CN114356173A - Message reply method and related device, electronic equipment and storage medium - Google Patents
Message reply method and related device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN114356173A CN114356173A CN202111478629.3A CN202111478629A CN114356173A CN 114356173 A CN114356173 A CN 114356173A CN 202111478629 A CN202111478629 A CN 202111478629A CN 114356173 A CN114356173 A CN 114356173A
- Authority
- CN
- China
- Prior art keywords
- message
- candidate
- reply
- user
- opposite
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/048—Interaction techniques based on graphical user interfaces [GUI]
- G06F3/0481—Interaction techniques based on graphical user interfaces [GUI] based on specific properties of the displayed interaction object or a metaphor-based environment, e.g. interaction with desktop elements like windows or icons, or assisted by a cursor's changing behaviour or appearance
- G06F3/0482—Interaction with lists of selectable items, e.g. menus
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F9/00—Arrangements for program control, e.g. control units
- G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
- G06F9/44—Arrangements for executing specific programs
- G06F9/451—Execution arrangements for user interfaces
Abstract
The application discloses a message reply method, a related device, electronic equipment and a storage medium, wherein the message reply method comprises the following steps: responding to the fact that a user copies the opposite-end message on the interactive interface, predicting based on the opposite-end message to obtain a plurality of candidate messages for replying the opposite-end message, and displaying the candidate messages on an expansion interface displayed on the same screen as the interactive interface; and selecting the candidate message as a first reply message to reply the opposite-end message based on the selection instruction of the user. According to the scheme, the communication efficiency can be improved, and meanwhile, the privacy disclosure risk and the consumption of electric quantity and memory are reduced.
Description
Technical Field
The present application relates to the field of information interaction technologies, and in particular, to a message reply method, a related apparatus, an electronic device, and a storage medium.
Background
With the rapid development of various industries, people have higher and higher demands on communication efficiency. Taking an office scene as an example, the communication efficiency is affected by the fact that the user needs to pay attention to work during the office process, the user does not have extra effort to return the opposite party by considering words and sentences, or taking daily life as an example, the user sometimes cuts a half day in the process of chatting with friends because of considering how to return the opposite party.
At present, users are generally required to collect user conversation information after providing barrier-free permission so as to provide answer for the users, and therefore the problems of privacy disclosure, high power consumption, high memory consumption and the like exist. In view of this, how to reduce the privacy disclosure risk and the power consumption and memory consumption while improving the communication efficiency becomes an urgent problem to be solved.
Disclosure of Invention
The technical problem mainly solved by the application is to provide a message reply method, a related device, an electronic device and a storage medium, which can reduce privacy disclosure risks and electric quantity and memory consumption while improving communication efficiency.
In order to solve the above technical problem, a first aspect of the present application provides a message reply method, including: responding to the fact that a user copies the opposite-end message on the interactive interface, predicting based on the opposite-end message to obtain a plurality of candidate messages for replying the opposite-end message, and displaying the candidate messages on an expansion interface displayed on the same screen as the interactive interface; and selecting the candidate message as a first reply message to reply the opposite-end message based on the selection instruction of the user.
In order to solve the above technical problem, a second aspect of the present application provides a message reply device, including: the message prediction module is used for responding to the fact that a user copies an opposite-end message on the interactive interface, predicting based on the opposite-end message to obtain a plurality of candidate messages for replying the opposite-end message, and displaying the candidate messages on an expansion interface displayed on the same screen as the interactive interface; and the message selection module is used for selecting the candidate message as a first reply message based on the selection instruction of the user so as to reply the opposite-end message.
In order to solve the above technical problem, a third aspect of the present application provides an electronic device, which includes a screen, a communication circuit, a memory and a processor, wherein the screen, the communication circuit and the memory are coupled to the processor, the memory stores program instructions, and the processor is configured to execute the program instructions to implement the message reply method in the first aspect.
In order to solve the above technical problem, a fourth aspect of the present application provides a computer-readable storage medium storing program instructions executable by a processor, the program instructions being configured to implement the message reply method in the first aspect.
According to the scheme, in response to the fact that the user is detected to copy the opposite-end message on the interactive interface, the candidate messages used for replying the opposite-end message are obtained based on the prediction of the opposite-end message, and the candidate messages are displayed on the expansion interface displayed on the same screen as the interactive interface, on the basis, the candidate messages are selected as the first reply message to reply the opposite-end message based on the selection instruction of the user, so that in the message reply process, on one hand, the candidate messages can be predicted according to the opposite-end message without manual input of the user, the user can reply only by selecting the candidate messages, the communication efficiency is favorably improved, on the other hand, the candidate messages are used as content transfer through the expansion interface, the dependence on barrier-free permission is not needed, and the privacy disclosure risk, the electric quantity and the memory consumption are favorably reduced. Therefore, the communication efficiency can be improved, and meanwhile, the privacy disclosure risk, the electric quantity and the memory consumption are reduced.
Drawings
FIG. 1 is a schematic flowchart of an embodiment of a message reply method according to the present application;
FIG. 2 is a schematic diagram of an embodiment of an intelligent reply function;
FIG. 3 is a diagram of one embodiment of a binding relationship between an input method and an extension interface;
FIG. 4 is a diagram of an embodiment in which no binding relationship exists between an input method and an extension interface;
FIG. 5 is a block diagram of an embodiment of a message ordering model;
FIG. 6 is a flow chart illustrating another embodiment of a message reply method according to the present application;
FIG. 7 is a process diagram of an embodiment of a message reply method of the present application;
FIG. 8 is a block diagram of an embodiment of a message reply apparatus according to the present application;
FIG. 9 is a block diagram of an embodiment of an electronic device of the present application;
FIG. 10 is a block diagram of an embodiment of a computer-readable storage medium of the present application.
Detailed Description
The following describes in detail the embodiments of the present application with reference to the drawings attached hereto.
In the following description, for purposes of explanation and not limitation, specific details are set forth such as particular system structures, interfaces, techniques, etc. in order to provide a thorough understanding of the present application.
The terms "system" and "network" are often used interchangeably herein. The term "and/or" herein is merely an association describing an associated object, meaning that three relationships may exist, e.g., a and/or B, may mean: a exists alone, A and B exist simultaneously, and B exists alone. In addition, the character "/" herein generally indicates that the former and latter related objects are in an "or" relationship. Further, the term "plurality" herein means two or more than two.
Referring to fig. 1, fig. 1 is a flowchart illustrating a message reply method according to an embodiment of the present application.
Specifically, the method may include the steps of:
step S11: and responding to the detection that the user copies the opposite-end message on the interactive interface, predicting based on the opposite-end message to obtain a plurality of candidate messages for replying the opposite-end message, and displaying the candidate messages on an expansion interface displayed on the same screen as the interactive interface.
