CN111953577A - Method, system and readable storage medium for preventing message from being mistakenly sent - Google Patents
Method, system and readable storage medium for preventing message from being mistakenly sent Download PDFInfo
- Publication number
- CN111953577A CN111953577A CN201910411526.1A CN201910411526A CN111953577A CN 111953577 A CN111953577 A CN 111953577A CN 201910411526 A CN201910411526 A CN 201910411526A CN 111953577 A CN111953577 A CN 111953577A
- Authority
- CN
- China
- Prior art keywords
- chat
- user
- network
- contact
- reinforcement learning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 38
- 230000002787 reinforcement Effects 0.000 claims abstract description 73
- 238000012544 monitoring process Methods 0.000 claims abstract description 26
- 238000012549 training Methods 0.000 claims description 24
- 238000012545 processing Methods 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 10
- 230000006870 function Effects 0.000 claims description 9
- 238000012795 verification Methods 0.000 claims description 7
- 230000002265 prevention Effects 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 238000013473 artificial intelligence Methods 0.000 description 4
- 230000004075 alteration Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 125000004122 cyclic group Chemical group 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 230000010485 coping Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000003203 everyday effect Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L51/00—User-to-user messaging in packet-switching networks, transmitted according to store-and-forward or real-time protocols, e.g. e-mail
- H04L51/04—Real-time or near real-time messaging, e.g. instant messaging [IM]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L51/00—User-to-user messaging in packet-switching networks, transmitted according to store-and-forward or real-time protocols, e.g. e-mail
- H04L51/04—Real-time or near real-time messaging, e.g. instant messaging [IM]
- H04L51/043—Real-time or near real-time messaging, e.g. instant messaging [IM] using or handling presence information
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L51/00—User-to-user messaging in packet-switching networks, transmitted according to store-and-forward or real-time protocols, e.g. e-mail
- H04L51/06—Message adaptation to terminal or network requirements
- H04L51/063—Content adaptation, e.g. replacement of unsuitable content
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Machine Translation (AREA)
- Information Transfer Between Computers (AREA)
Abstract
The invention provides a method, a device and a readable storage medium for preventing message misdistribution, which can automatically adjust parameters and search a reinforcement learning network of a network structure by obtaining in advance; obtaining a user chat record in advance, inputting the user chat record into the reinforcement learning network, and establishing an optimal relationship map between a user and a contact person; monitoring chat records in user chat software in real time, inputting the chat records into the reinforcement learning network to establish a chat topic model and obtaining a chat topic of a recent chat record; and according to the chat subject of the contact person in the user chat, determining whether the chat subject of the contact person in the user chat is matched with the relationship type of the contact person by combining the relationship type of the contact person in the optimal relationship map, and if not, sending an error prompt to prevent the user message from being mistakenly sent. The invention can effectively prevent the message from being mistakenly sent in the chat.
Description
Technical Field
The present invention relates to the field of communications technologies, and in particular, to a method, a system, and a readable storage medium for preventing message misdistribution.
Background
With the popularization of mobile intelligent devices, social software of chat tools is more and more commonly used, and most people use the chat software to communicate every day. People can communicate, send and receive files on the chat software. The chat software meets the requirements of life and work, reduces the problem of inconvenient communication, greatly facilitates life and improves the work efficiency. However, as the chat software is used more and more frequently, we may face the problem that a message may be misrouted, and when we are chatting with several people at the same time, or in a time emergency situation, or we originally select the intended contact, a group of messages suddenly bounces off the interface resulting in misordering. In summary, in various complex situations of the process of using the mobile intelligent device, message error is possible, and if the message error is some insignificant message, it will not cause any influence, but if some important working message or message needing protection and some important file are encountered, some unnecessary trouble may be brought. Obviously, it is necessary to take some intelligent measures and means to prevent the message from being sent, and there is no intelligent means to prevent the message from being sent.
Accordingly, the prior art is yet to be improved and developed.
Disclosure of Invention
In view of the defects of the prior art, the invention provides a technical scheme for preventing the message from being mistakenly sent in the chat software by using the AI technology, which can establish a relation map between a user and a contact according to the chat contact and the chat record of the user, train according to the linguistic data of the chat record to obtain an ideal training and learning network model, and continuously learn along with the increase of the data of the chat record, so that the data which is not matched with the learning type can be prompted or intercepted in the chat process, and the message can be prevented from being mistakenly sent in the chat process of the user.
