CN109960811B - Data processing method and device and electronic equipment - Google Patents
Data processing method and device and electronic equipment Download PDFInfo
- Publication number
- CN109960811B CN109960811B CN201910252940.2A CN201910252940A CN109960811B CN 109960811 B CN109960811 B CN 109960811B CN 201910252940 A CN201910252940 A CN 201910252940A CN 109960811 B CN109960811 B CN 109960811B
- Authority
- CN
- China
- Prior art keywords
- condition
- target
- target object
- question
- judging
- 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.)
- Active
Links
- 238000003672 processing method Methods 0.000 title claims abstract description 10
- 238000000034 method Methods 0.000 claims abstract description 30
- 238000012545 processing Methods 0.000 claims description 28
- 230000003993 interaction Effects 0.000 claims description 8
- 238000013139 quantization Methods 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 description 10
- 238000011156 evaluation Methods 0.000 description 9
- 230000007246 mechanism Effects 0.000 description 9
- 238000004422 calculation algorithm Methods 0.000 description 8
- 238000004891 communication Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 8
- 239000013598 vector Substances 0.000 description 8
- 230000006870 function Effects 0.000 description 7
- 238000003062 neural network model Methods 0.000 description 4
- 238000013507 mapping Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 241000590419 Polygonia interrogationis Species 0.000 description 2
- 210000004027 cell Anatomy 0.000 description 2
- 238000012512 characterization method Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 102100034761 Cilia- and flagella-associated protein 418 Human genes 0.000 description 1
- 101100439214 Homo sapiens CFAP418 gene Proteins 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000013210 evaluation model Methods 0.000 description 1
- 238000007477 logistic regression Methods 0.000 description 1
- 230000005055 memory storage Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application discloses a data processing method, a device and electronic equipment, wherein the method comprises the following steps: acquiring a target problem statement; determining a target object based on the target problem statement; judging whether the target object meets a first condition or not to obtain a judging result; based on the judging result, outputting feedback information; the feedback information comprises judgment information of the first condition, or the feedback information comprises answer information of the target question statement and judgment information of the first condition. Therefore, after receiving the question sentence, the method does not directly answer the question, but judges the question, for example, judges whether the corresponding target object meets the first condition, so that after the validity of the question is judged, whether to answer the question is selected, the efficiency of answering the question is improved, and the user experience is obviously improved.
Description
Technical Field
The present application relates to the field of intelligent customer service technologies, and in particular, to a data processing method, a data processing device, and an electronic device.
Background
In the intelligent customer service system, a user inputs a problem to be solved through the customer service system, the intelligent customer service solves the problem according to the problem of the user, and invalid problems can be set up because the technical level of the user or the knowledge of the product use are uneven, so that the customer service system can perform multiple rounds of invalid conversations, the efficiency of solving the problem of the user is reduced, and the use experience of the user is seriously influenced.
Therefore, a technical solution capable of determining the validity of the user problem is needed.
Disclosure of Invention
In view of the above, the present application provides a data processing method, including:
acquiring a target problem statement;
determining a target object based on the target problem statement;
Judging whether the target object meets a first condition or not to obtain a judging result;
based on the judging result, outputting feedback information;
The feedback information comprises judgment information of the first condition, or the feedback information comprises answer information of the target question statement and judgment information of the first condition.
In the above method, preferably, based on the determination result, outputting feedback information includes:
and outputting answer information of the target question sentence, and then outputting judgment information of the first condition.
In the above method, preferably, the first condition is generated based on the target object and the target question sentence.
In the above method, preferably, the determining whether the target object meets the first condition to obtain a determination result includes:
Acquiring a knowledge graph of the target object based on the object attribute of the target object;
and judging whether the knowledge graph of the target object meets a first condition or not to obtain a judging result.
The above method, preferably, further comprises:
obtaining the question type of the target question statement;
Judging whether the knowledge graph of the target object meets a first condition to obtain a judging result, wherein the judging step comprises the following steps:
If the problem type of the target problem statement belongs to a numerical value type, judging whether the knowledge graph of the target object meets a first condition or not through a first mode to obtain a judging result;
if the problem type of the target problem statement belongs to a non-numerical value type, judging whether the knowledge graph of the target object meets a first condition or not through a second mode to obtain a judging result.
In the above method, preferably, determining whether the knowledge graph of the target object meets the first condition in the first mode, to obtain a determination result includes:
generating an attribute judgment rule of the target object based on the knowledge graph of the target object;
and judging whether the attribute judging rule meets a first condition or not to obtain a judging result.
In the above method, preferably, determining whether the knowledge graph of the target object meets the first condition in the second mode, to obtain a determination result includes:
carrying out joint modeling on the knowledge graph of the target object and the first condition to obtain model data;
Carrying out quantization processing on the model data to obtain a scalar value;
And judging whether the scalar value is larger than or equal to a scalar threshold value or not so as to obtain a judging result.
In the above method, preferably, based on the determination result, outputting feedback information includes:
If the judging result represents that the target object meets the first condition, outputting answer information of the target question statement;
And if the judging result represents that the target object does not meet the first condition, outputting judging information of the first condition and answer information of the target question statement.