In one implementation scenario, the interactive interfaces may also be different depending on the specific application scenario. Taking a chat scenario as an example, the interactive interface may be a window interface provided by any communication program, and the communication program may include, but is not limited to: instant messaging software such as WeChat, QQ, iMessage and the like, and certainly, the communication program can also comprise non-instant messaging software such as but not limited to short message and the like; or, taking the online shopping scenario as an example, the interactive interface may also be an interactive window provided by any shopping program, for example, the shopping program may include but is not limited to: taobao, Jingdong and other electronic commercial software. Of course, in a real-world scenario, the interactive interface may also exist in an e-government scenario, an insurance consultation scenario, and so on, which can be analogized from the above description, and is not illustrated here.
In an implementation scenario, the technical solution of the embodiment of the present disclosure is integrated in an input method, and before detecting whether a user copies an opposite-end message, the input method may be ensured to start an intelligent reply function, and the steps in the embodiment of the present disclosure are executed under the condition that it is ensured that the input method has started the intelligent reply function. Referring to fig. 2, fig. 2 is a schematic diagram illustrating an arrangement of an embodiment of an intelligent reply function. As shown in fig. 2, a black bold rectangle is shown as an on/off option of the smart reply function, in the case of turning on the smart reply function, a first identifier may be displayed on the option to prompt a user that the smart reply function is currently turned on, the first identifier may be custom-set according to actual conditions, such as may be set as a "√" shape shown in fig. 2, or the option may be added with a background color such as "blue", which is not limited herein.
In a specific implementation scenario, in response to a user starting an intelligent reply function of an input method, a binding relationship can be established between the input method and an expansion interface, whether the user copies an opposite-end message or not is detected, and the expansion interface and the input interface of the input method are kept on the same screen for display under the condition that the binding relationship exists between the input method and the expansion interface. Specifically, in the case that the input method and the extension interface establish a binding relationship, the second identifier may be displayed on the input interface to prompt the user that the input method and the extension interface are currently in the binding relationship. Referring to fig. 3, fig. 3 is a schematic diagram of an embodiment when the input method and the extended interface have a binding relationship. As shown in FIG. 3, the second identifier may be set in the style of "thumbtack," and the "thumbtack" is indicated in a dark background when the input method is bound to the expansion interface. Of course, the second identifier may also be custom set in other styles as desired, such as may include but is not limited to: a "magnet" pattern, an "arrow" pattern, and the like, without limitation. In the mode, when the user starts the intelligent reply function of the input method, the binding relationship of the input French expansion interface is established by default, the subsequent operation of detecting whether the user copies the opposite-end message is executed, and when the input method and the expansion interface have the binding relationship, the expansion interface and the input interface of the input method are kept on the same screen for display, so that the user does not need to call the expansion interface repeatedly when copying the opposite-end message every time, and the memory consumption is reduced.
In one particular implementation scenario, the expansion interface may be a specific area that provides intelligent reply suggestions. As shown in fig. 3, the expansion interface may be disposed above the input method display interface. Of course, the extended interface may also be disposed below the input method display interface, or in other positions, which is not limited herein.
In a specific implementation scenario, after the binding relationship between the input method and the extension interface is established, the user can release the binding relationship according to the need of the user, in this case, in response to a release instruction of the user on the binding relationship, the binding relationship between the input method and the extension interface can be directly released, and in the case that the binding relationship does not exist between the input method and the extension interface, the extension interface and the input interface are kept on the same screen for display when meeting a preset trigger condition, and the preset trigger condition includes: and (3) detecting that the user copies the opposite-end message, namely, after the binding relationship between the input method and the expansion interface is released, only when the user copies the opposite-end message, the expansion interface is re-invoked. Specifically, in the case that the input method does not have a binding relationship with the extension interface, a third identifier may be displayed on the input interface to prompt the user that the input method does not currently have a binding relationship with the extension interface. Referring to fig. 4, fig. 4 is a schematic diagram of an embodiment when there is no binding relationship between the input method and the extension interface. As shown in FIG. 4, the third identifier may also be set in the style of "thumbtack," which is represented in a light background when there is no binding relationship between the input method and the expansion interface. Of course, the third identifier may also be customized to other styles as desired, such as may include but is not limited to: a "magnet" pattern, an "arrow" pattern, and the like, without limitation. In addition, as described above, in the case where the input method and the extended interface establish a binding relationship, the second identifier may be displayed on the input interface, and the release instruction may be a click instruction for the second identifier. Referring to fig. 3 and 4 in combination, after a click command of the user on the second identifier shown in fig. 3 is detected, it can be considered that the user needs to release the binding relationship between the input method and the expansion interface, and at this time, the style of "pushpin" changes from a dark color to a light color shown in fig. 4 to serve as a third identifier. Other cases may be analogized, and no one example is given here. In the above manner, in response to a release instruction of the user for the binding relationship, the binding relationship between the input method and the extension interface is released, and in the case that the binding relationship does not exist between the input method and the extension interface, the extension interface and the input interface are kept displaying on the same screen when meeting a preset trigger condition, and the preset trigger condition includes: the method and the device have the advantages that the copy of the opposite-end message by the user is detected, so that after the binding relationship is established, the user is also supported to self-define and remove the binding relationship, and the improvement of the user operation experience is facilitated.
In an implementation scenario, in the process of predicting candidate messages, a chat topic of an opposite-end message may be identified, a plurality of candidate interaction pairs related to the chat topic are extracted from a preset interaction library, each candidate interaction pair includes a first interaction message and a second interaction message replying to the first interaction message, on the basis, based on semantic similarity between the opposite-end message and the first interaction messages in a plurality of candidate interaction pairs, at least one candidate interaction pair to which the first interaction message belongs is selected, and the second interaction message in the selected candidate interaction pair is used as a candidate message. In the mode, the chat topics are identified firstly, then the candidate interaction pairs are screened, and finally the candidate messages are selected according to the semantic similarity, so that on one hand, the prediction interference of irrelevant topics can be avoided, the screening range is narrowed, the prediction efficiency is favorably improved, on the other hand, the candidate interaction pairs are selected according to the semantic similarity, the interaction messages irrelevant to the opposite-end message semantics can be eliminated on the semantic level, and the prediction precision is favorably improved. Therefore, the efficiency and the accuracy of candidate message prediction can be favorably improved.
