CN109658939A - A kind of telephonograph access failure reason recognition methods - Google Patents
A kind of telephonograph access failure reason recognition methods Download PDFInfo
- Publication number
- CN109658939A CN109658939A CN201910076136.3A CN201910076136A CN109658939A CN 109658939 A CN109658939 A CN 109658939A CN 201910076136 A CN201910076136 A CN 201910076136A CN 109658939 A CN109658939 A CN 109658939A
- Authority
- CN
- China
- Prior art keywords
- access failure
- telephonograph
- audio
- reason
- phone
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 70
- 230000011664 signaling Effects 0.000 claims abstract description 23
- 238000013145 classification model Methods 0.000 claims abstract description 9
- 238000004458 analytical method Methods 0.000 claims description 21
- 238000010276 construction Methods 0.000 claims description 15
- 238000012549 training Methods 0.000 claims description 10
- 238000006073 displacement reaction Methods 0.000 claims description 9
- 238000009825 accumulation Methods 0.000 claims description 8
- 238000000605 extraction Methods 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- 230000010354 integration Effects 0.000 claims description 5
- 230000008901 benefit Effects 0.000 claims description 4
- 230000005611 electricity Effects 0.000 claims description 2
- 238000005516 engineering process Methods 0.000 description 8
- 230000000694 effects Effects 0.000 description 4
- 230000006855 networking Effects 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 230000018109 developmental process Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000012216 screening Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000009411 base construction Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000010183 spectrum analysis 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
- G10L15/26—Speech to text systems
-
- 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
- G10L15/22—Procedures used during a speech recognition process, e.g. man-machine dialogue
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04M—TELEPHONIC COMMUNICATION
- H04M1/00—Substation equipment, e.g. for use by subscribers
- H04M1/64—Automatic arrangements for answering calls; Automatic arrangements for recording messages for absent subscribers; Arrangements for recording conversations
- H04M1/65—Recording arrangements for recording a message from the calling party
-
- 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
-
- 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
- H04M3/487—Arrangements for providing information services, e.g. recorded voice services or time announcements
Landscapes
- Engineering & Computer Science (AREA)
- Signal Processing (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Telephonic Communication Services (AREA)
- Telephone Function (AREA)
Abstract
The invention belongs to field of speech recognition, and in particular to a kind of phone access failure reason recognition methods.This method comprises: marking access failure reason by signaling；If failing to classify to obtain reason by signaling, audio fingerprint feature sequence is extracted from telephonograph to be identified, and retrieved in audio fingerprint database using the sequence；Matched fingerprint is such as found, then according to the access failure reason label in fingerprint key assignments, carries out access failure reason mark for phone to be identified；If failing to find matched fingerprint, audio content is identified as content of text by automatic speech recognition, file classification method is utilized based on the content, is classified in access failure reason document classification model, marks the telephonograph to be identified with the access failure causality classification result that classification obtains.This method can to recording file carry out identified off-line, can also convection type call voice identified, versatility with higher, the different application scene of applicable call center.
Description
Technical field
The invention belongs to speech processes fields, are related to audio-frequency fingerprint identification field and telephone outbound call business, and in particular to one
Kind phone access failure reason recognition methods.
Background technique
Active telephone outbound call is a kind of important value-added services businesses in telecommunications industry, carries such as public information and pushes away
It send, promoting service, all more important service content such as Communication with Customer and service follow-up.
In active telephone outbound call business, customer number number is most important service resources, the benefit of business can significantly by
To the restriction of coding resource utilization rate, especially for the number for failing access success, processing strategie is even more to directly influence
The operation level of business.Such as: if taking simple abandoning strategy to the code number of abandoned call, preciousness on the one hand can be wasted
Coding resource, on the other hand, for some critical outbound call services (such as Communication with Customer, complaint and feedback and service are followed up),
It is more likely to lead to great communication fault even business accident；In contrast, if rambunctiously taken repeatedly access failure number
The strategy of calling then not only will lead to valuable line resource and repeat to occupy with being deactivated, but will bring negative user experience,
The refusal of user is even further resulted in, resists and even complains.Therefore, failure cause is analyzed for loss call, according to
Reason takes different processing strategies, is a primary demand of outgoing call industry, and obtains the access failure reason of calling, then is real
The basis of this existing demand.
Obtaining the most direct mode of access failure is online progress telephone signaling analysis, i.e., tracking indicates to exhale in calling procedure
It is the signaling of state, and access failure original is judged by status signaling that access telecommunications gateway returns according in the case of phone access failure
Cause, this mode accuracy rate is high, does not need additional software and hardware resources substantially, networking switching equipment not high in networking complexity
Be arranged it is relatively simple in the case where have preferable effect.However, since the business of current telecommunications industry is multifarious, networking side
Formula is varied, while also needing often to take into account multiple business demand.In this case, the network environment of call center is often
Complex, networking signaling protocol also have different degrees of expansion and change, cause caller in many cases can not be steady
Determine the signaling in accurately being called, and the access failure prompt tone provided with audible can only be provided in exchange of media.
In this case, differentiate that access failure reason becomes unique selection based on access failure prompt tone.
In traditional lesser call center of scale, the operation of outgoing call is completed by call center's seat personnel,
While operation, seat personnel can listen the ring-back tone, busy tone and voice prompting debated in phone, accordingly to access failure reason into
Row Direct Mark.However, for fairly large call center, in order to more effectively realize management, it will usually which use will call
With the call policy of service cutting, i.e., dialled by outer calling platform is unified, again by call forwarding to idle seat after closing of the circuit
Seat personnel, in this case, the phone of access failure are not transferred to seat personnel, thus can only be led to by special staff
It crosses to listen and distinguishes that the mode recorded offline marks access failure reason, cause the significant wastage of manpower.Importantly, with artificial intelligence
Horizontal is constantly progressive, there has been by computer generation for the artificial outer paging system of intelligence, it is automatic by computer system
Access failure causality classification is carried out to access failure phone, manpower consumption can be greatly saved, reduces call center's operation cost, is improved
The working efficiency of call center.
