CN111246008A - Method, system and device for realizing telephone assistant - Google Patents

Abstract

The embodiment of the invention discloses a method, a system and a device for realizing telephone assistant, wherein the embodiment of the invention adopts a trained neural network model to construct a telephone assistant application, an incoming call of a mobile terminal is accessed into the telephone assistant application, the telephone assistant application performs voice recognition and semantic analysis on the content of the incoming call to obtain recognition and analysis information, and after multi-turn pre-conversation is performed on the voice of a user and the incoming call based on the recognition and analysis information, the obtained incoming call intention information is provided for the mobile terminal to be processed. The invention applies the neural network model of deep learning to the incoming call processing process, so that the incoming call can be processed intelligently without the setting of a user.

Description

Method, system and device for implementing a telephone assistant

technical field

The present invention relates to the field of communication technologies, and in particular, to a method, system and device for implementing a telephone assistant.

Background technique

With the development of communication technology, mobile terminals have now penetrated into our lives in an all-round way, bringing great convenience to our lives. But at the same time, mobile terminals also bring a lot of troubles to users. For example, making harassing calls is a channel and means for merchants to sell products. They frequently make calls to users to sell products, and even deceive users by making harassing calls. In short, harassing calls have seriously affected the life of users, and even brought hidden dangers to users' personal and property.

At present, there are many technologies similar to mobile terminal phone assistants to intercept harassing calls, such as: downloading and installing security applications in mobile terminals, intercepting certain harassing calls, or setting up some blacklists in mobile terminals to achieve the purpose of precise interception. And for truly meaningful calls, the phone assistant engages the user directly. However, the use of security applications to identify harassing calls sometimes cannot accurately help users identify harassing calls due to network or update problems; and because the phone numbers of harassing calls often change, the blacklist set in the mobile terminal cannot solve this type of harassment. question. Therefore, the telephone assistant set on the mobile terminal cannot bring a good experience to the user, and in most cases, the user is manually screened, which greatly wastes the user's time and experience.

Specifically, the mobile terminal intercepts and processes the harassing calls in the following ways.

The patent application with the patent publication number CN109688276A relates to a call filtering system and method based on artificial intelligence technology, which solves the problem of traditional anti-harassment interception and false interception, including: the terminal equipment of user one, the terminal equipment of user two, The terminal equipment of user 1 and the terminal equipment of user 2 both include communication modules, and the terminal equipment of user 2 has a call forwarding function. The call filtering system based on artificial intelligence technology also includes a telephone switch, and the telephone switch includes The automatic answering system includes a number verification module, a voiceprint recognition module, a speech recognition module, a dialogue system module, a speech synthesis module, a dialogue verification module and a storage module. The patent application details the process of intelligently intercepting harassing calls, but it verifies incoming calls through user preset values, and requires restarting the conversation with the user.

The patent application with the patent publication number CN109995925A relates to a method for identifying a harassing call, a terminal and a computer-readable storage medium. The method includes, when an incoming call is detected, acquiring the identification information of the incoming call, and calculating the identification information according to the identification information and the preset identification information. Incoming call characteristic value of incoming call; according to the incoming call characteristic value and the preset characteristic value, the incoming call is identified, which solves the problem that the existing harassing phone number identification method cannot accurately and timely identify the harassing phone number, which makes the user experience a sense of Not a high problem. It can be seen that the patent application uses preset feature values and marking information to harass and intercept incoming calls.

The patent application with the patent publication number CN106302942A relates to a method for intelligently intercepting harassing calls, by presetting a voice in the intelligent interception system to detect whether the incoming call is recorded or not, supplemented by the second connection of unfamiliar numbers, whitelist and blacklist , network query, key monitoring and other comprehensive measures to judge and intercept harassing calls, this method can independently identify group calls and incoming calls in the form of recording, intelligently search the caller number online, monitor the call in real time and record the whole process of unfamiliar calls, and the user has a key operation. Timely voice reminders can effectively resist recording fraud and group dial calls; ensure users are protected from the threat of harassment and fraudulent calls. It can be seen that the patent application is mainly aimed at the scene of recording fraud, and the recording fraud can be effectively intercepted by preset recording.