In a specific implementation scenario, the chat topic may be obtained by gradually identifying multiple layers of topic models from top to bottom, and the coverage of topics corresponding to topic models at upper layers is larger, and the coverage of topics corresponding to topic models at lower layers is smaller. Illustratively, the topic model corresponding to the first layer may include three topics of sports, finance and science, namely, the topic model corresponding to the first layer is used for distinguishing which one of the three topics the opposite end message belongs to, and based on this, three topic models may be provided in the second layer, wherein one topic model corresponding to the topic may include three topics of football, basketball and badminton, namely, the topic model is used for further distinguishing which one of football, basketball and badminton belongs to in case of determining to belong to the topic of "sports", another topic model corresponding to the second layer may include three topics of stocks, futures and funds, namely, the topic model is used for further distinguishing which one of stocks, futures and funds belongs to in case of determining to belong to the topic of "finance", and still another topic model corresponding to the topic model in the second layer may include information technology, The three subjects of the building engineering and the biotechnology are used for further distinguishing which one of the information technology, the building engineering and the biotechnology belongs to when the subject model belongs to the "science and technology" subject, and so on, a third layer, a fourth layer and the like can be further arranged, and the three subjects are not limited herein. It should be noted that the above example is only one possible setting manner of the multilayer theme model in the practical application process, and the specific setting manner of the multilayer theme model is not limited thereby. Specifically, through log analysis or manual sorting, one-to-one interactive corpora of the user can be collected in advance, the one-to-one interactive corpora can include a question Q and an answer a, then, the word segmentation can be performed on the data set (Q, a) and the stop word is removed, then LDA (late Dirichlet Allocation) clustering is performed, a parameter n is specified as a layer number, n different theme numbers are set, the higher the theme number is, the model can be more accurately mapped to a specific theme, and on the basis, n LDA models can be trained through the data set (Q, a) respectively, and then n layers of theme models can be obtained. It should be noted that the LDA model may be determined by Gibbs sampling, and the parameter determination process of the LDA model and the specific meaning of Gibbs sampling may refer to the technical details of the LDA model and Gibbs sampling, which are not described herein again. In the mode, the chat topics are gradually identified from top to bottom by utilizing the multilayer topic model, the coverage range of the topic model corresponding to the upper layer is larger, and the coverage range of the topic model corresponding to the lower layer is smaller, so that the accuracy of the chat topics can be improved.
In a specific implementation scenario, the preset interaction library may be pre-constructed based on a historical chat record of the user, and the preset interaction library includes interaction pairs of the user about several topics. Illustratively, pairs of interactions about "basketball", pairs of interactions about "football", etc. may be sampled from the user's historical chat records, not to mention one example at a time. In the mode, the preset interaction library is constructed in advance based on the historical chat records of the user, and comprises interaction pairs of the user about a plurality of subjects, so that the candidate messages finally predicted can meet the personalized chat requirements of the user.
In a specific implementation scenario, in order to improve the efficiency of semantic extraction, a semantic extraction model may be trained in advance, and the semantic extraction model may include, but is not limited to BERT (Bidirectional Encoder representation based on transforms), and the like, which is not limited herein. Taking the semantic extraction model as BERT as an example, a next sentence prediction task and a masking prediction task can be constructed in the training process, the text data is subjected to one-hot coding (namely one-hot) conversion in word granularity, characters in a sentence are randomly masked, and a top character (token) is reserved for learning of sentence semantic representation. After the BERT is trained and converged in the two tasks, the opposite-end message and the first interactive message can be respectively input into an embedding layer (embedding) of the BERT, after multi-layer transform processing, leading characters (token) are taken to respectively obtain a first semantic representation of the opposite-end message and a second semantic representation of each first interactive message, and on the basis, cosine similarity between the first semantic representation and each second semantic representation is calculated, so that semantic similarity between the opposite-end message and the first interactive message in a plurality of candidate interactive pairs can be obtained. It should be noted that the training process of the BERT model and the network structure of the BERT model may refer to the technical details of the BERT model, and are not described herein again.
In a specific implementation scenario, the first interaction messages may be sorted in order of semantic similarity from high to low, and a candidate interaction pair to which the first interaction message ranked in a previous preset ranking (e.g., the previous 4 ranking, the previous 5 ranking, etc.) belongs is selected, and a specific numerical value of the preset ranking may be set in a self-defined manner, which is not limited herein.
In one implementation scenario, referring to fig. 3 or 4 in combination, after the candidate message is obtained, the candidate message may be displayed on an expansion interface for selection by the user. For example, in the case that the peer message is "eaten up", the predicted candidate message may include: "eaten", "do you eat", "just eaten", "do you like eat". Other cases may be analogized, and no one example is given here.
In one implementation scenario, before the candidate message is displayed on the expansion interface, a number of reference features may also be extracted based on the candidate message, and the number of reference features may include: on the basis of the first reference feature related to the syntax structure of the candidate message, the second reference feature related to the current chat and the third reference feature related to the user, prediction is performed based on the plurality of reference features of the candidate message to obtain a probability value of the candidate message selected by the user, so that the candidate messages can be sorted firstly based on the corresponding probability value of each candidate message. Illustratively, the candidate messages may be sorted in order of high probability value. And when each candidate message is displayed on the expansion interface, each sorted candidate message can be displayed. According to the method, the syntactic characteristics of the candidate messages, the characteristics related to the current chat and the characteristics of the user are combined, the probability value of each candidate message selected by the user is predicted on the basis, different conversation requirements of different users in different chat scenes can be met, and the personalized requirements of message reply can be greatly met.
In a specific implementation scenario, the first reference feature may specifically include: at least one of the part of speech distribution characteristic, the grammatical structure characteristic and the punctuation habit characteristic of the candidate message. It should be noted that the parts of speech are distributedThe characteristics represent n different part-of-speech word distribution characteristics of places, names, verbs and the like in the candidate message, the grammatical structure characteristics represent m different grammatical structure characteristics such as a main predicate object, a definite form complement and the like presented by the candidate message, and the punctuation habit characteristics represent punctuation proportion in the candidate message. For ease of description, the part-of-speech distribution features may be denoted as
Grammatical structure features can be written as
Punctuation habit features can be recorded as
In the above manner, the first reference feature is set to include: at least one of the part-of-speech distribution characteristics, the syntactic structure characteristics and the punctuation habit characteristics of the candidate message can describe the syntactic structure of the candidate message as accurately as possible from different angles, and the accuracy of the characteristics is improved.