The application also provides a data processing device, which comprises:
a receiving unit for obtaining a target question sentence;
a determining unit, configured to determine a target object based on the target question sentence;
The judging unit is used for judging whether the target object meets a first condition or not so as to obtain a judging result;
The output unit is used for outputting feedback information based on the judging result;
Wherein the feedback information includes judgment information of the first condition, or the feedback information includes answer information of the target question sentence and judgment information of the first condition
The application also provides an electronic device, comprising:
the interaction device is used for acquiring target problem sentences;
Processing equipment, based on the target problem statement, determining a target object; judging whether the target object meets a first condition or not to obtain a judging result, and outputting feedback information through the interaction equipment based on the judging result;
The feedback information comprises judgment information of the first condition, or the feedback information comprises answer information of the target question statement and judgment information of the first condition.
According to the technical scheme, in the data processing method, the data processing device and the electronic equipment, after the target question statement is acquired, the corresponding target object is determined first, and then the corresponding feedback information is output based on the judgment result whether the target object corresponding to the target question statement meets the first condition or not, wherein the feedback information comprises the judgment information of the first condition or further comprises the answer information of the target question statement. Therefore, after receiving the question sentence, the method does not directly answer the question, but judges the question, for example, judges whether the corresponding target object meets the first condition, so that after the validity of the question is judged, whether to answer the question is selected, the efficiency of answering the question is improved, and the user experience is obviously improved.
Drawings
In order to more clearly illustrate the embodiments of the application or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart of a data processing method according to a first embodiment of the present application;
FIG. 2 is a diagram illustrating an application example of an embodiment of the present application;
FIG. 3 is a partial flow chart of a first embodiment of the present application;
FIG. 4 is a schematic diagram of a data processing apparatus according to a second embodiment of the present application;
Fig. 5 is a schematic structural diagram of an electronic device according to a third embodiment of the present application;
fig. 6 is a functional architecture diagram of an electronic device according to a third embodiment of the present application;
fig. 7 and 8 are diagrams illustrating other applications of the embodiment of the present application, respectively.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
As shown in fig. 1, a flowchart of a data processing method according to a first embodiment of the present application is applicable to an electronic device, such as a notebook or a server, which can provide question-answering services for an object or a product, such as an intelligent customer service application installed in the electronic device, and provide question-answering and technical communication for a user to use the product.
Specifically, the present embodiment may include the following steps:
step 101: obtaining a target problem statement.
The target question sentence may be a question sentence sent by a communication counterpart received by the input/output device on the electronic device. As shown in fig. 2, a user establishes a customer service connection with the customer service system of the server in this embodiment through his mobile phone, for example, the user opens a dialog box of the customer service system on his mobile phone, and the server customer service system responds to the opened dialog box to establish a communication connection.
Specifically, the target problem statement may be a problem statement for a certain object, such as a product, for example, a problem of how to use a product to swipe a card or how to detach a product shell is input in a dialog box of a mobile phone of the user, and the problem statement is transmitted through a connection between the mobile phone of the user and a server in the embodiment, and accordingly, the problem statement is obtained in the embodiment.
It should be noted that, data transmission can be performed between the user mobile phone and the server in this embodiment through wifi or mobile communication, or through wired network connection.
In this embodiment, after the target problem sentence is obtained, the target problem sentence may be preprocessed, for example, words such as calling, small talk or boring are filtered out, sentences with a length smaller than a certain threshold are filtered out, or multiple short sentences are combined into one long sentence, and so on.
Step 102: based on the target question statement, a target object is determined.
The target problem statement is a problem statement for a certain object, such as a product or a user, and in this embodiment, the target object corresponding to the target problem statement may be determined by performing character recognition or semantic recognition on the target problem statement. For example, the target question statement is "how does this phone swipe the card? In the embodiment, the character recognition algorithm can be utilized to recognize the target problem statement, the character of the mobile phone is analyzed, and the mobile phone with the A model of the target object corresponding to the mobile phone can be determined by combining the historical communication information, the preset information or the configuration information of the user to which the target problem statement belongs. Or for example, the target question sentence is "who is the actor", in this embodiment, the character recognition algorithm may be used to recognize the target question sentence, analyze the character of "the actor", and combine with the history communication information, the preset information, or the configuration information of the user to whom the target question sentence belongs, so as to determine that the target object corresponding to "the actor" is the person wearing blue clothes.
Step 103: and judging whether the target object meets the first condition or not to obtain a judging result.
The first condition may be a characteristic condition of the target object in the characterization target problem statement, and accordingly, in this embodiment, whether the target object meets the first condition is determined, so that whether the description of the target object in the target problem statement matches with the target object is determined, and whether the target problem statement is a valid problem statement is determined.
Step 104: and outputting feedback information based on the judgment result.
The feedback information may include: judging information of the first condition, namely whether the characteristics of the target object in the target problem statement are matched with the target object or not, namely whether the target problem statement is an effective problem statement or not can be characterized;
Or the feedback information may include: the answer information for the target question sentence and the judgment information for the first condition, that is, in this embodiment, not only the judgment information for the first condition may be output, but also the answer information for the target question sentence may be output regardless of whether the target question sentence is a valid question sentence in the judgment information for the first condition.