China mainland area, in the case where phone access failure, interchanger return telephone prompts sound be generally divided into
Lower two kinds of situations: one is the single-frequency automatic telephone switching network signal sounds defined in national standard GB 3380-82 standard, another
It is then the voice broadcast recording specified by operator, such as " phone that you dial is just busy ", " the phone out of reach that you dial "
Deng.In addition, being then the mixing of two kinds of prompt tones, such as voice prompting after jingle bell sound in more situations.For meeting national standard
Single-frequency switching network signal sound can be identified, for the latter, the most common means are logical by simple frequency spectrum analysis method
Artificial mark is crossed, that is, by manually according to the busy tone in recording, jingle bell sound and voice prompting, being marked to access failure reason
Note.Since this method needs manually to listen to complete access failure telephonograph, and one can only be handled in everyone same time
It takes on the telephone, therefore efficiency is extremely low.When call center's business expansion and the traffic rise to certain scale, this artificial mode
The efficiency and reliability of service operation can seriously be reduced.In this case, it is recorded by computer system efficient process access failure
Sound audio, and access failure reason is detected as inevitable choice.Taking this as an opportunity, the outgoing call access failure based on artificial intelligence technology
Reason detection becomes the technology for having significant application value.
Summary of the invention
To provide accurate phone access failure reason, the present invention provides a kind of numbers for the access failure phone analysis of causes
According to base construction method, include the following steps:
Obtain a certain number of access failure telephonographs；
Access failure reason mark is carried out to access failure telephonograph, obtains access failure telephonograph-access failure reason labeled data
Library；
Extraction audio fingerprint feature sequence in telephonograph is never turned on, and using corresponding access failure reason as key assignments, is obtained
Access failure telephonograph-audio fingerprint feature database.
A kind of database construction method for the access failure phone analysis of causes as described above, in which: to access failure electricity
Words recording carries out speech recognition and obtains text information, then carries out access failure reason text classification modeling to text information, obtains not
It connects prompt tone textual classification model and is stored in database.
A kind of database construction method for the access failure phone analysis of causes as described above, in which: according to online letter
It enables or voice content marks access failure reason.
A kind of database construction method for the access failure phone analysis of causes as described above, in which: the audio refers to
Audio-frequency fingerprint of the line characteristic sequence using the time-frequency domain differential code coding of fixed length as single frames voice signal.
A kind of database construction method for the access failure phone analysis of causes as described above, in which: for individually recording
Sound file combines the fingerprint characteristic sequence that each frame time information is constituted entire recording file with each frame audio-frequency fingerprint.
A kind of database construction method for the access failure phone analysis of causes as described above, in which: the database
Based on hashed table index structure, the index key assignments of the hashed table index structure is audio fingerprint feature, the hashed table index
The location contents that is indexed of structure is audio frame information corresponding to key assignments, and the audio frame information includes the recording of place access failure
Position in recording of text information and corresponding audio frame.
The invention also includes a kind of phone access failure reason recognition methods, comprising:
Access failure reason is marked by signaling；
If failing to classify to obtain reason by signaling, audio fingerprint feature sequence, and benefit are extracted from telephonograph to be identified
It is retrieved in audio fingerprint database with the sequence；Matched fingerprint is such as found, then according to not connecing in fingerprint key assignments
Logical reason label carries out access failure reason mark for phone to be identified；
If failing to find matched fingerprint, audio content is identified as by content of text by automatic speech recognition, is based on the text
This content utilizes file classification method, classifies in access failure reason document classification model, the access failure obtained with classification
Causality classification result marks the telephonograph to be identified.
The invention also includes another phone access failure reason recognition methods, comprising:
Audio-frequency fingerprint is extracted to access failure telephonograph frame by frame, obtains telephonograph audio-frequency fingerprint sequence to be identified;
To the telephonograph audio-frequency fingerprint sequence to be identified, single frames fuzzy search is carried out in audio fingerprint database, is obtained
The characteristic frame of candidate original recording audio fingerprint feature；
The characteristic frame of the candidate original recording audio fingerprint feature is integrated, and is referred to telephonograph audio to be identified
Line sequence carries out sequences match, obtains candidate original recording matching arrangement set;
By the orderly candidate original recording matching arrangement set in access failure telephonograph-access failure reason data library into
Row retrieval, it is candidate to select multiple access failure reasons；
The multiple access failure reason candidate is screened, most probable phone access failure reason is obtained.
A kind of phone access failure reason recognition methods as described above, in which: when carrying out the single frames fuzzy search, according to
Frequency band energy determination may be disturbed and several fingerprint digits of mismatch occurs, negated by number of bits combination and generate a series of expansions
Candidate fingerprint is filled, joint original fingerprint is retrieved together.
A kind of phone access failure reason recognition methods as described above, in which: carrying out the audio fingerprint feature integration
When with sequences match, the characteristic frame of the candidate original recording audio fingerprint feature and the spy of telephonograph to be matched are calculated first
Relative displacement of the frame in respective sequence is levied, when the accounting in the number of matches under identical relative displacement in original phone recording
When meeting threshold requirement, the characteristic frame with identical relative displacement is integrated into the segment in database in original telephonograph,
As candidate original recording matches sequence, and using accounting of the number of matches in original phone recording as described candidate former
The match hit rate score for the recording matching sequence that begins.