The patent application with the patent publication number CN109672786A relates to a method and device for answering an incoming call, including: step 101, when the terminal receives an incoming call, the voice assistant automatically connects the incoming call; step 103, the voice assistant records the incoming call according to the preset slot The caller number and/or the voice content of the caller number, fill in the relevant content into the slot value corresponding to the preset slot, and have a dialogue with the caller number. The method can realize the function of the voice assistant to intelligently handle incoming calls. The voice assistant can conduct multiple rounds of intelligent dialogue with the incoming call number, obtain the incoming call intention, and answer the caller number question as needed, so that the incoming caller can be answered in time for some urgent matters. Help users solve harassing calls, improve the humanization of smart calls, and improve user experience. It can be seen that in this patent application, all the call rights are given to the voice assistant, and the voice assistant talks with the incoming call on behalf of the user, and finally notifies the user of the content of the conversation.

It can be seen that when the above solution processes the incoming call to meet the user's answering needs, the incoming call can be verified according to the preset information, or the user can be prompted after the incoming call by using a recording phone or a voice assistant. Before or after participation and settings, causing inconvenience to users. Furthermore, the use of recorded calls or the use of voice assistants for incoming calls is pre-set corresponding to the corresponding incoming calls, which is not intelligent.

SUMMARY OF THE INVENTION

In view of this, embodiments of the present invention provide a method for implementing a telephone assistant, which can intelligently process incoming calls without requiring user settings to meet user needs.

Embodiments of the present invention also provide a system for implementing a telephone assistant, which can intelligently process incoming calls without requiring user settings to meet user needs.

Embodiments of the present invention also provide a device for implementing a telephone assistant, which can intelligently process incoming calls without requiring user settings to meet user needs.

The embodiments of the present invention are implemented as follows:

A method for implementing a telephone assistant, the method comprising:

Use the trained neural network model to build a phone assistant application;

Access the incoming call of the mobile terminal into the telephone assistant application, and the telephone assistant application obtains identification analysis information after performing voice recognition and semantic analysis on the content of the incoming call;

After the phone assistant application simulates the user's voice and performs multiple rounds of pre-conversation with the incoming call based on the identification and analysis information, and obtains the incoming call intention information, it provides it to the mobile terminal for processing.

Before the identification and analysis information is obtained after the phone assistant application performs speech recognition and semantic analysis on the content of the incoming call, the method further includes:

A call interception rule is set in the phone assistant application, and the incoming call is pre-filtered according to the incoming call interception rule. If the incoming call interception rule is met, it is directly intercepted; if the incoming call interception rule is not met, the incoming call is directly connected to the mobile terminal; If it cannot be determined whether the call interception rule is met, the step of obtaining the recognition analysis information after the phone assistant application performs voice recognition and semantic analysis on the content of the call is performed.

The processing provided to the mobile terminal includes:

When it is confirmed that the incoming call is a harassing incoming call or an incoming call within a preset threshold according to the incoming call intention information, hang up directly and notify the mobile terminal in the form of a text message;

When it is confirmed that the incoming call is a normal incoming call according to the incoming call intention information, the mobile terminal directly answers the call, and at the same time, the pre-chat content is displayed on the user interaction interface of the mobile terminal in the form of text.

The described use of the trained neural network model to build a telephone assistant application includes:

The neural network model is trained based on the deep learning automatic speech recognition ASR method and the deep learning natural language processing NLP method;

The identification and analysis information obtained after the phone assistant application performs speech recognition and semantic analysis on the content of the incoming call includes:

After the voice-to-text conversion of the incoming call of the mobile terminal is performed in the ASR manner, the language processing is performed in the NLP manner to obtain analysis information.

A system for implementing a telephone assistant, comprising: a mobile terminal, a caller device and an intelligent voice call assistant unit, wherein,

Caller equipment, used to initiate an incoming call to the mobile terminal;

The intelligent voice call assistant unit is used to construct a telephone assistant application by using the trained neural network model, and access the incoming call of the mobile terminal to the telephone assistant application, and the telephone assistant application performs voice recognition and semantic analysis on the content of the incoming call. Obtain identification and analysis information; after simulating user voice and incoming calls for multiple rounds of pre-conversation based on the identification and analysis information, obtain caller intention information, and provide it to the mobile terminal for processing;

The mobile terminal is configured to perform corresponding processing after receiving the incoming call intention information.

The intelligent voice call assistant unit is also used for setting a rule for intercepting incoming calls in the phone assistant application, pre-filtering incoming calls according to the rules for intercepting incoming calls, and directly intercepting incoming calls if they meet the rules for intercepting incoming calls; If the rule is met, the incoming call is directly connected to the mobile terminal; if it cannot be determined whether the call interception rule is met, the phone assistant application performs voice recognition and semantic analysis on the content of the incoming call to obtain identification analysis information.

The mobile terminal is also used to perform corresponding processing: when confirming that the incoming call is a harassing call or an incoming call within a preset threshold according to the incoming call intention information, hang up directly, and notify the mobile terminal in the form of a short message; confirm the incoming call according to the incoming call intention information. When it is a normal incoming call, the mobile terminal directly answers the call, and at the same time, the content of the pre-chat is displayed on the user interface of the mobile terminal in the form of text.

The intelligent voice call assistant unit is also used to construct a telephone assistant application by using the trained neural network model, including: training the neural network model based on the deep learning ASR method and the deep learning NLP method; the telephone assistant application responds to incoming calls. After the content is subjected to speech recognition and semantic analysis, the recognition analysis information includes: after the incoming call of the mobile terminal is converted from speech to text by ASR, language processing is performed by NLP to obtain the analysis information.

A device for realizing a telephone assistant, the device comprises: an intelligent voice answering module, an intelligent voice analysis decision-making module and an intelligent dialogue subject extraction module, wherein,

The intelligent voice answering module is used to perform voice recognition and semantic analysis on the incoming call of the connected mobile terminal, and then obtain the identification and analysis information and send it to the intelligent voice analysis decision-making module. Under the instruction of the intelligent voice analysis and decision-making module, simulate the user's voice and call multiple rounds of pre-conversation;

The intelligent voice analysis and decision-making module is used to make decisions based on the identification and analysis information, and instructs the intelligent voice answering module to simulate the user's voice and the incoming call for multiple rounds of pre-conversation; extract the content of the multi-round pre-conversation for decision-making, and obtain the incoming call intention information;

The intelligent dialogue subject extraction module is used for providing the pre-chat content to the mobile terminal in the form of text.

The intelligent voice answering module. It is also used to set the rules for intercepting incoming calls, and pre-filter the incoming calls according to the rules for intercepting incoming calls. If it meets the rules for intercepting incoming calls, it will be directly intercepted; If the rules for intercepting incoming calls are met, the steps of obtaining the identification and analysis information after the phone assistant application performs voice recognition and semantic analysis on the content of the incoming call;

The intelligent voice answering module is also used to perform voice recognition and semantic analysis on the content of the incoming call to obtain identification analysis information, including: after the incoming call of the mobile terminal is converted from speech to text in an ASR manner, and language processing is performed in an NLP manner. , to get analysis information.

As can be seen above, in the embodiment of the present invention, a trained neural network model is used to construct a telephone assistant application, and incoming calls from a mobile terminal are connected to the telephone assistant application, and the telephone assistant application performs speech recognition and semantic analysis on the content of the incoming call. The identification and analysis information is obtained, and based on the identification and analysis information, the user's voice is simulated for multiple rounds of pre-conversation with the incoming call, and the incoming call intention information is obtained, and then provided to the mobile terminal for processing. Since the present invention applies the deep learning neural network model to the processing of incoming calls, the incoming calls can be processed intelligently without user setting.

Description of drawings

FIG. 1 is a schematic process diagram of a method for implementing a telephone assistant provided by an embodiment of the present invention;

2 is a flowchart of a method for implementing a telephone assistant according to an embodiment of the present invention;

3 is a schematic structural diagram of a system for implementing a telephone assistant according to an embodiment of the present invention;

4 is a schematic structural diagram of a device for implementing a telephone assistant according to an embodiment of the present invention;

5 is a schematic diagram of the implementation of an intelligent voice call assistant unit provided by an embodiment of the present invention;

6 is a schematic diagram of a natural language processing process provided by an embodiment of the present invention;

7 is a schematic diagram of a speech synthesis process provided by an embodiment of the present invention;

8 is a schematic diagram of an intelligent speech analysis decision-making process provided by an embodiment of the present invention;

9 is a schematic diagram of a basic Recursive NN model provided by an embodiment of the present invention;

10 is a schematic diagram of slot-based identification provided by an embodiment of the present invention;

11 is a schematic diagram of a specific example of implementation of a telephone assistant provided by an embodiment of the present invention;

FIG. 12 is a schematic diagram of a second specific implementation example of a telephone assistant provided by an embodiment of the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be described in further detail below with reference to the accompanying drawings and embodiments.

In the embodiment of the present invention, a trained neural network model is used to construct a telephone assistant application, and incoming calls from a mobile terminal are connected to the telephone assistant application, and the telephone assistant application performs voice recognition and semantic analysis on the content of the incoming call to obtain identification analysis information , based on the identification and analysis information, after simulating the user's voice and the incoming call for multiple rounds of pre-conversation, the incoming call intention information is obtained, and then provided to the mobile terminal for processing.