In a specific implementation scenario, the second reference feature may specifically include: at least one of the conversation relationship characteristic, the chat scene characteristic and the chat time characteristic of the current chat. It should be noted that the conversation relationship feature represents a conversation relationship between the user and the opposite end (e.g., family, object, child, friend, etc.), the chat scenario feature represents an application (e.g., WeChat, QQ, nailing, etc.) used by the user in the current chat, and the chat time represents a time of the user in the current chat (e.g., morning, noon, afternoon, evening, etc., or a working day, a resting day, etc., or may be a specific year, month, day, hour, minute and second, which is not limited herein). For ease of description, the dialog relationship features may be written as
Chat scenario features may be noted
Chat time characteristics can be recorded as
In the above manner, the second reference feature is set to include: at least one of the conversation relationship characteristic, the chat scene characteristic and the chat time characteristic of the chat can describe the relevant characteristics of the chat as accurately as possible from different angles, and the characteristic accuracy is improved.
In a specific implementation scenario, the third reference feature may be configured to include: at least one of an age characteristic, a gender characteristic of the user. For ease of description, the third reference feature may be noted
In the above manner, the third reference feature is set to include: at least one of the age characteristic and the gender characteristic of the user can describe the user characteristic as accurately as possible from different angles, and the characteristic accuracy is improved.
In a specific implementation scenario, the aforementioned second reference feature may be a feature vector obtained by one-hot encoding (one-hot). For example, the second reference feature may be a feature vector with a preset dimension, and the preset dimension may be set according to an actual application, for example, the preset dimension may be set to 1000 dimensions, 1200 dimensions, and the like, which is not limited herein. In the above manner, the second reference feature is set as a feature vector obtained through the one-hot encoding, which can be beneficial to reducing the complexity of feature encoding.
In a specific implementation scenario, after obtaining each reference feature of the candidate message, the reference features may be interacted to obtain a plurality of interactive features, on this basis, prediction may be further performed based on the plurality of interactive features to obtain a probability value of the candidate message, after obtaining the probability value of each candidate message, the candidate messages may be ranked according to the probability value, and the ranked candidate messages may be displayed on the expansion interface. For example, to improve the efficiency of probability value prediction, a message ranking model, such as but not limited to an NN ranking model, may be trained in advance, where the network structure of the message ranking model is not limited. On the basis, the reference characteristics of the candidate message can be input into the message sorting model, and then the probability value of the candidate message selected by the user can be obtained. Referring to fig. 5, fig. 5 is a block diagram of an embodiment of a message ordering model. As shown in fig. 5, the message ordering model may include a dimension reduction layer, an interaction layer, and a prediction layer, where in the dimension reduction layer, each first reference feature is subjected to dimension reduction by at least one Fully connected layer (FC) to obtain a corresponding first dimension reduction feature, and the second reference features are subjected to region embedding dimension reduction and transverse splicing to obtain one second dimension reduction feature, and each third reference feature is subjected to dimension reduction by at least one Fully connected layer (FC) to obtain a corresponding third dimension reduction feature; further, in the interaction layer, the first dimension reduction feature, the second dimension reduction feature and the third dimension reduction feature may be interacted to obtain a plurality of interaction features. It should be noted that the interaction layer may be provided with hidden layer nodes with preset values (e.g., 8, 10, etc.), and after interaction, interaction features with preset values may be obtained. Illustratively, for each first dimension reduction feature, interaction with other first dimension reduction features, respective second dimension reduction features, and respective third dimension reduction features may be performed to obtain an interaction feature. With continued reference to fig. 5, the prediction layer may include a multi-layer perceptron and a normalization layer, and after the interactive features are processed by the multi-layer perceptron, the hidden layer features of the multi-layer perceptron may be taken and input to the normalization layer (e.g., sigmoid, etc.) for prediction, so as to obtain a probability value of the candidate message selected by the user. In addition, in the training process, sample messages marked with sample marks can be collected in advance, the sample messages comprise first sample messages manually input by a user on an interactive interface and second sample messages (for example, sample messages manually input by other users and the like), the first sample messages are marked with first marks (for example, marked with 1), the second sample messages are marked with second marks (for example, marked with 0), a plurality of sample reference features can be extracted based on the sample messages, a plurality of sample reference features are predicted by using a message sorting model, sample probability values of the sample messages input by the user are obtained, on the basis, the sample marks and the sample probability values of the sample messages can be processed by using a loss function such as binary cross entropy and the like, loss values of the message sorting model are obtained, and an optimization mode such as gradient descent and the like is adopted, network parameters of the message ordering model are adjusted based on the loss values. According to the mode, the plurality of interactive features are obtained by interacting the plurality of reference features, the probability value is obtained by predicting based on the plurality of interactive features, the reference features can be fully combined with other reference features in the prediction process, and the prediction accuracy is favorably improved.
Step S12: and selecting the candidate message as a first reply message to reply the opposite-end message based on the selection instruction of the user.
Specifically, referring to fig. 3 or fig. 4, after each candidate message is sorted, it may be displayed on the expansion interface: the user can select one of the candidate messages to reply to the opposite-end message by directly clicking, and exemplarily, the user can directly click the first candidate message "eaten" as the first reply message to reply to the opposite-end message. Other cases may be analogized, and no one example is given here.
According to the scheme, in response to the fact that the user is detected to copy the opposite-end message on the interactive interface, the candidate messages used for replying the opposite-end message are obtained based on the prediction of the opposite-end message, and the candidate messages are displayed on the expansion interface displayed on the same screen as the interactive interface, on the basis, the candidate messages are selected as the first reply message to reply the opposite-end message based on the selection instruction of the user, so that in the message reply process, on one hand, the candidate messages can be predicted according to the opposite-end message without manual input of the user, the user can reply only by selecting the candidate messages, the communication efficiency is favorably improved, on the other hand, the candidate messages are used as content transfer through the expansion interface, the dependence on barrier-free permission is not needed, and the privacy disclosure risk, the electric quantity and the memory consumption are favorably reduced. Therefore, the communication efficiency can be improved, and meanwhile, the privacy disclosure risk, the electric quantity and the memory consumption are reduced.
Referring to fig. 6, fig. 6 is a flowchart illustrating another embodiment of a message reply method according to the present application.
Specifically, the method may include the steps of:
step S61: and responding to the detection that the user copies the opposite-end message on the interactive interface, predicting based on the opposite-end message to obtain a plurality of candidate messages for replying the opposite-end message, and displaying the candidate messages on an expansion interface displayed on the same screen as the interactive interface.
Reference may be made to the related description in the foregoing embodiments, which are not repeated herein.
Step S62: and selecting the candidate message as the first reply message based on the selection instruction of the user.
Reference may be made to the related description in the foregoing embodiments, which are not repeated herein.
Step S63: and predicting based on the first reply message to obtain a second reply message which follows the first reply message when replying the opposite-end message.
