactive text to obtain the recognition result, the recognition time required by intention recognition is reduced, and the intention recognition efficiency is improved.
Description
Technical Field
Embodiments of the present invention relate to natural language processing, and in particular, to an intention recognition method, system, electronic device, and storage medium.
Background
Natural language processing is an important direction in the fields of computer science and artificial intelligence, and aims to realize effective communication between people and computers by using natural language. Intent recognition refers to a natural language processing technique that performs intent matching on interactive text and extracts part of key information.
The related intention identification method performs rule and vocabulary matching processing on each authorization field one by traversing all the authorization fields. With the continuous increase of authorization fields, the dimension of the matching behavior which needs to be traversed is also increased, and the identification needed by the intention identification method is also increased.
Therefore, the related intention identifying method has the following problems: when the authorized fields are more, the identification time is longer, and the identification speed is slower.
Disclosure of Invention
The embodiment of the invention aims to provide an intention identification method, so that the identification time required when more authorization fields exist is reduced in the intention identification process, and the identification efficiency is improved.
In order to solve the above technical problem, an embodiment of the present invention provides an intention identifying method, including: acquiring an interactive text to be identified; performing domain classification on the interactive text to obtain a classification result of the domain to which the interactive text belongs; performing dimension reduction processing on the first domain set comprising all authorized domains according to the classification result to obtain a second domain set related to the classification result; the number of the domains contained in the second domain set is smaller than that contained in the first domain set; and performing intention recognition on the interactive text according to the second domain set to obtain a recognition result.
Embodiments of the present invention also provide an intention recognition system, including: the acquisition module is used for acquiring the interactive text to be identified; the preprocessing module is used for carrying out domain classification on the interactive text to obtain a classification result of the domain to which the interactive text belongs; performing dimension reduction processing on the first domain set comprising all authorized domains according to the classification result to obtain a second domain set related to the classification result; the number of the domains contained in the second domain set is smaller than that of the domains contained in the first domain set; and the recognition module is used for performing intention recognition on the interactive text according to the second domain set to obtain a recognition result.
An embodiment of the present invention also provides an electronic device, including: at least one processor; a memory communicatively coupled to the at least one processor; the memory stores instructions executable by the at least one processor to enable the at least one processor to perform any of the intent recognition methods described above.
Embodiments of the present invention also provide a computer-readable storage medium storing a computer program, which when executed by a processor implements any of the above-described intention identifying methods.
In the embodiment of the invention, the interactive text is subjected to field classification, the first field set is subjected to dimensionality reduction according to the classification result to obtain the second field set, the second field set is subjected to intent matching with the interactive text to obtain the identification result, and because the second field set is obtained by dimensionality reduction of the first field set, the dimensionality required for traversal in matching of the second field set and the interactive text is reduced, the required identification time is reduced, and the intent identification efficiency is improved.
In addition, the method for classifying the interactive text to obtain the classification result of the interactive text belonging to the field comprises the following steps: inputting the interactive text into a pre-trained domain classification model to obtain a classification result output by the domain classification model; and the domain classification model is obtained by training according to historical experience data. In this embodiment, the classification result is obtained by inputting the interactive text into the domain classification model obtained by pre-training, and the training process of the domain classification model can be separated from the process of the domain classification model, so that the use of the model is not affected by the training of the domain classification model, and the intention recognition efficiency is improved.
In addition, after the intention recognition is performed on the interactive text according to the second domain set and the recognition result is obtained, the method further comprises the following steps: storing the interactive text and the recognition result, and using the interactive text and the recognition result as supplementary training data of the domain classification model; updating the domain classification model according to the supplementary training data; in the embodiment, the interactive text and the recognition result are stored and used as the supplementary training data, and the domain classification model is updated according to the supplementary training data, so that the historical recognition result can be used as a reference in the classification process, and the calculation result is closer to the actual recognition result after the domain classification model is trained, thereby improving the recognition accuracy.
