CN110472007A

CN110472007A - Information-pushing method, device, equipment and storage medium

Info

Publication number: CN110472007A
Application number: CN201910597333.XA
Authority: CN
Inventors: 黄伟军
Original assignee: Shenzhen Chase Technology Co Ltd
Current assignee: Shenzhen Chase Technology Co Ltd; Shenzhen Zhuiyi Technology Co Ltd
Priority date: 2019-07-04
Filing date: 2019-07-04
Publication date: 2019-11-19

Abstract

This application discloses a kind of information-pushing method, device, equipment and storage mediums, belong to field of computer technology.The described method includes: obtaining the target communication content between at least two personnel when at least two personnel communicate；Feature extraction is carried out to target communication content, obtains the characteristic information of target communication content；Characteristic information based on target communication content judges whether target communication content meets the trigger condition for showing target auxiliary information；When target communication content meets trigger condition, target auxiliary information is pushed to target terminal, target terminal is terminal corresponding to personnel at least two personnel.The efficiency that technical solution provided by the embodiments of the present application solves the problems, such as that contact staff provides service for user is lower.

Description

Information-pushing method, device, equipment and storage medium

Technical field

This application involves field of computer technology, are situated between more particularly to a kind of information-pushing method, device, equipment and storage Matter.

Background technique

Currently, in order to provide the user with more quality services, many enterprises, which are equipped with, directly services user Contact staff.In practical application, contact staff can provide business for user and consult by way of communicating with user It askes, business handling and business recommended etc. service.

Due to enterprise be the type of business that user provides be it is diversified, contact staff all can not grasp and remember Recall, therefore, in the related technology, during communicating with user, contact staff usually requires to look into according to the demand of user Database is ask, and provides service based on the auxiliary information inquired for user.

However, the time needed for contact staff inquires database is generally longer, this results in user to need in communication process The middle service for waiting longer time that can just obtain contact staff, this will affect contact staff and provides the efficiency of service for user.

Summary of the invention

Based on this, it is necessary to which the lower problem of the efficiency for providing service for user for contact staff provides a kind of information Method for pushing, device, equipment and storage medium.

In a first aspect, a kind of information-pushing method is provided, this method comprises:

When at least two personnel communicate, the target communication content between at least two personnel is obtained；To the mesh It marks Content of communciation and carries out feature extraction, obtain the characteristic information of the target communication content；Feature based on the target communication content Information judges whether the target communication content meets the trigger condition for showing target auxiliary information；When the target communication content meets When the trigger condition, the target auxiliary information is pushed to target terminal, which is one at least two personnel Terminal corresponding to personnel.

The characteristic information of the target communication content includes word speed information, semantic information, emotion in one of the embodiments, Information robs words at least one of information and key word information；

Wherein, which is used to indicate word speed of at least two personnel in communication, and the semantic information is for referring to Show the semanteme of the target communication content, which is used to indicate mood of at least two personnel in communication, this robs words Whether information is used to indicate at least two personnel in communication there is a phenomenon where robbing words, which is used to indicate the mesh Mark whether Content of communciation includes preset keyword.

The target communication content includes the communication of each personnel at least two personnel in one of the embodiments, Audio carries out feature extraction to the target communication content, obtains the characteristic information of the target communication content, comprising: will each be somebody's turn to do The communication audio input of personnel obtains the of audio emotion recognition neural network output into audio emotion recognition neural network One probabilistic information, first probabilistic information are used to indicate the probability that mood of each personnel in communication is target emotion；Root The emotion information is obtained according to first probabilistic information.

The emotion information is obtained according to first probabilistic information in one of the embodiments, comprising: utilize speech recognition The communication audio of each personnel is converted communication text by technology；By the communication text input of each personnel to text feelings In the other neural network of perception, the second probabilistic information of text emotion recognition neural network output, second probabilistic information are obtained It is used to indicate the probability that mood of each personnel in communication is the target emotion；According to first probabilistic information and this second Probabilistic information obtains the emotion information.

The emotion information is obtained according to first probabilistic information and second probabilistic information in one of the embodiments, Include: that first probabilistic information and second probabilistic information are input in comprehensive emotion recognition neural network, obtains the synthesis Emotion recognition neural network output third probabilistic information, the third probabilistic information be used to indicate each personnel communication when Mood is the probability of the target emotion；The emotion information is obtained according to the third probabilistic information.

The target communication content includes the communication of each personnel at least two personnel in one of the embodiments, Audio carries out feature extraction to the target communication content, obtains the characteristic information of the target communication content, comprising: obtains each Record length section corresponding to the communication audio of the personnel obtains at least two record length sections；Obtain at least two recording Overlapping time section in period between every two record length section；According to the overlapping time between the every two record length section Whether the duration of section is greater than duration threshold value, generates this and robs words information.

The target communication content includes the communication of each personnel at least two personnel in one of the embodiments, Audio carries out feature extraction to the target communication content, obtains the characteristic information of the target communication content, comprising: obtains each Long recording time corresponding to the communication audio of the personnel；By speech recognition technology, the communication audio of each personnel is converted To communicate text, and the number of words that the communication text for obtaining each personnel includes；The communication text for obtaining each personnel includes Number of words and each personnel communication audio corresponding to long recording time ratio, and the word speed is generated according to the ratio and is believed Breath.

At least two personnel include contact staff and user in one of the embodiments, which is the visitor Take terminal corresponding to personnel.

The target auxiliary information includes information relevant to professional knowledge and operation flow in one of the embodiments, Relevant information, exchanged with user if art information, need the information of business recommended to the user and to the contact staff's At least one of the quality information of service quality progress quality inspection.

Second aspect provides a kind of information push-delivery apparatus, which includes:

Module is obtained, it is logical for when at least two personnel communicate, obtaining the target between at least two personnel Interrogate content；

Extraction module obtains the feature letter of the target communication content for carrying out feature extraction to the target communication content Breath；

Judgment module judges whether the target communication content meets exhibition for the characteristic information based on the target communication content Show the trigger condition of target auxiliary information；

Pushing module, for it is auxiliary to push the target to target terminal when the target communication content meets the trigger condition Supplementary information, the target terminal are terminals corresponding to personnel at least two personnel.

The target communication content includes the communication of each personnel at least two personnel in one of the embodiments, Audio, the extraction module, comprising:

Acquisition submodule, for into audio emotion recognition neural network, obtaining the communication audio input of each personnel The first probabilistic information exported to the audio emotion recognition neural network, first probabilistic information are used to indicate each personnel and exist Mood when communication is the probability of target emotion；

Extracting sub-module, for obtaining the emotion information according to first probabilistic information.

The extracting sub-module in one of the embodiments, is specifically used for: speech recognition technology is utilized, by each the people The communication audio of member is converted into communication text；The communication text input of each personnel to text emotion is identified into neural network In, the second probabilistic information of text emotion recognition neural network output is obtained, which is used to indicate each be somebody's turn to do Mood of the personnel in communication is the probability of the target emotion；It is somebody's turn to do according to first probabilistic information and second probabilistic information Emotion information.

The extracting sub-module in one of the embodiments, is specifically used for: by first probabilistic information and second probability Information input obtains the third probability letter of synthesis emotion recognition neural network output into comprehensive emotion recognition neural network Breath, the third probabilistic information are used to indicate the probability that mood of each personnel in communication is the target emotion；According to this Three probabilistic informations obtain the emotion information.

The target communication content includes the communication of each personnel at least two personnel in one of the embodiments, Audio, the extraction module, is specifically used for: obtaining record length section corresponding to the communication audio of each personnel, obtains at least Two record length sections；Obtain the overlapping time section in at least two record lengths section between every two record length section；Root Whether the duration according to the overlapping time section between the every two record length section is greater than duration threshold value, generates this and robs words information.

The target communication content includes the communication of each personnel at least two personnel in one of the embodiments, Audio, the extraction module, is specifically used for: obtaining long recording time corresponding to the communication audio of each personnel；Known by voice The communication audio of each personnel is converted communication text by other technology, and the communication text for obtaining each personnel includes Number of words；Obtain the number of words and long recording time corresponding to each communication audio of the personnel that the communication text of each personnel includes Ratio, and the word speed information is generated according to the ratio.

The third aspect, provides a kind of computer equipment, including memory and processor, which is stored with computer Program, the computer program realize above-mentioned first aspect any information-pushing method when being executed by the processor.

Fourth aspect provides a kind of computer readable storage medium, is stored thereon with computer program, which is located Reason device realizes above-mentioned first aspect any information-pushing method when executing.

Technical solution bring beneficial effect provided by the embodiments of the present application includes at least:

By when at least two personnel communicate, obtaining the target communication content between at least two personnel, and Feature extraction is carried out to the target communication content, the characteristic information of target communication content is obtained, then, is sentenced according to this feature information Whether disconnected target communication content meets the trigger condition for showing target auxiliary information, if satisfied, then pushing away the target auxiliary information It send in terminal corresponding to a personnel at least two personnel, utilizes information push side provided by the embodiments of the present application Method automatically can provide contact staff to auxiliary required for service when contact staff and user communicate for user Information pushes to the terminal of contact staff, in this way, contact staff there is no need to during with user communication according to user's Demand inquires database, it is thus possible to improve contact staff provides the efficiency of service for user.

Detailed description of the invention

Fig. 1 is a kind of schematic diagram of the implementation environment of information-pushing method provided by the embodiments of the present application；

Fig. 2 is a kind of flow chart of information-pushing method provided by the embodiments of the present application；

Fig. 3 is the flow chart of another information-pushing method provided by the embodiments of the present application；

Fig. 4 is the flow chart of another information-pushing method provided by the embodiments of the present application；

Fig. 5 is the flow chart of another information-pushing method provided by the embodiments of the present application；

Fig. 6 is a kind of block diagram of information push-delivery apparatus provided by the embodiments of the present application；

Fig. 7 is a kind of block diagram of computer equipment provided by the embodiments of the present application.

Specific embodiment

To keep the purposes, technical schemes and advantages of the application clearer, below in conjunction with attached drawing to the application embodiment party Formula is described in further detail.

In practical application, in order to provide the user with more quality services, many enterprises be equipped with directly to user into The contact staff of row service.

In a kind of possible application scenarios, the place of a Relatively centralized is can be set as call center, visitor in enterprise The personnel of clothes intracardiac in a call can provide service for user.Specifically, phone has can be set in call center, and contact staff can To pass through the incoming call of phone listening user, alternatively, electricity is removed to user, during being communicated by phone and user, visitor The personnel of clothes can provide business consultation, business handling and business recommended etc. service for user.

In alternatively possible application scenarios, enterprise can be equipped with for contact staff and be equipped with instant messaging application program Terminal device, contact staff can use the terminal device, communicated by way of instant messaging with user, with During family is communicated, contact staff can provide business consultation, business handling and business recommended etc. for user Service.

Certainly, above-mentioned application scenarios are only exemplary, and are not intended to limit this application, and in practical application, may be used also There can be other contact staff and provide the scene of service for user, the embodiment of the present application just no longer repeats one by one herein.

In practical application, it is user by taking bank as an example that enterprise, which is that the type for the business that user provides is diversified, The business of offer may include credit card business, savings card business, Internet-based banking services etc..Since enterprise is the industry that user provides Service type is various, and therefore, the still shallow contact staff of contact staff, especially qualification can not grasp and remember to all business Recall.In view of this, in the related technology, during communicating with user, contact staff can look into according to the demand of user Database is ask, and provides service based on the auxiliary information inquired for user.

For example, user seeks advice from the annual fee of credit card in communication process to contact staff, and contact staff does not remember The annual fee of credit card is how many, at this point, contact staff can inquire database, obtains the year charge information of credit card by inquiry (that is to say auxiliary information described above), then, contact staff can be responded according to the year charge information of the credit card inquired The consulting of user.

In another example user proposes oneself to need to handle credit card in communication process to contact staff, and contact staff is simultaneously Do not remember that the process for handling credit card obtains doing for credit card by inquiry at this point, contact staff can inquire database It manages process (that is to say auxiliary information described above), then, contact staff can be user according to the process of handling inquired Handle credit card.

In another example user proposes query to contact staff or enterprise in communication process or is discontented with, and customer service people Member queries there is no reply user or discontented experience, at this point, contact staff can inquire database, by inquiring back Art (that is to say auxiliary information described above) if multiplexing family, then, contact staff can be according to the words art pair inquired User replys, to pacify the mood of user.

In another example contact staff needs recommended user during with user communication to handle credit card, and contact staff The recommendation information for not remembeing the credits card such as annual fee, the preferential activity of credit card, at this point, contact staff can inquire data Library obtains the recommendation information (that is to say auxiliary information described above) of credit card, then, contact staff by inquiry The business of credit card is handled to user's recommendation according to the recommendation information of the credit card inquired.

However, the time needed for contact staff inquires database is generally longer (being under normal conditions 10 seconds to 30 seconds), this On the one hand it will lead to the service that user needs to wait longer time that can just obtain contact staff in communication process, on the other hand It will lead to the negligible amounts of the user serviced in contact staff's unit time, therefore, the inquiry database that the relevant technologies provide Mode will affect contact staff and provide the efficiency of service for user.

The embodiment of the present application provides a kind of information-pushing method, and contact staff can be improved as user and provide the effect of service Rate.Mesh in the information-pushing method, when at least two personnel communicate, between available at least two personnel Content of communciation is marked, and feature extraction is carried out to the target communication content, obtains the characteristic information of target communication content, it then, can To judge whether target communication content meets the trigger condition of displaying target auxiliary information according to this feature information, if satisfied, then The target auxiliary information is pushed in terminal corresponding to a personnel at least two personnel, is implemented using the application The information-pushing method that example provides automatically can mention contact staff for user when contact staff and user communicate Push to the terminal of contact staff for auxiliary information required for servicing, in this way, contact staff there is no need to user communication During inquire database according to the demand of user, it is thus possible to improve contact staff provides the efficiency of service for user.

In the following, by being carried out briefly to implementation environment involved by information-pushing method provided by the embodiments of the present application It is bright.

Fig. 1 is a kind of schematic diagram of implementation environment provided by the embodiments of the present application.As shown in Figure 1, the implementation environment can be with Including server 101 and terminal 102, wherein server 101 and terminal 102 can be by wired or wirelessly carry out Communication, terminal 102 can be terminal corresponding to a personnel at least two personnel communicated, for example, should Terminal 102 can be terminal corresponding to the contact staff in the contact staff and user that are being communicated.

In implementation environment shown in Fig. 1, in the target communication between available at least two personnel of server 101 Hold, when determining that the target communication content meets the displaying condition for showing target auxiliary information information, server 101 can should Target auxiliary information is pushed to terminal 102, for terminal 102 by target auxiliary information show corresponding to it personnel (for example, Contact staff).

It should be pointed out that in the embodiment of the present application, terminal 102 can be desktop computer, tablet computer, notebook electricity Brain, smart phone, wearable device, mobile unit, MP3 player or MP4 player etc., server 101 can be one Server is also possible to the server cluster being made of multiple servers.

Referring to FIG. 2, it illustrates a kind of flow chart of information-pushing method provided by the embodiments of the present application, which is pushed away Delivery method can be applied in the server in implementation environment shown in Fig. 1.As shown in Fig. 2, the information-pushing method may include Following steps:

Step 201, when at least two personnel communicate, it is logical that server obtains the target between at least two personnel Interrogate content.

Optionally, which may include contact staff and user.

In some possible situations, for example, in the case where at least two personnel are communicated by phone, service The target communication content that device obtains may include the communication audio of each personnel at least two personnel.Optionally, this extremely In the case that few two personnel are communicated by phone, server can obtain the target by the interchanger in telephone network Content of communciation.

In other possible situations, for example, being communicated by way of instant messaging at least two personnel In the case where, the target communication content that server obtains may include the communication text of each personnel at least two personnel.

Step 202, server carry out feature extraction to target communication content, obtain the characteristic information of target communication content.

When target communication content is audio, the characteristic information of the target communication content may include word speed information, semanteme Information, emotion information rob words at least one of information and key word information.