Specifically, in order to improve the efficiency of message prediction, a message prediction model may be trained in advance, and the message prediction model may include, but is not limited to, a seq2seq model, where the network structure is not limited. On the basis, the first reply message can be input into the message prediction model to obtain a second reply message. Illustratively, referring to fig. 3 or 4 in combination, in the case where the user selects "eat" as the first reply message, the first reply message "eat" may be input to the message prediction model to find "when about a meal bar" which is the second reply message "eat" following the first reply message "eat" when the reply peer message "eat no". Other cases may be analogized, and no one example is given here.
In an implementation scenario, taking a message prediction model as a seq2seq model as an example, the message prediction model may include an encoding network (encoder) and a decoding network (decoder), and specific structures and operation processes of the encoding network and the decoding network may refer to technical details of the seq2seq model, which are not described herein again.
In one implementation scenario, sets of sample messages may be pre-collected based on a user's historical chat history, and each set of sample messages includes a first sample message and a second sample message following the first sample message. Illustratively, one set of sample messages may include a first sample message "i am on subway work today" and a second sample message "go home along the way to work" following the first sample message. On the basis, for each group of sample messages, the first sample message can be input into a message prediction model for message prediction, in the process of past prediction, prediction probability values of a plurality of preset characters can be obtained in each prediction, on the basis, the loss value of the message prediction model can be calculated based on the corresponding prediction probability value of the ith sample character in the second sample message in the ith prediction (wherein i is 1 to N, and N is the number of the sample characters in the second sample message), and the network parameters of the message prediction model are adjusted based on the loss value by adopting an optimization mode such as gradient descent. The specific calculation method of the loss may refer to technical details of a loss function such as cross entropy, and the specific adjustment process of the parameter may refer to technical details of an optimization method such as gradient descent, which are not described herein again.
In one implementation scenario, the first reply message and the second reply message may be sent in sequence to reply to the peer message. Referring to fig. 3 or fig. 4 in combination, after receiving the peer message "not eaten" the user may copy the peer message "not eaten", and after predicting and sorting, may obtain each candidate message shown in the extended interface, after the user clicks the candidate message "eaten", may take the candidate message "eaten" as the first reply message and send it to the peer, and predict, based on the first reply message "eaten", when about a bar "the second reply message" that follows the first reply message "eaten", and send the second reply message to the peer.
In one implementation scenario, the first reply message and the second reply message may also be sent simultaneously as a whole to reply to the peer message. Referring to fig. 3 or 4 in combination, after receiving the peer message "eat nothing", the user may obtain each candidate message shown in the extended interface by copying the peer message "eat nothing", predicting and sorting, after the user clicks the candidate message "eat," the candidate message "eat" may be taken as the first reply message, and predicting based on the first reply message "eat," to obtain when about each meal bar "of the second reply message" following the first reply message "eat," and to simultaneously transmit the first reply message "eat" and when about each meal bar "of the second reply message as a whole to the peer.
In an implementation scenario, please refer to fig. 7 in combination, and fig. 7 is a process diagram of an embodiment of a message reply method according to the present application. As shown in fig. 7, in response to the user starting the intelligent reply function of the input method, a binding relationship is established between the input method and the expansion interface, whether the user copies the opposite-end message is detected, based on which, when it is detected that the user copies the opposite-end message on the interactive interface, prediction is performed based on the opposite-end message, a plurality of candidate messages for replying the opposite-end message are obtained, and the candidate messages are displayed on the expansion interface displayed on the same screen as the interactive interface. In the process, a pre-constructed intelligent dialogue corpus model can be used for carrying out relevant prediction, the intelligent dialogue corpus model can be trained to obtain a topic identification model based on a question and answer corpus extracted in advance, so that a chat topic of an opposite-end message is identified and obtained by using the topic identification model, on the basis, a plurality of candidate interaction pairs relevant to the chat topic are extracted from a preset interaction library, each candidate interaction pair comprises a first interaction message and a second interaction message replying the first interaction message, on the basis of semantic similarity between the opposite-end message and the first interaction message in the candidate interaction pairs, at least one candidate interaction pair to which the first interaction message belongs is selected, and therefore the second interaction message in the selected candidate interaction pair is used as the candidate message. In addition, the intelligent dialogue corpus model further comprises a message ordering model, so that the candidate messages can be ordered by the message ordering model in combination with the reference features extracted from the candidate messages, and the ordered candidate messages are displayed on the expansion interface. Then, based on the selection instruction of the user, the candidate message may be selected as the first reply message, prediction is performed based on the first reply message, and a second reply message following the first reply message when the opposite-end message is replied is obtained, so that the first reply message and the second reply message may be sequentially sent to reply the opposite-end message, or the first reply message and the second reply message may also be simultaneously sent as a whole to reply the opposite-end message. And then, completing a round of message reply, further continuously detecting whether the user copies the opposite-end message, and starting a new round of message reply under the condition that the user copies the opposite-end message again.
According to the scheme, after the first reply message is obtained based on the user selection candidate message, prediction is further carried out based on the first reply message, and the second reply message which follows the first reply message when the opposite-end message is replied is obtained, so that the association reply can be further realized on the basis of intelligent reply, and the user experience can be further improved.
Referring to fig. 8, fig. 8 is a block diagram illustrating an embodiment of a message reply apparatus 80 according to the present application. The message reply device 80 includes: the message prediction module 81 is used for responding to the detection that the user copies the opposite-end message on the interactive interface, performing prediction based on the opposite-end message to obtain a plurality of candidate messages for replying the opposite-end message, and displaying the candidate messages on an expansion interface displayed on the same screen as the interactive interface; and a message selection module 82, configured to select the candidate message as the first reply message to reply to the peer message based on the selection instruction of the user.
According to the scheme, on one hand, the candidate messages can be obtained according to the opposite-end message prediction without manual input of the user, the user can reply only by selecting the candidate messages, the communication efficiency is favorably improved, on the other hand, the candidate messages are used as content transfer through the expansion interface, the barrier-free permission is not required, and the privacy disclosure risk and the electric quantity and memory consumption are favorably reduced. Therefore, the communication efficiency can be improved, and meanwhile, the privacy disclosure risk, the electric quantity and the memory consumption are reduced.
In some disclosed embodiments, the message replying device 80 includes a relationship binding module, configured to respond to that the user starts an intelligent replying function of the input method, establish a binding relationship between the input method and the extended interface, and detect whether the user copies the opposite-end message; and under the condition that the input method and the expansion interface have a binding relationship, the expansion interface and the input interface of the input method are kept on the same screen for display.
Therefore, when the user starts the intelligent reply function of the input method, the binding relationship of the input French extension interface is established by default, the subsequent operation of detecting whether the user copies the opposite-end message is executed, and when the input method and the extension interface have the binding relationship, the extension interface and the input interface of the input method are kept on the same screen for display, so that the user does not need to call the extension interface repeatedly when copying the opposite-end message every time, and the memory consumption is reduced.