In this way, since the present invention applies the deep learning neural network model to the processing of incoming calls, the incoming calls can be processed intelligently without user setting.

The phone assistant application provided by the embodiment of the present invention can simulate the user's voice and conduct multiple rounds of pre-conversation with the phone before the user answers the phone, and recognize the call intention information through deep learning technology, and intelligently judge whether to continue answering according to the call intention information, If you do not continue to answer the call, a message notification will be sent after hanging up, informing the intention of the call and the reason for hanging up; when the user of the mobile terminal needs to answer the call, the mobile terminal will pop up a normal answer window, and the content of the pre-chat will be displayed in text form on the In the pop-up window of the mobile terminal, it is convenient for the user to communicate seamlessly with the caller. The schematic diagram of the whole realization process is shown in Figure 1.

Specifically, when the mobile terminal receives an incoming call, it connects the incoming call to the set phone assistant application, and blacklist and whitelist can also be set in the phone assistant application, and the phone assistant application can set the blacklist and whitelist according to the set Preprocessing of incoming calls: if the incoming call number is in the blacklist, hang up the incoming call directly; if the incoming call number is in the white list, make the incoming call connect and reach the mobile terminal directly; if the incoming call number is neither in the blacklist nor in the white list In the list, the phone assistant application performs speech recognition and semantic analysis on the content of the incoming call to obtain identification and analysis information, and based on the identification and analysis information, simulates the user's voice and conducts multiple rounds of pre-conversation with the incoming call to obtain the incoming call intention information. When the phone assistant application confirms that the incoming call is a harassing call or an incoming call within a preset threshold according to the call intention information, it will hang up directly and notify the mobile terminal in the form of a text message at the same time; It will be handed over to the mobile terminal to answer directly, and the content of the pre-chat will be displayed on the user interface of the mobile terminal in the form of text, so that the user can know the purpose of the call in advance, and it is convenient for the user to communicate with the caller more effectively.

In this embodiment of the present invention, the phone assistant application may be called an intelligent voice call assistant, and may be set in a switch provided by a telecom operator, and the switch is set between the mobile terminal and the caller's device to process incoming calls.

In this way, after using the set phone assistant application, the mobile terminal can efficiently handle incoming calls according to the user's main wishes, directly hang up incoming calls that the user is not interested in, and let the user answer incoming calls that the user is interested in. The content of the pre-chat will be displayed on the user interface of the mobile terminal in the form of text, so that the user can know the purpose of the call in advance and make a more appropriate response.

2 is a flowchart of a method for implementing a telephone assistant provided by an embodiment of the present invention, and its specific steps are:

Step 201, using the trained neural network model to build a telephone assistant application;

Step 202: Access the incoming call of the mobile terminal to the telephone assistant application, and the telephone assistant application performs voice recognition and semantic analysis on the content of the incoming call to obtain identification analysis information;

Step 203: After the phone assistant application simulates the user's voice and conducts multiple rounds of pre-conversation with the incoming call based on the identification and analysis information, and obtains the incoming call intention information, it provides it to the mobile terminal for processing.

In the method, before the phone assistant application performs voice recognition and semantic analysis on the content of the incoming call to obtain recognition analysis information, the method further includes:

A call interception rule is set in the phone assistant application, and the incoming call is pre-filtered according to the incoming call interception rule. If the incoming call interception rule is met, it is directly intercepted; if the incoming call interception rule is not met, the incoming call is directly connected to the mobile terminal; If it cannot be determined whether the call interception rule is met, the step of obtaining the recognition analysis information after the phone assistant application performs voice recognition and semantic analysis on the content of the call is performed.

Here, the incoming call interception rule is a set call interception threshold or a priority call threshold, etc., or a blacklist and a whitelist set for the user ID, which perform direct pre-filtering operations.

In the method, the processing provided to the mobile terminal includes:

When it is confirmed that the incoming call is a harassing incoming call or an incoming call within a preset threshold according to the incoming call intention information, hang up directly and notify the mobile terminal in the form of a text message;

When it is confirmed that the incoming call is a normal incoming call according to the incoming call intention information, the mobile terminal directly answers the call, and at the same time, the content of the pre-chat is displayed on the user interface of the mobile terminal in the form of text.