In addition, according to the classification result, the dimension reduction processing is carried out on the first domain set including all authorized domains, so as to obtain a second domain set related to the classification result, and the method comprises the following steps: inputting the classification result and the first domain set into a domain dimension reduction model to obtain an output second domain set; the second domain set is a set of authorized domains in the first domain set that are closest to the classification result. In the embodiment, the dimension of the first field set is reduced through the field dimension reduction model, the field dimension needing to be traversed in intention matching is reduced, and the intention identification efficiency is improved.
Drawings
One or more embodiments are illustrated by way of example in the accompanying drawings, which correspond to the figures in which like reference numerals refer to similar elements and which are not to scale unless otherwise specified.
Fig. 1 is a flowchart of an example of an intention recognition method provided according to a first embodiment of the present invention;
fig. 2 is a flowchart of another example of an intention recognition method provided according to the first embodiment of the present invention;
FIG. 3 is a flowchart of an example of an intention recognition method provided according to a second embodiment of the present invention;
fig. 4 is a schematic structural diagram of an example of an intention recognition system provided according to a third embodiment of the present invention;
fig. 5 is a schematic structural view of another example of the intention recognition system provided according to the third embodiment of the invention;
fig. 6 is a schematic structural diagram of yet another example of the intention recognition system provided in accordance with the third embodiment of the present invention;
fig. 7 is a schematic structural diagram of an electronic device according to a fourth embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention more apparent, embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, it will be appreciated by those of ordinary skill in the art that numerous technical details are set forth in order to provide a better understanding of the present application in various embodiments of the present invention. However, the technical solution claimed in the present application can be implemented without these technical details and various changes and modifications based on the following embodiments. The following embodiments are divided for convenience of description, and should not constitute any limitation to the specific implementation manner of the present invention, and the embodiments may be mutually incorporated and referred to without contradiction.
A first embodiment of the present invention relates to an intention identifying method. The specific process is shown in fig. 1, and comprises the following steps:
102, performing domain classification on the interactive text to obtain a classification result of the domain to which the interactive text belongs;
103, performing dimension reduction processing on a first domain set comprising all authorized domains according to the classification result to obtain a second domain set related to the classification result; the number of the domains contained in the second domain set is smaller than that contained in the first domain set;
and 104, performing intention recognition on the interactive text according to the second domain set to obtain a recognition result.
In the embodiment, the interactive text is subjected to field classification, the first field set is subjected to dimensionality reduction according to the classification result to obtain the second field set, the second field set and the interactive text are subjected to intention matching to obtain the identification result, and the second field set is obtained by dimensionality reduction of the first field set, so that the number of dimensions required to be traversed by matching the second field set and the interactive text is reduced, the required identification time is shortened, and the intention identification efficiency is improved.
The intention recognition method in this embodiment is used for matching and recognizing the intention of an interactive statement in natural language processing, and for example, voice-type intelligent devices such as a smart phone, a smart speaker, a smart television, and a smart set-top box, voice interaction can be performed based on the intention recognition method in this embodiment. The device can recognize the voice information of the user by the intention recognition method in the embodiment of the application according to the voice information input by the user, so that the user intention in the voice information is recognized, and corresponding service is provided for the user according to the user intention.
Implementation details of the intention recognition method of the present embodiment are specifically described below, and the following description is provided only for the sake of understanding and is not necessary for implementing the present embodiment.
In step 101, the interactive text to be recognized is obtained, which may be directly obtained from the interactive text input by the user, or may be obtained from the interactive voice information of the user, the interactive voice of the user is recognized by using an automatic voice recognition technology, and the voice stream data is converted into text data to obtain the interactive text.
In step 102, the interactive text is subjected to domain classification, the interactive text is classified into the domain to which the interactive text is most likely to belong, and a classification result of the domain to which the interactive text belongs is obtained, wherein the domains which can be used for classification in the domain classification are all authorized domains of the device, namely, a first domain set, and the classification result can be one domain or a plurality of domains. For example, for a smart speaker device that can interact with a user by voice, the authorized fields are: music, audio books, speeches, pre-sleep stories, TV shows, radio shows, news, home, mother and baby, entertainment, science, education, that is, the first domain set of the smart speaker device includes the above domains, and when the interactive sentence of the user is "i want to listen to the western notes", the classification result for the sentence may be one domain of "audio books", or may be a plurality of domains of "music" and "audio books".