In some disclosed embodiments, the message replying device 80 includes a relationship unbinding module for unbinding the binding relationship between the input method and the extended interface in response to a unbinding instruction of the binding relationship by the user; under the condition that the input method and the expansion interface are not in a binding relationship, when the expansion interface meets a preset trigger condition, the expansion interface and the input interface are kept displaying on the same screen, and the preset trigger condition comprises the following steps: it is detected that the user duplicates the peer message.
Therefore, in response to a user's instruction for removing the binding relationship, the binding relationship between the input method and the extended interface is removed, and in the case that the input method and the extended interface do not have the binding relationship, the extended interface and the input interface are kept on the same screen for display when meeting a preset trigger condition, and the preset trigger condition includes: the method and the device have the advantages that the copy of the opposite-end message by the user is detected, so that after the binding relationship is established, the user is also supported to self-define and remove the binding relationship, and the improvement of the user operation experience is facilitated.
In some disclosed embodiments, the message prediction module 81 includes a topic selection sub-module, configured to identify a chat topic of an opposite-end message, and extract a plurality of candidate interaction pairs related to the chat topic from a preset interaction library, where each candidate interaction pair includes a first interaction message and a second interaction message that replies to the first interaction message; the message prediction module 81 includes an interaction selection sub-module, configured to select, based on semantic similarities between the peer message and first interaction messages in the candidate interaction pairs, at least one candidate interaction pair to which the first interaction message belongs; the message prediction module 81 includes a message selection sub-module for selecting the second interaction message of the selected candidate interaction pair as the candidate message.
Therefore, by identifying the chat theme, screening the candidate interaction pairs and selecting the candidate messages according to the semantic similarity, on one hand, the prediction interference of irrelevant themes can be avoided, the screening range is narrowed, the prediction efficiency is favorably improved, on the other hand, the candidate interaction pairs are selected according to the semantic similarity, the interaction messages irrelevant to the opposite-end message semantics can be eliminated on the semantic level, and the prediction precision is favorably improved. Therefore, the efficiency and the accuracy of candidate message prediction can be favorably improved.
In some disclosed embodiments, the preset interaction library is pre-constructed based on historical chat records of the user, and the preset interaction library comprises interaction pairs of the user about a plurality of topics; and/or the chat topic is obtained by gradually identifying from top to bottom by using a multi-layer topic model, and the more the upper layer topic model is covered by the topic, the less the lower layer topic model is covered by the topic.
Therefore, the preset interaction library is constructed in advance based on the historical chat records of the user, and comprises interaction pairs of the user about a plurality of subjects, so that the candidate messages finally predicted can meet the personalized chat requirements of the user; the chat topics are gradually identified from top to bottom by utilizing the multilayer topic models, the coverage range of the topic model corresponding to the topic on the upper layer is larger, and the coverage range of the topic model corresponding to the topic on the lower layer is smaller, so that the accuracy of the chat topics can be favorably improved.
In some disclosed embodiments, the message replying apparatus 80 further comprises a feature extraction module for extracting a number of reference features based on the candidate message; wherein the plurality of reference features comprises: the method comprises the steps of obtaining a first reference characteristic related to a syntax structure of a candidate message, a second reference characteristic related to the current chat and a third reference characteristic related to a user; the message replying device 80 further comprises a probability prediction module for predicting based on a plurality of reference features of the candidate message to obtain a probability value of the candidate message selected by the user; the message replying device 80 further includes a message sorting module for sorting the candidate messages based on the probability values corresponding to the candidate messages.
Therefore, by combining the syntactic characteristics of the candidate messages, the characteristics related to the current chat and the characteristics of the user, the probability value of each candidate message selected by the user is predicted on the basis, different conversation requirements of different users in different chat scenes can be met, and the individual requirements of message reply can be greatly met.
In some disclosed embodiments, the probability prediction module includes an interaction submodule configured to interact the plurality of reference features to obtain a plurality of interaction features; the probability prediction module comprises a prediction submodule used for predicting based on a plurality of interactive features to obtain a probability value.
Therefore, the interaction of the reference features is performed to obtain the interaction features, the probability value is obtained by predicting based on the interaction features, the reference features can be fully combined with other reference features in the prediction process, and the prediction accuracy is improved.
In some disclosed embodiments, the first reference feature comprises: at least one of the part of speech distribution characteristics, the grammatical structure characteristics and the punctuation habit characteristics of the candidate message; and/or, the second reference feature comprises: at least one of the conversation relationship characteristic, the chat scene characteristic and the chat time characteristic of the current chat; and/or, the third reference feature comprises: at least one of an age characteristic, a gender characteristic of the user; and/or the second reference feature is a feature vector obtained by independent hot coding; and/or displaying the sorted candidate messages on the expansion interface.
Accordingly, the first reference feature is arranged to comprise: at least one of the part-of-speech distribution characteristics, the grammatical structure characteristics and the punctuation habit characteristics of the candidate message can describe the syntactic structure of the candidate message as accurately as possible from different angles; and the second reference feature is arranged to include: at least one of the conversation relationship characteristic, the chat scene characteristic and the chat time characteristic of the chat can describe the relevant characteristics of the chat as accurately as possible from different angles; and the third reference feature is arranged to include: at least one of the age characteristic and the gender characteristic of the user can describe the user characteristic as accurately as possible from different angles, so that the characteristic accuracy can be improved. In addition, the second reference feature is set as a feature vector obtained through the one-hot encoding, so that the complexity of feature encoding can be favorably reduced. In addition, the sorted candidate messages are displayed on the expansion interface, so that the user can preferentially see the candidate messages which are selected with high probability, and the user experience is favorably improved.
In some disclosed embodiments, the message replying device 80 further includes an association reply module, configured to perform prediction based on the first reply message, and obtain a second reply message following the first reply message when replying to the peer message.
Therefore, after the first reply message is obtained based on the user selection candidate message, the prediction is further carried out based on the first reply message, and the second reply message which follows the first reply message when the opposite-end message is replied is obtained, so that the association reply can be further realized on the basis of the intelligent reply, and the user experience can be further promoted.
In some disclosed embodiments, the first reply message and the second reply message are sent in sequence to reply to the opposite-end message; or the first reply message and the second reply message are sent simultaneously as a whole to reply to the opposite-end message.
Therefore, the first reply message and the second reply message are sequentially sent to reply the opposite-end message, so that the association reply can be simultaneously carried out on the premise of improving the instantaneity of the message reply as much as possible, the user experience can be improved, the first reply message and the second reply message can be simultaneously sent as a whole to reply the opposite-end message, the message reply can be carried out together based on the intelligent reply and the association reply, and the communication efficiency can be improved.
