CN112241448A - Response information generation method, device, equipment and storage medium - Google Patents
Response information generation method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN112241448A CN112241448A CN201910649366.4A CN201910649366A CN112241448A CN 112241448 A CN112241448 A CN 112241448A CN 201910649366 A CN201910649366 A CN 201910649366A CN 112241448 A CN112241448 A CN 112241448A
- Authority
- CN
- China
- Prior art keywords
- attribute
- text word
- information
- response information
- value
- 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
- 230000004044 response Effects 0.000 title claims abstract description 188
- 238000000034 method Methods 0.000 title claims abstract description 92
- 239000013598 vector Substances 0.000 claims description 59
- 238000002372 labelling Methods 0.000 claims description 22
- 230000015654 memory Effects 0.000 claims description 20
- 238000012545 processing Methods 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 10
- 238000013528 artificial neural network Methods 0.000 claims description 8
- 230000002457 bidirectional effect Effects 0.000 claims description 8
- 230000000306 recurrent effect Effects 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 2
- 239000000126 substance Substances 0.000 claims 2
- 238000010586 diagram Methods 0.000 description 23
- 230000008569 process Effects 0.000 description 8
- 230000007246 mechanism Effects 0.000 description 6
- 238000012549 training Methods 0.000 description 6
- 230000001502 supplementing effect Effects 0.000 description 5
- 230000009286 beneficial effect Effects 0.000 description 4
- 238000010276 construction Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 230000006403 short-term memory Effects 0.000 description 2
- 238000007792 addition Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/01—Customer relationship services
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Finance (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Artificial Intelligence (AREA)
- Human Computer Interaction (AREA)
- Accounting & Taxation (AREA)
- Computational Linguistics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the invention provides a response information generation method, a response information generation device, response information generation equipment and a storage medium, wherein the method comprises the following steps: identifying an intention attribute in the query information sent by the first user terminal; extracting an attribute value corresponding to the intention attribute from first response information replied by the second user terminal aiming at the inquiry information; second response information is generated using the conversational template based on the attribute value and the intent attribute. The method is used for solving the problems that when the seller is a non-professional seller, a database for replying the buyer problem is difficult to construct and maintain, the buyer cannot be replied to ask a question in time, and the experience of the buyer is reduced. So as to reduce time and labor cost and improve user experience.
Description
Technical Field
The present invention relates to the field of data processing technologies, and in particular, to a method, an apparatus, a device, and a storage medium for generating response information.
Background
With the continuous development of terminal technology, customer service robots are widely used in various transaction platforms (e.g., e-commerce customer service robots). However, the application of this technology to second-hand trading markets is not yet mature.
The main reasons are as follows: (1) different from other trading platforms, the seller of the second-hand trading market is not a professional seller, the construction of the commodity database is not well understood, and the training cost is high; (2) a great deal of time is needed for maintaining the database of each commodity, and professional maintenance of the database is difficult for non-professional sellers; (3) the seller of the second-hand trading market usually evacuates to reply to the buyer, and cannot guarantee that the buyer can always interact with the buyer on line, and if the buyer receives the reply for too long time, the buyer usually loses the willingness of further interaction, and the user experience of the buyer is reduced.
Disclosure of Invention
One or more embodiments of the present invention describe a method, an apparatus, a device, and a storage medium for generating response information, so as to solve the problem that when a seller is a non-professional seller, it is difficult to construct and maintain a database for replying to a question of a buyer, so that the question of the buyer cannot be replied in time, and the experience of the user of the buyer is reduced.
According to a first aspect, there is provided a response information generating method, which may include:
identifying an intention attribute in the query information sent by the first user terminal;
extracting an attribute value corresponding to the intention attribute from first response information replied by the second user terminal aiming at the inquiry information;
second response information is generated using the conversational template based on the attribute value and the intent attribute.
According to a second aspect, there is provided a method of constructing a database, the method may comprise:
acquiring a plurality of pieces of inquiry information sent by a first user terminal and a plurality of pieces of first response information replied by a second user terminal aiming at the plurality of inquiry information;
with the response information generating method shown in the first aspect, a plurality of second response information is generated to construct a response information database.
According to a third aspect, there is provided a method of acquiring response information, which may include:
identifying a target intention attribute in inquiry information sent by a first user terminal;
from the response information database shown in the second aspect, second response information corresponding to the target intention attribute is taken as response information of the inquiry information;
and sending response information to the first user terminal.
According to a fourth aspect, there is provided a response information generating apparatus, which may include:
the identification module is used for identifying the intention attribute in the inquiry information sent by the first user terminal;
the extraction module is used for extracting an attribute value corresponding to the intention attribute from first response information replied by the second user terminal aiming at the inquiry information;
and the processing module is used for generating second response information by utilizing the conversational template based on the attribute value and the intention attribute.
According to a fifth aspect, there is provided an apparatus for constructing a database, the apparatus may include:
the system comprises an acquisition module, a processing module and a processing module, wherein the acquisition module is used for acquiring a plurality of pieces of inquiry information sent by a first user terminal and a plurality of pieces of first response information replied by a second user terminal aiming at the plurality of inquiry information;
generating means for generating a plurality of second response information to construct a response information database, using the response information generating means as shown in the fourth aspect.
According to a sixth aspect, there is provided an apparatus for acquiring response information, the apparatus may include:
the identification module is used for identifying the target intention attribute in the inquiry information sent by the first user terminal;
a processing module configured to use second response information corresponding to the target intention attribute as response information of the inquiry information from the response information constructed by the database constructing apparatus shown in the fifth aspect;
and the transceiver module is used for sending the response information to the first user terminal.
