CN110310663A - Words art detection method, device, equipment and computer readable storage medium in violation of rules and regulations - Google Patents
Words art detection method, device, equipment and computer readable storage medium in violation of rules and regulations Download PDFInfo
- Publication number
- CN110310663A CN110310663A CN201910411437.7A CN201910411437A CN110310663A CN 110310663 A CN110310663 A CN 110310663A CN 201910411437 A CN201910411437 A CN 201910411437A CN 110310663 A CN110310663 A CN 110310663A
- Authority
- CN
- China
- Prior art keywords
- violation
- art
- words art
- regulations
- rules
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 71
- 238000012549 training Methods 0.000 claims abstract description 53
- 238000010801 machine learning Methods 0.000 claims abstract description 13
- 239000013598 vector Substances 0.000 claims description 34
- 238000013178 mathematical model Methods 0.000 claims description 24
- 238000013528 artificial neural network Methods 0.000 claims description 13
- 238000009432 framing Methods 0.000 claims description 7
- 238000007689 inspection Methods 0.000 abstract description 16
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 239000010410 layer Substances 0.000 description 32
- 230000006854 communication Effects 0.000 description 11
- 238000004891 communication Methods 0.000 description 10
- 238000010586 diagram Methods 0.000 description 9
- 238000000034 method Methods 0.000 description 7
- 210000002569 neuron Anatomy 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 4
- 239000000284 extract Substances 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 238000013507 mapping Methods 0.000 description 3
- 238000004321 preservation Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000000151 deposition Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 239000011229 interlayer Substances 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/06—Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice
- G10L15/063—Training
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/08—Speech classification or search
- G10L15/16—Speech classification or search using artificial neural networks
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
- G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
- G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04M—TELEPHONIC COMMUNICATION
- H04M3/00—Automatic or semi-automatic exchanges
- H04M3/42—Systems providing special services or facilities to subscribers
- H04M3/50—Centralised arrangements for answering calls; Centralised arrangements for recording messages for absent or busy subscribers ; Centralised arrangements for recording messages
- H04M3/51—Centralised call answering arrangements requiring operator intervention, e.g. call or contact centers for telemarketing
- H04M3/5175—Call or contact centers supervision arrangements
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Signal Processing (AREA)
- Business, Economics & Management (AREA)
- Marketing (AREA)
- Evolutionary Computation (AREA)
- Telephonic Communication Services (AREA)
Abstract
The present invention relates to field of artificial intelligence, disclose a kind of violation words art detection method, comprising: using the dialog context comprising violation words art as training sample, are trained using machine learning mode, obtain talking about art identification model in violation of rules and regulations；Obtain the call audio in message registration library between customer service and client；Speech recognition, the dialog context of output character format are carried out to the call audio；The dialog context of speech recognition output is inputted the violation words art identification model to identify, exports recognition result；Based on the recognition result, determine in the dialog context with the presence or absence of words art in violation of rules and regulations；It talks about art in violation of rules and regulations if it exists, is then associated violation words art with corresponding call audio, and be saved in the message registration library.The invention also discloses a kind of violation words art detection device, equipment and computer readable storage mediums.The present invention realizes the automation quality inspection for talking about art in violation of rules and regulations, avoids the cumbersome time-consuming of artificial quality inspection, improves quality inspection working efficiency.
Description
Technical field
The present invention relates to field of artificial intelligence more particularly to a kind of violation words art detection method, device, equipment and meters
Calculation machine readable storage medium storing program for executing.
Background technique
In existing telephone customer service management, to promote service quality, generally it can all record to service calls, in order to rear
It is continuous that the service quality of phone customer service is spot-check.For example art is talked about in violation of rules and regulations with whether there is during customer communication, consequently facilitating
Customer service work is evaluated.And the sampling procedure of dialog context, if dialog context is more, uses usually by being accomplished manually
Artificial quality inspection mode must can have time and effort consuming, thus execution efficiency is not high.
Summary of the invention
The main purpose of the present invention is to provide a kind of violation words art detection method, device, equipment and computer-readable deposit
Storage media, it is intended to solve technical problem not high using the execution efficiency of artificial quality inspection mode in existing telephone customer service management.
To achieve the above object, the present invention provides a kind of violation words art detection method, and art detection method packet is talked about in the violation
Include following steps:
Using the dialog context comprising violation words art as training sample, it is trained, is obtained in violation of rules and regulations using machine learning mode
Talk about art identification model；
Obtain the call audio in message registration library between customer service and client；
Framing is carried out to the call audio, obtains multiple speech frames with timing；
Successively extracted according to timing the sound characteristic of the speech frame and generate comprising acoustic information multidimensional sound spy to
Amount；
The multidimensional sound characteristic vector is inputted preset acoustic model to handle, the corresponding phoneme letter of output speech frame
Breath；
Based on the phoneme information, preset dictionary is searched, exports the corresponding word of each phoneme information or word；
The corresponding word of each phoneme information or word are inputted preset language model according to output sequence to handle, output is single
Word or word are mutually related probability；
The word of the maximum probability of output or word are spliced into dialog context and the output of text formatting；
Dialog context between the customer service and client of speech recognition output is inputted the violation words art identification model to carry out
Words art identification in violation of rules and regulations, exports recognition result；
Based on the recognition result, determine in the dialog context with the presence or absence of words art in violation of rules and regulations；
It talks about art in violation of rules and regulations if it exists, is then associated violation words art with corresponding call audio, and be saved in described logical
It talks about in record storehouse.
Optionally, art detection method is talked about in the violation further include:
It talks about art in violation of rules and regulations if it exists, then exports the relevant information with the associated audio of conversing of violation words art, the related letter
Breath includes: customer service information, customer information and topic of call.
Optionally, described using the dialog context comprising violation words art as training sample, it is instructed using machine learning mode
Practice, obtaining words art identification model in violation of rules and regulations includes:
Using the dialog context comprising violation words art as training sample, the training sample is segmented, is obtained corresponding
Segment data；
All participle data are mapped to term vector to be trained；
The mathematical model based on Single hidden layer feedforward neural networks is constructed, and is the defeated of the mathematical model with the term vector
Enter, take the violation words art in the training sample as the output of the mathematical model, training is iterated to the mathematical model,
It obtains talking about art identification model in violation of rules and regulations.