In the method, the use of the trained neural network model to build a telephone assistant application includes:

The neural network model is trained based on the automatic speech recognition (ASR) method of deep learning and the natural language processing (NLP) method of deep learning;

The identification and analysis information obtained after the phone assistant application performs speech recognition and semantic analysis on the content of the incoming call includes:

After the voice-to-text conversion of the incoming call of the mobile terminal is performed in the ASR manner, the language processing is performed in the NLP manner to obtain analysis information.

3 is a schematic structural diagram of a system for implementing a telephone assistant according to an embodiment of the present invention, including: a mobile terminal, a caller device, and an intelligent voice call assistant unit, wherein,

Caller equipment, used to initiate an incoming call to the mobile terminal;

The intelligent voice call assistant unit is used to construct a telephone assistant application by using the trained neural network model, and access the incoming call of the mobile terminal to the telephone assistant application, and the telephone assistant application performs voice recognition and semantic analysis on the content of the incoming call. Obtain identification and analysis information; after simulating user voice and incoming calls for multiple rounds of pre-conversation based on the identification and analysis information, obtain caller intention information, and provide it to the mobile terminal for processing;

The mobile terminal is configured to perform corresponding processing after receiving the incoming call intention information.

In this system, the intelligent voice call assistant unit is further configured to set a rule for intercepting incoming calls in the phone assistant application, pre-filter incoming calls according to the rules for intercepting incoming calls, and directly intercept the incoming calls if it conforms to the rules for intercepting incoming calls; If the call interception rule is not met, the call is directly connected to the mobile terminal; if it cannot be determined whether the call interception rule is met, the phone assistant application performs voice recognition and semantic analysis on the content of the call to obtain identification analysis information.

In this system, the mobile terminal is also used to perform corresponding processing including: according to the call intention information, when confirming that the incoming call is a harassing call or an incoming call within a preset threshold, directly hang up and notify the mobile terminal in the form of a short message; When the incoming call intention information confirms that the incoming call is a normal incoming call, the mobile terminal directly answers the call, and at the same time, the content of the pre-chat is displayed on the user interface of the mobile terminal in the form of text.

In the system, the intelligent voice call assistant unit is also used to construct a telephone assistant application by using the trained neural network model, including: training the neural network model based on the deep learning ASR method and the deep learning NLP method; The phone assistant application performs speech recognition and semantic analysis on the content of the incoming call to obtain the identification analysis information, which includes: after the incoming call of the mobile terminal is converted from speech to text in the ASR mode, and language processing is performed in the NLP mode to obtain the analysis information.

In this system, the intelligent voice call assistant unit can be set in the switch provided by the telecom operator.

4 is a schematic structural diagram of a device for implementing a telephone assistant provided by an embodiment of the present invention. The device is also an intelligent voice call assistant unit, including: an intelligent voice answering module, an intelligent voice analysis and decision-making module, and an intelligent dialogue subject extraction module, wherein,

The intelligent voice answering module is used to perform voice recognition and semantic analysis on the incoming call of the connected mobile terminal, and then obtain the identification and analysis information and send it to the intelligent voice analysis decision-making module. Under the instruction of the intelligent voice analysis and decision-making module, simulate the user's voice and call multiple rounds of pre-conversation;

The intelligent voice analysis and decision-making module is used to make decisions based on the identification and analysis information, and instructs the intelligent voice answering module to simulate the user's voice and the incoming call for multiple rounds of pre-conversation; extract the content of the multi-round pre-conversation for decision-making, and obtain the incoming call intention information;

The intelligent dialogue subject extraction module is used for providing the pre-chat content to the mobile terminal in the form of text.

In the device, the intelligent voice answering module is also used to set the rules for intercepting incoming calls, and pre-filters incoming calls according to the rules for intercepting incoming calls. If it meets the rules for intercepting incoming calls, it will directly intercept them; Call the mobile terminal; if it cannot be determined whether the call interception rule is met, the phone assistant application performs voice recognition and semantic analysis on the content of the call to obtain identification analysis information.

In the device, the intelligent voice answering module is further configured to perform voice recognition and semantic analysis on the content of the incoming call to obtain the identification and analysis information, including: converting the incoming call of the mobile terminal to text by using ASR, and then using NLP. Perform language processing to obtain analytical information.

FIG. 5 is a schematic diagram of an implementation of an intelligent voice call assistant unit provided by an embodiment of the present invention. As shown in the figure, specifically, the intelligent voice answering module in the intelligent voice call assistant unit includes three parts: voice recognition, semantic analysis, and voice synthesis. After the caller number is verified by the number, it first enters the speech recognition, and the ASR technology based on deep learning can realize the conversion of speech into text. Then enter the NLP method of deep learning, perform natural speech understanding (NLU) and natural language generation (NLG) processing on the text, and infer the call intention information. At the same time, the obtained intent text is passed to the intelligent speech analysis and decision-making module, and the speech analysis and decision-making module performs the next step.