For example, if the judgment result indicates that the target object meets the first condition, that is, the target question sentence is a valid question sentence, then the target question sentence may be directly replied, that is, answer information of the target question sentence is output, and further, judgment information of the first condition may also be output;
If the judging result shows that the target object does not meet the first condition, namely the target problem statement is not an effective problem statement, judging information on the first condition can be directly output to prompt that the problem proposed by the affiliated user of the target problem statement does not accord with the corresponding target object, such as a product or a person;
Or if the judgment result shows that the target object does not meet the first condition, outputting answer information of the target question sentence, and outputting judgment information of the first condition, namely, even if the target question sentence is replied, reminding the user of the target question sentence that the question posed by the user does not meet the corresponding target object, such as a product or a person.
It can be seen from the foregoing technical solution that, in the data processing method provided in the first embodiment of the present application, after a target question sentence is obtained, a corresponding target object is first determined, and then corresponding feedback information is output based on a determination result of whether the target object corresponding to the target question sentence meets a first condition, where the feedback information includes determination information for the first condition or further includes answer information for the target question sentence. Therefore, in this embodiment, after receiving the question sentence, the answer is not directly answered, but is judged first, if the corresponding target object meets the first condition, so that after the validity of the question is judged, whether to answer the question is selected, and further the efficiency of answering the question is improved, and the user experience is obviously improved.
In one implementation manner, if the result of the judgment indicates that the target object does not meet the first condition, in this embodiment, answer information of the target question sentence may be output first, after the answer information is output, the judgment information of the first condition may be output, that is, in this embodiment, the answer to the target question sentence may be given priority, and even if the answer to the target question sentence is given priority, the judgment information of the first condition may be output immediately, so as to prompt that the question posed by the user of the target question sentence does not meet the corresponding target object.
In one implementation, the first condition may be generated based on the target object and the target question statement. That is, the first condition is generated based on the description information or the feature of the target object in the target question sentence. For example, "how does this handset swipe the card? Correspondingly, the target object is a mobile phone of model A, the target problem statement comprises words of mobile phone, card swiping and what, and correspondingly, the description characteristics of the mobile phone in the target problem statement can be determined by using algorithms such as word sense recognition and the like, namely: "A type cell phone can swipe card". Thus, the first condition generated based on the target question sentence and the target object in this embodiment is: the mobile phone of the A model can swipe the card;
For another example, the target question sentence is "who is the actor", in this embodiment, the character recognition algorithm may be used to identify the target question sentence, and analyze that the target object is a person wearing blue clothes, where the target question sentence includes "who", "actor" and "who", and accordingly, by using the word sense recognition algorithm, it may be determined that the description feature of "person wearing blue clothes" in the target question sentence is: "blue clothing person is an actor".
Accordingly, in this embodiment, whether the target object meets the first condition may be specifically determined by the following steps, as shown in fig. 3:
Step 301: and obtaining a knowledge graph of the target object based on the object attribute of the target object.
The object attribute of the target object refers to an attribute of the target object, such as a functional attribute or a configuration attribute, and the like, such as a processor version, a system version, a memory, a camera, and the like.
In this embodiment, a preset map construction algorithm may be used to construct a knowledge map of the target object based on the object attribute of the target object.
Step 302: and judging whether the knowledge graph of the target object meets the first condition or not to obtain a judging result.
Specifically, in this embodiment, the description feature of the target object and the knowledge spectrum of the target object are compared by the target problem statement in the first condition, so as to determine whether the knowledge spectrum of the target object matches the first condition, thereby obtaining a determination result.
Correspondingly, in this embodiment, when determining whether the knowledge graph of the target object meets the first condition, the corresponding determination manner may be determined by obtaining the question type of the target question statement, which specifically includes the following cases:
First, if the question type of the target question sentence belongs to the numerical value type, whether the knowledge graph of the target object meets the first condition can be judged through a first mode, so that a judging result is obtained.
In the second, if the question type of the target question sentence belongs to a non-numeric type, such as a character type, etc., then whether the knowledge graph of the target object meets the first condition can be judged in a second mode, so as to obtain a judgment result.
It can be seen that, in this embodiment, different manners are selected based on the question type of the target question sentence to realize the judgment between the knowledge graph and the first condition.
In the first aspect, the specific method may be:
Firstly, generating an attribute judgment rule of a target object based on a knowledge graph of the target object, and then judging whether the attribute judgment rule of the target object meets a first condition or not to obtain a judgment result.
For example, in this embodiment, based on a knowledge graph of the target object, a determination rule is preset on the basis of a preset standard template, for example, a rule with a memory of less than or equal to 32GB or a rule with a pixel of less than 200 ten thousand, and the like, and accordingly, in this embodiment, it is determined whether these rules of the target object are matched with the first condition, so as to obtain a determination result, if these rules are matched with the description feature of the target object in the target question sentence in the first condition, it indicates that the target question sentence is a valid sentence, and accordingly, answer information of the target question sentence can be directly output, and if these rules are not matched with the description feature of the target question in the first condition, it indicates that the target question sentence is an invalid sentence, and accordingly, answer information of the target question sentence can still be output, and also corresponding determination information of the first condition, i.e., the target question sentence, can be output. For example, the objective question statement is "do this handset can use 128G memory card? In this embodiment, a first condition is generated based on the sentence and the target object "the mobile phone" in the sentence, where the first condition includes "128G used memory" for describing the target object, in this embodiment, a knowledge graph of the target object may be constructed, and an attribute determination rule "memory less than 32GB" generated by the knowledge graph is compared with the first condition "128G used memory", where it is obvious that "memory less than 32GB" cannot be matched with "128G used memory", at this time, a determination result is obtained, a question answer "cannot be used" is output, and a determination information "invalid question" for the first condition is output.