Referring to fig. 9, fig. 9 is a schematic block diagram of an embodiment of an electronic device 90 according to the present application. The electronic device 90 comprises a screen 91, a communication circuit 92, a memory 93 and a processor 94, wherein the screen 91, the communication circuit 92 and the memory 93 are coupled to the processor 94, the memory 93 stores program instructions, and the processor 94 is configured to execute the program instructions to implement the steps in any of the above-mentioned message reply method embodiments. The electronic device 90 may include, but is not limited to: a mobile phone, a tablet computer, etc., without limitation. The screen 91 may display an interactive interface, an input interface, and an extended interface in the foregoing disclosed embodiments, so that the user may perform message interaction with the opposite terminal, and the communication circuit 92 may support communication protocols such as mobile communication (e.g., 4G, 5G, etc.), which is not limited herein.
In particular, the processor 94 is configured to control itself, the screen 91, the communication circuit 92, and the memory 93 to implement the steps in any of the above-described message reply method embodiments. The processor 94 may also be referred to as a CPU (Central Processing Unit). The processor 94 may be an integrated circuit chip having signal processing capabilities. The Processor 94 may also be a general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. In addition, the processor 94 may be collectively implemented by an integrated circuit chip.
According to the scheme, on one hand, the candidate messages can be obtained according to the opposite-end message prediction without manual input of the user, the user can reply only by selecting the candidate messages, the communication efficiency is favorably improved, on the other hand, the candidate messages are used as content transfer through the expansion interface, the barrier-free permission is not required, and the privacy disclosure risk and the electric quantity and memory consumption are favorably reduced. Therefore, the communication efficiency can be improved, and meanwhile, the privacy disclosure risk, the electric quantity and the memory consumption are reduced.
Referring to fig. 10, fig. 10 is a block diagram illustrating an embodiment of a computer-readable storage medium 100 according to the present application. The computer readable storage medium 100 stores program instructions 101 capable of being executed by a processor, the program instructions 101 being configured to implement the steps in any of the above-described message reply method embodiments.
According to the scheme, on one hand, the candidate messages can be obtained according to the opposite-end message prediction without manual input of the user, the user can reply only by selecting the candidate messages, the communication efficiency is favorably improved, on the other hand, the candidate messages are used as content transfer through the expansion interface, the barrier-free permission is not required, and the privacy disclosure risk and the electric quantity and memory consumption are favorably reduced. Therefore, the communication efficiency can be improved, and meanwhile, the privacy disclosure risk, the electric quantity and the memory consumption are reduced.
In some embodiments, functions of or modules included in the apparatus provided in the embodiments of the present disclosure may be used to execute the method described in the above method embodiments, and specific implementation thereof may refer to the description of the above method embodiments, and for brevity, will not be described again here.
The foregoing description of the various embodiments is intended to highlight various differences between the embodiments, and the same or similar parts may be referred to each other, and for brevity, will not be described again herein.
In the several embodiments provided in the present application, it should be understood that the disclosed method and apparatus may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, a division of a module or a unit is merely a logical division, and an actual implementation may have another division, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection of devices or units through some interfaces, and may be in an electrical, mechanical or other form.
Units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be substantially implemented or contributed to by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, a network device, or the like) or a processor (processor) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
Claims (13)
1. A method for replying to a message, comprising:
responding to the fact that a user copies an opposite-end message on an interactive interface, predicting based on the opposite-end message to obtain a plurality of candidate messages for replying the opposite-end message, and displaying the candidate messages on an expansion interface displayed on the same screen with the interactive interface;
and selecting the candidate message as a first reply message to reply the opposite-end message based on the selection instruction of the user.
2. The method of claim 1, wherein before the predicting based on the peer message in response to detecting that the user duplicates the peer message on the interactive interface, obtaining a number of candidate messages for replying to the peer message, the method further comprises:
responding to the intelligent reply function of the input method started by the user, establishing a binding relationship between the input method and the expansion interface, and detecting whether the user copies the opposite-end message;
and under the condition that the input method and the expansion interface have the binding relationship, the expansion interface and the input interface of the input method are kept on the same screen for display.
3. The method of claim 2, wherein after the establishing the binding relationship between the input method and the extended interface, the method further comprises:
responding to a releasing instruction of the user on the binding relationship, and releasing the binding relationship between the input method and the expansion interface;
under the condition that the binding relationship does not exist between the input method and the expansion interface, when the expansion interface meets a preset trigger condition, the expansion interface and the input interface are kept displaying on the same screen, and the preset trigger condition comprises: and detecting that the user copies the opposite-end message.
4. The method of claim 1, wherein the predicting based on the peer message to obtain candidate messages for replying to the peer message comprises:
identifying a chat topic of the opposite-end message, and extracting a plurality of candidate interaction pairs related to the chat topic from a preset interaction library, wherein each candidate interaction pair comprises a first interaction message and a second interaction message replying to the first interaction message;
selecting at least one candidate interaction pair to which the first interaction message belongs based on semantic similarity between the opposite-end message and the first interaction messages in the candidate interaction pairs respectively;
and using the second interaction message in the selected candidate interaction pair as the candidate message.
5. The method of claim 4, wherein the preset interaction library is pre-constructed based on historical chat records of the user, and the preset interaction library comprises interaction pairs of the user about a plurality of topics;
and/or the chat topic is obtained by gradually identifying from top to bottom by using a multi-layer topic model, and the more upper layers of topic models are larger in the coverage range of the corresponding topic, the more lower layers of topic models are smaller in the coverage range of the corresponding topic.
6. The method of claim 1, wherein after predicting based on the peer message, obtaining a plurality of candidate messages for replying to the peer message, and before displaying the plurality of candidate messages on the expansion interface displayed on the same screen as the interactive interface, the method further comprises:
extracting a plurality of reference features based on the candidate message; wherein the number of reference features includes: a first reference characteristic related to the syntactic structure of the candidate message, a second reference characteristic related to the current chat and a third reference characteristic related to the user;
predicting based on a plurality of reference features of the candidate message to obtain a probability value of the candidate message selected by the user;
and sorting the candidate messages based on the probability values corresponding to the candidate messages.
7. The method of claim 6, wherein the predicting based on the reference features of the candidate message to obtain a probability value of the candidate message selected by the user comprises:
interacting the plurality of reference features to obtain a plurality of interactive features;
and predicting based on the interactive features to obtain the probability value.