According to a seventh aspect, there is provided a system for constructing a database, comprising:
the input module is used for splitting inquiry information sent by a first user end and determining at least one first text word;
the word vector module is used for generating a word vector corresponding to each first text word in at least one first text word according to the at least one first text word;
the bidirectional recurrent neural network module is used for determining a context vector corresponding to the word vector corresponding to each first text word based on the word vector corresponding to each first text word; calculating a probability value corresponding to each first text word according to the context vector and the preset attention vector; and configuring a weight value of each first text word;
and the output module is used for determining the intention attribute in the query information according to the probability value corresponding to each first text word and the weight value of each first text word.
According to an eighth aspect, there is provided a system for acquiring response information, comprising:
the input module is used for splitting the first response information sent by the second user end and determining at least one second text word;
the word vector module is used for generating a word vector corresponding to each second word in at least one second text word according to the at least one second text word;
the bidirectional recurrent neural network module is used for determining a context vector corresponding to the word vector corresponding to each second text word based on the word vector corresponding to each second text word; calculating a probability value corresponding to each second text word according to the context vector and the preset attention vector; and configuring a weight value of each second text word;
and the label labeling module is used for labeling the initial part and the middle part of the attribute value in the second text word according to the probability value corresponding to each second text word and the weight value of each second text word so as to determine the attribute value in the second response information. According to a ninth aspect, there is provided a computing device comprising at least one processor and a memory, the memory being arranged to store computer program instructions, the processor being arranged to execute a program from the memory to control the computing device to implement a method as claimed in any of the first, second or third aspects.
According to a tenth aspect, there is provided a computer readable storage medium having a computer program stored thereon, wherein the computer program, if executed in a computer, causes the computer to perform the method as set forth in any one of the first, second or third aspects.
In the embodiment of the invention, the inquiry information of the buyer and the response information replied by the seller according to the inquiry information can be arranged according to the dialect template to obtain the response information, and the response information is beneficial to constructing and supplementing a corresponding response information database. Compared with the existing method for maintaining the database by professional personnel, the method provided by the embodiment of the invention can automatically construct the response information database, so that on one hand, for the seller, the threshold of the seller for answering the problem is reduced, professional training is not needed, and the time and labor cost are reduced; on the other hand, for the buyer, the problem that when the seller is a non-professional seller, the database for replying the buyer problem is difficult to construct and maintain, so that the question of the buyer cannot be replied in time, and the experience of the user of the buyer is reduced is solved.
Drawings
The present invention will be better understood from the following description of specific embodiments thereof taken in conjunction with the accompanying drawings, in which like or similar reference characters designate like or similar features.
FIG. 1 is a schematic diagram illustrating a structure for replying to a query message;
FIG. 2 illustrates a schematic diagram of a database constructed based on response information, according to one embodiment;
FIG. 3 shows a flow diagram of a reply information generation method according to one embodiment;
FIG. 4 illustrates a schematic diagram of determining a conversation template, according to one embodiment;
FIG. 5 illustrates a flow diagram of a method of building a database of response information, according to one embodiment;
FIG. 6 illustrates a flow diagram of a method of obtaining response information, according to one embodiment;
FIG. 7 is a block diagram that illustrates a method for building a database of response information based on response information, according to one embodiment;
FIG. 8 is a diagram illustrating a first model in a system for building a database, according to an embodiment;
FIG. 9 is a diagram illustrating a second model in a system for obtaining response information according to an embodiment;
fig. 10 shows a block diagram of the structure of a response information generating apparatus according to an embodiment;
fig. 11 is a block diagram showing the construction of the response information database apparatus according to an embodiment;
fig. 12 is a block diagram showing the configuration of an apparatus for acquiring response information according to an embodiment;
FIG. 13 illustrates a schematic structural diagram of a computing device, according to one embodiment.
Detailed Description
Features and exemplary embodiments of various aspects of the present invention will be described in detail below, and in order to make objects, technical solutions and advantages of the present invention more apparent, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not to be construed as limiting the invention. It will be apparent to one skilled in the art that the present invention may be practiced without some of these specific details. The following description of the embodiments is merely intended to provide a better understanding of the present invention by illustrating examples of the present invention.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any such measured relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
Currently, as shown in fig. 1, in most internet transaction platforms, a customer service robot may recognize a query intention of a buyer, and find and feed back an answer corresponding to the query intention from a product database. In the related art, the identification method of the query intention is already mature, but the building process of the commodity database is far behind. On one hand, most sellers in the second-hand market are non-professional sellers, and a database required by intelligent question answering is difficult to construct and maintain; here, the non-professional seller does not know the construction of the commodity database, and if relevant training is carried out, a large amount of manpower and material resources are wasted, so that the operation cost is increased; on the other hand, maintaining a database for each item requires additional time, making it difficult for a non-professional seller to reply to a buyer's question in a timely manner.
Therefore, in order to solve the above technical problems, embodiments of the present invention provide a method for generating response information, a method for constructing a database, a method for acquiring response information, an apparatus, a device and a storage medium, which are described in detail below.
Before explaining the response information generating method provided by the embodiment of the present invention, a flow structure for constructing a database based on response information is introduced.
FIG. 2 illustrates a schematic diagram of a database constructed based on response information, according to one embodiment.
As shown in fig. 2, query information transmitted by a first user (e.g., buyer) is received and intent attributes in the query information are identified. First response information of a second user end (such as a seller) aiming at the inquiry information is received, and an attribute value corresponding to the intention attribute is extracted from the first response information. And arranging the intention information and the attribute values corresponding to the intention information according to a dialect template to generate second response information. And constructing a response information database according to the plurality of generated second response information, so that when target inquiry information sent by the buyer is received, the second response information corresponding to the target inquiry information is found from the response information database, and the second response information corresponding to the target inquiry information is sent to the buyer, so that the problem proposed by the buyer can be automatically replied.
Here, in addition to the above-mentioned problem of automatically replying to the buyer, the embodiment of the present invention provides another possible example, that is, a plurality of target options are displayed, so that the seller can select whether to use the reply information generation method provided by the embodiment of the present invention or reply to the query information sent by the first user end by the seller, which is specifically as follows: before receiving the first response information of the second user end (for example: the seller) to the query information, the method may further include: the second user displays a plurality of selectable target options, wherein the target options may include: the method includes an option of automatically replying to the query message sent by the first user end (i.e., selecting to reply to the query message sent by the first user end through the reply message generation method provided by the embodiment of the invention) and/or an option of not automatically replying to the query message sent by the first user end (e.g., replying to the query message through a seller).
Therefore, through the embodiment, on one hand, the query information sent by the first user side is replied in a plurality of ways, so that the flexibility of the method provided by the embodiment of the invention in practical application is improved; on the other hand, the accuracy of replying inquiry information can be improved through manual replying of the buyer, so that the user experience is improved.
Besides the above mentioned matching answers from the first answer information answered by the seller, the second answer information may also be generated by matching answers in the merchandise information (such as pictures, videos, audios or texts).
Here, when a part of the product information is changed in the above-mentioned second response information, the corresponding second response information may be updated in the response information database.
For example: when the second response information includes "the parent shirt a has 10 pieces of yellow", and the attribute corresponding to the shirt a is updated from 10 pieces of yellow to 8 pieces of yellow, the second response information is updated, that is, the second response information includes 10 pieces of yellow shirts, to be modified to include 8 pieces of yellow shirts.
Therefore, the method provided by the embodiment of the invention can arrange the inquiry information of the buyer and the response information replied by the seller according to the inquiry information according to the dialect template to obtain the response information, and the response information is beneficial to constructing and supplementing a corresponding response information database. Compared with the existing method for maintaining the database by professional personnel, the method provided by the embodiment of the invention can automatically construct the response information database, so that on one hand, for the seller, the threshold of the seller for answering the problem is reduced, professional training is not needed, and the time and labor cost are reduced; on the other hand, for the buyer, the problem that when the seller is a non-professional seller, the database for replying the buyer problem is difficult to construct and maintain, so that the question of the buyer cannot be replied in time, and the experience of the user of the buyer is reduced is solved.
Based on the above scenario, the embodiment of the present invention specifically describes, with reference to fig. 3 to fig. 6, a response information generation method, a database construction method, and a response information acquisition method provided by the embodiment of the present invention in detail.
Fig. 3 shows a flow diagram of a reply information generation method according to one embodiment.
As shown in fig. 3, the method flow includes steps 310 to 330: first, step 310, identify the intention attribute in the first user end (e.g. buyer) sending inquiry information; secondly, in step 320, extracting an attribute value corresponding to the intention attribute from the first response information replied to the query information by the second user terminal (for example, the seller); then, step 330, a second response message is generated using the conversational template based on the attribute value and the intent attribute.
The above steps are described in detail below:
first, step 310 is involved: the identifying of the intention attribute in the query information sent by the first user end may specifically include: determining a context vector corresponding to each of a plurality of first text words in the query information according to the plurality of first text words; obtaining a probability value of each first text word based on the context vector and a preset attention vector; and determining the intention attribute in the inquiry information according to the probability value of each first text word and the preset weight value of each first text word.
Here, the probability value is used to represent the degree of importance of each first text word; the importance degree of each first text word is in a direct proportion relation with the preset weight value of each first text word.
Further, the embodiment of the present invention may determine the intention attribute in the query information according to the probability value of each first text word and the preset weight value of each first text word in two ways, which are respectively described below:
the first method is as follows: calculating an initial weight value of each first text word according to the probability value and a preset weight value of each first text word; selecting a key probability value with the probability value meeting a preset condition from the obtained probability values of the plurality of first text words, and distributing a key weight value higher than a preset weight value to the key probability value; according to the key probability value and the key weight value, recalculating the weight value of the key text word corresponding to the key probability value; wherein the plurality of first text words comprise key text words; and determining the intention attribute in the inquiry information according to the initial weight value and the weight value of the key text word.
For example: the plurality of text words includes: the text recognition method comprises the following steps of a first text word A, a first text word B and a first text word C, wherein the probability value of the first text word A is 10%, the probability value of the first text word B is 30%, the probability value of the first text word C is 60%, and the preset weight value is 0.1. At this time, an initial weight value of each first text word is calculated. Then, a key probability value meeting a preset condition, such as a probability value of 60% of the first text word C, is selected, and a key weight value higher than a preset weight value of 0.1 is assigned to the key probability value (for example, a weight value of 0.1 is accumulated on the basis of the preset weight value, or a key weight value of 0.2 is directly assigned to the key probability value), where the weight value of the first text word C is recalculated according to the probability value of 60% of the first text word C and the key weight value of 0.2.
The second method comprises the following steps: selecting a key probability value with a probability value meeting a preset condition and a common probability value without the preset condition from the obtained probability values of the plurality of first text words; calculating a weight value corresponding to the key probability value according to the key probability value and a preset key weight value; calculating a weight value corresponding to the common probability value according to the common probability value and a preset common weight value; wherein the preset key weight value is greater than the common weight value; determining an intention attribute in the query information according to the weight value corresponding to the key probability value and the weight value corresponding to the common probability value; the preset weight value of each first text word comprises a preset key weight value and a preset common weight value.
For example: the plurality of text words includes: a first text word a, a first text word B and a first text word C, the probability value for the first text word a being 10%, the probability value for the first text word B being 30% and the probability value for the first text word C being 60%. Selecting a key probability value (namely, the probability value of 60 percent of the first text word C) with a probability value meeting a preset condition, and selecting a common probability value (namely, the probability value of 10 percent of the first text word A and the probability value of 30 percent of the first text word B) with a probability value not meeting the preset condition. Here, the preset key weight value includes 0.2, and the preset normal weight value includes 0.1, whereby the weight value of the first text word C is calculated according to the probability value of 60% of the first text word C and the preset key weight value of 0.2; and respectively calculating the probability values of the first text word A and the first text word B according to the probability value of 10% of the first text word A, the probability value of 30% of the first text word B and a preset common weight value of 0.1.
It should be noted that, in addition to the above mentioned cases of only dividing into two types, the above mentioned preset weight value, the key weight value, the preset key weight value and the preset ordinary weight value may be respectively assigned with different weight values according to the percentage of each probability value. By way of example two, a probability value of 10% for the first text word a, a probability value of 30% for the first text word B, and a probability value of 60% for the first text word C may be assigned a weight value of 0.05, a weight value of 0.15 for the B probability value, and a weight value of 0.3 for the C probability value.
Secondly, step 320 is involved: the embodiment of the present invention provides 2 possible ways to extract the attribute value corresponding to the intention attribute, which are specifically as follows:
the first method is as follows: determining a context vector corresponding to each second text word in the plurality of second text words according to the plurality of second text words in the first response information; and extracting attribute values corresponding to the intention attributes through a label labeling algorithm according to the context vector corresponding to each second text word.
Further, the label labeling algorithm comprises: a sequence labeling algorithm and/or a BIO label algorithm; the sequence labeling algorithm comprises an original labeling algorithm and/or a combined labeling algorithm.
The second method comprises the following steps: an attribute value corresponding to the intention attribute is extracted from the first response information and the product description information related to the inquiry information.
Then, step 330 is involved: second response information is generated using the conversational template based on the attribute value and the intent attribute.
The embodiment of the invention can determine the dialogistic template by the following method:
as shown in fig. 4, identifying an intention attribute of each of a plurality of preset inquiry messages; extracting an attribute value corresponding to the intention attribute from each response message corresponding to each preset inquiry message; taking each response message after the attribute value is extracted as candidate dialect messages; determining a phonetics template according to the candidate phonetics information.
For example: "parent, no red, also red-blue and white in color. "where the attribute values are" no red, and red, blue, and white ", the attributes are extracted, and the remaining part is used as candidate dialogistic information.
Additionally, the phonetics template may be updated according to the frequency of use of the phonetics template and/or the readability of the phonetics template.
Therefore, in the embodiment of the invention, the inquiry information of the buyer and the response information replied by the seller according to the inquiry information can be arranged according to the conversation template to obtain the response information, and the response information is beneficial to constructing and supplementing a corresponding response information database. The problem of when the seller is non-professional seller, lead to unable in time reply buyer's questioning, appear reducing buyer user experience and feel is solved.
Here, based on the plurality of second response information determined as described above, a database of response information may be constructed so that when query information is received again, a query question issued by the buyer can be automatically replied to through the database of response information, as follows.
FIG. 5 illustrates a flow diagram of a method of building a database of response information, according to one embodiment.
As shown in fig. 5, the method flow includes steps 510 to 520: firstly, step 510, acquiring a plurality of query messages sent by a first user terminal and a plurality of first response messages replied by a second user terminal for the plurality of query messages; then, in step 520, a plurality of second response messages are generated by using the response message generating method shown in fig. 3 to construct a response message database.
Wherein, the response information database may include: a plurality of pieces of inquiry information and attribute values respectively corresponding to each of the plurality of pieces of inquiry information, or second response information.
Further, when the response information database only includes the plurality of query information and the attribute value corresponding to each of the plurality of query information, and when replying to the query information of the first user, the query information and the attribute value corresponding to each of the query information may be arranged according to the conversational template to generate second response information, and then the second response information is sent to the first user.
How this response information database is used will be described below, based on the manner in which the response information database is constructed as shown in fig. 5, and as shown below.
Fig. 6 shows a flow diagram of a method of obtaining reply information according to one embodiment.
As shown in fig. 6, the method flow includes steps 610-630: first, step 610: identifying a target intention attribute in inquiry information sent by a first user terminal; next, step 620: from the response information database shown in fig. 5, second response information corresponding to the target intention attribute is taken as response information of the inquiry information; then, step 130: and sending response information to the first user terminal.
In one example, the second response message may be automatically updated when a change occurs in a portion of the second response message (for example, the product-related product attribute value includes the product quantity, the product color, and the like), where the change is as follows:
determining a product associated with the query information; obtaining a product attribute value related to a product from a product information database; and updating the second response information according to the product attribute value.
In summary, in the embodiment of the present invention, the query information of the buyer and the response information replied by the seller for the query information can be arranged according to the tactical template to obtain the response information, and the response information is helpful for constructing and supplementing the corresponding response information database. Compared with the existing method for maintaining the database by professional personnel, the method provided by the embodiment of the invention can automatically construct the response information database, so that on one hand, for the seller, the threshold of the seller for answering the problem is reduced, professional training is not needed, and the time and labor cost are reduced; on the other hand, for the buyer, the problem that when the seller is a non-professional seller, the database for replying the buyer problem is difficult to construct and maintain, so that the question of the buyer cannot be replied in time, and the experience of the user of the buyer is reduced is solved.
For the convenience of understanding, the method provided by the embodiment of the present invention will be described in detail below by taking as an example the establishment of the response information database based on the question of the inquiry sent by the buyer (e.g.: what color is the piece of clothing.
Fig. 7 is a block diagram showing a method of establishing a response information database based on response information according to an embodiment.
As shown in fig. 7, first, step 701: an intent attribute in the first client (i.e., buyer) sending the query message is identified.
Specifically, for a buyer's challenge question, a corresponding challenge intent is identified. For example: what color is the garment? The intent of the color is identified. Further, the first model shown in fig. 8 can be used for identification, which is specifically as follows:
(1) an input layer that inputs a query message "what color is the clothing? ". The input layer is used to divide the query message into a plurality of first text words, i.e., "this", "piece", "clothing", "having", "what", and "color"; and converting the divided first text words into a number form and outputting a plurality of first text words existing in the number form.
(2) And the word vector layer is used for inputting a plurality of first text words in a digital form, generating a word vector corresponding to each first text word according to the plurality of first text words in the digital form and outputting a plurality of word vectors.
(3) And the bidirectional circulation neural network layer comprises a BilSTM (Bi-directional Long Short-Term Memory) part and an attention mechanism part.
The BilSTM part processes a plurality of word vectors output by the word vector layer and determines a context vector corresponding to each first text word. Capturing a preset attention vector in the keyword information and the attention mechanism part according to the plurality of context vectors, and calculating a probability value corresponding to each first text word.
For example, "what color is the piece of clothing? The text words of 'clothes', 'color' are captured, and probability values corresponding to the text words are calculated.
And an attention mechanism part for giving a weight to each keyword information and giving more weight values to the keywords.
(4) And the output layer predicts the intention attribute in the query information according to the probability value and the weight value corresponding to each first text word output by the bidirectional recurrent neural network layer.
Next, step 702: and extracting an attribute value corresponding to the intention attribute from first response information replied by the second user terminal aiming at the inquiry information.
Specifically, for the buyer's question, the second user end (i.e. seller) replies with the first reply information to the question information.
For example: buyer: "what color is this garment? ". Seller: "parent, no red, also red-blue and white. "where" red-blue and white "may be attribute values corresponding to the intention attribute.
Further, the second model shown in fig. 9 may be used to extract attribute values, which is specifically shown as follows:
(1) the input layer inputs first response information "parent, no red, also red-blue and white". The input layer is used for dividing the first response information into a plurality of two text words, namely 'no', 'red', 'present', 'red blue' and 'white'; and converting the divided second text words into a number form and outputting a plurality of second text words existing in the number form.
(2) And the word vector layer is used for inputting a plurality of second text words in a digital form, generating a word vector corresponding to each second text word according to the plurality of second text words in the digital form and outputting a plurality of word vectors.
(3) And the bidirectional circulation neural network layer comprises a BilSTM (Bi-directional Long Short-Term Memory) part and an attention mechanism part.
The BilSTM part processes a plurality of word vectors output by the word vector layer and determines a context vector corresponding to each second text word. And capturing the keyword information and the preset attention vector in the attention mechanism part according to the plurality of context vectors, and calculating a probability value corresponding to each keyword information.
For example, "parent, without red, red-blue and white" can capture the keywords "red-blue" and "white", and calculate the probability values corresponding to the keywords.
And an attention mechanism part for giving a weight to each keyword information and giving more weight values to the keywords.
(4) And the CRF layer determines an attribute value corresponding to the intention attribute through a CRF model according to the probability value and the weight value corresponding to each second text word output by the bidirectional recurrent neural network layer. Here, the most common BIO tag algorithm may be used.
Further, where B denotes a beginning portion, I denotes an intermediate portion, and O denotes an unnecessary portion. The CRF layer can automatically learn the transition relationships between successive words, such as the probability of transitioning from B to I and B to O being inconsistent, so that for "red-blue" it is known that "red" is the beginning part and "yellow" is the middle part, together the complete "red-blue" part. Attribute values corresponding to the intention attributes can be derived from the first response information through the output of the CRF layer.
Next, in step 703, a second response message is generated using the linguistic template based on the attribute value and the intention attribute.
Specifically, the intention attribute and the attribute value are filled in according to a predefined dialect template, and a complete answer is generated, for example, the template is 'parent', 'intention attribute' has 'attribute value' to obtain 'parent', and the color has red, blue and white. "
Here, the reply content, i.e., the second reply information is generated using the intention attribute and the attribute value determined in steps 701 and 702. A conversational template is employed to obtain continuous, smooth and meaningful answer information. On the basis, in order to enrich the diversity of generated texts, a required pattern making grammar template can be screened from historical data, for example: identifying an intention attribute of each preset inquiry message in a plurality of preset inquiry messages; extracting an attribute value corresponding to the intention attribute from each response message corresponding to each preset inquiry message; taking each response message after the attribute value is extracted as candidate dialect messages; determining a phonetics template according to the candidate phonetics information.
In one example, the phonetics template may be updated according to the frequency of use of the phonetics template and/or the readability of the phonetics template.
Then, step 704, based on the plurality of second response messages, constructs a response message database so that the next inquiry message sent by the buyer can be replied according to the response message database.
Therefore, the method provided by the embodiment of the invention can arrange the inquiry information of the buyer and the response information replied by the seller according to the inquiry information according to the dialect template to obtain the response information, and the response information is beneficial to constructing and supplementing a corresponding response information database. Compared with the existing method for maintaining the database by professional personnel, the method provided by the embodiment of the invention can automatically construct the response information database, so that on one hand, for the seller, the threshold of the seller for answering the problem is reduced, professional training is not needed, and the time and labor cost are reduced; on the other hand, for the buyer, the problem that when the seller is a non-professional seller, the database for replying the buyer problem is difficult to construct and maintain, so that the question of the buyer cannot be replied in time, and the experience of the user of the buyer is reduced is solved.
Fig. 10 shows a block diagram of the structure of a response information generating apparatus according to an embodiment.
As shown in fig. 10, the apparatus 100 may include:
an identifying module 1001, configured to identify an intention attribute in a query message sent by a first user end;
an extracting module 1002, configured to extract an attribute value corresponding to the intention attribute from first response information replied by the second user terminal for the query information;
a processing module 1003 for generating the second response information by using the conversational template based on the attribute value and the intention attribute.
The identifying module 1001 in this embodiment of the present invention may be specifically configured to determine, according to a plurality of first text words in the query information, a context vector corresponding to each of the plurality of first text words; obtaining a probability value of each first text word based on the context vector and a preset attention vector; and determining the intention attribute in the inquiry information according to the probability value of each first text word and the preset weight value of each first text word.
Here, the probability value is used to represent the degree of importance of each first text word; the importance degree of each first text word is in a direct proportion relation with the preset weight value of each first text word.
Further, the identification module 1001 in this embodiment of the present invention may be specifically configured to calculate an initial weight value of each first text word according to the probability value of each first text word and a preset weight value;
selecting a key probability value with the probability value meeting a preset condition from the obtained probability values of the plurality of first text words, and distributing a key weight value higher than a preset weight value to the key probability value; according to the key probability value and the key weight value, recalculating the weight value of the key text word corresponding to the key probability value; wherein the plurality of first text words comprise key text words; and determining the intention attribute in the inquiry information according to the initial weight value and the weight value of the key text word.
Or, the identification module 1001 may be specifically configured to select, from the obtained probability values of the plurality of first text words, a key probability value with a probability value satisfying a preset condition and a common probability value with a probability value not satisfying the preset condition; calculating a weight value corresponding to the key probability value according to the key probability value and a preset key weight value; calculating a weight value corresponding to the common probability value according to the common probability value and a preset common weight value; wherein the preset key weight value is greater than the common weight value; determining an intention attribute in the query information according to the weight value corresponding to the key probability value and the weight value corresponding to the common probability value; the preset weight value of each first text word comprises a preset key weight value and a preset common weight value.
The extracting module 1002 in this embodiment of the present invention may be specifically configured to determine, according to a plurality of second text words in the first response information, a context vector corresponding to each of the plurality of second text words; and extracting attribute values corresponding to the intention attributes through a label labeling algorithm according to the context vector corresponding to each second text word.
Here, the label labeling algorithm includes: a sequence labeling algorithm and/or a BIO label algorithm; the sequence labeling algorithm comprises an original labeling algorithm and/or a combined labeling algorithm.
In one example, the extracting module 1002 in the embodiment of the present invention may be specifically configured to extract an attribute value corresponding to the intention attribute from the first response information and the product description information related to the query information.
In addition, the apparatus 100 may further include a template generating module 1004 for identifying an intention attribute of each preset query message in the plurality of preset query messages; extracting an attribute value corresponding to the intention attribute from each response message corresponding to each preset inquiry message; taking each response message after the attribute value is extracted as candidate dialect messages; determining a phonetics template according to the candidate phonetics information.
The template generating module 1004 may be further configured to update the conversational template according to a frequency of using the conversational template and/or readability of the conversational template.
Fig. 11 shows a block diagram of a structure of an apparatus for constructing a response information database according to an embodiment.
As shown in fig. 11, the apparatus 110 may include:
an obtaining module 1101, configured to obtain a plurality of query messages sent by a first user and a plurality of first response messages replied by a second user for the plurality of query messages;
generating means 1102 for generating a plurality of second response information to construct a response information database by using the response information generating means shown in fig. 10.
Fig. 12 is a block diagram showing the configuration of an apparatus for acquiring response information according to an embodiment.
As shown in fig. 12, the apparatus 120 may include:
an identifying module 1201, configured to identify a target intention attribute in query information sent by a first user;
a processing module 1202 for taking the second response information corresponding to the target intention attribute as the response information of the inquiry information from the response information database constructed by the constructed response information database means shown in fig. 11;
the transceiver module 1203 is configured to send a response message to the first user equipment.
Additionally, the apparatus 120 can include an update module 1204 for determining a product associated with the query information; obtaining a product attribute value related to a product from a product information database; and updating the second response information according to the product attribute value.
FIG. 13 illustrates a schematic structural diagram of a computing device, according to one embodiment.
As shown in fig. 13, a block diagram of an exemplary hardware architecture of a computing device capable of implementing the anomaly point detection method and apparatus according to the embodiments of the present invention.
The apparatus may include a processor 1301 and a memory 1302 storing computer program instructions.
Specifically, the processor 1301 may include a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.
The processor 1301 realizes any one of the response information generation method, the method of constructing the database, and the method of acquiring the response information in the above embodiments by reading and executing the computer program instructions stored in the memory 1302.
The transceiver 1303 is mainly used to implement each apparatus in the embodiment of the present invention or communicate with other devices.
In one example, the device may also include a bus 1304. As shown in fig. 13, the processor 1301, the memory 1302, and the transceiver 1303 are connected via a bus 1304 to complete communication therebetween.
Embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, which, when the computer program is executed in a computer, causes the computer to perform the steps of a method for generating response information, a method for constructing a database, and a method for acquiring response information according to embodiments of the present invention.
It is to be understood that the invention is not limited to the particular arrangements and instrumentality described in the above embodiments and shown in the drawings. For convenience and brevity of description, detailed description of a known method is omitted here, and for the specific working processes of the system, the module and the unit described above, reference may be made to corresponding processes in the foregoing method embodiments, which are not described herein again.
It will be apparent to those skilled in the art that the method procedures of the present invention are not limited to the specific steps described and illustrated, and that various changes, modifications and additions, or equivalent substitutions and changes in the sequence of steps within the technical scope of the present invention are possible within the technical scope of the present invention as those skilled in the art can appreciate the spirit of the present invention.
Claims (20)
1. A response message generation method includes:
identifying an intention attribute in the query information sent by the first user terminal;
extracting an attribute value corresponding to the intention attribute from first response information replied by a second user terminal aiming at the inquiry information;
generating second response information using a conversational template based on the attribute value and the intent attribute.
2. The method of claim 1, wherein the identifying an intention attribute in the first user sending the query message comprises:
determining a context vector corresponding to each of a plurality of first text words in the query information according to the plurality of first text words;
obtaining a probability value of each first text word based on the context vector and a preset attention vector;
and determining the intention attribute in the inquiry information according to the probability value of each first text word and a preset weight value of each first text word.
3. The method of claim 2, wherein the probability value is used to characterize the degree of importance of the each first text word; wherein the content of the first and second substances,
the importance degree of each first text word is in a direct proportion relation with the preset weight value of each first text word.
4. The method of claim 2 or 3, wherein determining the intent attribute in the query message according to the probability value of each first text word and a preset weight value for each first text word comprises:
calculating an initial weight value of each first text word according to the probability value of each first text word and the preset weight value;
selecting a key probability value with the probability value meeting a preset condition from the obtained probability values of the first text words, and assigning a key weight value higher than the preset weight value to the key probability value;
according to the key probability value and the key weight value, recalculating the weight value of the key text word corresponding to the key probability value; wherein the plurality of first text words comprise key text words;
and determining the intention attribute in the inquiry information according to the initial weight value and the weight value of the key text word.
5. The method of claim 3, wherein determining the intent attribute in the query message based on the probability value for each first text word and a preset weight value for each first text word comprises:
selecting a key probability value with the probability value meeting a preset condition and a common probability value without meeting the preset condition from the obtained probability values of the plurality of first text words;
calculating a weight value corresponding to the key probability value according to the key probability value and a preset key weight value; calculating a weight value corresponding to the common probability value according to the common probability value and a preset common weight value; wherein the preset key weight value is greater than the common weight value;
determining an intention attribute in the inquiry information according to the weight value corresponding to the key probability value and the weight value corresponding to the common probability value;
wherein the preset weight value of each first text word comprises the preset key weight value and the preset common weight value.
6. The method of claim 1, wherein the extracting an attribute value corresponding to the intention attribute from the first reply information replied to by the second user terminal for the query information comprises:
determining a context vector corresponding to each second text word in the plurality of second text words according to the plurality of second text words in the first response information;
and extracting an attribute value corresponding to the intention attribute through a label labeling algorithm according to the context vector corresponding to each second text word.
7. The method of claim 6, wherein the label labeling algorithm comprises: a sequence labeling algorithm and/or a BIO label algorithm; wherein the content of the first and second substances,
the sequence labeling algorithm comprises an original labeling algorithm and/or a joint labeling algorithm.
8. The method of claim 1 or 6, wherein the extracting an attribute value corresponding to the intention attribute from the first reply information replied by the second user terminal for the query information comprises:
and extracting an attribute value corresponding to the intention attribute from the first response information and the product description information related to the inquiry information.
9. The method of claim 1, further comprising:
identifying an intention attribute of each preset inquiry message in a plurality of preset inquiry messages;
extracting an attribute value corresponding to the intention attribute from each piece of response information corresponding to each piece of preset inquiry information;
taking each response message after the attribute value is extracted as candidate dialect messages;
determining the phonetics template according to the candidate phonetics information.
10. The method of claim 1, further comprising:
updating the phonetics template according to a frequency of using the phonetics template and/or readability of the phonetics template.
11. A method of constructing a database, comprising:
acquiring a plurality of pieces of inquiry information sent by a first user terminal and a plurality of pieces of first response information replied by a second user terminal aiming at the plurality of inquiry information;
a plurality of second response messages are generated to construct a response message database using the response message generating method according to any one of claims 1 to 10.
12. A method for obtaining response information, comprising:
identifying a target intention attribute in inquiry information sent by a first user terminal;
from the response information database according to claim 11, second response information corresponding to the target intention attribute is made response information to the inquiry information;
and sending the response information to the first user terminal.
13. The method of claim 12, further comprising:
determining a product associated with the query information;
obtaining a product attribute value related to the product from a product information database;
and updating the second response information according to the product attribute value.
14. A response information generating apparatus, comprising:
the identification module is used for identifying the intention attribute in the inquiry information sent by the first user terminal;
the extraction module is used for extracting an attribute value corresponding to the intention attribute from first response information replied by a second user terminal aiming at the inquiry information;
a processing module for generating second response information using a conversational template based on the attribute value and the intent attribute.
15. An apparatus for constructing a database, comprising:
the system comprises an acquisition module, a processing module and a processing module, wherein the acquisition module is used for acquiring a plurality of pieces of inquiry information sent by a first user terminal and a plurality of pieces of first response information replied by a second user terminal aiming at the plurality of inquiry information;
generating means for generating a plurality of second response messages to construct a response message database, using the response message generating means according to claim 14.
16. An apparatus for acquiring response information, comprising:
the identification module is used for identifying the target intention attribute in the inquiry information sent by the first user terminal;
a processing module for using second response information corresponding to the target intention attribute as response information of the query information from the response information database constructed by the database constructing apparatus according to claim 11;
and the transceiver module is used for sending the response information to the first user terminal.
17. A system for building a database, comprising:
the input module is used for splitting inquiry information sent by a first user end and determining at least one first text word;
a word vector module, configured to generate, according to the at least one first text word, a word vector corresponding to each first text word in the at least one first text word;
a bidirectional recurrent neural network module, configured to determine, based on the word vector corresponding to each first text word, a context vector corresponding to the word vector corresponding to each first text word; calculating a probability value corresponding to each first text word according to the context vector and a preset attention vector; and configuring a weight value of each first text word;
and the output module is used for determining the intention attribute in the inquiry information according to the probability value corresponding to each first text word and the weight value of each first text word.
18. A system for obtaining response information, comprising:
the input module is used for splitting the first response information sent by the second user end and determining at least one second text word;
a word vector module, configured to generate, according to the at least one second text word, a word vector corresponding to each second word in the at least one second text word;
a bidirectional recurrent neural network module, configured to determine, based on the word vector corresponding to each second text word, a context vector corresponding to the word vector corresponding to each second text word; calculating a probability value corresponding to each second text word according to the context vector and a preset attention vector; and configuring a weight value of each second text word;
and the label labeling module is used for labeling the initial part and the middle part of the attribute value in the second text word according to the probability value corresponding to each second text word and the weight value of each second text word so as to determine the attribute value in the second response message.
19. A computing device, wherein the device comprises at least one processor and a memory, the memory being for storing computer program instructions, the processor being for executing the program of the memory to control the computing device to implement the method of any of claims 1-10, 11 or 12-13.
20. A computer-readable storage medium, on which a computer program is stored, wherein the computer program, if executed in a computer, causes the computer to perform the method of any of claims 1-10, 11 or 12-13.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910649366.4A CN112241448A (en) | 2019-07-18 | 2019-07-18 | Response information generation method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910649366.4A CN112241448A (en) | 2019-07-18 | 2019-07-18 | Response information generation method, device, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112241448A true CN112241448A (en) | 2021-01-19 |
Family
ID=74167904
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910649366.4A Pending CN112241448A (en) | 2019-07-18 | 2019-07-18 | Response information generation method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112241448A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114117021A (en) * | 2022-01-24 | 2022-03-01 | 北京数智新天信息技术咨询有限公司 | Method and device for determining reply content and electronic equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6278996B1 (en) * | 1997-03-31 | 2001-08-21 | Brightware, Inc. | System and method for message process and response |
| WO2018149326A1 (en) * | 2017-02-16 | 2018-08-23 | 阿里巴巴集团控股有限公司 | Natural language question answering method and apparatus, and server |
| CN108897867A (en) * | 2018-06-29 | 2018-11-27 | 北京百度网讯科技有限公司 | For the data processing method of knowledge question, device, server and medium |
| CN109285030A (en) * | 2018-08-29 | 2019-01-29 | 深圳壹账通智能科技有限公司 | Products Show method, apparatus, terminal and computer readable storage medium |
| US20190176820A1 (en) * | 2017-12-13 | 2019-06-13 | Humanising Autonomy Limited | Systems and methods for predicting pedestrian intent |
-
2019
- 2019-07-18 CN CN201910649366.4A patent/CN112241448A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6278996B1 (en) * | 1997-03-31 | 2001-08-21 | Brightware, Inc. | System and method for message process and response |
| WO2018149326A1 (en) * | 2017-02-16 | 2018-08-23 | 阿里巴巴集团控股有限公司 | Natural language question answering method and apparatus, and server |
| US20190176820A1 (en) * | 2017-12-13 | 2019-06-13 | Humanising Autonomy Limited | Systems and methods for predicting pedestrian intent |
| CN108897867A (en) * | 2018-06-29 | 2018-11-27 | 北京百度网讯科技有限公司 | For the data processing method of knowledge question, device, server and medium |
| CN109285030A (en) * | 2018-08-29 | 2019-01-29 | 深圳壹账通智能科技有限公司 | Products Show method, apparatus, terminal and computer readable storage medium |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114117021A (en) * | 2022-01-24 | 2022-03-01 | 北京数智新天信息技术咨询有限公司 | Method and device for determining reply content and electronic equipment |
| CN114117021B (en) * | 2022-01-24 | 2022-04-01 | 北京数智新天信息技术咨询有限公司 | Method and device for determining reply content and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108595494B (en) | Method and device for acquiring reply information | |
| CN104951428B (en) | User's intension recognizing method and device | |
| US10460029B2 (en) | Reply information recommendation method and apparatus | |
| WO2018040944A1 (en) | System, method, and device for identifying malicious address/malicious purchase order | |
| CN104933084A (en) | Method, apparatus and device for acquiring answer information | |
| CN104951456A (en) | Method, device and equipment used for obtaining answer information | |
| CN107592255B (en) | Information display method and equipment | |
| CN111291158B (en) | Information query method and device, electronic equipment and storage medium | |
| CN110347786B (en) | Semantic model tuning method and system | |
| CN109215630A (en) | Real-time speech recognition method, apparatus, equipment and storage medium | |
| CN106294676B (en) | A kind of data retrieval method of ecommerce government system | |
| CN112966081A (en) | Method, device, equipment and storage medium for processing question and answer information | |
| CN111368040B (en) | Dialogue processing method, model training method and related equipment | |
| CN111680165A (en) | Information matching method and device, readable storage medium and electronic equipment | |
| CN114186024A (en) | Recommendation method and device | |
| CN112241448A (en) | Response information generation method, device, equipment and storage medium | |
| CN111242713A (en) | Information pushing method and device | |
| US20220198487A1 (en) | Method and device for processing user interaction information | |
| CN116775815B (en) | Dialogue data processing method and device, electronic equipment and storage medium | |
| CN115660814A (en) | Risk prediction method and device, computer readable storage medium and electronic equipment | |
| JP2023554210A (en) | Sort model training method and apparatus for intelligent recommendation, intelligent recommendation method and apparatus, electronic equipment, storage medium, and computer program | |
| CN114297380A (en) | Data processing method, device, equipment and storage medium | |
| CN109074552A (en) | Knowledge based figure enhances contact card | |
| CN110334177B (en) | Semantic similarity model training and semantic similarity recognition methods and devices and electronic equipment | |
| CN111507773A (en) | Coupon management method, device 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 |