Optionally, art is talked about in violation of rules and regulations if it exists described, then be associated violation words art with corresponding call audio, and
It is saved in after the step in the message registration library, further includes:
Obtain the violation words art keyword inputted in preset query page；
Using violation words art keyword as querying condition, the message registration library is retrieved, to detect the message registration
Library whether there is the call audio with violation words art keyword association；
If it exists, then the call audio with violation words art keyword association is played.
Optionally, described using violation words art keyword as querying condition, the message registration library is retrieved, to detect
Message registration library, which is stated, with the presence or absence of the call audio with violation words art keyword association includes:
Character splicing is carried out to audio of respectively conversing in the message registration library associated violation words art, to form words art in violation of rules and regulations
Character string, and violation words art character string is passed in memory；
Using violation words art keyword as querying condition, the violation words art character string is retrieved, to detect the call
Record storehouse whether there is the call audio with violation words art keyword association.
Further, to achieve the above object, the present invention also provides a kind of violations to talk about art detection device, and art is talked about in the violation
Detection device includes:
Model training module, for using comprising violation words art dialog context as training sample, using machine learning mode
It is trained, obtains talking about art identification model in violation of rules and regulations；
Module is obtained, for obtaining the call audio in message registration library between customer service and client；
Speech recognition module obtains multiple speech frames with timing for carrying out framing to the call audio；According to when
Sequence successively extracts the sound characteristic of the speech frame and generates multidimensional sound spy's vector comprising acoustic information；By the multidimensional sound
Sound feature vector inputs preset acoustic model and is handled, the corresponding phoneme information of output speech frame；Based on the phoneme information,
Preset dictionary is searched, the corresponding word of each phoneme information or word are exported；According to output sequence by the corresponding word of each phoneme information or word
It inputs preset language model to be handled, exports single word or word is mutually related probability；By the word of the maximum probability of output or
Word is spliced into the dialog context of text formatting and output；
Art identification module is talked about, the dialog context between the customer service and client for exporting speech recognition inputs the violation
It talks about art identification model and carries out words art identification in violation of rules and regulations, export recognition result；Based on the recognition result, determine in the dialog context
Art is talked about with the presence or absence of violation；
It is associated with preserving module, for talking about art in violation of rules and regulations if it exists, is then closed violation words art with corresponding call audio
Connection, and be saved in the message registration library.
Optionally, art detection device is talked about in the violation further include:
Output module, for talking about art in violation of rules and regulations if it exists, then related the believing of output and the associated call audio of violation words art
Breath, the relevant information includes: customer service information, customer information and topic of call.
Optionally, the model training module is specifically used for:
Using the dialog context comprising violation words art as training sample, the training sample is segmented, is obtained corresponding
Segment data；
All participle data are mapped to term vector to be trained；
The mathematical model based on Single hidden layer feedforward neural networks is constructed, and is the defeated of the mathematical model with the term vector
Enter, take the violation words art in the training sample as the output of the mathematical model, training is iterated to the mathematical model,
It obtains talking about art identification model in violation of rules and regulations.
Optionally, art detection device is talked about in the violation further include:
Keyword obtains module, talks about art keyword for obtaining the violation inputted in preset query page；
Retrieval module, for the message registration library being retrieved, with detection using violation words art keyword as querying condition
The message registration library whether there is the call audio with violation words art keyword association；
Playing module, if for the message registration inventory with the violation words art keyword association call audio,
Then play the call audio with violation words art keyword association.
Optionally, the retrieval module is specifically used for:
Character splicing is carried out to audio of respectively conversing in the message registration library associated violation words art, to form words art in violation of rules and regulations
Character string, and violation words art character string is passed in memory；
Using violation words art keyword as querying condition, the violation words art character string is retrieved, to detect the call
Record storehouse whether there is the call audio with violation words art keyword association.
Further, to achieve the above object, the present invention also provides a kind of violations to talk about art detection device, and art is talked about in the violation
Detection device includes the violation words that memory, processor and being stored in can be run on the memory and on the processor
Art detects program, and the violation words art detection program realizes violation words as described in any one of the above embodiments when being executed by the processor
The step of art detection method.
Further, to achieve the above object, the present invention also provides a kind of computer readable storage medium, the computers
Words art detection program in violation of rules and regulations is stored on readable storage medium storing program for executing, the violation words art detection program is realized such as when being executed by processor
The step of violation words art detection method described in any of the above embodiments.
The present invention uses speech recognition technology, audio of conversing is identified as to the dialog context of text formatting, then to text
The dialog context of format carries out words art detection in violation of rules and regulations, if detecting words art in violation of rules and regulations, automatically by violation words art and corresponding conversation voice
Frequency is associated preservation, to realize the automation quality inspection for talking about art in violation of rules and regulations, avoids the cumbersome time-consuming of artificial quality inspection, improves matter
It examines working efficiency and saves cost.
Detailed description of the invention
Fig. 1 is the structural representation of the present invention device hardware running environment that words art detection device example scheme is related in violation of rules and regulations
Figure；
Fig. 2 is the flow diagram that the present invention talks about art detection method first embodiment in violation of rules and regulations；
Fig. 3 is the flow diagram that the present invention talks about art detection method second embodiment in violation of rules and regulations；
Fig. 4 is the functional block diagram that the present invention talks about art detection device first embodiment in violation of rules and regulations；
Fig. 5 is the functional block diagram that the present invention talks about art detection device second embodiment in violation of rules and regulations.
The embodiments will be further described with reference to the accompanying drawings for the realization, the function and the advantages of the object of the present invention.
Specific embodiment
It should be appreciated that described herein, specific examples are only used to explain the present invention, is not intended to limit the present invention.
The present invention provides a kind of violation words art detection device.
Referring to Fig.1, Fig. 1 is the present invention device hardware running environment that words art detection device example scheme is related in violation of rules and regulations
Structural schematic diagram.
As shown in Figure 1, violation words art detection device may include: processor 1001, such as CPU, communication bus 1002,
User interface 1003, network interface 1004, memory 1005.Wherein, communication bus 1002 is for realizing between these components
Connection communication.User interface 1003 may include display screen (Display), input unit such as keyboard (Keyboard), optional
User interface 1003 can also include standard wireline interface and wireless interface.Network interface 1004 optionally may include standard
Wireline interface, wireless interface (such as WI-FI interface).Memory 1005 can be high speed RAM memory, be also possible to stable
Memory (non-volatile memory), such as magnetic disk storage.Memory 1005 optionally can also be independently of aforementioned
The storage equipment of processor 1001.
It will be understood by those skilled in the art that the hardware configuration for talking about art detection device shown in Fig. 1 in violation of rules and regulations is not constituted
The restriction that art detection device is talked about to violation, may include than illustrating more or fewer components, perhaps combine certain components or
Different component layouts.
As shown in Figure 1, as may include operating system, net in a kind of memory 1005 of computer readable storage medium
Network communication module, Subscriber Interface Module SIM and violation words art detect program.Wherein, operating system is to manage and control words art in violation of rules and regulations
The program of detection device and software resource, support network communication module, Subscriber Interface Module SIM, in violation of rules and regulations words art detection program and its
The operation of his program or software；Network communication module is for managing and controlling network interface 1004；Subscriber Interface Module SIM is for managing
Reason and control user interface 1003.
In violation words art detection device hardware configuration shown in Fig. 1, after network interface 1004 is mainly used for connection system
Platform carries out data communication with system background；User interface 1003 is mainly used for connecting client (user terminal), carries out with client
Data communication；Words art detection device calls the violation stored in memory 1005 to talk about art and detects journey by processor 1001 in violation of rules and regulations
Sequence, and execute the operation of each embodiment of the following art detection method of words in violation of rules and regulations.
Art detection device hardware configuration is talked about based on above-mentioned violation, proposes that the present invention talks about each implementation of art detection method in violation of rules and regulations
Example.
It is the flow diagram that the present invention talks about art detection method first embodiment in violation of rules and regulations referring to Fig. 2, Fig. 2.In the present embodiment,
Detection method includes the following steps for the violation words art:
Step S10 is trained using the dialog context comprising violation words art as training sample using machine learning mode,
It obtains talking about art identification model in violation of rules and regulations；
In the present embodiment, to realize the automatic identification for talking about art in dialog context in violation of rules and regulations, it is therefore desirable to which training is corresponding in advance
Words art identification model in violation of rules and regulations.It is unlimited for trained specific implementation.
Optionally, in one embodiment, art identification model is talked about in violation of rules and regulations especially by following manner training:
(1) using the dialog context comprising violation words art as training sample, the training sample is segmented, is corresponded to
Participle data；
(2) all participle data are mapped to term vector to be trained；
(3) mathematical model based on Single hidden layer feedforward neural networks is constructed, and with the term vector for the mathematical model
Input, in the training sample violation words art be the mathematical model output, the mathematical model is iterated
Training obtains talking about art identification model in violation of rules and regulations.
Term vector (Word embedding) refer to by the word or expression from vocabulary be mapped to real number it is a kind of to
Measure representation.Usual words is all natural language form, but natural language can not be directly handled in machine learning techniques,
And needing to be converted to the words of natural language accessible mathematic(al) structure namely space vector form, any words all may be used
To be expressed as different vectors in space.For example, by all word ranks at a long character string, each word after sequence
A position will be corresponded to, then indicates some word with an array isometric with word quantity, the position where the word
Setting array value is just 1, and all positional values of other words are all 0, so as to which word is mapped as term vector.
Single hidden layer feedforward neural networks are a kind of special shapes of feedforward neural network.Feedforward neural network is artificial neuron
One kind of network.In such neural network, each neuron receives previous stage input since input layer, and is output to next
Grade, until output layer.Feedforward neural network uses a kind of one-way multilayer structure.Wherein each layer includes several neurons, together
It is not interconnected between one layer of neuron, the transmission of inter-layer information only carries out in one direction.Wherein first layer is referred to as defeated
Enter layer, the last layer is output layer, and centre is hidden layer.Hidden layer can be one layer, be also possible to multilayer.
In this alternative embodiment, Single hidden layer feedforward neural networks include: first layer: input layer (input layer), the
Two layers: hidden layer (hidden layer), third layer: output layer (output layer), wherein input layer is corresponding with dialog context
Term vector be training sample, output layer using in dialog context violation words art as training sample.By to building based on list
The mathematical model of hidden layer feedforward neural network carries out the training that iterates, by constantly adjusting input layer to hidden layer nerve
The weight of member, the threshold value of hidden neuron, hidden neuron to output layer neuron weight, to guarantee model output and it is expected
Output error within an acceptable range, and then completes model training and obtains that dialog context progress violation words art identification can be disobeyed
Rule words art identification model.
Step S20 obtains the call audio in message registration library between customer service and client；
It in the present embodiment, will record, form call audio and protect automatically with customer phone communication process in customer service
It is stored in message registration library.Call audio used in the detection of words art is either quasi real time, be also possible to history in violation of rules and regulations,
It is configured with specific reference to actual needs.
The selection mode of call audio for detecting every time is unlimited, such as is obtained in a set time or in the time limit every time
Call audio.Such as the call audio in 10 hours of acquisition, or the call audio in acquisition 1 day every time every time.
Optionally, when audio is conversed in storage, further storage information relevant to audio of conversing, such as customer service name,
Customer service work number, call time started, end time, customer name, customer telephone number, topic of call (such as individual deposit industry
Business, personal transferred account service etc.).
Step S30 carries out framing to the call audio, obtains multiple speech frames with timing；
In order to more effectively extract sound characteristic, therefore also need to be filtered call audio, the audio datas such as framing
Pretreatment, the sub-frame processing of the present embodiment are exactly sound to be divided into a bit of, and every segment is known as a frame speech frame, makes
Sub-frame processing is realized with mobile window function, obtains multiple speech frames with timing.
Step S40 successively extracts the sound characteristic of the speech frame according to timing and generates the multidimensional comprising acoustic information
Sound spy's vector；
Feature extraction is that voice signal is transformed into frequency domain from time domain, to provide suitable input feature vector for acoustic model
Vector.It is special that the present embodiment mainly uses linear prediction residue error (LPCC) and mel cepstrum coefficients (MFCC) algorithm to extract sound
Sign, and then each waveform speech frame is transformed into the multi-C vector comprising acoustic information.
The multidimensional sound characteristic vector is inputted preset acoustic model and handled by step S50, and output speech frame is corresponding
Phoneme information；
Acoustic model is the representation of knowledge to differences such as acoustics, phonetics, environmental variance, speaker's gender, accents.Acoustics
Model is obtained by being trained to voice data, and acoustic model each feature vector can exist according to calculation of Acoustic Characteristics
Probability score on acoustic feature, namely the sound characteristic of voice is established to the mapping relations between phoneme.