The technical scheme adopted by the invention for solving the technical problem is as follows:
a method for preventing message from missending is used for preventing the missending of messages in the chat process of users using chat software, and comprises the following steps:
obtaining a reinforcement learning network which can automatically adjust parameters and search a network structure for training in advance;
obtaining a user chat record in advance, inputting the user chat record into the reinforcement learning network, and establishing an optimal relationship map between a user and a contact person;
monitoring chat records in user chat software in real time, inputting the chat records into the reinforcement learning network to establish a chat topic model and obtaining a chat topic of a recent chat record;
and according to the chat subjects of the user and the contact persons, determining whether the chat subjects of the user and the contact persons are matched with the relationship types of the contact persons or not by combining the optimal relationship map, and if not, sending a message error prompt.
As a further improved technical scheme, the method for obtaining the reinforcement learning network model capable of automatically adjusting parameters and searching the network structure for training in advance specifically comprises the following steps:
pre-establishing a controller network consisting of an LSTM network structure;
the controller network generating a sub-network structure;
the sub-network structure conducts a plurality of steps of training on a training set with separated external input corpora to obtain a sub-network model;
inputting a verification set separated from an externally input corpus into the sub-network model to obtain an accuracy rate R, and feeding the accuracy rate R back to the controller network;
and the controller network adjusts the controller network parameter theta through the accuracy rate R, and circularly adjusts the sub-network to obtain the optimized reinforcement learning network.
As a further improved technical solution, the step of inputting the obtained chat records of the user to the reinforcement learning network in advance to establish the optimal relationship map between the user and the contact specifically includes the following steps:
extracting chat records with characters in the chat records of the users by using an unsupervised learning method to form an original corpus;
performing semantic unified processing on the original corpus to replace the original corpus with a corpus represented by a specific contact noun;
and inputting the corpus representation containing specific contact nouns into the reinforcement learning network, and forming an optimal relation map by training the reinforcement learning network.
As a further improved technical solution, the step of inputting the chat records in the real-time monitoring user chat software into the reinforcement learning network to establish the chat topic model specifically includes the following steps:
monitoring the latest chatting record data in the user chatting software in real time;
inputting the latest chatting record data of the user into the reinforcement learning network to learn a chatting subject model;
and obtaining the latest chatting record data and inputting the latest chatting record data into the chatting topic model to obtain the chatting topic.
As a further improved technical scheme, the monitoring of the chat records of the user in real time, determining whether the chat topics of the user and the contact are matched with the relationship types of the contact according to the chat topics of the user and the contact in combination with the optimal relationship map, and if not, sending a message error prompt specifically includes the following steps:
obtaining the chat topic type of the text of the message to be sent through the chat topic model;
acquiring the relationship type between a user and a certain contact through the optimal relationship map;
judging whether the chat subject type of the user and a contact is matched with the relationship type;
when the chat subject type and the relation type of the user and a certain contact are not matched, sending a message error prompt, and selecting whether to send according to the error prompt by the user; and when the chat subject type and the relationship type of the user and a certain contact are matched, not sending out an error prompt.
The invention also provides a device for preventing the message from being mistakenly sent, which is used for preventing the message from being mistakenly sent in the chatting process of the user by using the chatting software and comprises a reinforcement learning module, a relation map module, a chatting monitoring module and an analysis processing module;
the reinforcement learning module is used for obtaining automatic adjustment parameters in advance and searching a reinforcement learning network for training a network structure;
the relation map module is used for obtaining chat records of the users in advance and inputting the chat records into the reinforcement learning module to learn and establish an optimal relation map between the users and the contact persons;
the chat monitoring module is used for continuously monitoring the chat records in the user chat software, inputting the chat records into the reinforcement learning module to learn and establish a chat topic model between the user and the contact person and obtain the chat topic of the latest chat record;
the analysis processing module is used for acquiring the chat subject of the recent chat record of the user, and according to the chat subject of the user and the contact, determining whether the chat subject of the user and the contact is matched with the relationship type of the contact or not by combining the relationship type of the user and the contact in the relationship map module, and if not, sending a message error prompt.
As a further improved technical solution, the functions of the reinforcement learning module to automatically adjust parameters and search a network structure specifically include:
obtaining a controller network consisting of an LSTM network structure in advance;
the controller network generating a sub-network structure;
the sub-network structure trains a sub-network model in a plurality of steps on a training set separated from external input corpora;
inputting a verification set separated from an externally input corpus into the sub-network model to obtain an accuracy rate R, and feeding the accuracy rate R back to the controller network;
and the controller network adjusts the controller network parameter theta through the accuracy rate R, and further circularly adjusts the sub-network structure to obtain the optimized reinforcement learning network.
As a further improved technical solution, the function of the relationship graph module obtaining the optimal relationship graph between the user and the contact in advance specifically is as follows:
extracting chat records with characters in the chat records of the users by using an unsupervised learning method to form an original corpus;
performing semantic unified processing on the original corpus to replace the original corpus with a corpus represented by a specific contact noun;
and inputting the corpus representation containing specific contact nouns into the optimized reinforcement learning network of the reinforcement learning module, and forming an optimal relation map by training the reinforcement learning network.
As a further improved technical solution, the functions of the chat monitoring module monitoring the chat records of the user, learning to establish a chat topic model between the user and the contact, and obtaining the chat topic of the latest chat record are specifically as follows:
monitoring the latest chatting record data in the user chatting software in real time;
inputting the latest chatting record data of the user into a reinforcement learning network in the reinforcement learning module to learn a chatting subject model;
and obtaining the latest chatting record data and inputting the latest chatting record data into the chatting topic model to obtain the chatting topic.
As a further improved technical solution, the function of the analysis processing module for analyzing the relationship between the user chat record and the contact is specifically:
obtaining the chat topic type of the text of the message to be sent through the chat topic model;
acquiring the relationship type between a user and a certain contact through the optimal relationship map;
judging whether the chat subject type of the user and a contact is matched with the relationship type;
when the chat subject type and the relation type of the user and a certain contact are not matched, sending a message error prompt, and selecting whether to send according to the error prompt by the user; and when the chat subject type and the relationship type of the user and a certain contact are matched, not sending out an error prompt.
The present invention also provides a readable storage medium storing a program of a message mistransmission preventing method, which when executed by a processor, implements the above-described method steps of the message mistransmission preventing method.
Compared with the prior art, the invention adopts the neural network to train the chat subjects and the contacts in the chat records of the user to obtain the corresponding optimal subject network model, and integrates the relation map and the chat record data to judge whether the messages to be sent are mistaken in the chat of the user, so as to effectively prevent the messages from being mistaken in the chat.
Drawings
The embodiments of the invention will be further described with reference to the accompanying drawings, in which:
fig. 1 is a flowchart of a preferred embodiment of a method for preventing message misdistribution according to the present invention.
Fig. 2 is a schematic block diagram of a preferred embodiment of a message mistransmission prevention apparatus according to the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer and clearer, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The artificial intelligence is the current hot technology, is used more and more in the intelligent equipment, and the use of the artificial intelligence technology greatly facilitates the use of the intelligent equipment by people and also improves the working efficiency. In the artificial intelligence technology, it has become a common means to complete the operation of the intelligent device by using the neural network, and the neural network can establish an optimized network model in mass big data, so that the intelligent device can continuously form an intelligent coping strategy according to external data information. The chat APP installed in the intelligent device can meet the social requirement of people, convenience is provided for life connection of people, in order to solve unnecessary troubles caused by mistaken sending of messages in the chat software, the invention provides a solving mechanism for solving the problem that messages of the chat software are possibly mistaken sent, and the embarrassing situation that a user mistakenly sends messages in the chat software is effectively prevented by using artificial intelligence.
Fig. 1 is a flowchart of a preferred embodiment of a method for preventing message misdistribution according to the present invention, for preventing message misdistribution during a chat process of a user using chat software, including the following steps:
and step S100, obtaining a reinforcement learning network which can automatically adjust parameters and search a network structure for training in advance.
Specifically, the step of establishing the reinforcement learning network includes the following steps:
obtaining a controller network consisting of an LSTM network structure in advance; the LSTM (Long Short-Term Memory) is a neural network structure, is a special circulating neural network, and adopts the LSTM-based neural network as a controller network.
The controller network generating a sub-network structure;
the sub-network structure carries out a plurality of steps of training on a training set separated from external input corpora to obtain a sub-network model;
inputting a verification set separated from an externally input corpus into the sub-network model to obtain an accuracy rate R, and feeding the accuracy rate R back to the controller network;
and the controller network adjusts the controller network parameter theta through the accuracy rate R, so as to circularly adjust the sub-network structure and obtain the optimized reinforcement learning network.
Therefore, the reinforcement learning network is a cyclic iteration process, the controller generates a sub-network structure, trains the sub-network structure to obtain a model, and predicts the accuracy rate R on a verification set separated from the externally input corpus, the controller updates the controller network parameter theta through the accuracy rate R, the continuous updating of the controller network parameter theta also enables the generated sub-network structure to be better and better, finally, a set of best sub-network parameters and an optimal sub-network structure are obtained, and an ideal reinforcement learning network can be obtained subsequently according to various input parameters.