When the intelligent voice analysis and decision-making module makes a decision and needs to have a dialogue with the incoming call, it will send the reply text to the intelligent voice answering module. Here, the voice synthesis technology will convert the reply text into the reply voice. Speech synthesis itself is a relatively mature technology. You can choose a different voice to reply according to your own preferences, or you can record your own voice, so that the caller feels that he is communicating with himself.

FIG. 6 is a schematic diagram of a natural language processing process provided by an embodiment of the present invention; FIG. 7 is a schematic diagram of a speech synthesis process provided by an embodiment of the present invention. In the above process, the processes shown in FIG. 6 and FIG. 7 are applied.

The intelligent voice analysis and decision-making module provided by the embodiment of the present invention performs text analysis according to the intended text transmitted by the intelligent voice answering module. If it is a general sentence, the intelligent phone assistant can independently generate the reply text and send it to the intelligent voice answering module; If it is a sentence within the user's threshold range, the smart phone assistant will hang up directly and notify the mobile terminal in the form of a text message; if it is found to be a normal call after analysis, and the mobile terminal needs to be connected and processed by itself, it will initiate a call request to the mobile terminal and communicate with each other. The text is displayed on the incoming call user interface in the mobile terminal to achieve the effect of seamless conversation between the user and the caller. As shown in FIG. 8 , FIG. 8 is a schematic diagram of a decision-making process of intelligent speech analysis provided by an embodiment of the present invention.

In the intelligent dialogue subject extraction module provided by the embodiment of the present invention, when the user of the mobile terminal needs to answer in person, the intelligent voice dialogue module will send the text of the previous dialogue with the caller to the intelligent dialogue subject extraction module. The intelligent speech dialogue subject extraction module, based on the existing large-scale data set, can get the main idea of the dialogue text through some joint models. Here, RecNN+Viterbi is used as an example for description. As shown in FIG. 9 and FIG. 10 , FIG. 9 is a schematic diagram of a basic Recursive NN model provided by an embodiment of the present invention, and FIG. 10 is a schematic diagram of slot-based identification provided by an embodiment of the present invention.

The input of the incoming call is a single word vector (the subsequent optimization input is the word vector of the window), and each part of speech is regarded as a weight vector, so that the operation of each word in its path is a simple word vector and part of speech weight vector points Product operation. The square in the figure is the result of the dot product operation of the part-of-speech weight vector and the input vector. When a parent node has multiple child branches, it can be regarded as the sum operation of the product of each branch and the weight. For example, for the word "in", the path in the semantic analysis tree is "IN-PP-NP", and each output vector of the path is subjected to a weighted operation to obtain the path feature, and the concat of the path feature of the three words is used. Slot classification is performed as tri-path features to make a prediction of "in".

The following three specific examples are used to describe the embodiments of the present invention in detail

Example 1: Blocking of advertising calls

As shown in FIG. 11 , FIG. 11 is a schematic diagram of a specific implementation example of a telephone assistant provided by an embodiment of the present invention, including:

The first step, the phone assistant application receives an incoming call, and first enters the intelligent voice answering module;

In the second step, in the intelligent voice answering module, after being identified by ASR, enter the NLU module to analyze the sentence and understand its semantics;

The third step is to enter the NLG module, and translate the corresponding keyword text "real estate" according to the parsed semantics;

The fourth step, the intelligent voice answering module, sends the obtained intention text "real estate" to the intelligent voice analysis and decision-making module;

The fifth step, the intelligent voice analysis and decision-making module, according to the user's previous habits, the user wishes to filter out calls related to "real estate";

The sixth step, the intelligent voice analysis and decision-making module sends the reply text "Thank you, no need" to the intelligent voice answering module;

The seventh step, the intelligent voice answering module will reply the text "Thank you, no need", and use the voice synthesis technology to synthesize the user's voice to talk to the incoming call;

In the eighth step, the intelligent voice analysis and decision-making module hangs up the phone and sends information to the mobile terminal used by the user.