Step S60 is based on the phoneme information, searches preset dictionary, export the corresponding word of each phoneme information or word；
Dictionary is the corresponding phoneme index set of words, is the mapping between words and phoneme, by searching for dictionary, thus
Determine the corresponding word of each phoneme information or word.
The corresponding word of each phoneme information or word are inputted preset language model according to output sequence and handled by step S70,
It exports single word or word is mutually related probability；
Language model indicates the probability that a certain word sequence occurs, can be by being trained to obtain to text language data, language
Speech model can calculate voice signal according to linguistic characteristics and correspond to the probability of phrase sequence, namely establish the corresponding phoneme of text
The mapping relations of the phrase sequence formed to text.
The word of the maximum probability of output or word are spliced into dialog context and the output of text formatting by step S80；
After obtaining the probability that each possible corresponding word of call audio or phrase occur, by the word or word of maximum probability
It is spliced into the dialog context of text formatting and is exported as the result of speech recognition.
For example, it is assumed that having word content is that the call audio of " I is robot " obtains following characteristics by feature extraction
Vector [1 2345 6....10]；This feature vector input acoustic model is handled, corresponding phoneme is obtained, namely
[1 2 3 4 5 6....10]—>w o s i j i q i r n；Then again by searching for dictionary, it is corresponding to obtain a phoneme
Word, nest: w o；I: w o；It is: s i；Machine: j i；Device: q i；People: r n；Grade: j i；Bear: r n；Finally above-mentioned output is tied again
Fruit input language model is handled, and corresponding word or phrase sequence are obtained, as follows: I: 0.0786, be: 0.0546, I
It is: 0.0898, machine: 0.0967, robot: 0.6785；Compared by probability, determine each word or phrase most probably
Rate: I is: 0.0898, robot: and 0.6785, spliced output content is " I is robot ", completes the language of call audio
Sound identification.
In the present embodiment, speech recognition is carried out to call audio using speech recognition technology, and output character format is logical
Talk about content.
Optionally, to promote quality inspection efficiency, further the dialog context of text formatting is arranged respectively as in customer service call
Hold and client's dialog context.
Dialog context between the customer service and client of speech recognition output is inputted the violation words art identification by step S90
Model carries out words art identification in violation of rules and regulations, exports recognition result；
Step S100 is based on the recognition result, determines in the dialog context with the presence or absence of words art in violation of rules and regulations；
Words art refers to that the prescribed term of communication exchange, different industries or business use art if difference.Such as term of courtesy
Talk about art, business term words art etc..The present embodiment is unlimited for the setting of violation words art.Art is talked about either customer service is used alone
, it is also possible to what customer service was used in pairs based on the dialogue with client.
In violation of rules and regulations words art then refer to it is against regulation if art, such as remark of doubtful propriety or the art for not meeting business need
Language etc..The mode that the present embodiment talks about detection art in violation of rules and regulations is unlimited.For example it is detected or is based on by character match mode
Mathematical model carries out Classification and Identification.
In the present embodiment, in advance using comprising violation words art dialog context as training sample, using machine learning mode into
Row training obtains talking about art identification model in violation of rules and regulations；Then again that the dialog context between the customer service and client of speech recognition output is defeated
Enter the violation words art identification model and carry out words art identification in violation of rules and regulations, exports recognition result；It is finally based on the recognition result again, really
With the presence or absence of words art in violation of rules and regulations in the fixed dialog context.
The present embodiment carries out words art detection in violation of rules and regulations using disaggregated model matching way.Wherein, in advance to talk about art comprising violation
Dialog context as training sample, obtain to identify the dialog context comprising talking about art in violation of rules and regulations using the training of machine learning mode
Disaggregated model (namely in violation of rules and regulations words art identification model).For example, carrying out model training using supervised learning mode, wherein supervision
The training sample set of study, which requires to include, to be output and input, namely with dialog context be input, to talk about art in violation of rules and regulations as output.It is common
Supervised learning algorithm include regression analysis and statistical classification etc..
In the present embodiment, dialog context is detected using the disaggregated model that the training of machine learning mode obtains, if defeated
Result non-empty out, it is determined that there is words art in violation of rules and regulations in the dialog context of current detection；And if exporting result is sky, it is determined that current
There is no words arts in violation of rules and regulations in the dialog context of detection.
Step S110 talks about art in violation of rules and regulations if it exists, then is associated violation words art with corresponding call audio, and save
Into the message registration library.
In the present embodiment, there may be one or more words arts in violation of rules and regulations in the same call audio, or may not also deposit
Art is talked about in violation.If there is words art in violation of rules and regulations in testing result, will in violation of rules and regulations words art and corresponding call audio be associated be saved in it is logical
It talks about in record storehouse.For example, being recorded in message registration 1: talking about art a in violation of rules and regulations present in call audio A and call audio A；
It is recorded in message registration 2: there is words art b1, b2, b3 in violation of rules and regulations in call audio B and call audio B.
It is further alternative, it is defeated if detecting in current talking content there is words art in violation of rules and regulations in an alternative embodiment
Out with this in violation of rules and regulations words art it is associated call audio relevant information, the relevant information include: customer service information, customer information and
Topic of call.
In this alternative embodiment, if there is words art in violation of rules and regulations in dialog context testing result, further exports the presence and disobey
The relevant information of the call audio of rule words art, such as customer service information, such as customer service name, work number ID, call time started and knot
The beam time；Such as customer information, such as customer name, phone number etc.；Such as topic of call, such as about individual deposit business
Or personal transferred account service etc..Output exist in violation of rules and regulations words art call audio relevant information, can in order to be convenient for can be quick
Responsible person concerned is positioned, and then improves service quality, for example is criticized to as thing contact staff, apology is carried out to client and is laid equal stress on
It is new to answer customer issue, the processing such as right the wrong.
The present embodiment uses speech recognition technology, audio of conversing is identified as to the dialog context of text formatting, then to text
The dialog context of word format carries out words art in violation of rules and regulations and detects, if detecting words art in violation of rules and regulations, subtracts words art in violation of rules and regulations automatically and converses with corresponding
Audio is associated preservation, to realize the automation quality inspection for talking about art in violation of rules and regulations, avoids the cumbersome time-consuming of artificial quality inspection, improves
Quality inspection working efficiency and save cost.