And step S200, obtaining a user chat record in advance, inputting the user chat record into the reinforcement learning network, and establishing an optimal relationship map between the user and the contact person.
Specifically, the step of obtaining the optimal relationship map between the user and the contact in advance comprises the following steps:
extracting chat records with characters in the chat records of the users by using an unsupervised learning method to form an original corpus; for the original corpus, one chat record may be used as the original corpus, or several chat records may be combined into one original corpus.
Performing semantic unified processing on the original corpus to replace the original corpus with a corpus represented by a specific contact noun; the original corpus is processed by preprocessing the original corpus in a unified entity and pronoun replacement mode, wherein the unified entity is that two different nouns appear in a sentence, but the two different nouns point to a contact person, and the different nouns pointing to the same contact person need to be replaced by the same noun for representation; pronoun replacement is to replace pronouns such as "i", "you", "he" and "s" in the original corpus with specific contact noun representations to accurately determine the contact corresponding to the pronouns.
And inputting the corpus representation containing specific contact nouns into the reinforcement learning network to form an optimal relationship map. After the original corpora are preprocessed, data are input into the reinforcement learning network as parameters to construct an optimal relationship map between the user and a plurality of contacts, the original corpora are continuously increased along with the increase of chat records, the relationship map of characters is more and more perfect after the original corpora are continuously input into the reinforcement learning network, and the system can learn the relationship between the user and a certain contact through the relationship map of characters, such as the relationship of characters, including a colleague relationship, a classmate relationship, a friend relationship, a child-woman relationship, a sister relationship, a teacher-student relationship, a lovers relationship or a couple relationship.
Step S300, monitoring chat records in the user chat software in real time, inputting the chat records into the reinforcement learning network, establishing a chat topic model and obtaining a chat topic of the latest chat records.
Specifically, the establishment of the chat topic model according to the chat records between the user and the contact person comprises the following steps:
monitoring the latest chatting record data in the user chatting software in real time;
inputting the latest chatting record data of the user into the reinforcement learning network to learn a chatting subject model; with the increase of the chatting record data, the reason is established by the character relation graph, and the chatting subject model is more and more perfect.
And obtaining the latest chatting record data and inputting the latest chatting record data into the chatting subject model to obtain a chatting subject, wherein the chatting subject comprises subjects of work, dining, traveling, shopping and the like.
And step S400, according to the chat subjects of the user and the contact persons, determining whether the chat subjects of the user and the contact persons are matched with the relationship types of the contact persons or not by combining the optimal relationship map, and if not, sending a message error prompt.
Specifically, the judgment of monitoring the chat records of the user to prevent the message from being sent mistakenly comprises the following steps:
obtaining the chat topic type of the text of the message to be sent through the chat topic model;
acquiring the relationship type between a user and a certain contact through the optimal relationship map;
judging whether the chat subject type of the user and a contact is matched with the relationship type;
when the chat subject type and the relation type of the user and a certain contact are not matched, sending a message error prompt, and selecting whether to send according to the error prompt by the user; and when the chat subject type and the relationship type of the user and a certain contact are matched, not sending out an error prompt.
As shown in fig. 2, a schematic block diagram of a preferred embodiment of an apparatus for preventing misdistribution of messages during a user chatting using a chatting software according to the present invention, the apparatus 60 includes a reinforcement learning module 601, a relationship map module 602, a chatting monitoring module 603, and an analysis processing module 604.
The reinforcement learning module 601 is configured to obtain a reinforcement learning network capable of establishing an automatic adjustment parameter and searching a network structure for training in advance;
the relationship map module 602 is configured to obtain a chat record of a user in advance, input the chat record into the reinforcement learning module, learn to establish an optimal relationship map between the user and a contact;
the chat monitoring module 603 is configured to continuously monitor chat records in the user chat software, input the chat records into the reinforcement learning module, learn to establish a chat topic model between the user and the contact, and obtain a chat topic of the latest chat record;
the analysis processing module 604 is configured to obtain a chat topic of the recent chat record of the user, and determine whether the chat topic of the contact in the user chat matches the relationship type of the contact according to the chat topic of the user and the contact in the relationship map module in combination with the relationship type between the user and the contact, and send a message error prompt if the chat topic of the user and the contact in the user chat does not match the relationship type of the contact.
The reinforcement learning module 601 can automatically adjust parameters and search network structures, and the functions are specifically as follows:
obtaining a controller network consisting of an LSTM network structure in advance;
the controller network generating a sub-network structure;
the sub-network structure carries out a plurality of steps of training on a training set separated from external input corpora to obtain a sub-network model;
inputting a verification set separated from an externally input corpus into the sub-network model to obtain an accuracy rate R, and feeding the accuracy rate R back to the controller network;
and the controller network adjusts the controller network parameter theta through the accuracy rate R, and further circularly adjusts the sub-network structure to obtain the optimized reinforcement learning network.
In the same way as the preferred embodiment of the method, the reinforcement learning network is a cyclic iteration process, the controller generates a sub-network structure, the sub-network structure is trained to obtain a model, the accuracy rate R is predicted and generated on a training set separated from the externally input corpus, the controller updates the controller network parameter theta through the accuracy rate R, the continuous updating of the controller network parameter theta also enables the generated sub-network structure to be better and better, finally, a set of best sub-network parameters and an optimal sub-network structure are obtained, and the ideal reinforcement learning network can be obtained according to various input parameters.
The function of the relationship graph module 602 obtaining the optimal relationship graph between the user and the contact in advance specifically is as follows:
extracting chat records with characters in the chat records of the users by using an unsupervised learning method to form an original corpus; for the original corpus, one chat record may be used as the original corpus, or several chat records may be combined into one original corpus.
Performing semantic unified processing on the original corpus to replace the original corpus with a corpus represented by a specific contact noun; the original corpus is processed by preprocessing the original corpus in a unified entity and pronoun replacement mode, wherein the unified entity is that two different nouns appear in a sentence, but the two different nouns point to a contact person, and the different nouns pointing to the same contact person need to be replaced by the same noun for representation; pronoun replacement is to replace pronouns such as "i", "you", "he" and "s" in the original corpus with specific contact noun representations to accurately determine the contact corresponding to the pronouns.
And inputting the corpus representation containing specific contact nouns into the optimized reinforcement learning network of the reinforcement learning module to form an optimal relation map. After the original corpora are preprocessed, data are input into the reinforcement learning network as parameters to construct an optimal relationship map between the user and a plurality of contacts, the original corpora are continuously increased along with the increase of chat records, the relationship map of characters is more and more perfect after the original corpora are continuously input into the reinforcement learning network, and the system can learn the relationship between the user and a certain contact through the relationship map of characters, such as the relationship of characters, including a colleague relationship, a classmate relationship, a friend relationship, a child-woman relationship, a sister relationship, a teacher-student relationship, a lovers relationship or a couple relationship.
The chat monitoring module 603 continuously monitors the chat records of the user and learns to establish a chat topic model between the user and the contact person and obtain the chat topic of the latest chat record specifically comprises:
monitoring the latest chatting record data in the user chatting software in real time;
inputting the latest chatting record data of the user into a reinforcement learning network in the reinforcement learning module to learn a chatting subject model; with the increase of the chatting record data, the reason is established by the character relation graph, and the chatting subject model is more and more perfect.
And obtaining the latest chatting record data and inputting the latest chatting record data into the chatting subject model to obtain a chatting subject, wherein the chatting subject comprises subjects of work, dining, traveling, shopping and the like.
The function of the analysis processing module 604 for analyzing the relationship between the user chat record and the contact is specifically:
obtaining the chat topic type of the text of the message to be sent through the chat topic model;
acquiring the relationship type between a user and a certain contact through the optimal relationship map;
judging whether the chat subject type of the user and a contact is matched with the relationship type;
when the chat subject type and the relation type of the user and a certain contact are not matched, sending a message error prompt, and selecting whether to send according to the error prompt by the user; and when the chat subject type and the relationship type of the user and a certain contact are matched, not sending out an error prompt.
The present invention also provides a readable storage medium storing a program of a message mistransmission prevention method, which when executed by a processor, implements the above-described method steps of the message mistransmission prevention. The specific working principle is the same as the above preferred embodiment of the method for preventing message missending, and is not described herein again.
It should be understood that the above-mentioned embodiments are merely preferred examples of the present invention, and not restrictive, but rather, all the changes, substitutions, alterations and modifications that come within the spirit and scope of the invention as described above may be made by those skilled in the art, and all the changes, substitutions, alterations and modifications that fall within the scope of the appended claims should be construed as being included in the present invention.
Claims (11)
1. A method for preventing message from being mistakenly sent is used for preventing the message from being mistakenly sent in the process of chatting by a user by using chatting software, and is characterized by comprising the following steps:
obtaining a reinforcement learning network which can automatically adjust parameters and search a network structure for training in advance;
obtaining a user chat record in advance, inputting the user chat record into the reinforcement learning network, and establishing an optimal relationship map between a user and a contact person;
monitoring chat records in user chat software in real time, inputting the chat records into the reinforcement learning network to establish a chat topic model and obtaining a chat topic of a recent chat record;
and according to the chat subjects of the user and the contact persons, determining whether the chat subjects of the user and the contact persons are matched with the relationship types of the contact persons or not by combining the optimal relationship map, and if not, sending a message error prompt.
2. The method of claim 1, wherein the pre-obtaining of the reinforcement learning network trained by the auto-tuning parameters and the searching network structure comprises the following steps:
pre-establishing a controller network consisting of an LSTM network structure;
the controller network generating a sub-network structure;
the sub-network structure carries out a plurality of steps of training on a training set separated from external input corpora to obtain a sub-network model;
inputting a verification set separated from an externally input corpus into the sub-network model to obtain an accuracy rate R, and feeding the accuracy rate R back to the controller network;
and the controller network adjusts the controller network parameter theta through the accuracy rate R, and circularly adjusts the sub-network structure to obtain the optimized reinforcement learning network.
3. The method for preventing message misdistribution according to claim 1 or 2, wherein the step of inputting the pre-obtained chat records of the user into the reinforcement learning network to establish the optimal relationship map between the user and the contact specifically comprises the following steps:
extracting chat records with characters in the chat records of the users by using an unsupervised learning method to form an original corpus;
performing semantic unified processing on the original corpus to replace the original corpus with a corpus represented by a specific contact noun;
and inputting the corpus representation containing specific contact nouns into the reinforcement learning network, and forming an optimal relation map by training the reinforcement learning network.
4. The method as claimed in claim 3, wherein the step of monitoring chat records in the user chat software in real time and inputting the chat records into the reinforcement learning network to build a chat topic model and obtain the chat topic of the latest chat record includes the following steps:
monitoring the latest chatting record data in the user chatting software in real time;
inputting the latest chatting record data of the user into the reinforcement learning network to learn a chatting subject model;
and obtaining the latest chatting record data and inputting the latest chatting record data into the chatting topic model to obtain the chatting topic.
5. The method for preventing the message from being mistakenly sent according to claim 4, wherein the step of confirming whether the chat topic of the user and the contact is matched with the relationship type of the contact or not according to the chat topic of the user and the contact and combining the optimal relationship map, and if not, sending a message error prompt specifically comprises the following steps:
obtaining the chat topic type of the text of the message to be sent through the chat topic model;
acquiring the relationship type between a user and a certain contact through the optimal relationship map;
judging whether the chat subject type of the user and a contact is matched with the relationship type;
when the chat subject type and the relation type of the user and a certain contact are not matched, sending a message error prompt, and selecting whether to send according to the error prompt by the user; and when the chat subject type and the relationship type of the user and a certain contact are matched, not sending out an error prompt.
6. A device for preventing message from being mistakenly sent is used for preventing the message from being mistakenly sent in the chat process of a user using chat software, and is characterized in that the device for preventing message from being mistakenly sent comprises a reinforcement learning module, a relation map module, a chat monitoring module and an analysis processing module;
the reinforcement learning module is used for obtaining a reinforcement learning network which can automatically adjust parameters and search a network structure for training in advance;
the relation map module is used for obtaining chat records of the users in advance and inputting the chat records into the reinforcement learning module to learn and establish an optimal relation map between the users and the contact persons;
the chat monitoring module is used for continuously monitoring the chat records in the user chat software, inputting the chat records into the reinforcement learning module to learn and establish a chat topic model between the user and the contact person and obtain the chat topic of the latest chat record;
the analysis processing module is used for acquiring the chat subject of the recent chat record of the user, and according to the chat subject of the user and the contact, determining whether the chat subject of the user and the contact is matched with the relationship type of the contact or not by combining the relationship type of the user and the contact in the relationship map module, and if not, sending a message error prompt.
7. The apparatus for preventing message misdistribution according to claim 6, wherein the reinforcement learning module automatically adjusts parameters and searches network structure specifically:
obtaining a controller network consisting of an LSTM network structure in advance;
the controller network generating a sub-network structure;
the sub-network structure carries out a plurality of steps of training on a training set separated from external input corpora to obtain a sub-network model;
inputting a verification set separated from an externally input corpus into the sub-network model to obtain an accuracy rate R, and feeding the accuracy rate R back to the controller network;
and the controller network adjusts the controller network parameter theta through the accuracy rate R, and further circularly adjusts the sub-network structure to obtain the optimized reinforcement learning network.