In step 103, the dimension reduction processing is performed on the first domain set according to the classification result, and the obtained second domain set may be a set of a plurality of domains closest to the classification result. When the classification result includes a plurality of domains, the plurality of domains in the second domain set may be a plurality of authorized domains that are respectively closest to a certain domain in the classification result, or may be a plurality of authorized domains that are close to all domains in the classification result. Wherein, the authorization field is a field in which the device for intention identification is authorized to perform intention identification. The first set of realms and the second set of realms may be stored in a structure of a list. For example, when the classification result is the "voiced book" domain, the resulting second domain set may be the "voiced book", "say", "drama", "music" domain closest to the "voiced book" domain; when the classification result is the "music" field and the "audio book" field, the obtained second domain set may be the "music" and "drama" fields closest to the "music" field, or the "music", "audio book", and "pre-sleep story" fields closest to the "music" field and the "audio book" field.
In step 104, performing intent recognition on the interactive text according to the second domain set, and performing intent recognition through a regular expression word slot filling method, a characteristic character progressive intent deduction method or a combination of the two methods to obtain a recognition result. When the intention is identified, all the fields contained in the second field set need to be traversed, and each field is matched one by one.
In one example, as shown in fig. 2, includes:
and step 205, performing intention recognition on the interactive text according to the second domain set to obtain a recognition result.
In step 202, the first threshold is a preset value, and different first thresholds may be preset for different devices according to their uses. When the number of domains included in the first domain set is less than or equal to the first threshold, the apparatus does not perform step 203 and step 204, and directly performs step 205. That is, after obtaining the interactive text to be recognized (in step 201), it is necessary to determine whether the number of fields included in the first field set is greater than a first threshold, and perform field classification on the interactive text if the number of fields included in the first field set is greater than the first threshold to obtain a classification result of the field to which the interactive text belongs (in step 203), and perform dimension reduction processing on the first field set including all authorized fields according to the classification result to obtain a second field set related to the classification result; wherein, the number of domains contained in the second domain set is smaller than the number of domains contained in the first domain set (step 204), and the intention recognition is carried out on the interactive text according to the second domain set to obtain a recognition result (step 205).
In this embodiment, only when the number of fields included in the first field set is greater than the first threshold, the fields are classified and the dimensions of the first field set are reduced, because the field dimensions are reduced to reduce the number of fields to be traversed when performing the intent recognition, after the dimensions are reduced to obtain the second field set, the fields included in the second field set still need to be traversed, and the fields are matched one by one.
In one example, the number of domains included in the second domain set is equal to or less than the second threshold. The second threshold is a preset value, and different devices may be used differently, or different second thresholds may be preset in consideration of accuracy of identification.
The intention identification method in this embodiment may be used, for example, in an intelligent sound box, when an interactive statement of a user is "i want to listen to the westernist", the sound box converts the interactive statement of the user into an interactive text through voice processing, the sound box first determines the number of authorized fields, and when the number of fields included in the first field set is less than or equal to a first threshold, the intention identification is directly performed according to the interactive text and the first field set. If the number of the fields contained in the first field set is larger than a first threshold value, carrying out field classification on the interactive text 'i want to listen to the western journey', and assuming that the interactive sentence of the user most possibly corresponds to the 'vocal book' field. The device reduces the dimension of the first domain set containing 1000 domains according to the 'vocal book' domain and the first domain set to obtain a second domain set only containing four domains of 'music', 'before-sleep story', 'talking book' and 'vocal book'. And the equipment performs intention identification according to the interactive text 'I wants to listen to the West-run notes' and the second field set to obtain an identification result, and provides corresponding service for the user according to the identification result.
The steps of the above methods are divided for clarity, and the implementation may be combined into one step or split some steps, and the steps are divided into multiple steps, so long as the same logical relationship is included, which are all within the protection scope of the present patent; it is within the scope of the patent to add insignificant modifications to the algorithms or processes or to introduce insignificant design changes to the core design without changing the algorithms or processes.
A second embodiment of the present invention relates to an intention identifying method, as shown in fig. 3, including:
303, performing dimension reduction processing on a first domain set comprising all authorized domains according to the classification result to obtain a second domain set related to the classification result; the number of the domains contained in the second domain set is smaller than that contained in the first domain set;
and 304, performing intention recognition on the interactive text according to the second domain set to obtain a recognition result.
In step 302, a domain classification of the interactive text is rapidly performed by using a domain classification model trained in advance, including: and performing domain classification by using a convolutional neural network model or other machine learning models. Preferably, the interactive text may be classified using a text classification model TextCNN. The TextCNN model comprises a layer of convolution and a layer of pooling layer max-posing, and finally is connected with a softmax function to realize multi-domain classification. And the TextCNN model carries out vectorization processing on the interactive text, constructs word vectors, and obtains an output classification result through calculation of the convolution layer, the pooling layer and the softmax function. The TextCNN model in this embodiment is obtained by training according to historical empirical data, and when the device performs domain classification, the interactive text is input into the trained domain classification model to obtain a result output by the domain classification model, that is, a classification result. It can be seen that the process of training the model and using the model is not necessarily connected, the training process of the field classification model is separated from the using process of the field classification model, and the field classification model can respectively and independently complete the training process and the using process, so that the use of the classification model is not influenced when the classification model is trained.
In this embodiment, the classification result is obtained by inputting the interactive text into the domain classification model obtained by pre-training, and the training process of the domain classification model can be separated from the process of the domain classification model, so that the use of the model is not affected by the training of the domain classification model, and the intention recognition efficiency is improved.
In one example, after performing intent recognition on the interactive text according to the second domain set to obtain a recognition result, the method further includes:
storing the interactive text and the recognition result, and using the interactive text and the recognition result as supplementary training data of the domain classification model;
and updating the domain classification model according to the supplementary training data.
In practical use, the two steps occur after the step of 'performing intention recognition on the interactive text according to the second domain set to obtain a recognition result' for the nth intention recognition and before the same step of the N +1 th intention recognition.
In the embodiment, the interactive text and the recognition result are stored and used as the supplementary training data, and the domain classification model is updated according to the supplementary training data, so that the historical recognition result can be used as a reference in the classification process, and the calculation result is closer to the actual recognition result after the domain classification model is trained, thereby improving the recognition accuracy.
Furthermore, in order to avoid the problem that the updating frequency is too high, the effect difference of the domain classification model before and after each updating is small, so that the resource waste is caused, the domain classification model can be set to perform supplementary training on the domain classification model according to the historical interactive text and the recognition result at regular time, and the currently used domain classification model is updated. The time length of the timing update can be correspondingly set according to different purposes of the equipment or other factors.
In one example, step 303 comprises: inputting the classification result and the first domain set into a domain dimension reduction model to obtain an output second domain set; the second domain set is a set of authorized domains in the first domain set that are closest to the classification result.
The domain dimension reduction model is a model capable of performing classification, and preferably, the domain dimension reduction model can be a supervised or unsupervised machine learning classification model, for example, a proximity algorithm KNN, a clustering algorithm, a support vector machine model SVM algorithm, and the like, and the specific algorithm can be selected correspondingly according to the use purpose of different devices. And selecting a domain which is closest to or has the highest score with the domain of the interactive text classification result from the first domain set through an algorithm model as a domain contained in the second domain set. And the number of the domains contained in the second domain set is less than or equal to a second threshold value.
In the embodiment, the dimension of the first field set is reduced through the field dimension reduction model, the field dimension needing to be traversed in intention matching is reduced, and the intention identification efficiency is improved.
The intention recognition method in this embodiment may be used, for example, in an intelligent sound box, and when an interactive sentence of a user is "i want to listen to the western shorthand", the sound box converts the interactive sentence of the user into an interactive text through voice processing. And if the number of the fields contained in the first field set is larger than a first threshold value, inputting an interactive text 'i want to listen to the West travel notes' into a field classification model for field classification, judging that the interactive sentence of the user most possibly corresponds to the field of the 'voiced book' after the field classification model is trained according to the past recognition result of the user, and outputting the field of the 'voiced book' as a classification result. The device inputs the vocal book field and the first field set into a field dimension reduction model, the field dimension reduction model reduces the dimension of the first field set containing 1000 fields, and outputs a second field set containing only three fields of music, a before-sleep story and a book. And the equipment performs intention identification according to the interactive text 'I wants to listen to the West-run notes' and the second field set to obtain an identification result, and provides corresponding service for the user according to the identification result.
The steps of the above methods are divided for clarity, and the implementation may be combined into one step or split some steps, and the steps are divided into multiple steps, so long as the same logical relationship is included, which are all within the protection scope of the present patent; it is within the scope of the patent to add insignificant modifications to the algorithms or processes or to introduce insignificant design changes to the core design without changing the algorithms or processes.
A third embodiment of the present invention relates to an intention recognition system, as shown in fig. 4, including:
an obtaining module 401, configured to obtain an interactive text to be identified;
the preprocessing module 402 is configured to perform domain classification on the interactive text to obtain a classification result of a domain to which the interactive text belongs, and perform dimension reduction processing on a first domain set including all authorized domains according to the classification result to obtain a second domain set related to the classification result; the number of the domains contained in the second domain set is smaller than that of the domains contained in the first domain set;
and the identifying module 403 is configured to perform intent identification on the interactive text according to the second domain set to obtain an identification result.
In one example, as shown in FIG. 5, the pre-processing module 402 includes:
classification submodule 4021: the interactive text input device is used for inputting the interactive text into a pre-trained domain classification model to obtain a classification result output by the domain classification model; the domain classification model is obtained by training according to historical experience data;
dimension reduction submodule 4022: and the system is used for performing dimension reduction processing on the first domain set comprising all authorized domains according to the classification result output by the domain classification model to obtain a second domain set related to the classification result.
In one example, as shown in fig. 6, the preprocessing module 402 further includes:
offline submodule 4023: and the interactive text and the recognition result are stored and used as supplementary training data of the domain classification model, and the domain classification model is updated according to the supplementary training data.
It should be understood that this embodiment is a system example corresponding to the above embodiment, and that this embodiment can be implemented in cooperation with the above embodiment. The related technical details mentioned in the above embodiments are still valid in this embodiment, and are not described herein again in order to reduce repetition. Accordingly, the related-art details mentioned in the present embodiment can also be applied to the above-described embodiments.
It should be noted that each module referred to in this embodiment is a logical module, and in practical applications, one logical unit may be one physical unit, may be a part of one physical unit, and may be implemented by a combination of multiple physical units. In addition, in order to highlight the innovative part of the present invention, elements that are not so closely related to solving the technical problems proposed by the present invention are not introduced in the present embodiment, but this does not indicate that other elements are not present in the present embodiment.
A fourth embodiment of the present invention relates to an electronic apparatus, as shown in fig. 7, including:
at least one processor 701; a memory 702 communicatively coupled to the at least one processor; the memory 702 stores instructions executable by the at least one processor 701, and the instructions are executed by the at least one processor 701 to perform the intent recognition method.
The memory 702 and the processor 701 are coupled by a bus, which may comprise any number of interconnecting buses and bridges that couple one or more of the various circuits of the processor 701 and the memory 702. The bus may also connect various other circuits such as peripherals, voltage regulators, power management circuits, and the like, which are well known in the art, and therefore, will not be described any further herein. A bus interface provides an interface between the bus and the transceiver. The transceiver may be one element or a plurality of elements, such as a plurality of receivers and transmitters, providing a means for communicating with various other apparatus over a transmission medium. Information processed by processor 701 is transmitted over a wireless medium through an antenna, which receives the information and passes the information to processor 701.
The processor 701 is responsible for managing the bus and general processing and may also provide various functions including timing, peripheral interfaces, voltage regulation, power management, and other control functions. And memory 702 may be used to store information used by the processor in performing operations.
A fifth embodiment of the present invention relates to a computer-readable storage medium storing a computer program. The computer program realizes the above-described method embodiments when executed by a processor.
That is, as can be understood by those skilled in the art, all or part of the steps in the method for implementing the embodiments described above may be implemented by a program instructing related hardware, where the program is stored in a storage medium and includes several instructions to enable a device (which may be a single chip, a chip, or the like) or a processor (processor) to execute all or part of the steps of the method described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
It will be understood by those of ordinary skill in the art that the foregoing embodiments are specific examples for carrying out the invention, and that various changes in form and details may be made therein without departing from the spirit and scope of the invention in practice.
Claims (10)
1. An intent recognition method, comprising:
acquiring an interactive text to be identified;
performing domain classification on the interactive text to obtain a classification result of the domain to which the interactive text belongs;
performing dimension reduction processing on a first domain set comprising all authorized domains according to the classification result to obtain a second domain set related to the classification result; wherein the number of domains contained in the second domain set is less than the number of domains contained in the first domain set;
and performing intention recognition on the interactive text according to the second domain set to obtain a recognition result.
2. The method for recognizing the intention according to claim 1, wherein the domain classification of the interactive text to obtain the classification result of the domain to which the interactive text belongs comprises:
inputting the interactive text into a pre-trained domain classification model to obtain a classification result output by the domain classification model;
the domain classification model is obtained by training according to historical experience data.
3. The method of claim 2, wherein after the performing intent recognition on the interactive text according to the second domain set to obtain a recognition result, the method further comprises: storing the interactive text and the recognition result as supplementary training data of the domain classification model;
and updating the domain classification model according to the supplementary training data.
4. The method for identifying intent according to claim 1, wherein the performing dimension reduction on a first domain set including all authorized domains according to the classification result to obtain a second domain set related to the classification result comprises:
inputting the classification result and the first domain set into a domain dimension reduction model to obtain an output second domain set;
the second set of realms is a set of authorized realms in the first set of realms that are closest to the classification result.
5. The method of claim 1, wherein before the domain classification of the interactive text is performed to obtain a classification result of a domain to which the interactive text belongs, the method further comprises:
it is confirmed that the number of domains included in the first domain set is greater than the first threshold.
6. An intent recognition system, comprising:
the acquisition module is used for acquiring the interactive text to be identified;
the preprocessing module is used for carrying out domain classification on the interactive text to obtain a classification result of the domain to which the interactive text belongs, and carrying out dimension reduction processing on a first domain set comprising all authorized domains according to the classification result to obtain a second domain set related to the classification result; wherein the number of domains included in the second domain set is less than the number of domains included in the first domain set;
and the recognition module is used for performing intention recognition on the interactive text according to the second domain set to obtain a recognition result.
7. The intent recognition system of claim 6, wherein the preprocessing module comprises:
the classification submodule is used for inputting the interactive text into a pre-trained domain classification model to obtain a classification result output by the domain classification model; the domain classification model is obtained by training according to historical experience data;
and the dimension reduction submodule is used for carrying out dimension reduction processing on the first field set comprising all authorized fields according to the classification result output by the field classification model to obtain a second field set related to the classification result.
8. The intent recognition system of claim 7, wherein the preprocessing module further comprises:
an off-line submodule: and the interactive text and the recognition result are stored and used as supplementary training data of the domain classification model, and the domain classification model is updated according to the supplementary training data.
9. An electronic device, comprising:
at least one processor;
a memory communicatively coupled to the at least one processor;
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the intent recognition method of any of claims 1-5.
10. A computer-readable storage medium, storing a computer program, characterized in that the computer program, when executed by a processor, implements the intent recognition method of any of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011091700.8A CN112231474A (en) | 2020-10-13 | 2020-10-13 | Intention recognition method, system, electronic device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011091700.8A CN112231474A (en) | 2020-10-13 | 2020-10-13 | Intention recognition method, system, electronic device and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112231474A true CN112231474A (en) | 2021-01-15 |
Family
ID=74112529
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011091700.8A Pending CN112231474A (en) | 2020-10-13 | 2020-10-13 | Intention recognition method, system, electronic device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112231474A (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110347789A (en) * | 2019-06-14 | 2019-10-18 | 平安科技（深圳）有限公司 | Text is intended to intelligent method for classifying, device and computer readable storage medium |
| CN111161726A (en) * | 2019-12-24 | 2020-05-15 | 广州索答信息科技有限公司 | Intelligent voice interaction method, equipment, medium and system |
| CN111325037A (en) * | 2020-03-05 | 2020-06-23 | 苏宁云计算有限公司 | Text intention recognition method and device, computer equipment and storage medium |
| CN111597320A (en) * | 2020-05-26 | 2020-08-28 | 成都晓多科技有限公司 | Intention recognition device, method, equipment and storage medium based on hierarchical classification |
| CN111767384A (en) * | 2020-07-08 | 2020-10-13 | 上海风秩科技有限公司 | Man-machine conversation processing method, device, equipment and storage medium |
-
2020
- 2020-10-13 CN CN202011091700.8A patent/CN112231474A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110347789A (en) * | 2019-06-14 | 2019-10-18 | 平安科技（深圳）有限公司 | Text is intended to intelligent method for classifying, device and computer readable storage medium |
| CN111161726A (en) * | 2019-12-24 | 2020-05-15 | 广州索答信息科技有限公司 | Intelligent voice interaction method, equipment, medium and system |
| CN111325037A (en) * | 2020-03-05 | 2020-06-23 | 苏宁云计算有限公司 | Text intention recognition method and device, computer equipment and storage medium |
| CN111597320A (en) * | 2020-05-26 | 2020-08-28 | 成都晓多科技有限公司 | Intention recognition device, method, equipment and storage medium based on hierarchical classification |
| CN111767384A (en) * | 2020-07-08 | 2020-10-13 | 上海风秩科技有限公司 | Man-machine conversation processing method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111191016B (en) | Multi-round dialogue processing method and device and computing equipment | |
| US20230186912A1 (en) | Speech recognition method, apparatus and device, and storage medium | |
| WO2022057712A1 (en) | Electronic device and semantic parsing method therefor, medium, and human-machine dialog system | |
| CN112435656B (en) | Model training method, voice recognition method, device, equipment and storage medium | |
| CN112100349A (en) | Multi-turn dialogue method and device, electronic equipment and storage medium | |
| CN110853626B (en) | Bidirectional attention neural network-based dialogue understanding method, device and equipment | |
| WO2018039500A1 (en) | A speech recognition method and apparatus | |
| CN108364650B (en) | Device and method for adjusting voice recognition result | |
| Deena et al. | Recurrent neural network language model adaptation for multi-genre broadcast speech recognition and alignment | |
| CN106844587B (en) | It is a kind of for talking with the data processing method and device of interactive system | |
| Renkens | Capsule networks for low resource spoken language understanding | |
| CN113505198B (en) | Keyword-driven generation type dialogue reply method and device and electronic equipment | |
| CN111241814A (en) | Error correction method and device for voice recognition text, electronic equipment and storage medium | |
| CN114596844A (en) | Acoustic model training method, voice recognition method and related equipment | |
| CN116484879A (en) | Prompt message generation method and device, electronic equipment and storage medium | |
| CN112825114A (en) | Semantic recognition method and device, electronic equipment and storage medium | |
| CN118043885A (en) | Contrast twin network for semi-supervised speech recognition | |
| CN111274412A (en) | Information extraction method, information extraction model training device and storage medium | |
| CN115376495A (en) | Speech recognition model training method, speech recognition method and device | |
| CN111862961A (en) | Method and device for recognizing voice | |
| JP2021081713A (en) | Method, device, apparatus, and media for processing voice signal | |
| CN111625629A (en) | Task-based conversational robot response method, device, robot and storage medium | |
| CN111508481B (en) | Training method and device of voice awakening model, electronic equipment and storage medium | |
| CN113823265A (en) | Voice recognition method and device and computer equipment | |
| CN112257432A (en) | Self-adaptive intention identification method and device and electronic equipment |
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 |