It is the flow diagram that the present invention talks about art detection method second embodiment in violation of rules and regulations referring to Fig. 3, Fig. 3.Based on above-mentioned side
Method first embodiment, in the present embodiment, after above-mentioned steps S110, further includes:
Step S120 obtains the violation words art keyword inputted in preset query page；
Step S130 retrieves the message registration library, described in detection using violation words art keyword as querying condition
Message registration library whether there is the call audio with violation words art keyword association；
Step S140, and if it exists, then play the call audio with violation words art keyword association.
In the present embodiment, for the flexibility for further promoting quality inspection, therefore the inquiry further provided for for user search
The page.Query function provided in this embodiment passes through keyword match mode, search specifically using key word of the inquiry as querying condition
Dialog context, if exist in dialog context with the matched content of key word of the inquiry, it is automatic play it is matched with key word of the inquiry
Call audio corresponding to dialog context.
In the present embodiment, it is further provided query page, in order to which user flexibility is inspected by random samples.User can arbitrarily disobey
Rule words art retrieves violation message registration library as key word of the inquiry, thus what acquisition matched with target violation words art
Call audio simultaneously plays automatically, improves retrieval flexibility.
It is further alternative, in one embodiment of art detection method of words in violation of rules and regulations of the invention, to promote recall precision, above-mentioned step
Rapid S130 further comprises:
Character splicing is carried out to audio of respectively conversing in the message registration library associated violation words art, to form words art in violation of rules and regulations
Character string, and violation words art character string is passed in memory；
Using violation words art keyword as querying condition, the violation words art character string is retrieved, to detect the call
Record storehouse whether there is the call audio with violation words art keyword association.
In the present embodiment, is extracted from message registration library talk about art with the associated violation of each call audio in advance, then will mentioned
Each art of words in violation of rules and regulations taken carries out character splicing, to form violation words art character string, and words art character string is incoming interior in violation of rules and regulations by this
In depositing；Then again using the violation of input words art keyword as querying condition, violation words art character string is retrieved, is only needed primary
The retrieval of a plurality of violation words art in message registration library can be completed in search operaqtion, and then improves recall precision.
The present invention also provides a kind of violations to talk about art detection device.
It is the functional block diagram that the present invention talks about art detection device first embodiment in violation of rules and regulations referring to Fig. 4, Fig. 4.This implementation
In example, the violation words art detection device includes:
Model training module 10, for using comprising violation words art dialog context as training sample, using machine learning side
Formula is trained, and obtains talking about art identification model in violation of rules and regulations；
In one embodiment, the model training module 10 is specifically used for:
Using the dialog context comprising violation words art as training sample, the training sample is segmented, is obtained corresponding
Segment data；All participle data are mapped to term vector to be trained；Construct the mathematics based on Single hidden layer feedforward neural networks
Model, and be the input of the mathematical model with the term vector, with the violation words art in the training sample for the mathematics
The output of model is iterated training to the mathematical model, obtains talking about art identification model in violation of rules and regulations.
Module 20 is obtained, for obtaining the call audio in message registration library between customer service and client；
Speech recognition module 30 obtains multiple speech frames with timing for carrying out framing to the call audio；According to
Timing successively extracts the sound characteristic of the speech frame and generates multidimensional sound spy's vector comprising acoustic information；By the multidimensional
Sound characteristic vector inputs preset acoustic model and is handled, the corresponding phoneme information of output speech frame；Believed based on the phoneme
Breath, searches preset dictionary, exports the corresponding word of each phoneme information or word；According to output sequence by the corresponding word of each phoneme information or
Word inputs preset language model and is handled, and exports single word or word is mutually related probability；By the word of the maximum probability of output
Or word is spliced into dialog context and the output of text formatting；
Art identification module 40 is talked about, the dialog context input between the customer service and client for exporting speech recognition is described to be disobeyed
Rule words art identification model carries out words art identification in violation of rules and regulations, exports recognition result；Based on the recognition result, the dialog context is determined
In with the presence or absence of in violation of rules and regulations words art；
It is associated with preserving module 50, for talking about art in violation of rules and regulations if it exists, is then carried out violation words art with corresponding call audio
Association, and be saved in the message registration library.
Based on embodiment description identical with aforementioned present invention violation words art detection method, therefore the present embodiment is to separated
The embodiment content of rule words art detection device, which is not done, excessively to be repeated.
The present embodiment uses speech recognition technology, audio of conversing is identified as to the dialog context of text formatting, then to text
The dialog context of word format carries out words art in violation of rules and regulations and detects, if detecting words art in violation of rules and regulations, subtracts words art in violation of rules and regulations automatically and converses with corresponding
Audio is associated preservation, to realize the automation quality inspection for talking about art in violation of rules and regulations, avoids the cumbersome time-consuming of artificial quality inspection, improves
Quality inspection working efficiency and save cost.
It is the functional block diagram that the present invention talks about art detection device second embodiment in violation of rules and regulations referring to Fig. 5, Fig. 5.This implementation
In example, the violation words art detection device includes:
Keyword obtains module 60, talks about art keyword for obtaining the violation inputted in preset query page；
Retrieval module 70, for the message registration library being retrieved, with inspection using violation words art keyword as querying condition
Surveying the message registration library whether there is the call audio that art keyword association is talked about with the violation；
Playing module 80, if for the message registration inventory in the conversation voice with violation words art keyword association
Frequently, then the call audio with violation words art keyword association is played.
In the present embodiment, it is further provided query page, in order to which user flexibility is inspected by random samples.User can arbitrarily disobey
Rule words art retrieves violation message registration library as key word of the inquiry, thus what acquisition matched with target violation words art
Call audio simultaneously plays automatically, improves retrieval flexibility.
The present invention also provides a kind of computer readable storage mediums.
In the present embodiment, words art detection program in violation of rules and regulations, the violation words are stored on the computer readable storage medium
Art detection program realizes the step of violation words art detection method as described in the examples such as any of the above-described when being executed by processor.
Wherein, the method realized when words art detection program is executed by processor in violation of rules and regulations can refer to the art detection method of words in violation of rules and regulations of the invention
Each embodiment, therefore no longer excessively repeat.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can be realized by means of software and necessary general hardware platform, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Based on this understanding, technical solution of the present invention substantially in other words does the prior art
The part contributed out can be embodied in the form of software products, which is stored in a storage medium
In (such as ROM/RAM), including some instructions are used so that a terminal (can be mobile phone, computer, server or network are set
It is standby etc.) execute method described in each embodiment of the present invention.
The embodiment of the present invention is described with above attached drawing, but the invention is not limited to above-mentioned specific
Embodiment, the above mentioned embodiment is only schematical, rather than restrictive, those skilled in the art
Under the inspiration of the present invention, without breaking away from the scope protected by the purposes and claims of the present invention, it can also make very much
Form, it is all using equivalent structure or equivalent flow shift made by description of the invention and accompanying drawing content, directly or indirectly
Other related technical areas are used in, all of these belong to the protection of the present invention.
Claims (10)
1. art detection method is talked about in a kind of violation, which is characterized in that detection method includes the following steps for the violation words art:
It using the dialog context comprising violation words art as training sample, is trained using machine learning mode, obtains talking about art in violation of rules and regulations
Identification model；
Obtain the call audio in message registration library between customer service and client；
Framing is carried out to the call audio, obtains multiple speech frames with timing；
The sound characteristic of the speech frame is successively extracted according to timing and generates multidimensional sound spy's vector comprising acoustic information；
The multidimensional sound characteristic vector is inputted preset acoustic model to handle, the corresponding phoneme information of output speech frame；
Based on the phoneme information, preset dictionary is searched, exports the corresponding word of each phoneme information or word；
The corresponding word of each phoneme information or word preset language model is inputted according to output sequence to handle, export single word or
Word is mutually related probability；
The word of the maximum probability of output or word are spliced into dialog context and the output of text formatting；
Dialog context between the customer service and client of speech recognition output is inputted the violation words art identification model to carry out in violation of rules and regulations
Art identification is talked about, recognition result is exported；
Based on the recognition result, determine in the dialog context with the presence or absence of words art in violation of rules and regulations；
It talks about art in violation of rules and regulations if it exists, is then associated violation words art with corresponding call audio, and be saved in the call note
It records in library.
2. words art detection method in violation of rules and regulations as described in claim 1, which is characterized in that the violation words art detection method is also wrapped
It includes:
It talks about art in violation of rules and regulations if it exists, then exports the relevant information with the associated audio of conversing of violation words art, the relevant information packet
It includes: customer service information, customer information and topic of call.
3. words art detection method in violation of rules and regulations as described in claim 1, which is characterized in that in the call to talk about art comprising violation
Holding is training sample, is trained using machine learning mode, obtains words art identification model in violation of rules and regulations and includes:
Using the dialog context comprising violation words art as training sample, the training sample is segmented, corresponding participle is obtained
Data；
All participle data are mapped to term vector to be trained；
The mathematical model based on Single hidden layer feedforward neural networks is constructed, and take the term vector as the input of the mathematical model,
Take the violation words art in the training sample as the output of the mathematical model, training is iterated to the mathematical model, is obtained
Art identification model is talked about to violation.
4. art detection method is talked about in violation as claimed in any one of claims 1-3, which is characterized in that in the violation if it exists
Talk about art, then the step for the violation words art and corresponding call audio being associated, and being saved in the message registration library it
Afterwards, further includes:
Obtain the violation words art keyword inputted in preset query page；
Using violation words art keyword as querying condition, the message registration library is retrieved, is to detect the message registration library
The no call audio existed with violation words art keyword association；
If it exists, then the call audio with violation words art keyword association is played.
5. words art detection method in violation of rules and regulations as claimed in claim 4, which is characterized in that described to be with violation words art keyword
Querying condition retrieves the message registration library, whether there is and violation words art keyword with detecting the message registration library
Associated call audio includes:
Character splicing is carried out to audio of respectively conversing in the message registration library associated violation words art, to form words art character in violation of rules and regulations
String, and violation words art character string is passed in memory；
Using violation words art keyword as querying condition, the violation words art character string is retrieved, to detect the message registration
Library whether there is the call audio with violation words art keyword association.
6. art detection device is talked about in a kind of violation, which is characterized in that the violation talks about art detection device and includes:
Model training module, for being carried out using machine learning mode using the dialog context comprising violation words art as training sample
Training obtains talking about art identification model in violation of rules and regulations；
Module is obtained, for obtaining the call audio in message registration library between customer service and client；
Speech recognition module obtains multiple speech frames with timing for carrying out framing to the call audio；According to timing according to
The secondary sound characteristic for extracting the speech frame simultaneously generates multidimensional sound spy's vector comprising acoustic information；The multidimensional sound is special
Sign vector inputs preset acoustic model and is handled, the corresponding phoneme information of output speech frame；Based on the phoneme information, search
Preset dictionary exports the corresponding word of each phoneme information or word；The corresponding word of each phoneme information or word are inputted according to output sequence
Preset language model is handled, and exports single word or word is mutually related probability；The word of the maximum probability of output or word are spelled
It is connected in the dialog context of text formatting and output；
Art identification module is talked about, the dialog context between the customer service and client for exporting speech recognition inputs the violation and talks about art
Identification model carries out words art identification in violation of rules and regulations, exports recognition result；Based on the recognition result, determine in the dialog context whether
Art is talked about in the presence of violation；
It is associated with preserving module, for talking about art in violation of rules and regulations if it exists, is then associated violation words art with corresponding call audio, and
It is saved in the message registration library.
7. words art detection device in violation of rules and regulations as claimed in claim 6, which is characterized in that the model training module is specifically used for:
Using the dialog context comprising violation words art as training sample, the training sample is segmented, corresponding participle is obtained
Data；
All participle data are mapped to term vector to be trained；
The mathematical model based on Single hidden layer feedforward neural networks is constructed, and take the term vector as the input of the mathematical model,
Take the violation words art in the training sample as the output of the mathematical model, training is iterated to the mathematical model, is obtained
Art identification model is talked about to violation.
8. art detection device is talked about in violation as claimed in claims 6 or 7, which is characterized in that the violation words art detection device is also
Include:
Keyword obtains module, talks about art keyword for obtaining the violation inputted in preset query page；
Retrieval module, for the message registration library being retrieved, described in detection using violation words art keyword as querying condition
Message registration library whether there is the call audio with violation words art keyword association；
Playing module, if being broadcast for the message registration inventory in the call audio with violation words art keyword association
Put the call audio with violation words art keyword association.