A kind of phone access failure reason recognition methods as described above, in which: to the multiple access failure reason candidate into
When row screening, the accumulation of the match hit rate score and each candidate access failure reason of the candidate original recording matching sequence is utilized
Score obtains matching score, is ranked up with the matching score to candidate access failure reason, as highest scoring person is higher than in advance
The decision-making value of setting is then selected as final access failure reason result；The accumulation score refers to the orderly candidate
When original recording matching arrangement set is retrieved in the database, the similarity evaluation to the part matched.
A kind of phone access failure reason recognition methods as described above, in which: the setting method of the decision-making value are as follows: logical
One group of test data independently of training data is crossed, it is former to calculate the test data practical access failure in audio fingerprint database
The matching score of cause and other access failure reasons is distributed, and calculates threshold value according to optimal distinction principle.
The invention also includes a kind of phone access failure reason recognition methods, including one kind to be directed to the analysis of causes of access failure phone
Database sharing step；And the identification step to access failure telephonograph in library based on the data.
A kind of phone access failure reason recognition methods as described above, in which: it further include database update method, it is specific to wrap
Include following steps:
With newly-increased access failure telephonograph and access failure phone reason, Lai Gengxin access failure telephonograph-access failure reason mark
Infuse database；
With newly-increased access failure telephonograph and its audio-frequency fingerprint, Lai Gengxin access failure telephonograph-audio fingerprint feature data
Library；
With newly-increased access failure telephonograph and its text information, the textual classification model of Lai Gengxin access failure reason；
With newly-increased access failure telephonograph and its signaling information, the signaling table of Lai Gengxin access failure reason；
The newly-increased access failure telephonograph refers to identify the access failure telephonograph of access failure reason.
A kind of phone access failure reason recognition methods as described above, in which:
If there is the segment not being matched in the newly-increased access failure telephonograph, the segment is taken out as corresponding access failure
The training characteristics sequence of reason updates access failure telephonograph-audio fingerprint feature database.
Compared with the conventional method, the invention has the following advantages that
The present invention is based on audio fingerprint techniques to establish access failure reason audio database, comprehensively utilizes speech recognition technology and audio
Fingerprint technique, using initial stage, manually mark generates initial discrimination model and audio frequency feature library on a small quantity, later period iteration more new model
Strategy；On the one hand the audio fingerprint techniques for utilizing efficiently and accurately record to the access failure of the successful match in audio frequency feature library straight
It connects and provides court verdict, on the other hand using speech recognition keyword extraction and combine the method manually corrected, to failing to match
To access failure prompt tone classify, and by sorted audio be added audio frequency feature library.It by this method, can be few
The automatic discrimination of outer calling telephone access failure reason is efficiently and accurately realized under conditions of manual intervention.
By utilizing single frames fuzzy retrieval method, doubtful or similar audio can be never turned on into reason audio database
In screen, carry out subsequent integration and the matching of long sequence on this basis, reduce missing inspection to greatest extent, improve system
Recall rate.Audio to be measured is split as frame level characteristics and carries out fuzzy matching by this method, and successive constraint, has in no timing
Concurrent operations property improves system retrieval speed, provides technical foundation for real-time audio retrieval.
By can integrate multi-level clue using integration and sequences match method, expanded to from single frames matching longer
Frame sequence matching.On the one hand the fault-tolerance that system can be improved also has the audio for having noise jamming or channel effect preferable suitable
With property；On the other hand data mark cost can be reduced using lesser mark unit, increases the scope of application of system.
Save manually, cut operating costs: only need to manually carry out minimal amount of listen debates mark to this system during operation,
And it can uninterruptedly be handled with 24 hours.
Rapidly and efficiently: the automatic discrimination technology that this method uses speech recognition to combine with audio-frequency fingerprint, it is common in separate unit
Multi-pass access failure phone can be handled on the computer of configuration simultaneously, and the distinguishing speed often taken on the telephone significantly larger than manually is listened and debated
Speed.
This method can to recording file carry out identified off-line, can also convection type call voice identified, it is with higher
Versatility, the different application scene of applicable call center.
Detailed description of the invention
Fig. 1 is a kind of a kind of implementation of database construction method for the access failure phone analysis of causes provided by the invention
The flow chart of example；
Fig. 2 is a kind of another embodiment of database construction method for the access failure phone analysis of causes provided by the invention
Flow chart；
Fig. 3 is a kind of a kind of flow chart of embodiment of phone access failure reason recognition methods provided by the invention；
Fig. 4 is a kind of flow chart of another embodiment of phone access failure reason recognition methods provided by the invention.
Specific embodiment
The present invention is specifically described below with reference to embodiment.
Firstly, being explained as follows to the relevant technical term of the present invention:
1, " audio fingerprint techniques "
Audio fingerprint techniques (Audio fingerprinting technology) refer to a segment of audio through specific algorithm
In unique numerical characteristic extracted in the form of identifier, the sample sound of magnanimity or tracking and positioning sample for identification
Originally position in the database.Core algorithm of the audio-frequency fingerprint as content automatic identification technology, is widely used to music
Identification, content of copyright prison are broadcast, the fields such as content library duplicate removal and the interaction of the second screen of TV.
Audio fingerprint techniques will need the audio of identified content and foundation to refer to by extracting the data characteristics in sound
Completion is compared in line database.Identification process is not by the saving format of audio itself, coding mode, code rate and compress technique
It influences.The matching of audio-frequency fingerprint is the matching of high precision, independent of file meta information, watermarking and file cryptographic Hash.
The specific implementation of audio fingerprint feature includes a variety of methods, such as: hash method, cepstrum, wavelet analysis etc..
2, " automatic speech recognition ":
In recent years, development and big data language since especially 2009, by the deep learning research of machine learning field
The accumulation of material, the development that speech recognition technology is advanced by leaps and bounds.On the one hand, with machine learning field deep learning research quilt
It is introduced into voice recognition acoustic model training, the accuracy rate of acoustic model is greatly improved, becomes in the past 20 years
Most fast progress in terms of speech recognition technology.