8. The apparatus for preventing the message from being misdelivered according to claim 6 or 7, wherein the relationship graph module obtains the optimal relationship graph function between the user and the contact in advance specifically as follows:
extracting chat records with characters in the chat records of the users by using an unsupervised learning method to form an original corpus;
performing semantic unified processing on the original corpus to replace the original corpus with a corpus represented by a specific contact noun;
and inputting the corpus representation containing specific contact nouns into the optimized reinforcement learning network of the reinforcement learning module, and forming an optimal relation map by training the reinforcement learning network.
9. The apparatus for preventing the message from being mistakenly sent according to claim 8, wherein the chat monitoring module continuously monitors the chat records of the user and learns to establish the chat topic model between the user and the contact person and obtain the chat topic of the latest chat record specifically comprises:
monitoring the latest chatting record data in the user chatting software in real time;
inputting the latest chatting record data of the user into a reinforcement learning network in the reinforcement learning module to learn a chatting subject model;
and obtaining the latest chatting record data and inputting the latest chatting record data into the chatting topic model to obtain the chatting topic.
10. The apparatus for preventing the message from being mistakenly sent according to claim 9, wherein the function of the analysis processing module for analyzing the relationship between the chat records of the user and the contacts is specifically as follows:
obtaining the chat topic type of the text of the message to be sent through the chat topic model;
acquiring the relationship type between a user and a certain contact through the optimal relationship map;
judging whether the chat subject type of the user and a contact is matched with the relationship type;
when the chat subject type and the relation type of the user and a certain contact are not matched, sending a message error prompt, and selecting whether to send according to the error prompt by the user; and when the chat subject type and the relationship type of the user and a certain contact are matched, not sending out an error prompt.
11. A readable storage medium, characterized in that the readable storage medium stores a program of a message misdistribution prevention method, which when executed by a processor, implements the method steps of the message misdistribution prevention of any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910411526.1A CN111953577A (en) | 2019-05-16 | 2019-05-16 | Method, system and readable storage medium for preventing message from being mistakenly sent |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910411526.1A CN111953577A (en) | 2019-05-16 | 2019-05-16 | Method, system and readable storage medium for preventing message from being mistakenly sent |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111953577A true CN111953577A (en) | 2020-11-17 |
Family
ID=73336379
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910411526.1A Pending CN111953577A (en) | 2019-05-16 | 2019-05-16 | Method, system and readable storage medium for preventing message from being mistakenly sent |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111953577A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112396504A (en) * | 2021-01-21 | 2021-02-23 | 北京天通慧智科技有限公司 | E-commerce order intercepting method and device and electronic equipment |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104182535A (en) * | 2014-08-29 | 2014-12-03 | 苏州大学 | Method and device for extracting character relation |
| CN104462326A (en) * | 2014-12-02 | 2015-03-25 | 百度在线网络技术（北京）有限公司 | Person relation analyzing method as well as method and device for providing person information |
| CN104731842A (en) * | 2013-12-23 | 2015-06-24 | 国际商业机器公司 | Mapping relationships using electronic communications data |
| CN105099853A (en) * | 2014-04-25 | 2015-11-25 | 国际商业机器公司 | Erroneous message sending preventing method and system |
| CN105468605A (en) * | 2014-08-25 | 2016-04-06 | 济南中林信息科技有限公司 | Entity information map generation method and device |
| CN106375377A (en) * | 2016-08-25 | 2017-02-01 | 深圳市金立通信设备有限公司 | Data processing method and terminal |
| US20170068904A1 (en) * | 2015-09-09 | 2017-03-09 | Microsoft Technology Licensing, Llc | Determining the Destination of a Communication |
| US20170351855A1 (en) * | 2016-06-03 | 2017-12-07 | International Business Machines Corporation | Identifying sensitive information in a communication based on network communications history |
| CN108039995A (en) * | 2017-10-25 | 2018-05-15 | 努比亚技术有限公司 | Message sending control method, terminal and computer-readable recording medium |
| CN108566330A (en) * | 2018-03-21 | 2018-09-21 | 联想(北京)有限公司 | Information processing method and the first electronic equipment |
| CN109040430A (en) * | 2018-07-10 | 2018-12-18 | 麒麟合盛网络技术股份有限公司 | message display method and device |
| CN109643325A (en) * | 2017-05-26 | 2019-04-16 | 微软技术许可有限责任公司 | The recommending friends in automatic chatting |
| CN109716328A (en) * | 2016-12-15 | 2019-05-03 | 华为技术有限公司 | A kind of method and device of information alert |
-
2019
- 2019-05-16 CN CN201910411526.1A patent/CN111953577A/en active Pending
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104731842A (en) * | 2013-12-23 | 2015-06-24 | 国际商业机器公司 | Mapping relationships using electronic communications data |
| CN105099853A (en) * | 2014-04-25 | 2015-11-25 | 国际商业机器公司 | Erroneous message sending preventing method and system |
| CN105468605A (en) * | 2014-08-25 | 2016-04-06 | 济南中林信息科技有限公司 | Entity information map generation method and device |
| CN104182535A (en) * | 2014-08-29 | 2014-12-03 | 苏州大学 | Method and device for extracting character relation |
| CN104462326A (en) * | 2014-12-02 | 2015-03-25 | 百度在线网络技术（北京）有限公司 | Person relation analyzing method as well as method and device for providing person information |
| US20170068904A1 (en) * | 2015-09-09 | 2017-03-09 | Microsoft Technology Licensing, Llc | Determining the Destination of a Communication |
| US20170351855A1 (en) * | 2016-06-03 | 2017-12-07 | International Business Machines Corporation | Identifying sensitive information in a communication based on network communications history |
| CN106375377A (en) * | 2016-08-25 | 2017-02-01 | 深圳市金立通信设备有限公司 | Data processing method and terminal |
| CN109716328A (en) * | 2016-12-15 | 2019-05-03 | 华为技术有限公司 | A kind of method and device of information alert |
| CN109643325A (en) * | 2017-05-26 | 2019-04-16 | 微软技术许可有限责任公司 | The recommending friends in automatic chatting |
| CN108039995A (en) * | 2017-10-25 | 2018-05-15 | 努比亚技术有限公司 | Message sending control method, terminal and computer-readable recording medium |
| CN108566330A (en) * | 2018-03-21 | 2018-09-21 | 联想(北京)有限公司 | Information processing method and the first electronic equipment |
| CN109040430A (en) * | 2018-07-10 | 2018-12-18 | 麒麟合盛网络技术股份有限公司 | message display method and device |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112396504A (en) * | 2021-01-21 | 2021-02-23 | 北京天通慧智科技有限公司 | E-commerce order intercepting method and device and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR102338618B1 (en) | Method for providing chatting service with chatbot assisted by human agents | |
| CN103079008B (en) | Method and system for automatically generating replying suggestion according to content of short message | |
| US10193850B2 (en) | Discovering questions, directives, and requests from, and prioritizing replies to, a preferred message sender method and apparatus | |
| US20190243901A1 (en) | Method and System to Automatically Change or Update the Configuration or Setting of a Communication System | |
| CN105631022A (en) | Method and device for processing information | |
| CN107306306A (en) | Communicating number processing method and processing device | |
| US11757805B2 (en) | Systems and methods for mediation using NLP and machine learning techniques | |
| CN108763548A (en) | Collect method, apparatus, equipment and the computer readable storage medium of training data | |
| CN109716328A (en) | A kind of method and device of information alert | |
| CN110830661A (en) | Automatic dial testing method for intelligent voice customer service | |
| CN111953577A (en) | Method, system and readable storage medium for preventing message from being mistakenly sent | |
| CN108583629A (en) | A kind of railcar business fault handling method | |
| CN110120909A (en) | Transmission method and device, storage medium, the electronic device of message | |
| EP2169992B1 (en) | Detection of abnormal behaviour among users of mobile terminals in a telecommunications network | |
| CN107577673A (en) | Based on the robot interactive method and device monitored with detecting mechanism | |
| CN110265026A (en) | A kind of meeting shorthand system and meeting stenography method | |
| CN108111388B (en) | Communication system with intelligent agent | |
| CN113542509B (en) | Emergency processing method, device, storage medium and equipment | |
| CN109189912A (en) | The update method and device of user's consulting statement library | |
| CN110119715B (en) | Accompanying robot and emotion recognition method | |
| CN110263313A (en) | A kind of man-machine coordination edit methods for meeting shorthand | |
| CN116386276B (en) | Institution and home multimode safety monitoring system and method | |
| JP3783936B2 (en) | Interactive information providing apparatus, interactive information providing processing method, program, and recording medium | |
| CN110543556A (en) | Dialogue configuration method, storage medium and electronic equipment | |
| CN109617782A (en) | A kind of method for sending information, client, storage medium and equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20201117 |
|
| RJ01 | Rejection of invention patent application after publication |