9. art detection device is talked about in a kind of violation, which is characterized in that the violation words art detection device include memory, processor with
And it is stored in the violation words art detection program that can be run on the memory and on the processor, the violation words art detection
Program realizes the step of violation words art detection method according to any one of claims 1 to 5 when being executed by the processor.
10. a kind of computer readable storage medium, which is characterized in that be stored on the computer readable storage medium and talk about in violation of rules and regulations
Art detects program, and the violation words art detection program is realized according to any one of claims 1 to 5 when being executed by processor
The step of words art detection method in violation of rules and regulations.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910411437.7A CN110310663A (en) | 2019-05-16 | 2019-05-16 | Words art detection method, device, equipment and computer readable storage medium in violation of rules and regulations |
| PCT/CN2019/102261 WO2020228173A1 (en) | 2019-05-16 | 2019-08-23 | Illegal speech detection method, apparatus and device and computer-readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910411437.7A CN110310663A (en) | 2019-05-16 | 2019-05-16 | Words art detection method, device, equipment and computer readable storage medium in violation of rules and regulations |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110310663A true CN110310663A (en) | 2019-10-08 |
Family
ID=68074763
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910411437.7A Pending CN110310663A (en) | 2019-05-16 | 2019-05-16 | Words art detection method, device, equipment and computer readable storage medium in violation of rules and regulations |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN110310663A (en) |
| WO (1) | WO2020228173A1 (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111182162A (en) * | 2019-12-26 | 2020-05-19 | 深圳壹账通智能科技有限公司 | Telephone quality inspection method, device, equipment and storage medium based on artificial intelligence |
| CN111405128A (en) * | 2020-03-24 | 2020-07-10 | 中国—东盟信息港股份有限公司 | Call quality inspection system based on voice-to-text conversion |
| CN111598485A (en) * | 2020-05-28 | 2020-08-28 | 成都晓多科技有限公司 | Multi-dimensional intelligent quality inspection method, device, terminal equipment and medium |
| CN111738835A (en) * | 2020-06-22 | 2020-10-02 | 中国银行股份有限公司 | Monitoring method, device, equipment and storage medium |
| CN112183079A (en) * | 2020-09-07 | 2021-01-05 | 绿瘦健康产业集团有限公司 | Voice monitoring method, device, medium and terminal equipment |
| CN112380364A (en) * | 2020-11-17 | 2021-02-19 | 平安养老保险股份有限公司 | Method and system for file authentication |
| CN112466281A (en) * | 2020-10-13 | 2021-03-09 | 讯飞智元信息科技有限公司 | Harmful audio recognition decoding method and device |
| CN112671985A (en) * | 2020-12-22 | 2021-04-16 | 平安普惠企业管理有限公司 | Agent quality inspection method, device, equipment and storage medium based on deep learning |
| CN113038153A (en) * | 2021-02-26 | 2021-06-25 | 深圳道乐科技有限公司 | Financial live broadcast violation detection method, device and equipment and readable storage medium |
| CN113064983A (en) * | 2021-04-23 | 2021-07-02 | 深圳壹账通智能科技有限公司 | Semantic detection method and device, computer equipment and storage medium |
| CN113593553A (en) * | 2021-07-12 | 2021-11-02 | 深圳市明源云客电子商务有限公司 | Speech recognition method, speech recognition apparatus, speech management server, and storage medium |
| CN113641795A (en) * | 2021-08-20 | 2021-11-12 | 上海明略人工智能(集团)有限公司 | Method and device for dialectical statistics, electronic equipment and storage medium |
| WO2021238700A1 (en) * | 2020-05-28 | 2021-12-02 | 升智信息科技(南京)有限公司 | Recognition method employing speech statement to improve voice recognition accuracy |
| WO2022116438A1 (en) * | 2020-12-01 | 2022-06-09 | 平安科技（深圳）有限公司 | Customer service violation quality inspection method and apparatus, computer device, and storage medium |
| CN115396549A (en) * | 2021-05-25 | 2022-11-25 | 中国联合网络通信集团有限公司 | Method for processing illegal call service device and electronic apparatus |
| WO2023231785A1 (en) * | 2022-05-31 | 2023-12-07 | 支付宝(杭州)信息技术有限公司 | Data processing method, apparatus, and device |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107085700A (en) * | 2017-01-16 | 2017-08-22 | 安徽大学 | A kind of face identification method being combined based on rarefaction representation with neural networks with single hidden layer technology |
| CN107093431A (en) * | 2016-02-18 | 2017-08-25 | 中国移动通信集团辽宁有限公司 | A kind of method and device that quality inspection is carried out to service quality |
| CN107222865A (en) * | 2017-04-28 | 2017-09-29 | 北京大学 | The communication swindle real-time detection method and system recognized based on suspicious actions |
| CN108921342A (en) * | 2018-06-26 | 2018-11-30 | 圆通速递有限公司 | A kind of logistics customer churn prediction method, medium and system |
| CN109151218A (en) * | 2018-08-21 | 2019-01-04 | 平安科技（深圳）有限公司 | Call voice quality detecting method, device, computer equipment and storage medium |
| CN109147768A (en) * | 2018-09-13 | 2019-01-04 | 云南电网有限责任公司 | A kind of audio recognition method and system based on deep learning |
| CN109658923A (en) * | 2018-10-19 | 2019-04-19 | 平安科技（深圳）有限公司 | Voice quality detecting method, equipment, storage medium and device based on artificial intelligence |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10347244B2 (en) * | 2017-04-21 | 2019-07-09 | Go-Vivace Inc. | Dialogue system incorporating unique speech to text conversion method for meaningful dialogue response |
-
2019
- 2019-05-16 CN CN201910411437.7A patent/CN110310663A/en active Pending
- 2019-08-23 WO PCT/CN2019/102261 patent/WO2020228173A1/en active Application Filing
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107093431A (en) * | 2016-02-18 | 2017-08-25 | 中国移动通信集团辽宁有限公司 | A kind of method and device that quality inspection is carried out to service quality |
| CN107085700A (en) * | 2017-01-16 | 2017-08-22 | 安徽大学 | A kind of face identification method being combined based on rarefaction representation with neural networks with single hidden layer technology |
| CN107222865A (en) * | 2017-04-28 | 2017-09-29 | 北京大学 | The communication swindle real-time detection method and system recognized based on suspicious actions |
| CN108921342A (en) * | 2018-06-26 | 2018-11-30 | 圆通速递有限公司 | A kind of logistics customer churn prediction method, medium and system |
| CN109151218A (en) * | 2018-08-21 | 2019-01-04 | 平安科技（深圳）有限公司 | Call voice quality detecting method, device, computer equipment and storage medium |
| CN109147768A (en) * | 2018-09-13 | 2019-01-04 | 云南电网有限责任公司 | A kind of audio recognition method and system based on deep learning |
| CN109658923A (en) * | 2018-10-19 | 2019-04-19 | 平安科技（深圳）有限公司 | Voice quality detecting method, equipment, storage medium and device based on artificial intelligence |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111182162A (en) * | 2019-12-26 | 2020-05-19 | 深圳壹账通智能科技有限公司 | Telephone quality inspection method, device, equipment and storage medium based on artificial intelligence |
| CN111405128A (en) * | 2020-03-24 | 2020-07-10 | 中国—东盟信息港股份有限公司 | Call quality inspection system based on voice-to-text conversion |
| CN111405128B (en) * | 2020-03-24 | 2022-02-18 | 中国—东盟信息港股份有限公司 | Call quality inspection system based on voice-to-text conversion |
| WO2021238700A1 (en) * | 2020-05-28 | 2021-12-02 | 升智信息科技(南京)有限公司 | Recognition method employing speech statement to improve voice recognition accuracy |
| CN111598485A (en) * | 2020-05-28 | 2020-08-28 | 成都晓多科技有限公司 | Multi-dimensional intelligent quality inspection method, device, terminal equipment and medium |
| CN111738835A (en) * | 2020-06-22 | 2020-10-02 | 中国银行股份有限公司 | Monitoring method, device, equipment and storage medium |
| CN112183079A (en) * | 2020-09-07 | 2021-01-05 | 绿瘦健康产业集团有限公司 | Voice monitoring method, device, medium and terminal equipment |
| CN112466281A (en) * | 2020-10-13 | 2021-03-09 | 讯飞智元信息科技有限公司 | Harmful audio recognition decoding method and device |
| CN112380364A (en) * | 2020-11-17 | 2021-02-19 | 平安养老保险股份有限公司 | Method and system for file authentication |
| WO2022116438A1 (en) * | 2020-12-01 | 2022-06-09 | 平安科技（深圳）有限公司 | Customer service violation quality inspection method and apparatus, computer device, and storage medium |
| CN112671985A (en) * | 2020-12-22 | 2021-04-16 | 平安普惠企业管理有限公司 | Agent quality inspection method, device, equipment and storage medium based on deep learning |
| CN113038153A (en) * | 2021-02-26 | 2021-06-25 | 深圳道乐科技有限公司 | Financial live broadcast violation detection method, device and equipment and readable storage medium |
| CN113064983A (en) * | 2021-04-23 | 2021-07-02 | 深圳壹账通智能科技有限公司 | Semantic detection method and device, computer equipment and storage medium |
| CN113064983B (en) * | 2021-04-23 | 2024-04-26 | 深圳壹账通智能科技有限公司 | Semantic detection method, semantic detection device, computer equipment and storage medium |
| CN115396549A (en) * | 2021-05-25 | 2022-11-25 | 中国联合网络通信集团有限公司 | Method for processing illegal call service device and electronic apparatus |
| CN113593553A (en) * | 2021-07-12 | 2021-11-02 | 深圳市明源云客电子商务有限公司 | Speech recognition method, speech recognition apparatus, speech management server, and storage medium |
| CN113593553B (en) * | 2021-07-12 | 2022-05-24 | 深圳市明源云客电子商务有限公司 | Voice recognition method, voice recognition apparatus, voice management server, and storage medium |
| CN113641795A (en) * | 2021-08-20 | 2021-11-12 | 上海明略人工智能(集团)有限公司 | Method and device for dialectical statistics, electronic equipment and storage medium |
| WO2023231785A1 (en) * | 2022-05-31 | 2023-12-07 | 支付宝(杭州)信息技术有限公司 | Data processing method, apparatus, and device |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2020228173A1 (en) | 2020-11-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110310663A (en) | Words art detection method, device, equipment and computer readable storage medium in violation of rules and regulations | |
| CN112804400B (en) | Customer service call voice quality inspection method and device, electronic equipment and storage medium | |
| US8145482B2 (en) | Enhancing analysis of test key phrases from acoustic sources with key phrase training models | |
| US7606714B2 (en) | Natural language classification within an automated response system | |
| CN111128223B (en) | Text information-based auxiliary speaker separation method and related device | |
| US8311824B2 (en) | Methods and apparatus for language identification | |
| WO2021068843A1 (en) | Emotion recognition method and apparatus, electronic device, and readable storage medium | |
| CN102623011B (en) | Information processing apparatus, information processing method and information processing system | |
| CN110135879B (en) | Customer service quality automatic scoring method based on natural language processing | |
| CN112468659B (en) | Quality evaluation method, device, equipment and storage medium applied to telephone customer service | |
| CN109544104A (en) | A kind of recruitment data processing method and device | |
| KR20080024752A (en) | Dialog management apparatus and method for chatting agent | |
| CN112800743A (en) | Voice scoring model construction system and method based on specific field | |
| JP2020071675A (en) | Speech summary generation apparatus, speech summary generation method, and program | |
| CN114818649A (en) | Service consultation processing method and device based on intelligent voice interaction technology | |
| CN116631412A (en) | Method for judging voice robot through voiceprint matching | |
| CN113434670A (en) | Method and device for generating dialogistic text, computer equipment and storage medium | |
| JP2020071676A (en) | Speech summary generation apparatus, speech summary generation method, and program | |
| JP6605105B1 (en) | Sentence symbol insertion apparatus and method | |
| JP4067481B2 (en) | Telephone reception system | |
| CN107886940A (en) | Voiced translation processing method and processing device | |
| CN115641850A (en) | Method and device for recognizing ending of conversation turns, storage medium and computer equipment | |
| CN112002306B (en) | Speech class recognition method and device, electronic equipment and readable storage medium | |
| US11196864B1 (en) | Analyzing voice response to telephone call to assign appropriate agent | |
| Lee et al. | A study on natural language call routing |
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 |