Example 2: Friends call through

As shown in FIG. 12 , FIG. 12 is a schematic diagram of a second specific implementation example of a telephone assistant provided by an embodiment of the present invention, including:

The first step, the phone assistant application receives an incoming call, and first enters the intelligent voice answering module;

In the second step, in the intelligent voice answering module, after being identified by ASR, enter the NLU module to analyze the sentence and understand its semantics;

The third step is to enter the NLG module, and translate the corresponding keyword text "Xiao Liang, are you free tonight" for the parsed semantics;

The fourth step, the intelligent voice answering module, sends the obtained intention text "Xiao Liang, are you free tonight" to the intelligent voice analysis and decision-making module;

The fifth step, the intelligent speech analysis decision-making module, according to the previous extensive training, the text is a common greeting;

The sixth step, the intelligent voice analysis and decision-making module responds autonomously, and sends the reply text "Is something wrong" to the intelligent voice answering module;

The seventh step, the intelligent voice answering module will reply the text "Is something wrong", and use the voice synthesis technology to synthesize the user's voice and the incoming call;

The eighth step, the caller initiates the second round of calls, "I'm on a business trip in Nanjing, and I'll go to the school in the evening to find Xiaowei and Huasheng for a drink together";

The ninth step, the intelligent voice answering module, extracts the intent text "in the evening, look for Xiaowei near the school, Huasheng drinks";

The tenth step, the intelligent speech analysis decision-making module, analyzes that it is a question-and-answer term, which requires the user to answer independently;

The eleventh step, the intelligent speech analysis and decision-making module sends the dialogue text "Xiao Liang, are you free tonight, in the evening, find Xiao Wei, Huasheng for a drink" to the intelligent dialogue subject extraction module;

The twelfth step, the intelligent dialogue subject extraction module, according to the pre-communication intention text, through the deep learning model, to obtain the main idea;

In the thirteenth step, the intelligent voice analysis and decision-making module transfers the call to the mobile terminal and displays the main idea. The user of the mobile terminal can know the intention of the incoming call in advance, make a pre-judgment, and communicate with the incoming call seamlessly.

Example 3: Active phone reminders recorded in the memo

The specific process of this example is:

The first step, according to the user's memo record, the phone assistant application knows that the user and his friend Xiaoliang are going to eat together on Friday;

The second step, on Thursday, the user has not called Xiaoliang to mention the meal;

The third step, the phone assistant application automatically broadcasts the user's phone call on Thursday to remind him of his meal on Friday;

The fourth step, the phone assistant application inquires whether the user needs to talk to Xiaoliang;

In the fifth step, when the user agrees to call, the phone assistant application actively dials Xiaoliang's phone and agrees to have a meal.

It can be seen that the voice assistant application provided by the embodiment of the present invention obtains the corresponding keyword text according to the voice recognition of incoming calls and NLP technology, and compares it with the existing text thresholds in the database, so as to achieve filtering advertising promotion and user-defined The purpose of interception can also help the user to pre-communicate the purpose of the call before answering the normal call, which is convenient for the user to talk more effectively. The smart phone assistant greatly facilitates the user and enhances the user experience.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the present invention. within the scope of protection.

Claims (10)

1. A method for implementing a telephone assistant, the method comprising:

adopting a trained neural network model to construct a telephone assistant application;

accessing an incoming call of a mobile terminal into the telephone assistant application, and performing voice recognition and semantic analysis on the content of the incoming call by the telephone assistant application to obtain recognition analysis information;

the telephone assistant application simulates the user voice to carry out multiple rounds of pre-conversation with the incoming call based on the identification and analysis information, and after the incoming call intention information is obtained, the incoming call intention information is provided for the mobile terminal to be processed.

2. The method of claim 1, wherein before the phone assistant application performs speech recognition and semantic analysis on the incoming call content to obtain recognition analysis information, the method further comprises:

setting an incoming call intercepting rule in the telephone assistant application, pre-filtering an incoming call according to the incoming call intercepting rule, and directly intercepting the incoming call if the incoming call intercepting rule is met; if the incoming call interception rule is not met, directly connecting the incoming call to the mobile terminal; and if the incoming call interception rule can not be determined to be met, executing the step that the telephone assistant application carries out voice recognition and semantic analysis on the incoming call content to obtain recognition analysis information.

3. The method of claim 1, wherein said providing the mobile terminal with processing comprises:

when the incoming call is confirmed to be a harassing call or a call within a preset threshold value according to the incoming call intention information, the incoming call is directly hung up and is notified to the mobile terminal in a short message form;

and when the incoming call is confirmed to be a normal incoming call according to the incoming call intention information, the mobile terminal answers directly, and simultaneously the pre-conversation content is displayed on a user interaction interface of the mobile terminal in a text form.