In the second mode, specifically, it may be:
Firstly, carrying out joint modeling on a knowledge graph of a target object and a first condition, for example, taking a product triplet of the target object and description characteristics of the target object in the first condition as input data, establishing an LSTM model based on attention mechanism, operating the model to obtain model data, then carrying out quantization processing on the model data to obtain a scalar value, judging whether the scalar value is larger than or equal to a scalar threshold value, and obtaining a judging result, wherein if the scalar value is larger than the scalar threshold value, the answer information of the target problem statement is directly output, and if the scalar value is not larger than the scalar threshold value, the judgment information of the first condition is output, namely the judgment information for considering that the target problem statement is invalid. For example, "how does this handset swipe the card? Correspondingly, the target object is a mobile phone of model A, in this embodiment, the knowledge graph of the mobile phone of model A and the "mobile phone card swiping of model A" in the first condition are used as model input data, and the built LSTM model based on attention mechanism is operated to obtain model data and quantized scalar values thereof. Thus, if the scalar value is 0.7, which is greater than the scalar threshold value of 0.5, the obtained judgment result indicates that the target question sentence is a valid question, and answer information for the target question sentence is output at this time: if the scalar value is 0.3 and is obviously smaller than the scalar threshold value of 0.5, a judgment result is obtained at the moment, and correspondingly, after the answer information of the target question statement, namely the operation flow of the mobile phone for swiping the card, is output, judgment information of a first condition, namely the mobile phone does not support a swiping function, can be output.
As shown in fig. 4, a schematic structural diagram of a data processing apparatus according to a second embodiment of the present application is provided, and the apparatus may be applied to an electronic device, such as a notebook or a server, which may provide question-answering services for an object or a product, such as an intelligent customer service application installed in the electronic device, to provide answer questions, technical communications, etc. for a user to use the product.
In this embodiment, the data processing apparatus may include the following structure:
a receiving unit 401 for obtaining a target question sentence.
A determining unit 402, configured to determine a target object based on the target question statement.
A judging unit 403, configured to judge whether the target object meets a first condition, so as to obtain a judgment result.
Wherein the first condition is generated based on the target object and the target question sentence.
Specifically, the determining unit 403 determines whether the target object meets the first condition, so as to obtain a determination result, which may specifically be:
and obtaining a knowledge graph of the target object based on the object attribute of the target object, and then judging whether the knowledge graph of the target object meets a first condition or not to obtain a judging result.
It should be noted that, the judging unit 403 may specifically judge, by obtaining the question type of the target question sentence, based on whether the question type belongs to the numerical type or belongs to the non-numerical type in a corresponding manner, for example:
If the problem type of the target problem statement belongs to a numerical value type, judging whether the knowledge graph of the target object meets a first condition through a first mode to obtain a judging result, wherein the first mode can be as follows: and generating an attribute judgment rule of the target object based on the knowledge graph of the target object, and judging whether the attribute judgment rule meets a first condition or not to obtain a judgment result.
If the problem type of the target problem statement belongs to a non-numerical value type, judging whether the knowledge graph of the target object meets a first condition through a second mode to obtain a judging result, wherein the second mode can be as follows: carrying out joint modeling on the knowledge graph of the target object and the first condition to obtain model data; carrying out quantization processing on the model data to obtain a scalar value; and judging whether the scalar value is larger than or equal to a scalar threshold value or not so as to obtain a judging result.
And an output unit 404, configured to output feedback information based on the determination result.
The feedback information comprises judgment information of the first condition, or the feedback information comprises answer information of the target question statement and judgment information of the first condition.
Specifically, based on the determination result, the output unit 404 may specifically implement the following manner when outputting feedback information: and outputting answer information of the target question sentence, and then outputting judgment information of the first condition.
For example, if the judgment result indicates that the target object satisfies the first condition, the output unit 404 outputs answer information of the target question sentence;
if the judgment result indicates that the target object does not meet the first condition, the output unit 404 outputs answer information of the target question sentence, and then outputs judgment information of the first condition.
It can be seen that, in the data processing apparatus provided in the second embodiment of the present application, after the target question sentence is obtained, the corresponding target object is first determined, and then the corresponding feedback information is output based on the determination result that whether the target object corresponding to the target question sentence meets the first condition, where the feedback information includes the determination information for the first condition or further includes the answer information for the target question sentence. Therefore, in this embodiment, after receiving the question sentence, the answer is not directly answered, but is judged first, if the corresponding target object meets the first condition, so that after the validity of the question is judged, whether to answer the question is selected, and further the efficiency of answering the question is improved, and the user experience is obviously improved.
It should be noted that, the implementation and illustration of each unit in the apparatus of this embodiment may refer to the corresponding content in the foregoing, and will not be described in detail herein.
Referring to fig. 5, a schematic structural diagram of an electronic device according to a third embodiment of the present application may be a device capable of performing data processing, such as a computer or a server, where the electronic device may provide question-answering services for an object or a product, for example, an intelligent customer service application is installed in the electronic device, so as to provide question-answering and technical communication for a user to use the product.
Specifically, the electronic device in this embodiment may include the following structure:
An interaction device 501 for obtaining a target question sentence.
The interaction device 501 may be an input/output device, or may be a device connected to the input/output device, so as to obtain a target problem statement, for example, after a user inputs the target problem statement on a touch screen of the server, or after the user inputs the target problem statement through his own mobile phone, the server in this embodiment receives the target problem statement sent by the user mobile phone through the interaction device 501.
A processing device 502 that determines a target object based on the target question statement; whether the target object meets the first condition is judged, so that a judging result is obtained, and feedback information is output through the interaction device 501 based on the judging result.
The feedback information comprises judgment information of the first condition, or the feedback information comprises answer information of the target question statement and judgment information of the first condition.
Specifically, the processing device 502 determines whether the target object meets the first condition, so as to obtain a determination result, which may specifically be:
and obtaining a knowledge graph of the target object based on the object attribute of the target object, and then judging whether the knowledge graph of the target object meets a first condition or not to obtain a judging result.
It should be noted that, after obtaining the question type of the target question sentence, the processing device 502 may determine, based on whether the question type belongs to a numerical type or a non-numerical type, in a corresponding manner, for example:
If the problem type of the target problem statement belongs to a numerical value type, judging whether the knowledge graph of the target object meets a first condition through a first mode to obtain a judging result, wherein the first mode can be as follows: and generating an attribute judgment rule of the target object based on the knowledge graph of the target object, and judging whether the attribute judgment rule meets a first condition or not to obtain a judgment result.
If the problem type of the target problem statement belongs to a non-numerical value type, judging whether the knowledge graph of the target object meets a first condition through a second mode to obtain a judging result, wherein the second mode can be as follows: carrying out joint modeling on the knowledge graph of the target object and the first condition to obtain model data; carrying out quantization processing on the model data to obtain a scalar value; and judging whether the scalar value is larger than or equal to a scalar threshold value or not so as to obtain a judging result.
Specifically, based on the determination result, the processing device 502 may specifically implement the following manner when outputting feedback information: and outputting answer information of the target question sentence, and then outputting judgment information of the first condition.
For example, if the determination result characterizes that the target object satisfies the first condition, the processing device 502 outputs answer information of the target question sentence;
If the judging result indicates that the target object does not meet the first condition, the processing device 502 outputs answer information of the target question statement, and then outputs judging information of the first condition.
It can be seen that, in the electronic device provided in the third embodiment of the present application, after the target problem statement is obtained, the corresponding target object is first determined, and then the corresponding feedback information is output based on the determination result that whether the target object corresponding to the target problem statement meets the first condition, where the feedback information includes the determination information for the first condition or further includes the answer information for the target problem statement. Therefore, in this embodiment, after receiving the question sentence, the answer is not directly answered, but is judged first, if the corresponding target object meets the first condition, so that after the validity of the question is judged, whether to answer the question is selected, and further the efficiency of answering the question is improved, and the user experience is obviously improved.
It should be noted that, the implementation and illustration of each structure in the electronic device of this embodiment may refer to the corresponding content in the foregoing, and will not be described in detail herein.
The following exemplifies the practical application of the technical solution in this embodiment:
First, the problem to be solved by this embodiment is: and carrying out validity evaluation on the questions raised by the user. In this embodiment, features that can represent the capability of software and hardware of a product, such as a CPU model, a system version, a baseband frequency, and an inductor, are extracted from a product knowledge graph as implicit preconditions of a problem. Then, the user questions are classified into numerical questions and non-numerical questions. Wherein numerical problems are handled by a template or rule matching system, and non-numerical problems are handled by an LSTM neural network model based on the Attention mechanism. The neural network model receives the product attribute triples and the user question marks as input, and an Attention mechanism in the model is used for automatically finding out the alignment relation between the user question marks and words/phrases in the product attribute triples.
The key points of the implementation scheme in the embodiment are as follows:
1. The method comprises the steps of constructing a product knowledge graph, wherein the key points comprise features for reflecting the capability of software and hardware of the product, and the features comprise a CPU, a system version, a network system, a memory, a camera, an inductor and the like. Constructing an enabling relation graph of the features and the product functions, so as to assist in the validity evaluation of the following user question;
2. classifying user problems, namely classifying the user problems into two types of numerical type and non-numerical type, wherein the numerical type problems comprise maximum expansion memory supported by a mobile phone, maximum resolution supported by the mobile phone for shooting and the like;
3. The natural language processing technology is adopted to link the products contained in the user problems with the product knowledge graph, so that the semantic information of the user question is enriched, and a foundation is laid for the intelligent evaluation;
4. For the non-numerical problem, an LSTM neural network based on an Attention mechanism is adopted to jointly model the user problem and the product attribute characteristics, and the corresponding relation between the user question and the product attribute characteristics is automatically found, for example, the user problem is 'how to open 5G', and then the user question is linked with the communication module characteristics of the corresponding product.
Based on this, the present embodiment has the advantage that:
In the embodiment, the validity of the problem proposed by the user is intelligently evaluated, so that the polling times when the problem of the user is invalid can be greatly reduced, and the question and answer quality is improved. Specifically, firstly, user problem information and product attribute information are comprehensively considered, and the implicit relation between user questions and product attributes is found through pattern matching and a neural network based on an attribute mechanism, so that the defect that the traditional intelligent customer service lacks problem related knowledge information is overcome. Secondly, by constructing a product knowledge graph and combining product attribute characteristics according to the inherent semantic information of the problem, the accuracy of judging the validity of the problem is greatly improved. Third, the dialogue can be terminated as early as possible for invalid questions, thereby improving customer satisfaction. Fourth,: the scheme provided by the embodiment can be continuously adjusted, self-learned and perfected in the use of the system, and the evaluation accuracy of the effectiveness of the user problem can be gradually perfected along with the advancement of the system;
In addition, the proposal provided by the embodiment does not need to manually label a large amount of data, automatically digs from the real data, and has higher degree of automation. The intelligent customer service system and the intelligent customer service method expand the knowledge capability of intelligent customer service, optimize the service quality of intelligent customer service, improve the satisfaction of users, and have important significance for applications such as user understanding and personalized customer service.
Specifically, the technical solution in this embodiment may be implemented by a plurality of functional modules, as shown in fig. 6, and the main functions and workflows of each module are described with reference to the implementation flowchart in fig. 7:
1. Training corpus mining module: from the actual customer service history questions such as the history questions stored in the database, the customer service inquiry sentences after the user puts out the questions, such as the user inquiry "how to take a bus with a mobile phone? ", then customer service might ask" do you's cell phone support NFC? ". Through the additional question, the implicit association between the user question and the product function can be obtained, and the answer of the user can be used as a learning sample of the finished product function characteristics and the validity of the user question.
2. And a pretreatment module: the method mainly comprises three processes, namely, filtering out calling, coldness and boring and the like; second, filtering short sentences with the length smaller than a certain threshold value; third, multiple consecutive phrases are merged, i.e., into one if the user's question is entered multiple times.
3. And a product knowledge graph construction module: the product knowledge graph can embody the software and hardware capability of the product, such as the attributes of CPU, system version, network system, memory, camera, sensor and the like. And constructing an enabling relation graph of the features and the product functions, so as to assist the validity evaluation of the following user question. The product attribute information in the knowledge graph is shown in table 1.
TABLE 1 product Property characterization
4. The user problem processing module at least comprises a classification module and an entity link module: firstly, the user questions are parallelized by using a classification module, wherein the classification module adopts a common classification algorithm such as fastText to divide questions into two types of numerical questions and non-numerical questions, and then uses an entity linking module to link products contained in the user questions with product attribute feature triplets as shown in table 2.
Furthermore, the numerical problems can be compared through the logic processing module, if the problem of the user is how to shoot 4K video, and the corresponding mobile phone supported cameras support 1080P video shooting at most, the logic processing module can determine that the problem of the user is an invalid problem. For non-numerical problems, then, decisions need to be made through the LSTM neural network model based on the Attention mechanism.
TABLE 2 user question type example
| Problem ID | User problems | Question type |
| QA1 | Is MotoZ support for shooting 4K video? | Numerical value type |
| QA2 | Is MotoZ a 128G memory card available? | Numerical value type |
| QA3 | Is MotoZ available to brush a bus? | Non-numerical value type |
| QA4 | Is MotoZ G supported by MotoZ? | Non-numerical value type |
The numerical problem is processed through a preset standard template matching method, and the standard template is set based on a regular expression. In this embodiment, the numerical question may be first matched with the standard template, where the matching structure is a Trie. A set of decision rules are then predefined for standard templates and product attributes, such as user question "MotoZ can use 128G memory card? The corresponding template is "$product|$support|$num|$memory Card", the Product knowledge graph is queried through entity links to obtain < Moto Z3, the maximum storage expansion capacity and 2TB >, the corresponding judgment rule is "$num is less than or equal to 2TB", and the comparison result can be obtained through unit normalization, so that the user problem can be known to be an effective problem.
5. Neural network module
In this embodiment, the numerical problem is determined by the logic processing module, and the non-numerical problem is determined by the neural network module of the LSTM model. The LSTM model specifically comprises the following operation processes:
In this embodiment, a user question is taken as an assumption, and a product attribute triplet obtained in a knowledge graph is taken as a precondition, specifically, the attribute triplet in the product knowledge graph is taken as the precondition. The present model performs joint modeling on the assumptions and preconditions obtained above, and the specific flowchart is shown in the schematic diagram of fig. 8:
The user question and product attribute triplets are processed through two different LSTM neural networks, wherein the memory storage of the LSTM neural network processing the user question is initialized by the memory of the last neuron of the LSTM processing the product attribute triplets (as shown at C5). In this embodiment, a neural network model is used to vectorize and encode the user question and the product attribute triplet, and then an input vector x is obtained through a linear projection layer.
LSTM neural networks incorporating the Attention mechanism are used to achieve alignment of user questions with specific features of product attribute triples. That is, the LSTM does not need to record all information of the product attribute triples, but only needs to keep attribute feature information related to the user question, so that on one hand, the correspondence between the user question and the product attribute feature is realized, and on the other hand, the storage pressure of the neural network is also reduced.
Let Y ε R K*L denote a matrix of first LSTM neural network output vectors [ h1, h2, ], hL ], L denote the number of words contained in the product attribute triplet, and k is a dimension of the hyper-parametric representation word vector. In addition, let e L∈RL be the full 1 vector, h N be the final output vector of the two LSTM neural networks. The attribute mechanism obtains a weight vector alpha and a weight vector gamma of the product attribute feature through the following formulas (1), (2) and (3).
α＝softmax(wTM),α∈RL (2)
γ＝YαT,γ∈Rk (3)
Where W y、Wh∈Rk*k is the projection matrix to be trained, W εR k is the parameter vector, and W T represents the transpose. The final question and product attribute triplet representation is derived from the nonlinear function tanh acting on h N and γ, namely company (4):
h*＝tanh(WprN+WxhN),h*∈Rk (4)
thus, h * can obtain a scalar value through the logistic regression process, and the user problem is considered to be valid when the scalar value is greater than or equal to a certain threshold value, such as 0.5, or invalid.
6. Self-learning optimization module: in this embodiment, after the validity evaluation is performed on the user problem, the result is fed back to the customer service personnel, and the customer service personnel can determine whether the customer problem evaluation is correct. If customer service considers that the evaluation result is correct, that is, in this embodiment, the corresponding relationship between the user question and the product attribute and the determination rule are learned, for example, the user question is "how to use MotoZ as the air conditioner remote controller", then in this embodiment, the corresponding relationship between "air conditioner remote controller" and whether the product attribute "supports infrared" can be obtained through learning. Otherwise, if the customer service considers that the evaluation is wrong, there is a possibility that the user question and the product attribute do not have a corresponding relationship or the wrong corresponding relationship is learned or the wrong determination method is learned. Therefore, in this embodiment, a large number of samples with correct mapping relationships and determination methods and samples without mapping relationships or with wrong mapping and erroneous determination can be obtained by this method, and then the evaluation model is retrained, so as to realize self-learning optimization of the system.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative elements and steps are described above generally in terms of functionality in order to clearly illustrate the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. The software modules may be disposed in Random Access Memory (RAM), memory, read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (8)
1. A data processing method, comprising:
acquiring a target problem statement;
determining a target object based on the target problem statement;
Judging whether the target object meets a first condition to obtain a judging result, including: constructing a knowledge graph of the target object based on the object attribute of the target object; judging whether the knowledge graph of the target object meets a first condition or not to obtain a judging result; wherein the object properties include: the functional attribute or the configuration attribute is generated based on the target object and the target problem statement, and the first condition is a characteristic condition of the target object aiming at the characteristic target object in the target problem statement;
based on the judging result, outputting feedback information;
The feedback information comprises judgment information of the first condition, or the feedback information comprises answer information of the target question statement and judgment information of the first condition.
2. The method of claim 1, outputting feedback information based on the determination result, comprising:
and outputting answer information of the target question sentence, and then outputting judgment information of the first condition.
3. The method of claim 1, further comprising:
obtaining the question type of the target question statement;
Judging whether the knowledge graph of the target object meets a first condition to obtain a judging result, wherein the judging step comprises the following steps:
If the problem type of the target problem statement belongs to a numerical value type, judging whether the knowledge graph of the target object meets a first condition or not through a first mode to obtain a judging result;
if the problem type of the target problem statement belongs to a non-numerical value type, judging whether the knowledge graph of the target object meets a first condition or not through a second mode to obtain a judging result.
4. The method according to claim 3, wherein determining whether the knowledge graph of the target object meets the first condition in the first manner to obtain the determination result includes:
generating an attribute judgment rule of the target object based on the knowledge graph of the target object;
and judging whether the attribute judging rule meets a first condition or not to obtain a judging result.
5. The method according to claim 3, wherein determining whether the knowledge graph of the target object meets the first condition by the second method to obtain the determination result includes:
carrying out joint modeling on the knowledge graph of the target object and the first condition to obtain model data;
Carrying out quantization processing on the model data to obtain a scalar value;
And judging whether the scalar value is larger than or equal to a scalar threshold value or not so as to obtain a judging result.
6. The method of claim 1, outputting feedback information based on the determination result, comprising:
If the judging result represents that the target object meets the first condition, outputting answer information of the target question statement;
And if the judging result represents that the target object does not meet the first condition, outputting judging information of the first condition and answer information of the target question statement.
7.A data processing apparatus comprising:
a receiving unit for obtaining a target question sentence;
a determining unit, configured to determine a target object based on the target question sentence;
The judging unit is configured to judge whether the target object meets a first condition, so as to obtain a judgment result, and includes: constructing a knowledge graph of the target object based on the object attribute of the target object; judging whether the knowledge graph of the target object meets a first condition or not to obtain a judging result; wherein the object properties include: the functional attribute or the configuration attribute is generated based on the target object and the target problem statement, and the first condition is a characteristic condition of the target object aiming at the characteristic target object in the target problem statement;
The output unit is used for outputting feedback information based on the judging result;
The feedback information comprises judgment information of the first condition, or the feedback information comprises answer information of the target question statement and judgment information of the first condition.
8. An electronic device, comprising:
the interaction device is used for acquiring target problem sentences;
Processing equipment, based on the target problem statement, determining a target object; judging whether the target object meets a first condition to obtain a judging result, including: constructing a knowledge graph of the target object based on the object attribute of the target object; judging whether the knowledge graph of the target object meets a first condition or not to obtain a judging result; wherein the object properties include: the first condition is generated based on the target object and the target problem statement, the first condition is a characteristic condition of the target object aiming at the target problem statement, and feedback information is output through the interaction equipment based on the judging result;
The feedback information comprises judgment information of the first condition, or the feedback information comprises answer information of the target question statement and judgment information of the first condition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910252940.2A CN109960811B (en) | 2019-03-29 | 2019-03-29 | Data processing method and device and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910252940.2A CN109960811B (en) | 2019-03-29 | 2019-03-29 | Data processing method and device and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109960811A CN109960811A (en) | 2019-07-02 |
| CN109960811B true CN109960811B (en) | 2024-04-26 |
Family
ID=67025360
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910252940.2A Active CN109960811B (en) | 2019-03-29 | 2019-03-29 | Data processing method and device and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109960811B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110457021B (en) * | 2019-07-12 | 2023-05-16 | 招商局金融科技有限公司 | Index data processing method, device and computer readable storage medium |
| CN110705274B (en) * | 2019-09-06 | 2023-03-24 | 电子科技大学 | Fusion type word meaning embedding method based on real-time learning |
| CN111966706B (en) * | 2020-08-19 | 2023-08-22 | 中国银行股份有限公司 | Official micro-reply method and device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105068661A (en) * | 2015-09-07 | 2015-11-18 | 百度在线网络技术（北京）有限公司 | Man-machine interaction method and system based on artificial intelligence |
| CN105843875A (en) * | 2016-03-18 | 2016-08-10 | 北京光年无限科技有限公司 | Smart robot-oriented question and answer data processing method and apparatus |
| US9542447B1 (en) * | 2015-10-13 | 2017-01-10 | International Business Machines Corporation | Supplementing candidate answers |
| CN108446286A (en) * | 2017-02-16 | 2018-08-24 | 阿里巴巴集团控股有限公司 | A kind of generation method, device and the server of the answer of natural language question sentence |
| CN108491443A (en) * | 2018-02-13 | 2018-09-04 | 上海好体信息科技有限公司 | By computer-implemented and method and computer system of user session |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101306667B1 (en) * | 2009-12-09 | 2013-09-10 | 한국전자통신연구원 | Apparatus and method for knowledge graph stabilization |
-
2019
- 2019-03-29 CN CN201910252940.2A patent/CN109960811B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105068661A (en) * | 2015-09-07 | 2015-11-18 | 百度在线网络技术（北京）有限公司 | Man-machine interaction method and system based on artificial intelligence |
| US9542447B1 (en) * | 2015-10-13 | 2017-01-10 | International Business Machines Corporation | Supplementing candidate answers |
| CN105843875A (en) * | 2016-03-18 | 2016-08-10 | 北京光年无限科技有限公司 | Smart robot-oriented question and answer data processing method and apparatus |
| CN108446286A (en) * | 2017-02-16 | 2018-08-24 | 阿里巴巴集团控股有限公司 | A kind of generation method, device and the server of the answer of natural language question sentence |
| CN108491443A (en) * | 2018-02-13 | 2018-09-04 | 上海好体信息科技有限公司 | By computer-implemented and method and computer system of user session |
Non-Patent Citations (2)
| Title |
|---|
| 基于知识图谱的智能客服***研究;饶竹一 等;《电力信息与通信技术》;20170731;第15卷(第7期);第41-45页，图3 * |
| 聊天机器人与自动问答技术;mjs360电商技术;《https://blog.csdn.net/gfyufffb/article/details/83782028》;20181106;第1-6页，表1 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109960811A (en) | 2019-07-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20220044094A1 (en) | Method and apparatus for constructing network structure optimizer, and computer-readable storage medium | |
| CN111061856B (en) | Knowledge perception-based news recommendation method | |
| CN109960811B (en) | Data processing method and device and electronic equipment | |
| CN112732911A (en) | Semantic recognition-based conversational recommendation method, device, equipment and storage medium | |
| CN107071193A (en) | The method and apparatus of interactive answering system accessing user | |
| CN112632244A (en) | Man-machine conversation optimization method and device, computer equipment and storage medium | |
| CN109726291B (en) | Loss function optimization method and device of classification model and sample classification method | |
| CN112418059A (en) | Emotion recognition method and device, computer equipment and storage medium | |
| CN113344596A (en) | Customer service processing method, device, equipment and computer readable storage medium | |
| CN112995414A (en) | Behavior quality inspection method, device, equipment and storage medium based on voice call | |
| CN115270807A (en) | Method, device and equipment for judging emotional tendency of network user and storage medium | |
| CN111160443A (en) | Activity and user identification method based on deep multitask learning | |
| CN116758379B (en) | Image processing method, device, equipment and storage medium | |
| CN117593752A (en) | PDF document input method, PDF document input system, storage medium and electronic equipment | |
| CN117688165A (en) | Multi-edge collaborative customer service method, device, equipment and readable storage medium | |
| CN113705242A (en) | Intelligent semantic matching method and device for education consultation service | |
| CN112184317A (en) | Waste mobile phone pricing method based on value-preserving rate and discrete neural network | |
| CN113220828A (en) | Intention recognition model processing method and device, computer equipment and storage medium | |
| CN108154165A (en) | Love and marriage object matching data processing method, device, computer equipment and storage medium based on big data and deep learning | |
| CN109033413B (en) | Neural network-based demand document and service document matching method | |
| CN116467416A (en) | Multi-mode dialogue emotion recognition method and system based on graphic neural network | |
| CN111062452A (en) | Method and equipment for performing enhanced training aiming at Gaussian classification neural network | |
| CN111081221A (en) | Training data selection method and device, electronic equipment and computer storage medium | |
| CN108897894A (en) | A kind of problem generation method | |
| CN114420109B (en) | Voice gender joint recognition method and device, electronic 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |