CN117648440A - Intention classification method, device, equipment and medium based on multi-channel aggregation - Google Patents
Intention classification method, device, equipment and medium based on multi-channel aggregation Download PDFInfo
- Publication number
- CN117648440A CN117648440A CN202311613675.9A CN202311613675A CN117648440A CN 117648440 A CN117648440 A CN 117648440A CN 202311613675 A CN202311613675 A CN 202311613675A CN 117648440 A CN117648440 A CN 117648440A
- Authority
- CN
- China
- Prior art keywords
- intention
- matching
- recommended
- category
- recommendation
- 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
- 230000002776 aggregation Effects 0.000 title claims abstract description 62
- 238000004220 aggregation Methods 0.000 title claims abstract description 62
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000012545 processing Methods 0.000 claims abstract description 17
- 239000013598 vector Substances 0.000 claims description 33
- 238000004590 computer program Methods 0.000 claims description 18
- 238000000605 extraction Methods 0.000 claims description 16
- 230000004931 aggregating effect Effects 0.000 claims description 6
- 238000013473 artificial intelligence Methods 0.000 abstract description 7
- 238000005065 mining Methods 0.000 abstract description 3
- 238000005516 engineering process Methods 0.000 description 9
- 230000008569 process Effects 0.000 description 8
- 230000011218 segmentation Effects 0.000 description 7
- 238000000926 separation method Methods 0.000 description 6
- 230000002457 bidirectional effect Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000010606 normalization Methods 0.000 description 4
- 230000009467 reduction Effects 0.000 description 4
- 230000004044 response Effects 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000001514 detection method Methods 0.000 description 3
- RWSOTUBLDIXVET-UHFFFAOYSA-N Dihydrogen sulfide Chemical compound S RWSOTUBLDIXVET-UHFFFAOYSA-N 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 description 2
- 239000010931 gold Substances 0.000 description 2
- 229910052737 gold Inorganic materials 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000003058 natural language processing Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The invention relates to the technical field of artificial intelligence, in particular to an intention classification method, device, equipment and medium based on multi-channel aggregation. Acquiring N recommendation channels in which an object to be recommended is located, determining content information of the object to be recommended recorded in the recommendation channels, extracting keywords from the content information to obtain keywords of the recommendation channels, selecting a matching intention class set matched with the keywords from preset intention classes, carrying out aggregation processing on the matching intention classes in the N matching intention class sets to obtain an intention aggregation result, carrying out scoring processing on the matching intention classes in the intention aggregation result to obtain a scoring result in each matching intention class, and determining a target intention class of the object to be recommended according to the scoring result. The intention category of the user to be recommended in the single recommendation channel is extracted, then the intention category in the multiple recommendation channels is aggregated, and the intention mining classification is performed again according to the aggregated result, so that the precision of the intention classification is improved.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to an intention classification method, device, equipment and medium based on multi-channel aggregation.
Background
In the financial field, when financial institutions such as banks, securities and insurance popularize in financial products or other businesses, telephone sales is a common promotion mode, in the prior art, an outbound person of a telephone center contacts a customer through a telephone, but because the content records fed back by the outbound person to the customer are not uniform, the real intention category of the customer cannot be obtained, the next contact customer cannot be effectively adjusted, the customer is disturbed many times, the customer satisfaction degree is reduced, and the performance is reduced, so that how to improve the classification precision of the customer intention becomes a problem to be solved urgently.
Disclosure of Invention
Based on the above, it is necessary to provide a method, a device and a medium for classifying intention based on multi-channel aggregation, so as to solve the problem of low precision of intention of clients.
A first aspect of an embodiment of the present application provides an intent classification method based on multi-channel aggregation, the intent classification method including:
Acquiring an object to be recommended of an insurance product to be recommended, extracting image information of the object to be recommended, and determining N recommendation channels in which the object to be recommended is positioned according to the image information, wherein N is an integer greater than 1;
determining content information of the object to be recommended recorded by the recommendation channel aiming at any recommendation channel, extracting keywords from the content information to obtain keywords of the recommendation channel, and selecting a matching intention category set matched with the keywords from preset intention categories;
traversing all recommendation channels, determining N matching intention class sets of the N recommendation channels, and carrying out aggregation treatment on the matching intention classes in the N matching intention class sets to obtain an intention aggregation result;
and scoring the matching intention categories in the intention aggregation result to obtain scoring results in each matching intention category, and determining the target intention category of the object to be recommended according to the scoring results.
A second aspect of embodiments of the present application provides an intent classification device based on multi-channel aggregation, the intent classification device including:
the acquisition module is used for acquiring an object to be recommended of an insurance product to be recommended, extracting image information of the object to be recommended, and determining N recommendation channels in which the object to be recommended is positioned according to the image information, wherein N is an integer greater than 1;
The extraction module is used for determining content information of the object to be recommended recorded by the recommendation channel aiming at any recommendation channel, extracting keywords from the content information to obtain keywords of the recommendation channel, and selecting a matching intention class set matched with the keywords from preset intention classes;
the aggregation module is used for traversing all the recommendation channels, determining N matching intention class sets of the N recommendation channels, and carrying out aggregation processing on the matching intention classes in the N matching intention class sets to obtain an intention aggregation result;
the scoring module is used for scoring the matching intention categories in the intention aggregation result to obtain a scoring result in each matching intention category, and determining the target intention category of the object to be recommended according to the scoring result.
In a third aspect, an embodiment of the present invention provides a terminal device, where the terminal device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, and where the processor implements the method for classifying intention according to the first aspect when executing the computer program.
In a fourth aspect, embodiments of the present invention provide a computer readable storage medium storing a computer program which, when executed by a processor, implements the method of intent classification as described in the first aspect.
Compared with the prior art, the invention has the beneficial effects that:
obtaining an object to be recommended of an insurance product to be recommended, extracting image information of the object to be recommended, determining N recommendation channels in which the object to be recommended is located according to the image information, wherein N is an integer greater than 1, determining content information of the object to be recommended recorded by any recommendation channel, extracting keywords from the content information to obtain keywords of the recommendation channels, selecting a matching intention class set matched with the keywords from preset intention classes, traversing all the recommendation channels, determining N matching intention class sets of the N recommendation channels, performing aggregation processing on the matching intention classes in the N matching intention class sets to obtain an intention aggregation result, performing scoring processing on the matching intention classes in the intention aggregation result to obtain scoring results in each matching intention class, and determining a target intention class of the object to be recommended according to the scoring result. In the method, the intention category of the user to be recommended in the single recommendation channel is extracted, then the intention category in the multiple recommendation channels is aggregated, and the intention mining classification is performed again according to the aggregated result, so that the precision of the intention classification is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments of the present invention will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic view of an application environment of an intent classification method based on multi-channel aggregation according to an embodiment of the present invention;
FIG. 2 is a flow chart of an intent classification method based on multi-channel aggregation according to an embodiment of the present invention;
FIG. 3 is a schematic structural diagram of an intent classification device based on multi-channel aggregation according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a terminal device according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be understood that the terms "comprises" and/or "comprising," when used in this specification and the appended claims, 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 should also be understood that the term "and/or" as used in the present specification and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes such combinations.
As used in the present description and the appended claims, the term "if" may be interpreted as "when..once" or "in response to a determination" or "in response to detection" depending on the context. Similarly, the phrase "if a determination" or "if a [ described condition or event ] is detected" may be interpreted in the context of meaning "upon determination" or "in response to determination" or "upon detection of a [ described condition or event ]" or "in response to detection of a [ described condition or event ]".
Furthermore, the terms "first," "second," "third," and the like in the description of the present specification and in the appended claims, are used for distinguishing between descriptions and not necessarily for indicating or implying a relative importance.
Reference in the specification to "one embodiment" or "some embodiments" or the like means that a particular feature, structure, or characteristic described in connection with the embodiment is included in one or more embodiments of the invention. Thus, appearances of the phrases "in one embodiment," "in some embodiments," "in other embodiments," and the like in the specification are not necessarily all referring to the same embodiment, but mean "one or more but not all embodiments" unless expressly specified otherwise. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless expressly specified otherwise.
The embodiment of the invention can acquire and process the related data based on the artificial intelligence technology. Among these, artificial intelligence (Artificial Intelligence, AI) is the theory, method, technique and application system that uses a digital computer or a digital computer-controlled machine to simulate, extend and extend human intelligence, sense the environment, acquire knowledge and use knowledge to obtain optimal results.
Artificial intelligence infrastructure technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a robot technology, a biological recognition technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and other directions.
It should be understood that the sequence numbers of the steps in the following embodiments do not mean the order of execution, and the execution order of the processes should be determined by the functions and the internal logic, and should not be construed as limiting the implementation process of the embodiments of the present invention.
In order to illustrate the technical scheme of the invention, the following description is made by specific examples.
The intent classification method based on multi-channel aggregation provided by the embodiment of the invention can be applied to an application environment as shown in fig. 1, wherein a client communicates with a server. The client includes, but is not limited to, a handheld computer, a desktop computer, a notebook computer, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (personal digital assistant, PDA), and other terminal devices. The server may be implemented as a stand-alone server or as a cluster of servers generated by multiple servers.
Referring to fig. 2, a flow chart of a multi-channel aggregation-based intent classification method according to an embodiment of the present invention is provided, where the multi-channel aggregation-based intent classification method may be applied to a server in fig. 1, and the server is connected to a corresponding client, as shown in fig. 2, and the multi-channel aggregation-based intent classification method may include the following steps.
S201: acquiring an object to be recommended of an insurance product to be recommended, extracting image information of the object to be recommended, and determining N recommendation channels in which the object to be recommended is positioned according to the image information, wherein N is an integer greater than 1.
In step S201, an object to be recommended of an insurance product to be recommended is obtained, wherein the object to be recommended is a user registered on the insurance recommendation platform or a user called out, image information of the object to be recommended is extracted, and N recommendation channels in which the object to be recommended is located are determined according to the image information, and the N recommendation channels are called out calls or short messages of the insurance recommendation platform through which the object to be recommended browses or the insurance product to be recommended is passively received by the object to be recommended.
In this embodiment, an object to be recommended of an insurance product to be recommended is obtained, and portrait information of the object to be recommended is extracted, where the portrait information of the user is unique identification information of the user, and may be a user account name, for example, zhangsan, lisi, and the like; numbering information is also possible, such as 000001, 000002, etc., as examples; the account name and number information may be combined, such as Zhang three 000001, lifour 000002, etc., and the identity information of the user, etc., and it should be understood that the user portrait information is only illustrated for convenience and is not limited to the present invention.
According to the portrait information, N recommendation channels in which the object to be recommended is located are determined, wherein the recommendation channels comprise recommendation platforms which can be actively known by the object to be recommended and channels which passively acquire the insurance product to be recommended, and whether browsing is performed on the corresponding recommendation platform can be determined from the recommendation platforms which can be actively known by the object to be recommended. If browsing is performed on the recommendation platform of the insurance product to be recommended, determining that the recommendation platform is a recommendation channel where the object to be recommended is located, for example, jin Guangu, a supervision platform, etc., the channel in which the insurance product to be recommended is passively acquired may be a channel in which the insurance product information to be recommended is acquired through a phone call, for example, an AI recommendation channel, etc.
S202: for any recommendation channel, determining content information of an object to be recommended recorded by the recommendation channel, extracting keywords from the content information to obtain keywords of the recommendation channel, and selecting a matching intention class set matched with the keywords from preset intention classes.
In step S202, for any recommendation channel, content information of an object to be recommended recorded by the recommendation channel is determined, wherein the content information is recorded information about an insurance product to be recommended of the object to be recommended. Extracting keywords from the content information to obtain keywords of the recommendation channel, and selecting a matched intention category set matched with the keywords from preset intention categories. The preset intention category is a preset category of purchasing intention of the insurance product to be recommended, and the matching intention category set is the purchasing intention of the object to be recommended, which is mined according to the keywords, on the corresponding insurance product to be recommended.
In this embodiment, content information of an object to be recommended recorded in a recommendation channel is determined, keyword extraction is performed on the content information, and a matching intention category set matched with the keyword is selected from preset intention categories. The preset intention category can be business handling intention, paying intention, planned handling intention and the like.
Optionally, extracting keywords from the content information to obtain keywords of the recommendation channel, including:
determining the data type of a recommended channel;
if the data type is voice type, performing voice recognition on the content information to obtain a voice recognition result, and extracting keywords from the voice recognition result to obtain keywords of a recommendation channel;
and if the data type is voice type and text type, extracting keywords from the content information to obtain keywords of the recommendation channel.
In this embodiment, when extracting keywords from content information, different methods are used to extract keywords according to data types of different recommendation channels, where the data types may be a voice type and a text type, if the data types are voice types, performing voice recognition on the content information to obtain a voice recognition result, and performing keyword extraction on the voice recognition result to obtain keywords of the recommendation channels, and if the data types are voice types and text types, performing keyword extraction on the content information to obtain keywords of the recommendation channels.
It should be noted that, the voice recognition is performed on the content information to obtain a voice recognition result, where when the voice recognition is performed, voice in the content information may be subjected to voice separation processing to obtain separated voice, and the voice recognition is performed on the separated voice to obtain a voice recognition result.
It should be noted that, because the voice collected by the phone platform generally has noise, including noise in the background environment and noise generated in the recording process of the front-end communication device (such as a phone), the voice carrying noise can affect the accuracy of voice recognition when performing voice recognition, so that noise reduction processing needs to be performed on the content information, so that purer voice is extracted from the historical outbound voice as far as possible to be used as the voice to be recognized, and the recognition result is more accurate when the voice recognition is performed based on the voice to be recognized later. Among them, the method of noise reduction includes, but is not limited to, using spectral subtraction, EEMD decomposition algorithm, SVD singular value algorithm, etc.
And performing voice separation processing on the voice after noise reduction to obtain separated voice, and performing voice recognition by using the voice after voice separation so as to improve voice recognition accuracy. In this embodiment, voice recognition is performed on the voice after noise reduction by using a voice print recognition model, and voice separation is performed according to the voice print recognition result. And carrying out voiceprint recognition on the noise-reduced voice by adopting a voiceprint recognition model. Here, the voiceprint recognition model may include three residual error networks, and the voiceprint recognition model performs voiceprint recognition on the noise-reduced voice to perform voice separation, so as to obtain a voice separation result.
And performing voice recognition on the separated voice to obtain a voice recognition result. In this embodiment, a trained speech recognition model is used to perform speech recognition, where the trained speech recognition model may be a deep learning model, and the trained speech recognition model includes a classification recognition network and an attention recognition network. And extracting keywords from the voice recognition result to obtain keywords of the recommended channel.
And if the data type is voice type and text type, extracting keywords from the content information to obtain keywords of the recommendation channel. In this embodiment, text recognition is performed by using a preset Bert model, keywords in the recognized text are extracted, word segmentation is performed on the recognized text by using a preset word segmentation strategy in the preset Bert model, and a corresponding word segmentation feature sequence is obtained, wherein the word segmentation strategy is used for segmenting sentences in a character element table of the recognized text according to semantics, and Chinese characters of one word are segmented together when segmentation is performed. Alternatively, word segmentation may be implemented using a neural network training based model. And marking the parts of speech of the grapheme expression by using a preset part of speech marking strategy to obtain a part of speech feature sequence corresponding to the grapheme expression, wherein the preset part of speech marking strategy is used for predicting the part of speech of each word in the grapheme expression data.
It should be noted that the preset Bert model is based on a bidirectional coding representation of a transducer, which is a model for improving the training speed of the model by using an attention mechanism in the natural language field, and the preset Bert model constructs a multi-layer bidirectional coding network by using a transducer structure. The preset Bert model is composed of coding parts in a plurality of transformers, a coding unit of one Transformer is generated by superposition of a multi-head attention and layer normalization, the multi-head attention is composed of a plurality of self-attentions, the layer normalization is used for normalizing 0 mean 1 variance of a certain layer of neural network nodes, and a structure of the Transformer can be used for predicting mask characters (token) through text context, so that bidirectional relations of character vectors are captured.
It should be noted that, in practical application, the pretrained Bert model may include a plurality of feature extraction layers, each feature extraction layer has a coding unit, in a relatively large pretrained Bert model, there are 24 feature extraction layers, each layer has 16 attribute, the feature vector has a dimension of 1024, in a relatively small pretrained Bert model, there are 12 feature extraction layers, each layer has 12 attribute, and the feature vector has a dimension of 768. For example, a preset Bert model of 12 feature extraction layers is exemplified: layer1 through layer4 are lower layers and are learned to be lexical features such as: whether the word is a verb or an adjective, which characters the word is composed of, etc., layers_5 to layer_8 are middle layers, and syntactic features are learned, such as: the number of words in a sentence, the dependency relationship between words in a sentence, and the like, and layers_9 to layer_12 are high-level, and learned are semantic features such as: what the semantics of the sentence expression are, which are keywords in the sentence, etc., thereby obtaining the corresponding keywords.
S203: traversing all the recommendation channels, determining N matching intention class sets of the N recommendation channels, and carrying out aggregation treatment on the matching intention classes in the N matching intention class sets to obtain an intention aggregation result.
In step S203, the intention categories of the objects to be recommended in all the channels to be recommended are aggregated to obtain an intention aggregation result, so that the final intention category of the objects to be recommended is determined according to the aggregation result.
In this embodiment, intention categories of different recommendation channels are aggregated to obtain intention information of an object to be recommended about different insurance products to be recommended, for example, the intention category of purchasing safe and good insurance obtained in an outbound recommendation channel is business handling, the intention category obtained in a supervision recommendation channel is the intention category of policy loan insurance is payment intention, and the like. The intention category in the outbound channel and the intention category of the supervision channel are aggregated, and the obtained intention aggregation result is { peace blessing |outbound|1, policy loan |supervision|1 }.
Optionally, aggregating the matching intent categories in the N matching intent category sets to obtain an intent aggregation result, including:
removing the same matching intention category in the N matching intention category sets to obtain a removed matching intention category;
And aggregating the removed matching intention categories to obtain intention aggregation results containing all the matching intention categories in the N matching intention category sets.
In this embodiment, in the process of aggregating the intention categories, possibly different recommendation channels include intention categories of the same insurance product, for example, in an outbound channel, the intention category of purchasing safe and good insurance is business transaction, and in a supervision recommendation channel, the intention category of purchasing safe and good insurance is business transaction, then the intention category of the insurance product in one recommendation channel is reserved, and the obtained intention aggregation result is { safe and good |outbound|1, safe and good |supervision|0 }, wherein 1 is to aggregate the insurance product in the outbound recommendation channel, and 0 is the intention category of the same insurance product and insurance product also included in the supervision recommendation channel.
S204: and scoring the matching intention categories in the intention aggregation result to obtain scoring results in each matching intention category, and determining the target intention category of the object to be recommended according to the scoring results.
In step S204, scoring processing is performed on the matching intent categories in the intent aggregation result to obtain a scoring result in each matching intent category, where the scoring processing is to set different scores for different intent categories, so as to determine the final intent of the object to be recommended according to the scoring result.
In this embodiment, the matching intent categories in the intent aggregation result are scored according to a preset scoring rule to obtain a scoring result in each matching intent category, where different intent categories may be scored because the same insurance product to be recommended includes different intent categories, and an intent category with a high score is selected as a final intent category, and scoring may be performed according to preset key elements, for example, key elements such as "subject+predicate+object" in each matching intent category may be extracted, an intent set of "WHO) +do (transact/know/plan) +what (business/product)" is output, the output intent set and the preset intent set are scored, and when the scoring is performed, a distance between the output "WHO) +do (transact/know/plan) +what (business/product)" and a preset intent text in the preset intent set is calculated, and when the distance is large, the scoring result is small, and when the distance is small, the scoring result is large. And determining a target intention category of the object to be recommended according to the grading result, wherein the target intention category is the intention category corresponding to the highest grading.
Optionally, scoring the matching intent categories in the intent aggregation result to obtain scoring results in each matching intent category, including:
performing word vector coding on the matching intention category by using a preset word vector model to obtain a coded vector corresponding to the matching intention category;
according to the preset coding vector corresponding to the coding vector and the preset intention category, calculating the distance between the coding vector and the preset coding vector, and according to the distance, grading the matching intention category to obtain grading results in each matching intention category.
In this embodiment, the scoring may be performed according to the distance between the preset intent category and the matching intent category in the preset intent set, for example, the matching intent category including peaceful bless in the intent aggregation result is business handling and delivering deposit, the preset intent set includes "i want to purchase peaceful bless" and "i handle peaceful bless business", etc., the distance between the matching intent category and the preset intent category in the preset intent set is calculated, and the scoring result in each matching intent category is determined according to the size of the distance.
In another embodiment, semantic recognition can be performed on the matching intention category, the distance between the matching intention category and the preset intention category in the preset intention set is calculated according to the semantic of the preset intention category in the preset intention set and the grading result in each matching intention category is determined according to the distance.
When the semantic recognition is performed, a preset feature extraction model can be used for extracting semantic features of the matching intention category and the preset intention category, wherein the preset feature extraction model is a trained semantic extraction model, and a preset word vector model is used for encoding the matching intention category and the preset intention category to obtain encoding features of the matching intention category and encoding features of the preset intention category. The word2vec is a natural language processing tool, and the word2vec is used for converting words in natural language into word vectors which can be understood by a computer. The conventional word vector is easy to be afflicted by dimension disasters, and any two words are isolated and cannot reflect the relationship between the words, so that word2vec is adopted to obtain the word vector in the embodiment, and the similarity between the words can be reflected by calculating the distance between the vectors. word2vec mainly adopts two models of Skip-Gram and CBOW, and the Skip-Gram is used for realizing word vector conversion in the embodiment, and the Skip-Gram model mainly predicts context words through central words.
In another embodiment, performing semantic recognition on the matching intention category and the preset intention category to obtain a semantic recognition result, and may further include: word segmentation processing is carried out on the matched intention category and the preset intention category to obtain a plurality of segmented words in the matched intention category and the preset intention category, word vector extraction is carried out on the segmented words to obtain word vectors of the segmented words, and semantic recognition is carried out on the word vectors of the segmented words by using a preset bidirectional long-short-time memory network to obtain a semantic recognition result.
Optionally, scoring the matching intent categories according to the distance to obtain scoring results in each matching intent category, including:
normalizing the distance to obtain a normalized distance;
taking the inverse of the normalized distance as the scoring result in each matching intent category.
In this embodiment, the distance is normalized to obtain a normalized distance, and the inverse of the normalized distance is taken as the scoring result in each matching intention category, where, when normalization is performed, a corresponding weight value can be set according to the intention category of different recommendation channels, for example, the supervision log is manually understood and processed, which is more valuable and more credible than other recommendation channels. The matching intention category obtained through the supervision recommendation channel can be set with a larger weight value, and the set weight value and the normalized distance are combined to be used as the final normalized distance.
Optionally, after the intention classification method, the method further includes:
determining a target conversation matched with the target intention category from a preset conversation set according to the target intention category;
determining a target recommendation channel corresponding to the target intention category according to the service type of each recommendation channel;
and distributing the target conversation to the user to be recommended through the target recommendation channel.
In this embodiment, a target conversation matching with a target intention category is determined from a preset conversation set according to the target intention category, and a target recommendation channel corresponding to the target intention category is determined according to the service type of each recommendation channel, for example, if the target intention category of the object to be recommended is a service resource, the target conversation may be served to the user to be recommended through a gold manager recommendation channel, if the target intention category of the object to be recommended is a service transaction, the target conversation may be served to the user to be recommended through a gold manager recommendation channel or a manual customer service recommendation channel, and if the target intention category of the object to be recommended is a fee-paying intention, the target conversation may be served to the user to be recommended through a supervision platform recommendation channel.
Obtaining an object to be recommended of an insurance product to be recommended, extracting image information of the object to be recommended, determining N recommendation channels in which the object to be recommended is located according to the image information, wherein N is an integer greater than 1, determining content information of the object to be recommended recorded by any recommendation channel, extracting keywords from the content information to obtain keywords of the recommendation channels, selecting a matching intention class set matched with the keywords from preset intention classes, traversing all the recommendation channels, determining N matching intention class sets of the N recommendation channels, performing aggregation processing on the matching intention classes in the N matching intention class sets to obtain an intention aggregation result, performing scoring processing on the matching intention classes in the intention aggregation result to obtain scoring results in each matching intention class, and determining a target intention class of the object to be recommended according to the scoring result. In the method, the intention category of the user to be recommended in the single recommendation channel is extracted, then the intention category in the multiple recommendation channels is aggregated, and the intention mining classification is performed again according to the aggregated result, so that the precision of the intention classification is improved.
Referring to fig. 3, fig. 3 is a schematic structural diagram of an intention classification device based on multi-channel aggregation according to an embodiment of the invention. The terminal in this embodiment includes units for executing the steps in the embodiment corresponding to fig. 2. Refer specifically to fig. 2 and the related description in the embodiment corresponding to fig. 2. For convenience of explanation, only the portions related to the present embodiment are shown. As shown in fig. 3, the intention classifying device 30 includes: the system comprises an acquisition module 31, an extraction module 32, an aggregation module 33 and a scoring module 34.
The acquiring module 31 is configured to acquire an object to be recommended of the insurance product to be recommended, extract image information of the object to be recommended, and determine N recommendation channels in which the object to be recommended is located according to the image information, where N is an integer greater than 1.
The extracting module 32 is configured to determine content information of an object to be recommended recorded by a recommendation channel for any recommendation channel, extract keywords from the content information to obtain keywords of the recommendation channel, and select a matching intention category set matching the keywords from preset intention categories.
The aggregation module 33 is configured to traverse all the recommendation channels, determine N sets of matching intention categories of the N recommendation channels, and aggregate matching intention categories in the N sets of matching intention categories to obtain an intention aggregation result.
The scoring module 34 is configured to score the matching intent categories in the intent aggregation result, obtain a scoring result in each matching intent category, and determine a target intent category of the object to be recommended according to the scoring result.
Optionally, the acquiring module 31 further includes:
and the first determining module is used for determining the target voice matched with the target intention category from the preset voice set according to the target intention category.
And the second determining module is used for determining a target recommendation channel corresponding to the target intention category according to the service type of each recommendation channel.
And the dispatch module is used for dispatching the target conversation to the user to be recommended through the target recommendation channel.
Optionally, the extracting module 32 includes:
and the determining unit is used for determining the data type of the recommended channel.
And the first judging unit is used for carrying out voice recognition on the content information if the data type is voice type to obtain a voice recognition result, and carrying out keyword extraction on the voice recognition result to obtain keywords of the recommended channel.
And the second judging unit is used for extracting keywords from the content information if the data type is voice type and text type, and obtaining keywords of the recommended channel.
Optionally, the aggregation module 33 includes:
the rejecting unit is used for rejecting the same matching intention category in the N matching intention category sets to obtain a rejected matching intention category.
The obtaining unit is used for aggregating the rejected matching intention categories to obtain intention aggregation results containing all the matching intention categories in the N matching intention category sets.
Optionally, the scoring module 34 includes:
the encoding unit is used for carrying out word vector encoding on the matching intention category by using a preset word vector model to obtain an encoding vector corresponding to the matching intention category.
The computing unit is used for computing the distance between the coding vector and the preset coding vector according to the preset coding vector corresponding to the coding vector and the preset intention category, and scoring the matching intention category according to the distance to obtain a scoring result in each matching intention category.
Optionally, the computing unit includes:
and the normalization subunit is used for normalizing the distance to obtain the normalized distance.
And the scoring result determining subunit is used for taking the reciprocal of the normalized distance as the scoring result in each matching intention category.
It should be noted that, because the content of information interaction and execution process between the above units is based on the same concept as the method embodiment of the present invention, specific functions and technical effects thereof may be referred to in the method embodiment section, and will not be described herein.
Fig. 4 is a schematic structural diagram of a terminal device according to an embodiment of the present invention. As shown in fig. 4, the terminal device of this embodiment includes: at least one processor (only one shown in fig. 4), a memory, and a computer program stored in the memory and executable on the at least one processor, the processor executing the computer program performing any of the individual multi-channel aggregation-based intent classification method steps described above.
The terminal device may include, but is not limited to, a processor, a memory. It will be appreciated by those skilled in the art that fig. 4 is merely an example of a terminal device and is not limiting of the terminal device, and that the terminal device may comprise more or less components than shown, or may combine some components, or different components, e.g. may further comprise a network interface, a display screen, input means, etc.
The processor may be a CPU, but may also be other general purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), off-the-shelf programmable gate arrays (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory includes a readable storage medium, an internal memory, etc., where the internal memory may be a memory of the terminal device, and the internal memory provides an environment for the operation of an operating system and computer readable instructions in the readable storage medium. The readable storage medium may be a hard disk of the terminal device, and in other embodiments may be an external storage device of the terminal device, for example, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash Card (Flash Card), etc. that are provided on the terminal device. Further, the memory may also include both an internal storage unit of the terminal device and an external storage device. The memory is used to store an operating system, application programs, boot loader (BootLoader), data, and other programs such as program codes of computer programs, and the like. The memory may also be used to temporarily store data that has been output or is to be output.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-described division of the functional units and modules is illustrated, and in practical application, the above-described functional distribution may be performed by different functional units and modules according to needs, i.e. the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-described functions. The functional units and modules in the embodiment may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit, where the integrated units may be implemented in a form of hardware or a form of a software functional unit. In addition, the specific names of the functional units and modules are only for distinguishing from each other, and are not used for limiting the protection scope of the present invention. The specific working process of the units and modules in the above device may refer to the corresponding process in the foregoing method embodiment, which is not described herein again. The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the present invention may implement all or part of the flow of the method of the above-described embodiment, and may be implemented by a computer program to instruct related hardware, and the computer program may be stored in a computer readable storage medium, where the computer program, when executed by a processor, may implement the steps of the method embodiment described above. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, executable files or in some intermediate form, etc. The computer readable medium may include at least: any entity or device capable of carrying computer program code, a recording medium, a computer Memory, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), an electrical carrier signal, a telecommunications signal, and a software distribution medium. Such as a U-disk, removable hard disk, magnetic or optical disk, etc. In some jurisdictions, computer readable media may not be electrical carrier signals and telecommunications signals in accordance with legislation and patent practice.
The present invention may also be implemented by a computer program product for implementing all or part of the steps of the method embodiments described above, when the computer program product is run on a terminal device, causing the terminal device to execute the steps of the method embodiments described above.
In the foregoing embodiments, the descriptions of the embodiments are emphasized, and in part, not described or illustrated in any particular embodiment, reference is made to the related descriptions of other embodiments.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
In the embodiments provided in the present invention, it should be understood that the disclosed apparatus/terminal device and method may be implemented in other manners. For example, the apparatus/terminal device embodiments described above are merely illustrative, e.g., the division of modules or units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection via interfaces, devices or units, which may be in electrical, mechanical or other forms.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed over a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
The above embodiments are only for illustrating the technical solution of the present invention, and are not limiting; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention, and are intended to be included in the scope of the present invention.
Claims (10)
1. An intent classification method based on multi-channel aggregation, characterized in that the intent classification method comprises:
acquiring an object to be recommended of an insurance product to be recommended, extracting image information of the object to be recommended, and determining N recommendation channels in which the object to be recommended is positioned according to the image information, wherein N is an integer greater than 1;
Determining content information of the object to be recommended recorded by the recommendation channel aiming at any recommendation channel, extracting keywords from the content information to obtain keywords of the recommendation channel, and selecting a matching intention category set matched with the keywords from preset intention categories;
traversing all recommendation channels, determining N matching intention class sets of the N recommendation channels, and carrying out aggregation treatment on the matching intention classes in the N matching intention class sets to obtain an intention aggregation result;
and scoring the matching intention categories in the intention aggregation result to obtain scoring results in each matching intention category, and determining the target intention category of the object to be recommended according to the scoring results.
2. The method for classifying intention according to claim 1, wherein the extracting the keyword from the content information to obtain the keyword of the recommendation channel comprises:
determining the data type of the recommendation channel;
if the data type is voice type, performing voice recognition on the content information to obtain a voice recognition result, and extracting keywords from the voice recognition result to obtain keywords of the recommendation channel;
And if the data type is a voice type and is a text type, extracting keywords from the content information to obtain keywords of the recommendation channel.
3. The intention classification method according to claim 1, wherein the aggregating the matching intention categories in the N matching intention category sets to obtain an intention aggregation result includes:
removing the same matching intention category in the N matching intention category sets to obtain a removed matching intention category;
and aggregating the rejected matching intention categories to obtain intention aggregation results containing all the matching intention categories in the N matching intention category sets.
4. The intent classification method as claimed in claim 1, wherein scoring the matching intent categories in the intent aggregate result to obtain scoring results in each matching intent category includes:
performing word vector coding on the matching intention category by using a preset word vector model to obtain a coding vector corresponding to the matching intention category;
calculating the distance between the coding vector and the preset coding vector according to the preset coding vector corresponding to the coding vector and the preset intention category, and grading the matching intention category according to the distance to obtain grading results in each matching intention category.
5. The intent classification method as recited in claim 4, wherein scoring the matching intent categories based on the distance results in scoring results in each matching intent category comprising:
normalizing the distance to obtain a normalized distance;
taking the inverse of the normalized distance as a scoring result in each matching intention category.
6. The intent classification method as recited in any of claims 1 to 5, further comprising, after the intent classification method:
determining a target conversation matched with the target intention category from a preset conversation set according to the target intention category;
determining a target recommendation channel corresponding to the target intention category according to the service type of each recommendation channel;
and distributing a target conversation to the user to be recommended through the target recommendation channel.
7. An intent classification device based on multi-channel aggregation, the intent classification device comprising:
the acquisition module is used for acquiring an object to be recommended of an insurance product to be recommended, extracting image information of the object to be recommended, and determining N recommendation channels in which the object to be recommended is positioned according to the image information, wherein N is an integer greater than 1;
The extraction module is used for determining content information of the object to be recommended recorded by the recommendation channel aiming at any recommendation channel, extracting keywords from the content information to obtain keywords of the recommendation channel, and selecting a matching intention class set matched with the keywords from preset intention classes;
the aggregation module is used for traversing all the recommendation channels, determining N matching intention class sets of the N recommendation channels, and carrying out aggregation processing on the matching intention classes in the N matching intention class sets to obtain an intention aggregation result;
the scoring module is used for scoring the matching intention categories in the intention aggregation result to obtain a scoring result in each matching intention category, and determining the target intention category of the object to be recommended according to the scoring result.
8. The intent classification device of claim 7, wherein the intent classification device further comprises:
the first determining module is used for determining a target conversation matched with the target intention category from a preset conversation set according to the target intention category;
the second determining module is used for determining a target recommendation channel corresponding to the target intention category according to the service type of each recommendation channel;
And the dispatch module is used for dispatching the target conversation to the user to be recommended through the target recommendation channel.
9. A terminal device, characterized in that it comprises a processor, a memory and a computer program stored in the memory and executable on the processor, which processor implements the method of intent classification as claimed in any of the claims 1 to 6 when executing the computer program.
10. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the intention classification method of any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311613675.9A CN117648440A (en) | 2023-11-27 | 2023-11-27 | Intention classification method, device, equipment and medium based on multi-channel aggregation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311613675.9A CN117648440A (en) | 2023-11-27 | 2023-11-27 | Intention classification method, device, equipment and medium based on multi-channel aggregation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117648440A true CN117648440A (en) | 2024-03-05 |
Family
ID=90044481
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311613675.9A Pending CN117648440A (en) | 2023-11-27 | 2023-11-27 | Intention classification method, device, equipment and medium based on multi-channel aggregation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117648440A (en) |
-
2023
- 2023-11-27 CN CN202311613675.9A patent/CN117648440A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110163478B (en) | Risk examination method and device for contract clauses | |
| CN110597964B (en) | Double-recording quality inspection semantic analysis method and device and double-recording quality inspection system | |
| CN110597952A (en) | Information processing method, server, and computer storage medium | |
| CN111291570A (en) | Method and device for realizing element identification in judicial documents | |
| CN112487824A (en) | Customer service speech emotion recognition method, device, equipment and storage medium | |
| CN110610003B (en) | Method and system for assisting text annotation | |
| US11934556B2 (en) | Identifying sensitive content in electronic files | |
| CN112632248A (en) | Question answering method, device, computer equipment and storage medium | |
| CN117278675A (en) | Outbound method, device, equipment and medium based on intention classification | |
| CN111274390A (en) | Emotional reason determining method and device based on dialogue data | |
| CN117611335A (en) | Financial risk identification method, apparatus, electronic device and storage medium | |
| CN111783424A (en) | Text clause dividing method and device | |
| CN112581297A (en) | Information pushing method and device based on artificial intelligence and computer equipment | |
| CN113505293B (en) | Information pushing method and device, electronic equipment and storage medium | |
| CN117648440A (en) | Intention classification method, device, equipment and medium based on multi-channel aggregation | |
| CN115080744A (en) | Data processing method and device | |
| CN114118062A (en) | Customer feature extraction method and device, electronic equipment and storage medium | |
| CN113870478A (en) | Rapid number-taking method and device, electronic equipment and storage medium | |
| CN113807920A (en) | Artificial intelligence based product recommendation method, device, equipment and storage medium | |
| CN113064983A (en) | Semantic detection method and device, computer equipment and storage medium | |
| CN111782601A (en) | Electronic file processing method and device, electronic equipment and machine readable medium | |
| CN114818683B (en) | Operation and maintenance method and device based on mobile terminal | |
| CN117556050B (en) | Data classification and classification method and device, electronic equipment and storage medium | |
| CN117057333A (en) | Method, device, equipment and medium for mining conversation template of financial business | |
| US20230131066A1 (en) | Composite extraction systems and methods for artificial intelligence platform |
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 |