4. The method of claim 1, wherein the building the telephony assistant application using the trained neural network model comprises:

training a neural network model based on an Automatic Speech Recognition (ASR) mode of deep learning and a Natural Language Processing (NLP) mode of deep learning;

the telephone assistant application performs voice recognition and semantic analysis on incoming call content to obtain recognition analysis information, and the recognition analysis information comprises the following steps:

and after the incoming call of the mobile terminal is converted from voice to text in an ASR mode, language processing is carried out in an NLP mode to obtain analysis information.

5. A system for implementing a telephone assistant, comprising: a mobile terminal, a caller device and an intelligent voice call assistant unit, wherein,

the caller equipment is used for initiating a call to the mobile terminal;

the intelligent voice call assistant unit is used for constructing a telephone assistant application by adopting a trained neural network model, accessing an incoming call of the mobile terminal into the telephone assistant application, and performing voice recognition and semantic analysis on the content of the incoming call by the telephone assistant application to obtain recognition analysis information; after multiple pre-conversations between the voice of the user and the incoming call are simulated based on the recognition analysis information, the incoming call intention information is obtained and provided for the mobile terminal to be processed;

and the mobile terminal is used for carrying out corresponding processing after receiving the incoming call intention information.

6. The apparatus of claim 5,

the intelligent voice call assistant unit is also used for setting an incoming call intercepting rule in the telephone assistant application, pre-filtering an incoming call according to the incoming call intercepting rule, and directly intercepting the incoming call if the incoming call intercepting rule is met; if the incoming call interception rule is not met, directly connecting the incoming call to the mobile terminal; and if the incoming call interception rule can not be determined to be met, executing the step that the telephone assistant application carries out voice recognition and semantic analysis on the incoming call content to obtain recognition analysis information.

7. The apparatus of claim 6,

the mobile terminal is further configured to perform corresponding processing including: when the incoming call is confirmed to be a harassing call or a call within a preset threshold value according to the incoming call intention information, the incoming call is directly hung up and is notified to the mobile terminal in a short message form; and when the incoming call is confirmed to be a normal incoming call according to the incoming call intention information, the mobile terminal answers directly, and simultaneously the pre-conversation content is displayed on a user interaction interface of the mobile terminal in a text form.

8. The apparatus of claim 6, wherein the intelligent voice call assistant unit, further configured to build a telephony assistant application using the trained neural network model, comprises: training a neural network model based on an ASR mode of deep learning and an NLP mode of deep learning; the telephone assistant application performs voice recognition and semantic analysis on incoming call content to obtain recognition analysis information, and the recognition analysis information comprises the following steps: and after the incoming call of the mobile terminal is converted from voice to text in an ASR mode, language processing is carried out in an NLP mode to obtain analysis information.

9. An apparatus for implementing a telephone assistant, the apparatus comprising: an intelligent voice answering module, an intelligent voice analysis decision module and an intelligent dialogue subject extraction module, wherein,

the intelligent voice answering module is used for carrying out voice recognition and semantic analysis on an incoming call of the accessed mobile terminal, obtaining recognition analysis information and sending the recognition analysis information to the intelligent voice analysis decision module, and simulating the voice of a user to carry out multiple rounds of pre-conversation with the incoming call under the indication of the intelligent voice analysis decision module;

the intelligent voice analysis decision module is used for making a decision based on the recognition analysis information and indicating the intelligent voice answering module to simulate the voice of a user and carry out multiple rounds of pre-conversation with incoming calls; extracting multi-round pre-conversation contents to make a decision to obtain incoming call intention information;

and the intelligent conversation subject extraction module is used for providing the pre-conversation content to the mobile terminal in a text mode.

10. The apparatus of claim 9, wherein the intelligent voice response module. The system is also used for setting an incoming call intercepting rule, pre-filtering an incoming call according to the incoming call intercepting rule, and directly intercepting the incoming call if the incoming call intercepting rule is met; if the incoming call interception rule is not met, directly connecting the incoming call to the mobile terminal; if the incoming call interception rule can not be determined to be met, executing the step that the telephone assistant application carries out voice recognition and semantic analysis on incoming call content to obtain recognition analysis information;

the intelligent voice answering module is also used for carrying out voice recognition and semantic analysis on the incoming call content to obtain recognition analysis information, and comprises the following steps: and after the incoming call of the mobile terminal is converted from voice to text in an ASR mode, language processing is carried out in an NLP mode to obtain analysis information.

Images