8. The method of claim 6, wherein the first reference feature comprises: at least one of part-of-speech distribution characteristics, grammatical structure characteristics and punctuation habit characteristics of the candidate message;
and/or, the second reference feature comprises: at least one of the conversation relationship characteristic, the chat scene characteristic and the chat time characteristic of the current chat;
and/or, the third reference feature comprises: at least one of an age characteristic, a gender characteristic of the user;
and/or the second reference feature is a feature vector obtained by one-hot coding;
and/or displaying each candidate message after the sequence on the expansion interface.
9. The method of claim 1, wherein after the selecting the candidate message as the first reply message based on the user's selection instruction, the method further comprises:
and predicting based on the first reply message to obtain a second reply message which follows the first reply message when replying the opposite-end message.
10. The method according to claim 9, wherein the first reply message and the second reply message are sent in sequence to reply to the peer message;
or, the first reply message and the second reply message are sent simultaneously as a whole to reply to the opposite-end message.
11. A message reply apparatus, comprising:
the message prediction module is used for responding to the fact that a user copies the opposite-end message on an interactive interface, predicting based on the opposite-end message to obtain a plurality of candidate messages for replying the opposite-end message, and displaying the candidate messages on an expansion interface displayed on the same screen as the interactive interface;
and the message selection module is used for selecting the candidate message as a first reply message based on the selection instruction of the user so as to reply the opposite-end message.
12. An electronic device comprising a screen, communication circuitry, a memory and a processor, the screen, the communication circuitry and the memory being coupled to the processor, the memory having stored therein program instructions for execution by the processor to implement the message reply method of any of claims 1 to 10.
13. A computer-readable storage medium, characterized in that program instructions executable by a processor for implementing the message reply method according to any one of claims 1 to 10 are stored.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111478629.3A CN114356173A (en) | 2021-12-06 | 2021-12-06 | Message reply method and related device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111478629.3A CN114356173A (en) | 2021-12-06 | 2021-12-06 | Message reply method and related device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114356173A true CN114356173A (en) | 2022-04-15 |
Family
ID=81096745
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111478629.3A Pending CN114356173A (en) | 2021-12-06 | 2021-12-06 | Message reply method and related device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114356173A (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106325682A (en) * | 2016-09-13 | 2017-01-11 | 宇龙计算机通信科技(深圳)有限公司 | Method and system for replying messages |
| CN110311856A (en) * | 2019-06-28 | 2019-10-08 | 上海连尚网络科技有限公司 | Instant communicating method, equipment and computer readable storage medium |
| CN110929122A (en) * | 2018-09-20 | 2020-03-27 | 北京搜狗科技发展有限公司 | Data processing method and device and data processing device |
| CN111382242A (en) * | 2018-12-28 | 2020-07-07 | 北京搜狗科技发展有限公司 | Information providing method, device and readable medium |
| CN112003778A (en) * | 2020-07-17 | 2020-11-27 | 北京百度网讯科技有限公司 | Message processing method, device, equipment and computer storage medium |
| CN112631435A (en) * | 2019-09-24 | 2021-04-09 | 北京搜狗科技发展有限公司 | Input method, device, equipment and storage medium |
| CN113039758A (en) * | 2018-12-19 | 2021-06-25 | 深圳市欢太科技有限公司 | Method and related device for automatically replying information |
| CN113037932A (en) * | 2021-02-26 | 2021-06-25 | 北京百度网讯科技有限公司 | Reply message generation method and device, electronic equipment and storage medium |
-
2021
- 2021-12-06 CN CN202111478629.3A patent/CN114356173A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106325682A (en) * | 2016-09-13 | 2017-01-11 | 宇龙计算机通信科技(深圳)有限公司 | Method and system for replying messages |
| CN110929122A (en) * | 2018-09-20 | 2020-03-27 | 北京搜狗科技发展有限公司 | Data processing method and device and data processing device |
| CN113039758A (en) * | 2018-12-19 | 2021-06-25 | 深圳市欢太科技有限公司 | Method and related device for automatically replying information |
| CN111382242A (en) * | 2018-12-28 | 2020-07-07 | 北京搜狗科技发展有限公司 | Information providing method, device and readable medium |
| CN110311856A (en) * | 2019-06-28 | 2019-10-08 | 上海连尚网络科技有限公司 | Instant communicating method, equipment and computer readable storage medium |
| CN112631435A (en) * | 2019-09-24 | 2021-04-09 | 北京搜狗科技发展有限公司 | Input method, device, equipment and storage medium |
| CN112003778A (en) * | 2020-07-17 | 2020-11-27 | 北京百度网讯科技有限公司 | Message processing method, device, equipment and computer storage medium |
| CN113037932A (en) * | 2021-02-26 | 2021-06-25 | 北京百度网讯科技有限公司 | Reply message generation method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11334635B2 (en) | Domain specific natural language understanding of customer intent in self-help | |
| CN110178132B (en) | Method and system for automatically suggesting content in a messaging application | |
| US10380249B2 (en) | Predicting future trending topics | |
| CN110166802B (en) | Bullet screen processing method and device and storage medium | |
| WO2022052817A1 (en) | Search processing method and apparatus, and terminal and storage medium | |
| US10021061B1 (en) | Message presentation management in a social networking environment | |
| Ngejane et al. | Digital forensics supported by machine learning for the detection of online sexual predatory chats | |
| CN113127746B (en) | Information pushing method based on user chat content analysis and related equipment thereof | |
| CN112015928A (en) | Information extraction method and device of multimedia resource, electronic equipment and storage medium | |
| CN108304374A (en) | Information processing method and related product | |
| CN113392205A (en) | User portrait construction method, device and equipment and storage medium | |
| CN109558531A (en) | News information method for pushing, device and computer equipment | |
| CN113573128A (en) | Audio processing method, device, terminal and storage medium | |
| WO2024021685A1 (en) | Reply content processing method and media content interactive content interaction method | |
| CN116775815B (en) | Dialogue data processing method and device, electronic equipment and storage medium | |
| CN115168568B (en) | Data content identification method, device and storage medium | |
| CN114356173A (en) | Message reply method and related device, electronic equipment and storage medium | |
| CN114218364A (en) | Question-answer knowledge base expansion method and device | |
| CN116016421A (en) | Method, computing device readable storage medium, and computing device for facilitating media-based content sharing performed in a computing device | |
| CN113901337A (en) | Application recommendation method, device, equipment and computer-readable storage medium | |
| CN114363664A (en) | Method and device for generating video collection title | |
| CN109947947B (en) | Text classification method and device and computer readable storage medium | |
| CN114547435A (en) | Content quality identification method, device, equipment and readable storage medium | |
| CN116704423B (en) | Hierarchical video character social interaction identification method, system, equipment and medium | |
| CN115481619A (en) | Interactive point pair extraction method, related equipment and storage medium |
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 |