When target communication content is text, the characteristic information of the target communication content may include semantic information, emotion At least one of information and key word information.

Wherein, which is used to indicate word speed of at least two personnel in communication, which is used to indicate The semanteme of target communication content, the emotion information are used to indicate mood of at least two personnel in communication, this is robbed words information and uses Whether there is a phenomenon where robbing words when at least two personnel that indicate are in communication, which is used to indicate target communication content It whether include preset keyword.

The embodiment of the present application provides one kind and optionally " carries out feature extraction to target communication content, obtain target communication The technical solution of the semantic information of content ", the technical solution may include the following three steps:

A1, server obtain the target text of pending feature extraction.

When target communication content is audio, that is to say, when target communication content include at least two personnel it is each When the communication audio of personnel, server can use speech recognition technology, and each personnel's for including by target communication content is logical News audio is converted into text, and the text that conversion obtains is retrieved as to the target text of pending feature extraction.

When target communication content is text, that is to say, when target communication content include at least two personnel it is each When the communication text of personnel, server can directly by the communication text for each personnel that target communication content includes be retrieved as to Carry out the target text of feature extraction.

Target text is input in intent classifier neural network by A2, server, and it is defeated to obtain the intent classifier neural network Intention probabilistic information out, the intention probabilistic information are used to indicate the corresponding probability for being intended to target intention of target text.

Wherein, target intention described above may include that multiple business are intended to be intended to (that is to say with a refusal identification The intention unrelated with business), refusal identification is intended to and each business is intended to, it is intended that the intention of Classification Neural output Probabilistic information can indicate the corresponding probability for being intended to the intention of target text.For example, target intention may include business It is intended to A, business is intended to B and refusal identification is intended to C, it is intended that the intention probabilistic information of Classification Neural output can indicate: mesh The corresponding corresponding probability and target text for being intended to business and being intended to B of probability, target text for being intended to business and being intended to A of mark text This corresponding probability for being intended to refusal identification and being intended to C.

The intention probabilistic information that A3, server are exported according to intent classifier neural network determines the corresponding meaning of target text Figure, and according to the corresponding semantic information for being intended to generate target communication content of target text.

The highest intention of the probability for being intended to probabilistic information instruction can be determined as the corresponding intention of target text by server, For example, the intention probabilistic information exported by intent classifier neural network is it is found that the corresponding business that is intended to of target text is intended to A's Probability highest, the corresponding probability for being intended to refusal identification intention C of target text is minimum, then business can be intended to A by server It is determined as the corresponding intention of target text.

For example, if target text is " why that credit card that I did last year will receive annual fee ", according to intent classifier mind The intention probabilistic information exported through network can determine the corresponding business meaning for being intended to " consulting credit card annual fee " of target text Figure, server can be intended to generate the semantic information of target communication content according to the business of " the consulting credit card annual fee ".

In another example if target text is " ", it can be with according to the intention probabilistic information of intent classifier neural network output It determines that target text is corresponding and is intended to refusal identification intention, that is to say that the intention server unrelated with business can be according to refusal Identification is intended to generate the semantic information of target communication content.

The embodiment of the present application provides one kind and optionally " carries out feature extraction to target communication content, obtain target communication The technical solution of the key word information of content ", the technical solution may include following four step:

B1, server obtain the target text of pending feature extraction.

Similarly, details are not described herein for the embodiment of the present application for the technical process of the technical process of step B1 and step A1.

B2, server segment target text, obtain multiple words that target text includes.

Multiple word searching keyword databases that B3, server include according to target text, determine that target text includes Multiple words whether be stored in keyword database.

B4, server lead to according to whether multiple words that target text includes are stored in generation target in keyword database Interrogate the key word information of content.

Feature extraction is carried out to target communication content, obtain the word speed information of target communication content, emotion information and robs words The technical process of information is detailed in the application hereafter, is not described herein.

Step 203, server judge whether target communication content meets displaying based on the characteristic information of target communication content The trigger condition of target auxiliary information.

Wherein, target auxiliary information may include information relevant to professional knowledge, information relevant with operation flow, with Art information, the information of needs business recommended to the user and the service quality progress quality inspection to contact staff if user exchanges At least one of quality information.

Information relevant to professional knowledge refers to the information that contact staff can be assisted to respond user's consulting；With Business Stream The relevant information of journey refers to the information that can be used in that contact staff is assisted to be user's transacting business；Art is believed if exchanging with user Breath refers to the information that contact staff can be assisted to be exchanged with user；The matter of quality inspection is carried out to the service quality of contact staff Inspection information is the information for characterizing the service quality that contact staff is the service that user provides, which can be used for referring to Show whether contact staff refers to violated word during with user communication, the too drastic phenomenon of mood whether occur, whether occur Whether the phenomenon that robbing words there is the too fast phenomenon of word speed etc..

Wherein, which can be that technical staff is preset, which may include regular expression Deng.For example, triggering shows that the trigger condition of this target auxiliary information of credit card rate can be " semantic information=credit card Rate by stages ".

Step 204, when target communication content meets trigger condition, server to target terminal push target auxiliary letter Breath.

Wherein, target terminal is terminal corresponding to a personnel at least two personnel, and optionally, the target is whole End can be the corresponding terminal of contact staff.

In information-pushing method provided by the embodiments of the present application, by obtaining when at least two personnel communicate Target communication content between at least two personnel, and feature extraction is carried out to the target communication content, obtain target communication The characteristic information of content then judges whether target communication content meets according to this feature information and shows target auxiliary information Trigger condition, if satisfied, the target auxiliary information is then pushed to end corresponding to a personnel at least two personnel It, can be when contact staff and user communicate, automatically using information-pushing method provided by the embodiments of the present application in end Contact staff is provided to the terminal for servicing required auxiliary information and pushing to contact staff for user, in this way, contact staff is just It does not need to inquire database according to the demand of user during with user communication, it is thus possible to improve contact staff is User provides the efficiency of service.

Further, target auxiliary information can be quality information, during contact staff and user communicate The service quality to contact staff quality inspection can be carried out in real time to the mode of the corresponding terminal push quality information of contact staff, And prompting contact staff improves the service quality of oneself in real time.In this way, in a first aspect, can effectively guarantee contact staff to The service quality that family is serviced, second aspect, do not need the special quality inspection personnel of business organization afterwards to contact staff and Content of communciation between user carries out quality inspection, and so as to reduce human cost, the third aspect only can be right compared to artificial quality inspection For the mode that contact staff is inspected by random samples, it is ensured that all carry out quality inspection to the service quality of all contact staff.

On the basis of embodiment described above, when target communication content includes each personnel at least two personnel When communicating audio, the embodiment of the present application provides one kind and optionally " carries out feature extraction to target communication content, it is logical to obtain target The method of the emotion information of news content ", in the following, the embodiment of the present application will be briefly described this method, as shown in figure 3, This method may comprise steps of:

The communication audio input for each personnel that target communication content is included by step 301, server to audio emotion is known In other neural network, the first probabilistic information of audio emotion recognition neural network output is obtained.

Audio emotion recognition neural network can use the acoustics prosodic features of the communication audio of each personnel to everyone Mood of the member in communication is identified, thus the first probabilistic information of output.Wherein, acoustics prosodic features may include pitch, Intensity, sound quality, sound spectrum and cepstrum etc., it is target emotion that the first probabilistic information, which is used to indicate mood of each personnel in communication, Probability.

Wherein, target emotion may include that normal mood, positive mood and negative sense mood, the first probabilistic information can indicate Mood of each personnel in communication is the probability of each above-mentioned mood, be that is to say, the first probabilistic information can indicate: each Mood of the personnel in communication is positive probability that the mood of the probability of reason thread, each personnel in communication is positive mood and every Mood of a personnel in communication is negative the probability of reason thread.

Step 302, server obtain emotion information according to the first probabilistic information.

In one possible implementation, server directly can generate emotion information using first probabilistic information.

In alternatively possible implementation, server can use speech recognition technology for target communication content and include The communication audio of each personnel be converted into communication text, obtain the communication text of each personnel, then, server can will be every The communication text input of a personnel obtains the of text emotion identification neural network output into text emotion identification neural network Two probabilistic informations, wherein it is the general of above-mentioned target emotion that the second probabilistic information, which is used to indicate mood of each personnel in communication, Rate that is to say that the second probabilistic information can indicate: each personnel communication when mood be positive reason thread probability, everyone Mood of the member in communication is that the mood of the probability and each personnel of positive mood in communication is negative the probability of reason thread, To after the second probabilistic information, server can obtain the emotion information according to the first probabilistic information and the second probabilistic information.

Optionally, first probabilistic information and second probabilistic information can be input to comprehensive emotion recognition mind by server Through the third probabilistic information in network, obtaining comprehensive emotion recognition neural network output, wherein the third probabilistic information is for referring to Show that mood of each personnel in communication is the probability of target emotion, that is to say, third probabilistic information can indicate: each personnel Mood in communication be positive probability that the mood of the probability of reason thread, each personnel in communication is positive mood and everyone Mood of the member in communication is negative the probability of reason thread, and after obtaining third probabilistic information, server can be directly using should Third probabilistic information generates the emotion information.

The each personnel that can be reflected in the way of third probabilistic information generation emotion information with integrated communication audio are existed Mood of each personnel that mood and communication text when communication are reflected in communication, so that finally obtained emotion information institute Mood of each personnel of reflection in communication is more accurate.

It should be pointed out that target communication content be text when, that is to say, target communication content include this at least two In a personnel when the communication text of each personnel, server can not execute the technical process for obtaining the first probabilistic information, and only The technical process for obtaining the second probabilistic information is executed, and directly generates the emotion information using second probabilistic information.

On the basis of embodiment described above, when target communication content includes each personnel at least two personnel When communicating audio, the embodiment of the present application provides one kind and optionally " carries out feature extraction to target communication content, it is logical to obtain target News content robs words information " method, in the following, the embodiment of the present application will be briefly described this method, as shown in figure 4, This method may comprise steps of:

When step 401, server obtain recording corresponding to the communication audio for each personnel that target communication content includes Between section, obtain at least two record length sections.

In practical application, the communication audio of each personnel is corresponding with a record length section, which is used for The period that instruction personnel speak, in step 401, the communication for each personnel that the available target communication content of server includes Record length section corresponding to audio, to obtain at least two record length sections.

When step 402, server obtain the overlapping in at least two record lengths section between every two record length section Between section.

Wherein, the overlapping time section between two record length sections refers to the time that two record length sections all include Section, for example, record length section A is 10:01 to 10:03, record length section B is 10:02 to 10:04, then record length section A and record Overlapping time section between sound period B is the period that the record length section A and record length section B include, and be that is to say 10:02 to 10:03.

Whether step 403, server are greater than duration according to the duration of the overlapping time section between every two record length section Words information is robbed in threshold value, generation.

In practical application, the duration of the overlapping time section between two record length sections is greater than duration threshold value and illustrates two people The duration that member speaks simultaneously is longer, in this case, the phenomenon that robbing words has occurred between two personnel, conversely, two recording The duration of overlapping time section between period illustrates two personnel less than or equal to duration threshold value while the duration spoken is shorter, In this case, there is no robbing words phenomenon between two personnel.

In view of this, in step 403, server can be according to the overlapping time section between every two record length section Whether duration is greater than duration threshold value, and words information is robbed in generation.

On the basis of embodiment described above, when target communication content includes each personnel at least two personnel When communicating audio, the embodiment of the present application provides one kind and optionally " carries out feature extraction to target communication content, it is logical to obtain target The method of the word speed information of news content ", in the following, the embodiment of the present application will be briefly described this method, as shown in figure 5, This method may comprise steps of:

When step 501, server obtain recording corresponding to the communication audio for each personnel that target communication content includes It is long.

In practical application, the communication audio of each personnel is corresponding with a long recording time, which is used to indicate The duration that personnel speak.

Step 502 passes through speech recognition technology, and the communication audio of each personnel is converted communication text by server, and Obtain the number of words that the communication text of each personnel includes.

Step 503, server obtain the number of words and the communication audio institute of each personnel that the communication text of each personnel includes The ratio of corresponding long recording time, and word speed information is generated according to the ratio.

Wherein, which is used to characterize personnel's number of words described within the unit time.

Referring to FIG. 6, it illustrates a kind of block diagram of information push-delivery apparatus 600 provided by the embodiments of the present application, the information Driving means 600 can be configured in server.As shown in fig. 6, the information push-delivery apparatus 600 may include: acquisition module 601, extraction module 602, judgment module 603 and pushing module 604.

Wherein, module 601 is obtained, for obtaining between at least two personnel when at least two personnel communicate Target communication content.

Extraction module 602 obtains the feature of the target communication content for carrying out feature extraction to the target communication content Information.

Judgment module 603 judges whether the target communication content is full for the characteristic information based on the target communication content Foot shows the trigger condition of target auxiliary information.

Pushing module 604, for pushing the target to target terminal when the target communication content meets the trigger condition Auxiliary information, the target terminal are terminals corresponding to personnel at least two personnel.

In one embodiment of the application, the characteristic information of the target communication content include word speed information, semantic information, Emotion information robs words at least one of information and key word information；

In one embodiment of the application, which includes each personnel at least two personnel Communicate audio, the extraction module 602, including acquisition submodule and extracting sub-module.

Wherein, acquisition submodule, for by the communication audio input of each personnel to audio emotion recognition nerve net In network, the first probabilistic information of audio emotion recognition neural network output is obtained, which is used to indicate each Mood of the personnel in communication is the probability of target emotion.

The extracting sub-module, for obtaining the emotion information according to first probabilistic information.

In one embodiment of the application, which is specifically used for: speech recognition technology is utilized, it will be each The communication audio of the personnel is converted into communication text；The communication text input of each personnel to text emotion is identified into nerve net In network, the second probabilistic information of text emotion recognition neural network output is obtained, which is used to indicate each Mood of the personnel in communication is the probability of the target emotion；It is obtained according to first probabilistic information and second probabilistic information The emotion information.

In one embodiment of the application, which is specifically used for: by first probabilistic information and this second Probabilistic information is input in comprehensive emotion recognition neural network, obtains the third probability of synthesis emotion recognition neural network output Information, the third probabilistic information are used to indicate the probability that mood of each personnel in communication is the target emotion；According to this Third probabilistic information obtains the emotion information.

In one embodiment of the application, which includes each personnel at least two personnel Audio is communicated, which is specifically used for: obtaining record length section corresponding to the communication audio of each personnel, obtain To at least two record length sections；Obtain the overlapping time in at least two record lengths section between every two record length section Section；Whether it is greater than duration threshold value according to the duration of the overlapping time section between the every two record length section, generates this and rob words letter Breath.

In one embodiment of the application, which includes each personnel at least two personnel Audio is communicated, which is specifically used for: obtaining long recording time corresponding to the communication audio of each personnel；Pass through Speech recognition technology converts communication text for the communication audio of each personnel, and obtains the communication text of each personnel Including number of words；Obtain number of words and each record corresponding to the communication audio of the personnel that the communication text of each personnel includes The ratio of sound duration, and the word speed information is generated according to the ratio.

In one embodiment of the application, which includes contact staff and user, which is Terminal corresponding to the contact staff.

In one embodiment of the application, which includes information relevant to professional knowledge and business The relevant information of process, exchanged with user if art information, need the information of business recommended to the user and to customer service people The service quality of member carries out at least one of the quality information of quality inspection.

Information pushing meanss provided by the embodiments of the present application, may be implemented above method embodiment, realization principle and skill Art effect is similar, and details are not described herein.

Specific about information push-delivery apparatus limits the restriction that may refer to above for information-pushing method, herein not It repeats again.Modules in above- mentioned information driving means can be realized fully or partially through software, hardware and combinations thereof.On Stating each module can be embedded in the form of hardware or independently of in the processor in computer equipment, can also store in a software form In memory in computer equipment, the corresponding operation of the above modules is executed in order to which processor calls.

In one embodiment of the application, a kind of computer equipment is provided, which can be server, Its internal structure chart can be as shown in Figure 7.The computer equipment includes the processor and memory connected by system bus.Its In, the processor of the computer equipment is for providing calculating and control ability.The memory of the computer equipment includes non-volatile Property storage medium, built-in storage.The non-volatile memory medium is stored with operating system and computer program.The built-in storage is The operation of operating system and computer program in non-volatile memory medium provides environment.The computer program is held by processor To realize a kind of information-pushing method when row.

It will be understood by those skilled in the art that structure shown in Fig. 7, only part relevant to application scheme is tied The block diagram of structure does not constitute the restriction for the computer equipment being applied thereon to application scheme, specific computer equipment It may include perhaps combining certain components or with different component layouts than more or fewer components as shown in the figure.

In one embodiment of the application, a kind of computer equipment is provided, which can be server, The computer equipment includes memory and processor, is stored with computer program in memory, which executes computer journey It is performed the steps of when sequence

The target communication content includes the communication audio of each personnel at least two personnel, at one of the application In embodiment, processor is also performed the steps of when executing computer program by the communication audio input of each personnel to sound In frequency emotion recognition neural network, the first probabilistic information of audio emotion recognition neural network output, first probability are obtained Information is used to indicate the probability that mood of each personnel in communication is target emotion；It is somebody's turn to do according to first probabilistic information Emotion information.

In one embodiment of the application, processor also performs the steps of when executing computer program and utilizes voice The communication audio of each personnel is converted communication text by identification technology；By the communication text input of each personnel to text In this emotion recognition neural network, the second probabilistic information of text emotion recognition neural network output, second probability are obtained Information is used to indicate the probability that mood of each personnel in communication is the target emotion；According to first probabilistic information and it is somebody's turn to do Second probabilistic information obtains the emotion information.

In one embodiment of the application, processor execute computer program when also perform the steps of by this first Probabilistic information and second probabilistic information are input in comprehensive emotion recognition neural network, obtain the synthesis emotion recognition nerve net The third probabilistic information of network output, it is the target feelings which, which is used to indicate mood of each personnel in communication, The probability of thread；The emotion information is obtained according to the third probabilistic information.

The target communication content includes the communication audio of each personnel at least two personnel, at one of the application In embodiment, processor is also performed the steps of when executing computer program corresponding to the communication audio for obtaining each personnel Record length section, obtain at least two record length sections；Obtain every two record length in at least two record lengths section Overlapping time section between section；Whether it is greater than duration threshold according to the duration of the overlapping time section between the every two record length section Value, generates this and robs words information.

The target communication content includes the communication audio of each personnel at least two personnel, at one of the application In embodiment, processor is also performed the steps of when executing computer program corresponding to the communication audio for obtaining each personnel Long recording time；By speech recognition technology, communication text is converted by the communication audio of each personnel, and obtain and each should The number of words that the communication text of personnel includes；Obtain number of words and each communication of the personnel that the communication text of each personnel includes The ratio of long recording time corresponding to audio, and the word speed information is generated according to the ratio.

Computer equipment provided by the embodiments of the present application, implementing principle and technical effect and above method embodiment class Seemingly, details are not described herein.

In one embodiment of the application, a kind of computer readable storage medium is provided, computer is stored thereon with Program performs the steps of when computer program is executed by processor

The target communication content includes the communication audio of each personnel at least two personnel, at one of the application It is also performed the steps of in embodiment, when computer program is executed by processor by the communication audio input of each personnel extremely In audio emotion recognition neural network, the first probabilistic information of audio emotion recognition neural network output is obtained, this is first general Rate information is used to indicate the probability that mood of each personnel in communication is target emotion；It is obtained according to first probabilistic information The emotion information.

In one embodiment of the application, is also performed the steps of when computer program is executed by processor and utilize language The communication audio of each personnel is converted communication text by sound identification technology；Extremely by the communication text input of each personnel Text emotion identifies in neural network, obtains the second probabilistic information of text emotion recognition neural network output, this is second general Rate information is used to indicate the probability that mood of each personnel in communication is the target emotion；According to first probabilistic information and Second probabilistic information obtains the emotion information.

Also performed the steps of in one embodiment of the application, when computer program is executed by processor by this One probabilistic information and second probabilistic information are input in comprehensive emotion recognition neural network, obtain the synthesis emotion recognition nerve The third probabilistic information of network output, it is the target which, which is used to indicate mood of each personnel in communication, The probability of mood；The emotion information is obtained according to the third probabilistic information.

The target communication content includes the communication audio of each personnel at least two personnel, at one of the application In embodiment, the communication audio institute that each personnel that obtain also are performed the steps of when computer program is executed by processor is right The record length section answered obtains at least two record length sections；When obtaining every two recording in at least two record lengths section Between overlapping time section between section；Whether it is greater than duration according to the duration of the overlapping time section between the every two record length section Threshold value generates this and robs words information.

The target communication content includes the communication audio of each personnel at least two personnel, at one of the application In embodiment, the communication audio institute that each personnel that obtain also are performed the steps of when computer program is executed by processor is right The long recording time answered；By speech recognition technology, communication text is converted by the communication audio of each personnel, and obtain each The number of words that the communication text of the personnel includes；The number of words that the communication text of each personnel of acquisition includes is logical with each personnel's The ratio of long recording time corresponding to audio is interrogated, and the word speed information is generated according to the ratio.

Computer readable storage medium provided in this embodiment, implementing principle and technical effect and above method embodiment Similar, details are not described herein.

Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with Relevant hardware is instructed to complete by computer program, the computer program can be stored in a non-volatile computer In read/write memory medium, the computer program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, To any reference of memory, storage, database or other media used in each embodiment provided herein, Including non-volatile and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM (PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include Random access memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms, Such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram (DDRSDRAM), enhancing Type SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..

Each technical characteristic of embodiment described above can be combined arbitrarily, for simplicity of description, not to above-mentioned reality It applies all possible combination of each technical characteristic in example to be all described, as long as however, the combination of these technical characteristics is not deposited In contradiction, all should be considered as described in this specification.

The several embodiments of the application above described embodiment only expresses, the description thereof is more specific and detailed, but simultaneously The limitation to claim therefore cannot be interpreted as.It should be pointed out that coming for those of ordinary skill in the art It says, without departing from the concept of this application, various modifications and improvements can be made, these belong to the protection of the application Range.Therefore, the scope of protection shall be subject to the appended claims for the application patent.

Claims

1. a kind of information-pushing method, which is characterized in that the described method includes:

When at least two personnel communicate, the target communication content between at least two personnel is obtained；

Feature extraction is carried out to the target communication content, obtains the characteristic information of the target communication content；

Characteristic information based on the target communication content judges whether the target communication content meets and shows target auxiliary letter The trigger condition of breath；

When the target communication content meets the trigger condition, the target auxiliary information is pushed to target terminal, it is described Target terminal is terminal corresponding to personnel at least two personnel.

2. the method according to claim 1, wherein the characteristic information of the target communication content includes word speed letter Breath, emotion information, robs words at least one of information and key word information at semantic information；

Wherein, the word speed information is used to indicate word speed of at least two personnel in communication, and institute's semantic information is used for Indicate that the semanteme of the target communication content, the emotion information are used to indicate feelings of at least two personnel in communication Thread, it is described rob words information be used to indicate at least two personnel communication when whether there is a phenomenon where rob words, the keyword Information is used to indicate whether the target communication content includes preset keyword.

3. according to the method described in claim 2, it is characterized in that, the target communication content includes at least two personnel In each personnel communication audio, it is described to the target communication content carry out feature extraction, obtain the target communication The characteristic information of content, comprising:

By the communication audio input of each personnel into audio emotion recognition neural network, the audio emotion recognition is obtained First probabilistic information of neural network output, first probabilistic information are used to indicate mood of each personnel in communication For the probability of target emotion；

The emotion information is obtained according to first probabilistic information.

4. according to the method described in claim 3, it is characterized in that, described obtain the emotion according to first probabilistic information Information, comprising:

Using speech recognition technology, communication text is converted by the communication audio of each personnel；

By the communication text input of each personnel into text emotion identification neural network, the text emotion identification is obtained Second probabilistic information of neural network output, second probabilistic information are used to indicate mood of each personnel in communication For the probability of the target emotion；

The emotion information is obtained according to first probabilistic information and second probabilistic information.

5. according to the method described in claim 4, it is characterized in that, described according to first probabilistic information and described second general Rate information obtains the emotion information, comprising:

First probabilistic information and second probabilistic information are input in comprehensive emotion recognition neural network, obtained described The third probabilistic information of comprehensive emotion recognition neural network output, the third probabilistic information are used to indicate each personnel and exist Mood when communication is the probability of the target emotion；

The emotion information is obtained according to the third probabilistic information.

6. according to the method described in claim 2, it is characterized in that, the target communication content includes at least two personnel In each personnel communication audio, it is described to the target communication content carry out feature extraction, obtain the target communication The characteristic information of content, comprising:

Record length section corresponding to the communication audio of each personnel is obtained, at least two record length sections are obtained；

Obtain the overlapping time section in at least two record lengths section between every two record length section；

Whether it is greater than duration threshold value according to the duration of the overlapping time section between the every two record length section, is robbed described in generation Talk about information.

7. according to the method described in claim 2, it is characterized in that, the target communication content includes at least two personnel In each personnel communication audio, it is described to the target communication content carry out feature extraction, obtain the target communication The characteristic information of content, comprising:

Obtain long recording time corresponding to the communication audio of each personnel；

By speech recognition technology, communication text is converted by the communication audio of each personnel, and obtain each people The number of words that the communication text of member includes；

Obtain recording corresponding to the communication audio of number of words and each personnel that the communication text of each personnel includes The ratio of duration, and the word speed information is generated according to the ratio.

8. method according to any one of claims 1 to 7, which is characterized in that at least two personnel include contact staff And user, the target terminal are terminal corresponding to the contact staff.

9. according to the method described in claim 8, it is characterized in that, the target auxiliary information includes relevant to professional knowledge Information, information relevant to operation flow, exchanged with user if art information, need business recommended to the user information and At least one of the quality information of quality inspection is carried out to the service quality of the contact staff.

10. a kind of information push-delivery apparatus, which is characterized in that described device includes:

Module is obtained, for obtaining the target communication between at least two personnel when at least two personnel communicate Content；

Pushing module, for pushing the target to target terminal when the target communication content meets the trigger condition Auxiliary information, the target terminal are terminals corresponding to personnel at least two personnel.

11. a kind of computer equipment, which is characterized in that including memory and processor, the memory is stored with computer journey Sequence, which is characterized in that the computer program realizes letter as described in any one of claim 1 to 9 when being executed by the processor Cease method for pushing.

12. a kind of computer readable storage medium, which is characterized in that be stored thereon with computer program, which is characterized in that the journey Information-pushing method as described in any one of claim 1 to 9 is realized when sequence is executed by processor.