On the other hand, the speech recognition decoder of mainstream has used the solution based on finite state machine (WFST) mostly at present
Code network, which, which can share sound word collection Unified Set language model, dictionary and acoustics, becomes a big decoding net
Network substantially increases decoded speed, provides the foundation for the real-time application of speech recognition.
It, at present can be from multiple in addition, With the fast development of internet and the popularization and application of the mobile terminals such as mobile phone
Channel obtains the corpus in terms of a large amount of texts or voice, this is provided for the training of language model and acoustic model in speech recognition
Resource abundant, makes it possible the general extensive language model of building and acoustic model.
3, Text Classification
The definition of text classification problem is that corresponding class is selected in predefined category label according to the content of a document
Not.The basic step of Chinese Text Categorization is Chinese word segmentation, feature extraction, training pattern, prediction classification, based on statistics
Text classification be typically necessary the relatively good corpus marked as training set, model is trained, using model to not
The text of classification is classified.Common statistics feature mainly has chi-square statistics, information gain, mutual information, probability ratio, hands over
Pitch the methods of entropy.
Fig. 1 is a kind of database construction method for the access failure phone analysis of causes provided by the invention, including as follows
Step:
Step S11: a certain number of access failure telephonographs are obtained；
Step S12: access failure reason mark is carried out to access failure telephonograph, obtains access failure telephonograph-access failure reason
Mark database.When mark, can by telephony access platform by signaling differentiate access failure phone the reason of, can also be according to access failure
Prompt tone (such as the contents such as " shutdown ", " line be busy ") differentiates the reason of access failure phone.
Step S13: extraction audio fingerprint feature sequence in telephonograph is never turned on, and with corresponding access failure reason
As key assignments, access failure telephonograph-audio fingerprint feature database is obtained.
Access failure telephonograph may include the information such as access failure prompt tone and ring back tone, CRBT, to obtain more preferably
Effect can remove the information such as CRBT, only retain access failure prompt tone as access failure telephonograph.
Wherein, audio fingerprint feature sequence can be encoded using the time-frequency domain differential code of fixed length as single frames voice signal
Audio-frequency fingerprint.
For single recording file, each frame time information can also be combined with each frame audio-frequency fingerprint and be constituted entire recording file
Fingerprint characteristic sequence.
More preferably, as shown in Fig. 2, step S24 can also be carried out: carrying out voice to access failure phone or access failure prompt tone
Identification obtains corresponding text information, and text information is stored in database.Also reason is not answered using telephonograph is corresponding
Text classification is carried out to text information.It is that every a kind of access failure reason extracts text classification based on identification text when text classification
Feature forms access failure prompt tone textual classification model.
More preferably, database is established based on hashed table index structure.The index key assignments of hashed table index structure is audio
Fingerprint characteristic, being indexed location contents is audio frame information corresponding to key assignments；Wherein audio frame information includes place access failure
The position of the text information of recording and corresponding audio frame in recording.
As shown in figure 3, a kind of phone access failure reason recognition methods, specifically comprises the following steps:
Step S31: for access failure telephonograph to be identified, access failure reason is marked by signaling；
Step S32: for the access failure telephonograph to be identified for failing to classify by signaling, from access failure phone to be identified
Audio fingerprint feature sequence is extracted in recording, is retrieved in audio fingerprint database；Matched fingerprint is such as found, then basis
Access failure reason label in fingerprint key assignments carries out access failure reason mark for access failure telephonograph to be identified；
Step S33: if do not found matched fingerprint, audio content is identified as by content of text by automatic speech recognition, is based on
Text content utilizes file classification method, classifies in access failure reason document classification model, is obtained not with classification
It connects causality classification result and marks the access failure telephonograph to be identified.
File classification method can be bayes method, traditional decision-tree, neural network method etc..If the text of telephonograph
This length is shorter, then can obtain preferable effect using Naive Bayes Classification method.
A kind of database update method for the access failure phone analysis of causes is given in Fig. 3, is specifically included as follows
Step:
Step S34: with newly-increased access failure telephonograph and access failure phone reason, Lai Gengxin access failure telephonograph-do not connect
Logical reason mark database；
Step S35: with newly-increased access failure telephonograph and its audio-frequency fingerprint, Lai Gengxin access failure telephonograph-audio-frequency fingerprint
Property data base；
Step S36: with newly-increased access failure telephonograph and its text information, the text classification mould of Lai Gengxin access failure reason
Type；
Step S37: with newly-increased access failure telephonograph and its signaling information, the signaling table of Lai Gengxin access failure reason.
Wherein, newly-increased access failure telephonograph refers to identify the access failure telephonograph of access failure reason.
If there is the segment not being matched in newly-increased access failure telephonograph, it is former as corresponding access failure to take out the segment
The training characteristics sequence of cause updates access failure telephonograph-audio fingerprint feature database.
Fig. 4 is a kind of phone access failure reason recognition methods, is specifically comprised the following steps:
Step S41: extracting audio-frequency fingerprint to access failure telephonograph frame by frame, obtains telephonograph audio-frequency fingerprint sequence to be identified.
Step S42: it to the telephonograph audio-frequency fingerprint sequence to be identified of generation, is carried out in audio fingerprint database frame by frame
Retrieval, obtains the characteristic frame of candidate original recording audio fingerprint feature.
Retrieval can be fuzzy search, such as may be disturbed according to frequency band energy determination and several fingerprints of mismatch occur
Digit is negated by number of bits combination and generates a series of expansion candidate fingerprints, and joint original fingerprint is retrieved together, to reduce
Because of mismatch caused by local interference.
Step S43: integrating the characteristic frame of obtained candidate original recording audio fingerprint feature, and with it is to be identified
Telephonograph audio-frequency fingerprint sequence carries out sequences match, obtains candidate original recording matching arrangement set.
More preferably, when carrying out the audio fingerprint feature integration with sequences match, candidate original recording sound is calculated first
The relative displacement of the characteristic frame of frequency fingerprint characteristic and the characteristic frame of telephonograph to be matched in respective sequence, when identical opposite
When accounting of the number of matches in original phone recording under displacement meets threshold requirement, by the feature with identical relative displacement
Frame is integrated into the segment in database in original telephonograph, i.e., candidate original recording matches sequence, and by the number of matches
Match hit rate score of the accounting as the candidate original recording matching sequence in original phone recording.
Step S44: by obtained orderly candidate original recording matching arrangement set in access failure telephonograph-access failure
Reason data is retrieved in library, and it is candidate to select multiple access failure reasons.
Step S45: screening obtained multiple access failure reason candidates, and it is former to obtain most probable phone access failure
Cause.
When screening to multiple access failure reason candidates, the match hit of each candidate original recording matching sequence is utilized
The accumulation score of rate score and each candidate access failure reason obtains matching score, and being ranked up to candidate access failure reason (can be with
It is added, is multiplied with accumulation score with match hit rate score, the methods of weighted sum obtains the matching score for sequence), such as
Highest scoring person is higher than preset decision-making value, then is selected as final access failure reason result.
Wherein, accumulation score refers to examines the orderly candidate original recording matching arrangement set in the database
Suo Shi, the similarity evaluation to the part that can be matched.
Preset decision-making value can be empirical value, can also use following setting methods: by one group independently of
The test data of training data calculates the test data practical access failure reason in audio fingerprint database and does not connect with other
The matching score distribution of logical reason, calculates threshold value according to optimal distinction principle.
A kind of phone access failure reason recognition methods may include a kind of database for the access failure phone analysis of causes
Construction step；And the identification step to access failure telephonograph based on the database.
Wherein, database sharing step, comprising: obtain a certain number of access failure telephonographs；Access failure phone is recorded
Sound carries out access failure reason mark, obtains access failure telephonograph-access failure reason mark database；Never turn on telephonograph
Middle extraction audio fingerprint feature sequence, and using corresponding access failure reason as key assignments, it obtains access failure telephonograph-audio and refers to
Line property data base；
Wherein, identification step, comprising: access failure reason is marked by signaling；If failing to classify to obtain reason by signaling, from
Audio fingerprint feature sequence is extracted in telephonograph, is retrieved in audio fingerprint database；Matched fingerprint is such as found, then
According to the access failure reason label in fingerprint key assignments, access failure reason mark is carried out for phone；If failing to find matched fingerprint,
The identification content text segmented by automatic speech recognition, and file classification method is utilized, in access failure reason document point
It is retrieved in class model, marks the telephonograph with the access failure causality classification that retrieval obtains.
In addition, further including the update step to database.
Although the present invention is disclosed as above with preferred embodiment, it is not for limiting the present invention, any this field skill
Art personnel without departing from the spirit and scope of the present invention, can make possible variation and modification, therefore guarantor of the invention
Shield range should be subject to the range that the claims in the present invention are defined.
Claims (15)
1. a kind of database construction method for the access failure phone analysis of causes, includes the following steps:
Obtain a certain number of access failure telephonographs；
Access failure reason mark is carried out to access failure telephonograph, obtains access failure telephonograph-access failure reason labeled data
Library；
Extraction audio fingerprint feature sequence in telephonograph is never turned on, and using corresponding access failure reason as key assignments, is obtained
Access failure telephonograph-audio fingerprint feature database.
2. a kind of database construction method for the access failure phone analysis of causes as described in claim 1, it is characterised in that:
Speech recognition is carried out to access failure telephonograph and obtains text information, then the text classification of access failure reason is carried out to text information and is built
Mould obtains access failure prompt tone textual classification model and is stored in database.
3. a kind of database construction method for the access failure phone analysis of causes as claimed in claim 2, it is characterised in that:
Access failure reason is marked according to online signaling or voice content.
4. a kind of database construction method for the access failure phone analysis of causes as claimed in claim 1 or 2, feature exist
In: audio-frequency fingerprint of the audio fingerprint feature sequence using the time-frequency domain differential code coding of fixed length as single frames voice signal.
5. a kind of database construction method for the access failure phone analysis of causes as claimed in claim 1 or 2, feature exist
In: for single recording file, the fingerprint characteristic that each frame time information is constituted entire recording file is combined with each frame audio-frequency fingerprint
Sequence.
6. a kind of database construction method for the access failure phone analysis of causes as claimed in claim 1 or 2, feature
Be: the database is based on hashed table index structure, and the index key assignments of the hashed table index structure is audio fingerprint feature,
The location contents that is indexed of the hashed table index structure is audio frame information corresponding to key assignments, and the audio frame information includes
The position of the text information of place access failure recording and corresponding audio frame in recording.
7. a kind of phone access failure reason recognition methods, comprising:
Access failure reason is marked by signaling；
If failing to classify to obtain reason by signaling, audio fingerprint feature sequence, and benefit are extracted from telephonograph to be identified
It is retrieved in audio fingerprint database with the sequence；Matched fingerprint is such as found, then according to not connecing in fingerprint key assignments
Logical reason label carries out access failure reason mark for phone to be identified；
If failing to find matched fingerprint, audio content is identified as by content of text by automatic speech recognition, is based on the text
This content utilizes file classification method, classifies in access failure reason document classification model, the access failure obtained with classification
Causality classification result marks the telephonograph to be identified.
8. a kind of phone access failure reason recognition methods, comprising:
Audio-frequency fingerprint is extracted to access failure telephonograph frame by frame, obtains telephonograph audio-frequency fingerprint sequence to be identified;
To the telephonograph audio-frequency fingerprint sequence to be identified, single frames fuzzy search is carried out in audio fingerprint database, is obtained
The characteristic frame of candidate original recording audio fingerprint feature；
The characteristic frame of the candidate original recording audio fingerprint feature is integrated, and is referred to telephonograph audio to be identified
Line sequence carries out sequences match, obtains candidate original recording matching arrangement set;
By the orderly candidate original recording matching arrangement set in access failure telephonograph-access failure reason data library into
Row retrieval, it is candidate to select multiple access failure reasons；
The multiple access failure reason candidate is screened, most probable phone access failure reason is obtained.
9. a kind of phone access failure reason recognition methods as claimed in claim 8, it is characterised in that: it is fuzzy to carry out the single frames
When retrieval, it may be disturbed according to frequency band energy determination and several fingerprint digits of mismatch occur, negated by number of bits combination
A series of expansion candidate fingerprints are generated, joint original fingerprint is retrieved together.
10. a kind of phone access failure reason recognition methods as claimed in claim 8 or 9, it is characterised in that: described in progress
When audio fingerprint feature integration is with sequences match, calculate first the characteristic frame of the candidate original recording audio fingerprint feature with to
Relative displacement of the characteristic frame of telephonograph in respective sequence is matched, when the number of matches under identical relative displacement is original
When accounting in telephonograph meets threshold requirement, the characteristic frame with identical relative displacement is integrated into original electricity in database
Segment in words recording, as candidate original recording match sequence, and the accounting in original phone recording by the number of matches
Than the match hit rate score as the candidate original recording matching sequence.
11. a kind of phone access failure reason recognition methods as claimed in claim 10, it is characterised in that: do not connect to the multiple
When logical reason candidate screens, do not connect using the match hit rate score and each candidate of the candidate original recording matching sequence
The accumulation score of logical reason obtains matching score, is ranked up with the matching score to candidate access failure reason, most such as score
High person is higher than preset decision-making value, then is selected as final access failure reason result；The accumulation score refers to institute
When stating orderly candidate original recording matching arrangement set and being retrieved in the database, the similarity of the part matched is commented
Valence.
12. a kind of phone access failure reason recognition methods as claimed in claim 11, it is characterised in that: the decision-making value
Setting method are as follows: by one group of test data independently of training data, calculate the test data in audio fingerprint data
The matching score of practical access failure reason and other access failure reasons is distributed in library, calculates threshold value according to optimal distinction principle.
13. a kind of phone access failure reason recognition methods, including a kind of database sharing for the access failure phone analysis of causes
Step；And the identification step to access failure telephonograph in library based on the data.
14. a kind of phone access failure reason recognition methods as described in claim 3 or 13, it is characterised in that: further include data
Library update method, specifically comprises the following steps:
With newly-increased access failure telephonograph and access failure phone reason, Lai Gengxin access failure telephonograph-access failure reason mark
Infuse database；
With newly-increased access failure telephonograph and its audio-frequency fingerprint, Lai Gengxin access failure telephonograph-audio fingerprint feature data
Library；
With newly-increased access failure telephonograph and its text information, the textual classification model of Lai Gengxin access failure reason；
With newly-increased access failure telephonograph and its signaling information, the signaling table of Lai Gengxin access failure reason；
The newly-increased access failure telephonograph refers to identify the access failure telephonograph of access failure reason.
15. a kind of phone access failure reason recognition methods as claimed in claim 14, it is characterised in that:
If there is the segment not being matched in the newly-increased access failure telephonograph, the segment is taken out as corresponding access failure
The training characteristics sequence of reason updates access failure telephonograph-audio fingerprint feature database.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910076136.3A CN109658939B (en) | 2019-01-26 | 2019-01-26 | Method for identifying reason of call record non-connection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910076136.3A CN109658939B (en) | 2019-01-26 | 2019-01-26 | Method for identifying reason of call record non-connection |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109658939A true CN109658939A (en) | 2019-04-19 |
| CN109658939B CN109658939B (en) | 2020-12-01 |
Family
ID=66121641
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910076136.3A Active CN109658939B (en) | 2019-01-26 | 2019-01-26 | Method for identifying reason of call record non-connection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109658939B (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110290280A (en) * | 2019-05-28 | 2019-09-27 | 同盾控股有限公司 | A kind of recognition methods of the SOT state of termination, device and storage medium |
| CN110379444A (en) * | 2019-08-30 | 2019-10-25 | 北京太极华保科技股份有限公司 | Detection method and device, the electronic equipment of telephone state are judged by preposition media |
| CN110995938A (en) * | 2019-12-13 | 2020-04-10 | 上海优扬新媒信息技术有限公司 | Data processing method and device |
| CN111193940A (en) * | 2019-12-09 | 2020-05-22 | 腾讯科技（深圳）有限公司 | Audio playing method and device, computer equipment and computer readable storage medium |
| CN111368527A (en) * | 2020-02-28 | 2020-07-03 | 上海汇航捷讯网络科技有限公司 | Key value matching method |
| CN111510566A (en) * | 2020-03-16 | 2020-08-07 | 深圳追一科技有限公司 | Method and device for determining call label, computer equipment and storage medium |
| CN111601002A (en) * | 2020-04-10 | 2020-08-28 | 北京捷通华声科技股份有限公司 | Client number state matching method and device, electronic equipment and storage medium |
| CN111629108A (en) * | 2020-04-27 | 2020-09-04 | 北京青牛技术股份有限公司 | Real-time identification method of call result |
| CN112562654A (en) * | 2020-12-02 | 2021-03-26 | 天津车之家数据信息技术有限公司 | Audio classification method and computing device |
| CN112954118A (en) * | 2021-02-22 | 2021-06-11 | 上海井星信息科技有限公司 | Null number detection method and system based on frequency spectrum high point |
| CN113269935A (en) * | 2021-07-16 | 2021-08-17 | 融讯伟业（北京）科技有限公司 | Visual weighing method and weighing system based on screen-free weighing device |
| CN113312509A (en) * | 2021-06-22 | 2021-08-27 | 中国平安财产保险股份有限公司 | Telephone recall method, telephone recall device, computer equipment and storage medium |
| WO2022073507A1 (en) * | 2020-10-09 | 2022-04-14 | 深圳壹账通智能科技有限公司 | Method, apparatus, electronic device, and storage medium for distinguishing type of non-connected telephone call |
| WO2022206719A1 (en) * | 2021-03-31 | 2022-10-06 | 北京智齿博创科技有限公司 | Outbound call failure result detection method based on freeswitch and asr |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1768422A1 (en) * | 2004-06-22 | 2007-03-28 | ZTE Corporation | Method for protecting incoming calls in personalized ring back tone service |
| US7783021B2 (en) * | 2005-01-28 | 2010-08-24 | Value-Added Communications, Inc. | Digital telecommunications call management and monitoring system |
| US20110286585A1 (en) * | 2002-08-08 | 2011-11-24 | Stephen Lee Hodge | Telecommunication Call Management And Monitoring System With Voiceprint Verification |
| US8331919B1 (en) * | 2009-04-24 | 2012-12-11 | Nuance Communications, Inc. | System, method, and software program product for tracking call failures on a wireless phone |
| US8781098B2 (en) * | 2007-04-04 | 2014-07-15 | At&T Intellectual Property Ii, L.P. | System and method for prompt modification based on caller hang ups in IVRs |
| US9247057B2 (en) * | 2014-01-15 | 2016-01-26 | Verizon Patent And Licensing Inc. | System and method for providing proactive service assurance in emergency networks |
| CN106254696A (en) * | 2016-08-02 | 2016-12-21 | 北京京东尚科信息技术有限公司 | Outgoing call result determines method, Apparatus and system |
| CN107293307A (en) * | 2016-03-31 | 2017-10-24 | 阿里巴巴集团控股有限公司 | Audio-frequency detection and device |
| CN107580149A (en) * | 2017-08-28 | 2018-01-12 | 携程旅游网络技术（上海）有限公司 | The recognition methods of outgoing call failure cause, device, electronic equipment, storage medium |
| CN107622773A (en) * | 2017-09-08 | 2018-01-23 | 科大讯飞股份有限公司 | Audio feature extraction method and device and electronic equipment |
| CN107911557A (en) * | 2017-11-30 | 2018-04-13 | 维沃移动通信有限公司 | The processing method and mobile terminal of a kind of missed call |
| CN108124065A (en) * | 2017-12-05 | 2018-06-05 | 浙江鹏信信息科技股份有限公司 | A kind of method junk call content being identified with disposal |
| CN108399913A (en) * | 2018-02-12 | 2018-08-14 | 北京容联易通信息技术有限公司 | High robust audio fingerprinting method and system |
| CN108924370A (en) * | 2018-07-23 | 2018-11-30 | 携程旅游信息技术（上海）有限公司 | Call center's outgoing call speech waveform analysis method, system, equipment and storage medium |
| CN109151220A (en) * | 2018-09-11 | 2019-01-04 | 中国—东盟信息港股份有限公司 | A kind of communication session call failure scene analysis system |
-
2019
- 2019-01-26 CN CN201910076136.3A patent/CN109658939B/en active Active
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110286585A1 (en) * | 2002-08-08 | 2011-11-24 | Stephen Lee Hodge | Telecommunication Call Management And Monitoring System With Voiceprint Verification |
| EP1768422A1 (en) * | 2004-06-22 | 2007-03-28 | ZTE Corporation | Method for protecting incoming calls in personalized ring back tone service |
| US7783021B2 (en) * | 2005-01-28 | 2010-08-24 | Value-Added Communications, Inc. | Digital telecommunications call management and monitoring system |
| US8781098B2 (en) * | 2007-04-04 | 2014-07-15 | At&T Intellectual Property Ii, L.P. | System and method for prompt modification based on caller hang ups in IVRs |
| US8331919B1 (en) * | 2009-04-24 | 2012-12-11 | Nuance Communications, Inc. | System, method, and software program product for tracking call failures on a wireless phone |
| US9247057B2 (en) * | 2014-01-15 | 2016-01-26 | Verizon Patent And Licensing Inc. | System and method for providing proactive service assurance in emergency networks |
| CN107293307A (en) * | 2016-03-31 | 2017-10-24 | 阿里巴巴集团控股有限公司 | Audio-frequency detection and device |
| CN106254696A (en) * | 2016-08-02 | 2016-12-21 | 北京京东尚科信息技术有限公司 | Outgoing call result determines method, Apparatus and system |
| CN107580149A (en) * | 2017-08-28 | 2018-01-12 | 携程旅游网络技术（上海）有限公司 | The recognition methods of outgoing call failure cause, device, electronic equipment, storage medium |
| CN107622773A (en) * | 2017-09-08 | 2018-01-23 | 科大讯飞股份有限公司 | Audio feature extraction method and device and electronic equipment |
| CN107911557A (en) * | 2017-11-30 | 2018-04-13 | 维沃移动通信有限公司 | The processing method and mobile terminal of a kind of missed call |
| CN108124065A (en) * | 2017-12-05 | 2018-06-05 | 浙江鹏信信息科技股份有限公司 | A kind of method junk call content being identified with disposal |
| CN108399913A (en) * | 2018-02-12 | 2018-08-14 | 北京容联易通信息技术有限公司 | High robust audio fingerprinting method and system |
| CN108924370A (en) * | 2018-07-23 | 2018-11-30 | 携程旅游信息技术（上海）有限公司 | Call center's outgoing call speech waveform analysis method, system, equipment and storage medium |
| CN109151220A (en) * | 2018-09-11 | 2019-01-04 | 中国—东盟信息港股份有限公司 | A kind of communication session call failure scene analysis system |
Non-Patent Citations (2)
| Title |
|---|
| SHARMISTHA SARKAR DAS ET AL: "APPLICATION OF AUTOMATIC SPEECH RECOGNITION IN CALL CLASSIFICATION", 《2002 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS,SPEECH,AND SIGNAL PROCESSING》 * |
| 周壮隆等: "基于信令分析的呼叫失败用户提醒方案", 《广东通信技术》 * |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110290280A (en) * | 2019-05-28 | 2019-09-27 | 同盾控股有限公司 | A kind of recognition methods of the SOT state of termination, device and storage medium |
| CN110379444A (en) * | 2019-08-30 | 2019-10-25 | 北京太极华保科技股份有限公司 | Detection method and device, the electronic equipment of telephone state are judged by preposition media |
| CN111193940A (en) * | 2019-12-09 | 2020-05-22 | 腾讯科技（深圳）有限公司 | Audio playing method and device, computer equipment and computer readable storage medium |
| CN110995938A (en) * | 2019-12-13 | 2020-04-10 | 上海优扬新媒信息技术有限公司 | Data processing method and device |
| CN110995938B (en) * | 2019-12-13 | 2022-04-26 | 度小满科技(北京)有限公司 | Data processing method and device |
| CN111368527A (en) * | 2020-02-28 | 2020-07-03 | 上海汇航捷讯网络科技有限公司 | Key value matching method |
| CN111368527B (en) * | 2020-02-28 | 2023-06-20 | 上海汇航捷讯网络科技有限公司 | Key value matching method |
| CN111510566A (en) * | 2020-03-16 | 2020-08-07 | 深圳追一科技有限公司 | Method and device for determining call label, computer equipment and storage medium |
| CN111510566B (en) * | 2020-03-16 | 2021-05-28 | 深圳追一科技有限公司 | Method and device for determining call label, computer equipment and storage medium |
| CN111601002A (en) * | 2020-04-10 | 2020-08-28 | 北京捷通华声科技股份有限公司 | Client number state matching method and device, electronic equipment and storage medium |
| CN111629108A (en) * | 2020-04-27 | 2020-09-04 | 北京青牛技术股份有限公司 | Real-time identification method of call result |
| WO2022073507A1 (en) * | 2020-10-09 | 2022-04-14 | 深圳壹账通智能科技有限公司 | Method, apparatus, electronic device, and storage medium for distinguishing type of non-connected telephone call |
| CN112562654A (en) * | 2020-12-02 | 2021-03-26 | 天津车之家数据信息技术有限公司 | Audio classification method and computing device |
| CN112954118A (en) * | 2021-02-22 | 2021-06-11 | 上海井星信息科技有限公司 | Null number detection method and system based on frequency spectrum high point |
| WO2022206719A1 (en) * | 2021-03-31 | 2022-10-06 | 北京智齿博创科技有限公司 | Outbound call failure result detection method based on freeswitch and asr |
| CN113312509A (en) * | 2021-06-22 | 2021-08-27 | 中国平安财产保险股份有限公司 | Telephone recall method, telephone recall device, computer equipment and storage medium |
| CN113269935B (en) * | 2021-07-16 | 2021-11-30 | 融讯伟业（北京）科技有限公司 | Visual weighing method and weighing system based on screen-free weighing device |
| CN113269935A (en) * | 2021-07-16 | 2021-08-17 | 融讯伟业（北京）科技有限公司 | Visual weighing method and weighing system based on screen-free weighing device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109658939B (en) | 2020-12-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109658939A (en) | A kind of telephonograph access failure reason recognition methods | |
| AU2019200722B2 (en) | Call detail record analysis to identify fraudulent activity and fraud detection in interactive voice response systems | |
| CN106550155B (en) | Swindle sample is carried out to suspicious number and screens the method and system sorted out and intercepted | |
| CN109600752B (en) | Deep clustering fraud detection method and device | |
| CN107798032B (en) | Method and device for processing response message in self-service voice conversation | |
| CN109819127B (en) | Method and system for managing crank calls | |
| US8219404B2 (en) | Method and apparatus for recognizing a speaker in lawful interception systems | |
| CN1852354A (en) | Method and device for collecting user behavior characteristics | |
| CN108154425A (en) | Method is recommended by the Xian Xia trade companies of a kind of combination community network and position | |
| CN104462600A (en) | Method and device for achieving automatic classification of calling reasons | |
| WO2020159917A1 (en) | Unsupervised keyword spotting and word discovery for fraud analytics | |
| CN103514165A (en) | Method and device for identifying persons mentioned in conversation | |
| CN106936997B (en) | A kind of rubbish voice recognition methods and system based on social networks map | |
| CN113095852B (en) | Self-service distribution and service system through intelligent voice | |
| CN105827787B (en) | number marking method and device | |
| CN111131593A (en) | Crank call identification method and device | |
| CN116665676B (en) | Semantic recognition method for intelligent voice outbound system | |
| CN115408553A (en) | System for optimizing and generating call center service | |
| CN114722191A (en) | Automatic call clustering method and system based on semantic understanding processing | |
| CN114491010A (en) | Training method and device of information extraction model | |
| CN111916057A (en) | Language identification method and device, electronic equipment and computer readable storage medium | |
| CN116915903A (en) | Call voice processing method, device, server and readable storage medium | |
| CN113345419B (en) | Dialect accent based speech translation method, system and readable storage medium | |
| CN110955754A (en) | Model construction method for repeated incoming call analysis and recognition | |
| CN114006982B (en) | Harassment number identification method based on classification gradient lifting algorithm |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |