CN111353024B - Method, device and system for generating comment text - Google Patents
Method, device and system for generating comment text Download PDFInfo
- Publication number
- CN111353024B CN111353024B CN201811474376.0A CN201811474376A CN111353024B CN 111353024 B CN111353024 B CN 111353024B CN 201811474376 A CN201811474376 A CN 201811474376A CN 111353024 B CN111353024 B CN 111353024B
- Authority
- CN
- China
- Prior art keywords
- text
- comment
- user
- review
- hidden state
- 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
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
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06Q30/015—Providing customer assistance, e.g. assisting a customer within a business location or via helpdesk
- G06Q30/016—After-sales
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Business, Economics & Management (AREA)
- Artificial Intelligence (AREA)
- Marketing (AREA)
- Biomedical Technology (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Accounting & Taxation (AREA)
- Economics (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Human Computer Interaction (AREA)
- Machine Translation (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a method, a device and a system for generating a review text. Wherein, the method comprises the following steps: the method comprises the following steps of constructing a back evaluation model based on a preset corpus, wherein the preset corpus comprises at least one of the following: a user history comment text, a merchant history comment-back text and a commodity description information text; receiving a current review text of a user; and generating a comment-back text corresponding to the current comment text of the user by using the comment-back model. The method and the device solve the technical problem that in the related art, when the comment content of the user is replied, the attention of the user to merchants is reduced due to the fact that the feedback information is often inconsistent with the comment content of the user.
Description
Technical Field
The invention relates to the technical field of text processing, in particular to a method, a device and a system for generating a review text.
Background
In the correlation technique, the trade company can show own commodity and sell commodity for the user in the online shop, the user can purchase commodity through the online shop, this kind of transaction mode, the user can't experience the quality of commodity on the spot, also can't judge the quality of commodity, can only be after purchasing, take the goods after, after the commodity of trying, just can know the instruction of commodity, the user can make certain evaluation to commodity after taking the commodity, different users take the commodity of different quality, the evaluation of making is also different. Under the circumstance, a merchant often needs to reply to the evaluation content of the user, for example, thank you for purchase, welcome the next time, and timely feedback is given for the evaluation with quality problems, but when replying to the comment content of the user, text reply content is often preset, and the reply content is easily irrelevant to the comment content of the user, for example, in current application software such as panning and the like, the merchant can preset a plurality of reply templates, and after the user sends out a comment, if corresponding characters are detected, the templates can be directly sent to the user, so that the comment text is replied in a manner, although certain feedback can be given to the user, because the feedback information is often inconsistent with the comment content of the user, the attention of the user to the merchant is reduced, and the use interest of the user is reduced.
In view of the above problems, no effective solution has been proposed.
Disclosure of Invention
The embodiment of the invention provides a method, a device and a system for generating a comment-back text, which are used for at least solving the technical problem that in the related art, when comment content of a user is replied, attention of the user to a merchant is reduced because feedback information is often inconsistent with the comment content of the user.
According to an aspect of the embodiments of the present invention, there is provided a method for generating a review text, including: the method comprises the following steps of constructing a back evaluation model based on a preset corpus, wherein the preset corpus comprises at least one of the following: a user history comment text, a merchant history comment-back text and a commodity description information text; receiving a current comment text of a user; and generating a comment-back text corresponding to the current comment text of the user by using the comment-back model.
According to another aspect of the embodiments of the present invention, there is also provided a comment text generating apparatus including: the building unit is used for building a review model based on a preset corpus, wherein the preset corpus comprises at least one of the following: a user history comment text, a merchant history comment-back text and a commodity description information text; the receiving unit is used for receiving the current comment text of the user; and the generating unit is used for generating the comment-back text corresponding to the current comment text of the user by using the comment-back model.
According to another aspect of the embodiment of the present invention, there is also provided a storage medium, where the storage medium includes a stored program, and when the program runs, the apparatus on which the storage medium is located is controlled to execute the method for generating the comment text described in any one of the above.
According to another aspect of the embodiments of the present invention, there is also provided a system for generating a review text, including: a processor; and a memory coupled to the processor for providing instructions to the processor for processing the following processing steps: step 1, establishing a review model based on a preset corpus, wherein the preset corpus comprises at least one of the following: a user history comment text, a merchant history comment-back text and a commodity description information text; step 2, receiving a current comment text of a user; and 3, generating a comment-back text corresponding to the current comment text of the user by using the comment-back model.
In the embodiment of the present invention, a review model is constructed based on a preset corpus, wherein the preset corpus includes at least one of the following: the method comprises the steps of receiving a user historical comment text, a merchant historical review text and a commodity description information text, receiving a user current comment text, and generating a review text corresponding to the user current comment text by using a review model. In the embodiment, various reply contents with rich information and pertinence can be automatically generated by combining the constructed review model with the comment text of the user and the description information of the commodity, the reply contents are adaptive to the text contents of the current comment of the user, so that the user can more deeply know the commodity information corresponding to the purchased commodity and the review contents, the attention of the user to merchants is improved, and the technical problem that the attention of the user to the merchants is reduced due to the fact that the feedback information is often inconsistent with the contents of the comment of the user when the comment contents of the user are replied in the related technology is solved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention and do not constitute a limitation of the invention. In the drawings:
fig. 1 shows a hardware configuration block diagram of a computer terminal (or mobile device) for implementing a method of generating a review text;
FIG. 2 is a schematic diagram of a network terminal for implementing the method for generating the review text;
fig. 3 is a flowchart of a method for generating a comment text according to a first embodiment of the present invention;
FIG. 4 is a flowchart of an alternative method for building a comment model using corpus to be trained according to an embodiment of the present invention;
FIG. 5 is a block diagram of an alternative review model according to an embodiment of the invention;
FIG. 6 is a schematic diagram of an alternative review text generation apparatus according to an embodiment of the present invention;
FIG. 7 is a schematic diagram of an automatic review system according to an embodiment of the present invention;
fig. 8 is a block diagram of a computer terminal according to an embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
First, some terms or terms appearing in the description of the embodiments of the present application are applicable to the following explanations:
RNN, current Neural Network, recurrent Neural Network.
A Seq2Seq framework, which is a variation of RNN and is also called an Encode-Decoder model, encodes input data into a context vector to obtain a context vector, and then decodes the context vector by using another RNN network to obtain a decoding sequence. In the application, the Seq2Seq framework is applied to parsing of the review text, and the automatic review text is output for a parsing sequence.
Example 1
There is also provided, in accordance with an embodiment of the present invention, an embodiment of a method for generating a review text, it should be noted that the steps illustrated in the flowchart of the accompanying drawings may be performed in a computer system such as a set of computer-executable instructions, and that, although a logical order is illustrated in the flowchart, in some cases, the steps illustrated or described may be performed in an order different than here.
The method provided by the first embodiment of the present application may be executed in a mobile terminal, a computer terminal, or a similar computing device. Fig. 1 shows a hardware configuration block diagram of a computer terminal (or mobile device) for implementing the method of generating a review text. As shown in fig. 1, the computer terminal 10 (or mobile device 10) may include one or more (shown as 102a, 102b, … …,102 n) processors 102 (the processors 102 may include, but are not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission device 106 for communication functions. Besides, the method can also comprise the following steps: a display, an input/output interface (I/O interface), a Universal Serial Bus (USB) port (which may be included as one of the ports of the I/O interface), a network interface, a power source, and/or a camera. It will be understood by those skilled in the art that the structure shown in fig. 1 is only an illustration and is not intended to limit the structure of the electronic device. For example, the computer terminal 10 may also include more or fewer components than shown in FIG. 1, or have a different configuration than shown in FIG. 1.
It should be noted that the one or more processors 102 and/or other data processing circuitry described above may be referred to generally herein as "data processing circuitry". The data processing circuitry may be embodied in whole or in part in software, hardware, firmware, or any combination thereof. Further, the data processing circuit may be a single stand-alone processing module, or incorporated in whole or in part into any of the other elements in the computer terminal 10 (or mobile device). As referred to in the embodiments of the application, the data processing circuit acts as a processor control (e.g. selection of a variable resistance termination path connected to the interface).
The memory 104 may be used to store software programs and modules of application software, such as program instructions/data storage devices corresponding to the method for generating the comment text in the embodiment of the present invention, and the processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, implements the method for generating the comment text described above. The memory 104 may include high speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include memory located remotely from the processor 102, which may be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The transmission device 106 is used to receive or transmit data via a network. Specific examples of the network described above may include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission device 106 includes a Network adapter (NIC) that can be connected to other Network devices through a base station to communicate with the internet. In one example, the transmission device 106 can be a Radio Frequency (RF) module, which is used to communicate with the internet in a wireless manner.
The display may be, for example, a touch screen type Liquid Crystal Display (LCD) that may enable a user to interact with a user interface of the computer terminal 10 (or mobile device).
Fig. 1 shows a block diagram of a hardware structure, which may be taken as an exemplary block diagram of not only the computer terminal 10 (or the mobile device) but also the server, and in an alternative embodiment, fig. 2 shows an embodiment of using the computer terminal 10 (or the mobile device) shown in fig. 1 as a sending end/receiving end in a block diagram. As shown in FIG. 2, the computer terminal 10 (or mobile device) may be connected or electronically connected to one or more servers 20 (e.g., a security server, a data storage server, etc. in an alternative embodiment, the computer terminal 10 (or mobile device) may be any mobile computing device, etc. the data network connection may be a local area network connection, a wide area network connection, an Internet connection, or other type of data network connection the computer terminal 10 (or mobile device) may execute to connect to a network service executed by a server (e.g., a data storage server) or a group of servers.
Under the operating environment, the application provides a method for generating the review text as shown in fig. 3. Fig. 3 is a flowchart of a method for generating a comment text according to a first embodiment of the present invention. As shown in fig. 3, the generating method includes the following steps:
step S302, a review model is built based on a preset corpus, wherein the preset corpus comprises at least one of the following: a user history comment text, a merchant history review text and a commodity description information text;
step 302, receiving a current comment text of a user;
and step S306, generating a review text corresponding to the current review text of the user by using the review model.
Through the steps, the comment model is built based on the preset corpus, wherein the preset corpus comprises at least one of the following: the method comprises the steps of receiving a user historical comment text, a merchant historical comment-back text and a commodity description information text, receiving a user current comment text, and generating a comment-back text corresponding to the user current comment text by using a comment-back model. In the embodiment, various reply contents with rich information and pertinence can be automatically generated by combining the constructed review model with the user review text and the description information of the commodity, the reply contents are adaptive to the text contents of the current reviews of the user, so that the user can more deeply know the information of the commodity and the review information of the merchant, the attention of the user to the merchant is improved, and the technical problem that the attention of the user to the merchant is reduced due to the fact that the feedback information is often inconsistent with the contents of the user reviews when the review contents of the user are replied in the related technology is solved.
The above steps will be explained below.
According to the embodiment of the invention, the reply text can be automatically generated based on the current comment content of the user, and the reply content is more targeted.
The method and the device can be applied to E-commerce scenes, and the E-commerce scenes comprise at least one of the following: e-commerce marketing scene in television shopping and E-commerce marketing scene in internet shopping.
The embodiment of the invention can analyze the text based on the Seq2Seq framework so as to output the corresponding comment-back text.
In the present invention, three types of text are involved, including: the method comprises a comment text, a comment-back text and a commodity description information text, wherein the comment text can be understood as a text generated by a user for commenting commodities, generally speaking, after the user purchases commodities through online application, a certain comment is often published based on the purchased commodities, and the comment content may include: text, pictures, voice, video, etc.; for the comment back text, it can be understood as the text generated by the merchant for the content of each user's comment text reply, which would generally include: thank you, intimate and nickname languages, reply to questions in the comment text, reply to praise languages in the comment text, and the like; the description information text of the product can be understood as description information of the product, including information describing the material, the type, the comfort level, the quality, the size (including the volume, the length, the width, etc.) of the product, etc. generally, the user will make a comment on the product summary and the feeling of using the product with respect to the purchased product, and the merchant will also reply the description content of the product related to the type, the material, etc. of the product.
The comment texts related to the application comprise two types, wherein the first type is a user history comment text, the user indicates a plurality of unspecified history users, each comment text of the history users is analyzed, and then a review model is constructed; the second type is a current comment text of a user, the user indicates the user who gives the comment text at present, the comment text of the user can be analyzed in real time by using a comment-back model, and then a corresponding comment-back text is generated. That is, the time of the user history comment text is before the time point of the user currently commenting on the text, and the user in the user history comment text is also different from the user in the user currently commenting on the text.
Step S302, a review model is constructed based on a preset corpus, wherein the preset corpus comprises at least one of the following: a user history comment text, a merchant history comment-back text and a commodity description information text;
in an embodiment of the present invention, the preset corpus may include at least one of the following: the system comprises a user historical comment text, a merchant historical review text and a commodity description information text. The user history comment text can be a comment text of the user to the commodity in the history process, and the merchant history comment-back text can be comment-back content of each user comment text of the merchant in the history process. The user history comment text, the merchant history review text and the commodity description information text can be obtained from a server or a database where a comment log for recording the content of a comment area is located, and the type and the specific model of the specific server and the specific model of the database are not limited.
In an optional example of this embodiment, the constructing the review model based on the preset corpus includes: preprocessing a preset corpus, and establishing an association relation between a user historical comment text and a merchant historical review text through commodity attribute information to obtain a corpus to be trained; and adopting the corpus to be trained to construct a review model. In the embodiment of the invention, after the preset corpus is preprocessed, the commodity attribute information in a comment conversation can be obtained, wherein the commodity attribute information is the commodity attribute information which is commonly contained in comment contents and comment back contents of the user and the merchant about commodities in the transaction behavior.
In the invention, the preprocessing mode can be preprocessing the text corpus, for example, preprocessing the text word segmentation, special symbol and noise removal and the like on the user history comment text and the merchant history comment text, after the preprocessing, extracting commodity attribute words (such as color difference, pure cotton and size) by performing data analysis on the user history comment text, and using the commodity attribute words for screening and filtering the merchant reply text content, namely, the user history comment text and the merchant history comment text at least refer to one commodity attribute content together, so that the commodity attribute information in the comment conversation can be screened out during the screening. For example, in a commodity transaction, a user reviews: "there is a place to break, sew up oneself, just not so thick just can be suitable for when the day is not cold in this shop to buy two trousers this trousers how the yards are the same, but short a big section. Ask the shop owner that he says that the code is slightly smaller and cannot be slightly inclined to the same size on the chip! Is too late to use when calculating, and is fit with a bar! "and the merchant reviews: "really is sorry, the clothes of the small shop can not achieve the effect expected by the relatives, and people can disapprove. The single-layer flannel is medium-thickness, the three layers of sandwich cotton are the thickest woolen cloth, and the thickened portions of the sandwich cotton can be seen if the sandwich cotton is thin. And in order to guarantee the client right, the small shop gives a premium additionally, and if the premium is not suitable, the game can be refunded. If there are problems, please contact us in time, and certainly provide satisfied service for you, thank you ". After preprocessing, the commodity attribute information of 'not so thick' in the user comment and 'medium thickness in the merchant review', the 'thickest' in the three-layer cotton inclusion is corresponding to can be obtained.
The embodiment indicates that at least one commodity attribute information between two texts can be obtained after the user history comment text and the merchant history comment-back text are preprocessed, in the application, after the commodity attribute information corresponding to each comment conversation is obtained by preprocessing each comment conversation in this way, then the association relation can be established between the user history comment text and the merchant history comment-back text through the commodity attribute information, and the linguistic data to be trained is obtained.
In the embodiment of the present application, after obtaining the corpus to be trained, the corpus to be trained can be used to construct the review model,
fig. 4 is a flowchart of an alternative method for constructing a review model using corpus to be trained according to an embodiment of the present invention, as shown in fig. 4, the method includes:
step S401, each word vector in the historical comment text of the user is coded into a first hidden state corresponding to a time step, and a first coding result is obtained;
step S403, encoding the word vector corresponding to each key value pair in the commodity description information text into a second hidden state corresponding to the time step to obtain a second encoding result;
step S405, according to the first coding result, the second coding result and the obtained third hidden state, obtaining the word vector corresponding to each generation step, and constructing a review model, wherein the third hidden state is the hidden state corresponding to the word vector obtained by prediction in the previous step of each generation step.
Through the steps, each word vector in the historical comment text of the user can be coded into a first hidden state corresponding to a time step to obtain a first coding result, then the word vector corresponding to each key value in the commodity description information text is coded into a second hidden state corresponding to the time step to obtain a second coding result, finally according to the first coding result, the second coding result and the obtained third hidden state, the word vector corresponding to each generation step is obtained, and a comment model is constructed.
The review model in the embodiment of the invention can realize automatic review text review aiming at the current review text of the user, has stronger pertinence and effectiveness, and leads merchants and users to be more satisfied.
Optionally, for step S401, each word vector in the user history comment text is encoded into a first hidden state corresponding to the time step, so as to obtain a first encoding result. Optionally, in the embodiment of the present invention, two layers of bidirectional recurrent neural networks (BiRNN) with Gate Recurrent Units (GRU) may be used to encode each word vector, in the present invention, each word vector may be input as a sequence, the word vector is input into the bidirectional recurrent neural networks, and the first implicit state is obtained by outputting the word vector.
Alternatively, for step S403, the word vector corresponding to each key-value pair in the commodity description information text is encoded into the second hidden state of the corresponding time step, so as to obtain a second encoding result. The step can be that after the commodity description information text is preprocessed, the word vector corresponding to each obtained key value pair is coded, and the coding obtains the hidden state corresponding to the time step. Similar to step S401, in the embodiment of the present invention, a two-layer bidirectional recurrent neural network (BiRNN) with a Gate Recurrent Unit (GRU) may also be used to encode the word vector corresponding to each key value pair in the commodity description information text, and in the present invention, the word vector corresponding to each key value pair may be input as a sequence, and the word vector is input into the bidirectional recurrent neural network, and output to obtain the second hidden state.
In an embodiment of the present invention, the key-value pair may indicate information of the article, for example, the key-value pair indicates a pair of data of "attribute-value".
Alternatively, for step S405, obtaining the word vector corresponding to each generating step according to the first encoding result, the second encoding result, and the third implicit state includes: generating a user history comment text vector based on the third hidden state of each generation step and the first hidden state of the corresponding time step; generating a commodity description information text vector based on the third hidden state of each generation step, the second hidden state of the corresponding time step and the user history comment text vector; fusing a user history comment text vector and a commodity description information text vector into a middle vector; and generating a word vector corresponding to each generation step based on the third implicit state and the intermediate vector of each generation step.
In this embodiment of the present invention, the third hidden state is a hidden state corresponding to a word vector obtained through prediction in a previous step of each generation step, that is, before each generation step, the third hidden state may be obtained through decoding the encoded word vector.
After the first hidden state, the second hidden state and the third hidden state are obtained, a user history comment text vector and a commodity description information text vector can be generated. When generating the text vector of the user historical text comment, the text vector of the user historical text comment may be generated based on the first hidden state and the third hidden state generated in step S401; and generating a commodity description information text vector based on the second hidden state and the third hidden state generated in the step S403 and the user history text comment text vector.
The above embodiment illustrates how to generate the text vector of the user history comment and the text vector of the product description information, and then the text vector of the user history comment and the text vector of the product description information may be fused into an intermediate vector, and a word vector corresponding to each generation step may be generated based on the third hidden state and the intermediate vector of each generation step.
Fig. 5 is a schematic diagram of a framework of an optional review model according to an embodiment of the present invention, as shown in fig. 5, which may include a review document encoder 51, a reply text decoder 52, a product description encoder 53, a door multi-mode unit 54, and a copy module 55, wherein each module has the following purposes:
a comment file encoder 51 for encoding each word vector xi in the input user history comment text as a hidden state corresponding to the time step
(corresponding to the first implicit state in the above-described embodiment). Two-layer bidirectional recurrent neural network (BiRNN) with Gate Recurrent Unit (GRU) is used for coding.
A commodity description encoder 53 for encoding a word vector (t) corresponding to each key value pair (e.g., attribute-value) of the input commodity description information k ，z k ) Encoding as implicit states corresponding to time steps
(corresponding to the second implicit state in the above embodiment), similarly to the comment text encoder, two layers of BiRNN with GRU are used for encoding as well.
The reply text decoder 52, once the user history comment text and the item description information are encoded, a two-layer RNN decoder with GRUs can be used to generate the reply text content. At each generation step, the reply text decoder 52 receives the word vector y predicted at the previous step i And generates implicit states
(corresponds to the third implicit state in the above-described embodiment).
Implicit states generated based on comment file encoder 51
And generated by the reply text decoder 52>
Constructing an attention machine system to generate a text vector->
(corresponding to the user history comment text vector in the above-described embodiment). Implicit status based on the generation of a product description encoder 53>
And generated by the reply text decoder 52>
And a context vector +>
Constructing an attention machine system and generating a commodity description vector->
(corresponding to the item description information text vector in the above embodiment).
A gate multi-mode unit 54 for generating a text vector
And a product description vector->
Fused into an intermediate vector c i 。
The copy module 55 may improve the coverage of the review text content by using part of the content from the user history review text and the product description information when the review text is generated. Each step i, is based on the implicit state of the decoder at the input
And intermediate vector c i Here, theTwo score functions are involved to respectively reference contents from user historical comment texts and commodity description information, and the word vector y of the current step is generated by comprehensive consideration i (corresponding to the generation of the word vector corresponding to each generation step based on the third implicit state and the intermediate vector for each generation step in the above-described embodiments).
Through the embodiment, when the review model is generated and trained, the word vector can be analyzed by using the historical review text of the user, the historical review text of the merchant and the commodity description information as input information, and the word vector describing the current step is generated by generating the hidden state.
And step S304, generating a review text corresponding to the current review text of the user by using the review model.
The step S304 indicates that the review text can be automatically generated by using the review model, and the automation degree and the reply content are more targeted.
Optionally, after the comment-back text corresponding to the current comment text of the user is generated by using the comment-back model, the method further includes: and in the automatic reply mode, returning the comment text to a target user submitting the current comment text of the user. After a certain user sends the comment text, the user is determined to be the target user, then the comment-back text is generated by using the comment-back model and is sent to the target user, and the target user can view the comment-back text through a terminal and the like. For example, the user utters comment text as: good quality and good style. The review text can be automatically generated through the review model: thank you for appreciation and liking, cutting of each inch of fabric is accompanied by dripping of tax. Without thinking about the production of clothes, people tend to pay one hundred percent of effort and hearts, hope to bring a more comfortable wearing experience to people, hope that the fashion elegance can be informed of your lifetime, and give your hearty care. * Starship store expecting your next light! "; or, the comment text sent by the user is: "describe the size too much different from the actual size, make me buy a big size! The user is very troublesome to change, i wear a bar for others! "the review text can be automatically generated through the review model: "lovely customers, thank you for your attention to the flag store. Because the design concept and the elasticity of the fabric of each style are different, the corresponding sizes are possibly slightly different, a user is advised to measure the basic size of the user and then to buy by referring to the page size information, and the user can also consult customer service in advance to hope that the user buys goods pleasantly! ".
As another optional example of the present invention, after generating the comment-back text corresponding to the current comment text of the user by using the comment-back model, the method further includes: under a manual intervention mode, adjusting the review content of the review text according to the currently received correction information to obtain the revised review text; and returning the revised review text to the target user.
The manual intervention mode indicates that the contents of the comment-back text are changed through human actions or operations (indicating manual intervention), and the above example indicates that the merchant can modify the comment-back text generated by the comment-back model through manual intervention, adjust the comment-back text adapted to the comment text, and return the modified text to the user. The manual intervention mode is set, so that the accuracy and timeliness of the replied content are guaranteed, the use habit of a merchant is better met, and the merchant can autonomously combine the review mode and the text of the manual intervention mode to generate a proper review text.
By the embodiment, data can be preprocessed based on the collected user history comment text, merchant history review text and commodity description information, an automatic review model is constructed, then the current comment text of the user is received, the review text can be generated for the new comment text through the automatic review model, the merchant can further correct the comment text and then reply the comment text to the target user, and therefore the merchant can generate the desired review text by using the review model and some inquiry requirements of the user are met.
It should be noted that for simplicity of description, the above-mentioned method embodiments are shown as a series of combinations of acts, but those skilled in the art will recognize that the present invention is not limited by the order of acts, as some steps may occur in other orders or concurrently in accordance with the invention. Further, those skilled in the art should also appreciate that the embodiments described in the specification are preferred embodiments and that the acts and modules referred to are not necessarily required by the invention.
Through the above description of the embodiments, those skilled in the art can clearly understand that the method for generating the review text according to the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but the former is a better implementation mode in many cases. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal device (such as a mobile phone, a computer, a server, or a network device) to execute the method according to the embodiments of the present invention.
Example 2
According to an embodiment of the present invention, there is also provided a review text generation apparatus for implementing the review text generation method, and fig. 6 is a schematic diagram of an optional review text generation apparatus according to an embodiment of the present invention, as shown in fig. 6, the apparatus includes: a building unit 61, a receiving unit 63, a generating unit 65, wherein,
the building unit 61 is configured to build a review model based on a preset corpus, where the preset corpus includes at least one of the following: a user history comment text, a merchant history comment-back text and a commodity description information text;
a receiving unit 63, configured to receive a current comment text of a user;
and the generating unit 65 is used for generating the comment-back text corresponding to the current comment text of the user by using the comment-back model.
The generating device of the review text can construct a review model based on a preset corpus through the constructing unit 61, wherein the preset corpus includes at least one of the following: the user historical comment text, the merchant historical review text and the commodity description information text are received through the receiving unit 63, and the review text corresponding to the user current comment text is generated through the generating unit 65 by using the review model. In the embodiment, various reply contents with rich information and pertinence can be automatically generated by combining the constructed review model with the comment text of the user and the description information of the commodity, the reply contents are adaptive to the text contents of the current comment of the user, so that the user can more deeply know the related information of the commodity, the attention of the user to the merchant is promoted, the attention of the user to the merchant is improved, and the technical problem that the attention of the user to the merchant is reduced due to the fact that the feedback information is often inconsistent with the comment contents of the user when the comment contents of the user are replied in the related technology is solved.
Optionally, the construction unit includes: the system comprises a preprocessing module, a query module and a training module, wherein the preprocessing module is used for preprocessing a preset corpus and establishing an association relation between a user history comment text and a merchant history comment text through commodity attribute information to obtain a corpus to be trained; and the building module is used for building a review model by adopting the corpus to be trained.
Another optional, the building block comprises: the first coding submodule is used for coding each word vector in the historical comment text of the user into a first hidden state corresponding to the time step to obtain a first coding result; the second coding module is used for coding the word vector corresponding to each key value pair in the commodity description information text into a second hidden state corresponding to the time step to obtain a second coding result; and the obtaining sub-module is used for obtaining the word vector corresponding to each generation step according to the first coding result, the second coding result and the obtained third hidden state, and constructing the review model, wherein the third hidden state is the hidden state corresponding to the word vector obtained by prediction in the previous step of each generation step.
In the embodiment of the present invention, the obtaining sub-module includes: the first generation submodule is used for generating a user history comment text vector based on the third hidden state of each generation step and the first hidden state of the corresponding time step; the second generation submodule is used for generating a commodity description information text vector based on the third hidden state of each generation step, the second hidden state of the corresponding time step and the user historical comment text vector; the fusion sub-module is used for fusing the text vector of the historical comment of the user and the text vector of the commodity description information into an intermediate vector; and the third generation submodule is used for generating a word vector corresponding to each generation step based on the third implicit state and the intermediate vector of each generation step.
As another optional example of the present invention, the apparatus for generating a review text further includes: and the first returning unit is used for returning the comment text to a target user submitting the current comment text of the user in an automatic returning mode after the comment text corresponding to the current comment text of the user is generated by using the comment back model.
As another optional example of the present invention, the apparatus for generating a review text further includes: after the review text corresponding to the current review text of the user is generated by using the review model, the adjusting unit is used for adjusting the review content of the review text according to the currently received correction information in a manual intervention mode to obtain the corrected review text; and the second returning unit is used for returning the revised review text to the target user.
As another optional example of the present invention, the apparatus for generating the review text is applied to an e-market scene, where the e-market scene includes at least one of: e-commerce marketing scene in television shopping and E-commerce marketing scene in internet shopping.
It should be noted here that the above-mentioned constructing unit 61, receiving unit 63 and generating unit 65 correspond to steps S302 to S306 in embodiment 1, and the two modules are the same as the examples and application scenarios realized by the corresponding steps, but are not limited to the disclosure of the above-mentioned embodiment one. It should be noted that the modules described above as part of the apparatus may be run in the computer terminal 10 provided in the first embodiment.
In an embodiment of the present invention, there is further provided a review system, and fig. 7 is a schematic diagram of an automatic review system according to an embodiment of the present invention, as shown in fig. 7, the review system includes: a system foreground interactive interface 71 and a system background management interface 73, wherein,
and the system foreground interactive interface 71 comprises an automatic reply mode 711 and a manual intervention mode 713, the merchant can set the automatic reply mode to directly reply the model generation content to the target buyer, and the merchant can also select the manual intervention mode to further modify the generation content and feed the modified generation content back to the user.
And a system background management interface 73, which includes a model training setup interface 732 and a system management interface 734, where the model training setup interface can reset parameters and train a model after data is added, and the system management interface is used for an administrator to know the current operation state of the system more conveniently.
The automatic review system corresponds to the review text generation device, can provide different interfaces for the user and the merchant respectively by using the automatic review system, and can automatically generate and train a review model.
Through the automatic review system, the merchant can use the review model to automatically send the review text, the review text of the user is responded in a targeted manner, the use habit of the merchant is better met, and the use rate of the merchant is improved.
Example 3
The embodiment of the invention can provide a computer terminal which can be any computer terminal device in a computer terminal group. Optionally, in this embodiment, the computer terminal may also be replaced with a terminal device such as a mobile terminal.
Optionally, in this embodiment, the computer terminal may be located in at least one network device of a plurality of network devices of a computer network.
In this embodiment, the computer terminal may execute program codes of the following steps in the method for generating the review text: the method comprises the following steps of constructing a back evaluation model based on a preset corpus, wherein the preset corpus comprises at least one of the following: a user history comment text, a merchant history review text and a commodity description information text; receiving a current comment text of a user; and generating a comment-back text corresponding to the current comment text of the user by using the comment-back model.
Alternatively, fig. 8 is a block diagram of a computer terminal according to an embodiment of the present invention. As shown in fig. 8, the computer terminal a may include: one or more processors, memory, and network interfaces, I/O interfaces, input/output interfaces, keyboards, displays.
The memory may be used to store software programs and modules, such as program instructions/modules corresponding to the method and apparatus for generating a review text in the embodiments of the present invention, and the processor executes various functional applications and data processing by operating the software programs and modules stored in the memory, that is, implements the method for generating a review text described above. The memory may include high speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory may further include memory remotely located from the processor, and these remote memories may be connected to terminal a through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The processor can call the information and application program stored in the memory through the transmission device to execute the following steps: the method comprises the following steps of constructing a review model based on a preset corpus, wherein the preset corpus comprises at least one of the following: a user history comment text, a merchant history comment-back text and a commodity description information text; receiving a current comment text of a user; and generating a review text corresponding to the current comment text of the user by using the review model.
Optionally, the processor may further execute the program code of the following steps: preprocessing a preset corpus, and establishing an association relation between a user history comment text and a merchant history comment text through commodity attribute information to obtain a corpus to be trained; and adopting the corpus to be trained to construct a review model.
Optionally, the processor may further execute the program code of the following steps: coding each word vector in the historical comment text of the user into a first hidden state corresponding to a time step to obtain a first coding result; coding the word vector corresponding to each key value pair in the commodity description information text into a second hidden state corresponding to the time step to obtain a second coding result; and obtaining a word vector corresponding to each generation step according to the first coding result, the second coding result and an obtained third hidden state, and constructing a review model, wherein the third hidden state is a hidden state corresponding to the word vector obtained by prediction in the previous step of each generation step.
Optionally, the processor may further execute the program code of the following steps: generating a commodity description information text vector based on the third hidden state of each generation step, the second hidden state of the corresponding time step and the user history comment text vector; fusing a user history comment text vector and a commodity description information text vector into a middle vector; and generating a word vector corresponding to each generation step based on the third implicit state and the intermediate vector of each generation step.
Optionally, the processor may further execute the program code of the following steps: after the comment-back text corresponding to the current comment text of the user is generated by using the comment-back model, the comment-back text is returned to a target user submitting the current comment text of the user in an automatic reply mode.
Optionally, the processor may further execute the program code of the following steps: after the comment-back text corresponding to the current comment text of the user is generated by using the comment-back model, in a manual intervention mode, the comment-back content of the comment-back text is adjusted according to the currently received correction information, and the corrected comment-back text is obtained; and returning the revised review text to the target user.
By adopting the embodiment of the invention, the comment-back model is constructed based on the preset corpus, the current comment text of the user is received, and the comment-back model is used for generating the comment-back text corresponding to the current comment text of the user. In the embodiment, various, rich-information and targeted reply contents can be automatically generated by combining the constructed review model with the user review text and the description information of the commodity, the reply contents are adaptive to the text contents of the current reviews of the user, so that the user can more deeply know the relevant information of the commodity, the attention of the merchant to the user is improved, the attention of the user to the merchant is improved, and the technical problem that the attention of the user to the merchant is reduced due to the fact that the feedback information is often inconsistent with the contents of the user reviews when the comment contents of the user are replied in the related technology is solved.
It can be understood by those skilled in the art that the structure shown in fig. 8 is only an illustration, and the computer terminal may also be a terminal device such as a smart phone (e.g., an Android phone, an iOS phone, etc.), a tablet computer, a palmtop computer, a Mobile Internet Device (MID), a PAD, and the like.
Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by a program instructing hardware associated with the terminal device, where the program may be stored in a computer-readable storage medium, and the storage medium may include: flash disks, read-Only memories (ROMs), random Access Memories (RAMs), magnetic or optical disks, and the like.
Example 4
The embodiment of the invention also provides a storage medium. Optionally, in this embodiment, the storage medium may be configured to store a program code executed by the method for generating a review text provided in the first embodiment.
Optionally, in this embodiment, the storage medium may be located in any one of computer terminals in a computer terminal group in a computer network, or in any one of mobile terminals in a mobile terminal group.
Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps: the method comprises the following steps of constructing a back evaluation model based on a preset corpus, wherein the preset corpus comprises at least one of the following: a user history comment text, a merchant history review text and a commodity description information text; and generating a comment-back text corresponding to the current comment text of the user by using the comment-back model.
Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps: preprocessing a preset corpus, and establishing an association relation between a user history comment text and a merchant history comment text through commodity attribute information to obtain a corpus to be trained; and adopting the corpus to be trained to construct a review model.
Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps: coding each word vector in the historical comment text of the user into a first hidden state corresponding to a time step to obtain a first coding result; coding the word vector corresponding to each key value pair in the commodity description information text into a second hidden state corresponding to the time step to obtain a second coding result; and obtaining a word vector corresponding to each generation step according to the first coding result, the second coding result and an obtained third hidden state, and constructing a review model, wherein the third hidden state is a hidden state corresponding to the word vector obtained by prediction in the previous step of each generation step.
According to another aspect of the embodiments of the present invention, there is also provided a system for generating a review text, including: a processor; and a memory coupled to the processor for providing instructions to the processor for processing the following processing steps: step 1, constructing a review model based on a preset corpus, wherein the preset corpus comprises at least one of the following: a user history comment text, a merchant history comment-back text and a commodity description information text; step 2, receiving a current comment text of a user; and 3, generating a comment-back text corresponding to the current comment text of the user by using the comment-back model.
The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
In the above embodiments of the present invention, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.
In the embodiments provided in the present application, it should be understood that the disclosed technology can be implemented in other ways. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units is only one type of division of logical functions, and there may be other divisions when actually implemented, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, units or modules, and may be in an electrical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit may be implemented in the form of hardware, or may also be implemented in the form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention, which is substantially or partly contributed by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic or optical disk, and other various media capable of storing program codes.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and amendments can be made without departing from the principle of the present invention, and these modifications and amendments should also be considered as the protection scope of the present invention.
Claims (7)
1. A method for generating a review text, comprising:
the method comprises the following steps of constructing a back evaluation model based on a preset corpus, wherein the preset corpus comprises at least one of the following: a user history comment text, a merchant history comment-back text and a commodity description information text;
receiving a current comment text of a user;
generating a comment-back text corresponding to the current comment text of the user by using the comment-back model;
wherein, the step of constructing the review model based on the preset corpus comprises the following steps:
preprocessing the preset corpus, and establishing an association relation between the user historical comment text and the merchant historical review text through commodity attribute information to obtain a corpus to be trained; adopting the corpus to be trained to construct the review model;
wherein, adopting the corpus to be trained to construct the review model comprises:
coding each word vector in the user history comment text into a first hidden state corresponding to a time step to obtain a first coding result; coding the word vector corresponding to each key value pair in the commodity description information text into a second hidden state corresponding to the time step to obtain a second coding result; obtaining a word vector corresponding to each generation step according to the first coding result, the second coding result and an obtained third hidden state, and constructing the review model, wherein the third hidden state is a hidden state corresponding to a word vector obtained by prediction in the previous step of each generation step;
wherein, according to the first encoding result, the second encoding result, and the third hidden state, obtaining the word vector corresponding to each generation step includes:
generating a user history comment text vector based on the third hidden state of each generation step and the first hidden state of the corresponding time step; generating a commodity description information text vector based on the third hidden state of each generation step, the second hidden state of the corresponding time step and the user history comment text vector; fusing the user history comment text vector and the commodity description information text vector into an intermediate vector; and generating a word vector corresponding to each generation step based on the third implicit state of each generation step and the intermediate vector.
2. The method of claim 1, after generating a comment-back text corresponding to the user's current comment text using the comment-back model, further comprising:
and in an automatic reply mode, returning the comment-back text to a target user submitting the current comment text of the user.
3. The method of claim 1, after generating a comment-back text corresponding to the user's current comment text using the comment-back model, further comprising:
adjusting the review content of the review text according to the currently received correction information in a manual intervention mode to obtain the revised review text;
and returning the revised review text to the target user.
4. The method of claim 1, wherein the method is applied to e-commerce scenarios comprising at least one of:
e-commerce marketing scenes in television shopping and E-commerce marketing scenes in internet shopping.
5. An apparatus for generating a review text, comprising:
the building unit is used for building a feedback evaluation model based on a preset corpus, wherein the preset corpus comprises at least one of the following: a user history comment text, a merchant history comment-back text and a commodity description information text;
the receiving unit is used for receiving the current comment text of the user;
the generating unit is used for generating a comment-back text corresponding to the current comment text of the user by using the comment-back model;
wherein the construction unit comprises: the preprocessing module is used for preprocessing the preset corpus and establishing an association relation between the user historical comment text and the merchant historical review text through commodity attribute information to obtain a corpus to be trained; the building module is used for building the comment model by adopting the corpus to be trained;
wherein the building block comprises: the first coding submodule is used for coding each word vector in the historical comment text of the user into a first hidden state corresponding to the time step to obtain a first coding result; the second coding submodule is used for coding the word vector corresponding to each key value pair in the commodity description information text into a second hidden state corresponding to the time step to obtain a second coding result; the obtaining sub-module is used for obtaining a word vector corresponding to each generation step according to the first coding result, the second coding result and an obtained third hidden state, and constructing the review model, wherein the third hidden state is a hidden state corresponding to a word vector obtained by prediction in the previous step of each generation step;
wherein the acquisition submodule comprises: the first generation submodule is used for generating a user history comment text vector based on the third hidden state of each generation step and the first hidden state of the corresponding time step; the second generation submodule is used for generating a commodity description information text vector based on the third hidden state of each generation step, the second hidden state of the corresponding time step and the user history comment text vector; the fusion submodule is used for fusing the user history comment text vector and the commodity description information text vector into an intermediate vector; and the third generation submodule is used for generating a word vector corresponding to each generation step based on the third implicit state of each generation step and the intermediate vector.
6. A storage medium, characterized in that the storage medium comprises a stored program, wherein when the program runs, a device where the storage medium is located is controlled to execute the method for generating the comment text according to any one of claims 1 to 4.
7. A system for generating a review text, comprising:
a processor; and
a memory coupled to the processor for providing instructions to the processor for processing the following processing steps:
step 1, constructing a review model based on a preset corpus, wherein the preset corpus comprises at least one of the following: a user history comment text, a merchant history comment-back text and a commodity description information text;
step 2, receiving a current comment text of a user;
step 3, generating a comment-back text corresponding to the current comment text of the user by using the comment-back model;
wherein, constructing a review model based on the preset corpus comprises:
preprocessing the preset corpus, and establishing an association relation between the user historical comment text and the merchant historical review text through commodity attribute information to obtain a corpus to be trained; adopting the corpus to be trained to construct the review model;
wherein, adopting the corpus to be trained to construct the review model comprises:
coding each word vector in the user history comment text into a first hidden state corresponding to the time step to obtain a first coding result; coding the word vector corresponding to each key value pair in the commodity description information text into a second hidden state corresponding to the time step to obtain a second coding result; obtaining a word vector corresponding to each generation step according to the first coding result, the second coding result and an obtained third hidden state, and constructing the review model, wherein the third hidden state is a hidden state corresponding to a word vector obtained by prediction in the previous step of each generation step;
wherein, according to the first encoding result, the second encoding result, and the third hidden state, obtaining the word vector corresponding to each generation step includes:
generating a user history comment text vector based on the third hidden state of each generation step and the first hidden state of the corresponding time step; generating a commodity description information text vector based on the third hidden state of each generation step, the second hidden state of the corresponding time step and the user history comment text vector; fusing the user history comment text vector and the commodity description information text vector into an intermediate vector; generating a word vector corresponding to each generation step based on the intermediate vector and the third hidden state of each generation step.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811474376.0A CN111353024B (en) | 2018-12-04 | 2018-12-04 | Method, device and system for generating comment text |
| PCT/CN2019/121853 WO2020114324A1 (en) | 2018-12-04 | 2019-11-29 | Method, apparatus, and system for generating review responses |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811474376.0A CN111353024B (en) | 2018-12-04 | 2018-12-04 | Method, device and system for generating comment text |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111353024A CN111353024A (en) | 2020-06-30 |
| CN111353024B true CN111353024B (en) | 2023-04-18 |
Family
ID=70975197
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811474376.0A Active CN111353024B (en) | 2018-12-04 | 2018-12-04 | Method, device and system for generating comment text |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN111353024B (en) |
| WO (1) | WO2020114324A1 (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111914566A (en) * | 2020-07-30 | 2020-11-10 | 混沌时代(北京)教育科技有限公司 | Automatic comment generation method |
| CN113744012A (en) * | 2020-08-10 | 2021-12-03 | 北京沃东天骏信息技术有限公司 | Information processing method, device and storage medium |
| CN112507112B (en) * | 2020-12-07 | 2023-07-25 | 中国平安人寿保险股份有限公司 | Comment generation method, comment generation device, comment generation equipment and storage medium |
| CN113313510A (en) * | 2021-04-25 | 2021-08-27 | 北京搜狗科技发展有限公司 | Data processing method, device and medium |
| CN114840658B (en) * | 2022-07-06 | 2023-03-31 | 浙江口碑网络技术有限公司 | Evaluation reply method, electronic device, and computer storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102789449A (en) * | 2011-05-20 | 2012-11-21 | 日电（中国）有限公司 | Method and device for evaluating comment text |
| CN105847116A (en) * | 2016-03-28 | 2016-08-10 | 乐视控股（北京）有限公司 | Comment reply information processing method and apparatus |
| CN106294373A (en) * | 2015-05-15 | 2017-01-04 | 阿里巴巴集团控股有限公司 | Review information answering method and device |
| CN108228794A (en) * | 2017-12-29 | 2018-06-29 | 三角兽（北京）科技有限公司 | Apparatus for management of information, information processing unit and automatically reply/comment method |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103810170B (en) * | 2012-11-06 | 2018-04-27 | 腾讯科技（深圳）有限公司 | Intercommunion platform file classification method and device |
| CN104951542A (en) * | 2015-06-19 | 2015-09-30 | 百度在线网络技术（北京）有限公司 | Method and device for recognizing class of social contact short texts and method and device for training classification models |
| CN106649345A (en) * | 2015-10-30 | 2017-05-10 | 微软技术许可有限责任公司 | Automatic session creator for news |
| CN107040450B (en) * | 2016-07-20 | 2018-06-01 | 平安科技（深圳）有限公司 | Automatic reply method and device |
-
2018
- 2018-12-04 CN CN201811474376.0A patent/CN111353024B/en active Active
-
2019
- 2019-11-29 WO PCT/CN2019/121853 patent/WO2020114324A1/en active Application Filing
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102789449A (en) * | 2011-05-20 | 2012-11-21 | 日电（中国）有限公司 | Method and device for evaluating comment text |
| CN106294373A (en) * | 2015-05-15 | 2017-01-04 | 阿里巴巴集团控股有限公司 | Review information answering method and device |
| CN105847116A (en) * | 2016-03-28 | 2016-08-10 | 乐视控股（北京）有限公司 | Comment reply information processing method and apparatus |
| CN108228794A (en) * | 2017-12-29 | 2018-06-29 | 三角兽（北京）科技有限公司 | Apparatus for management of information, information processing unit and automatically reply/comment method |
Non-Patent Citations (1)
| Title |
|---|
| 李学明 ; 张朝阳 ; 佘维军 ; .基于用户回复内容观点支持度的评论有用性计算.计算机应用.(第10期),2767-2776. * |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2020114324A1 (en) | 2020-06-11 |
| CN111353024A (en) | 2020-06-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111353024B (en) | Method, device and system for generating comment text | |
| CN105607756B (en) | Information recommendation method and device | |
| Shankar et al. | Mobile shopper marketing: Key issues, current insights, and future research avenues | |
| CN104769627B (en) | Method and apparatus for opposite end auxiliary shopping | |
| JP6634019B2 (en) | Technologies that facilitate electronic trading | |
| JP6619801B2 (en) | System and method for facilitation cross order | |
| KR102206000B1 (en) | Optimized item attribute comparison | |
| EP3079116A1 (en) | System and method for generating recommendations | |
| US20150120386A1 (en) | System and method for identifying purchase intent | |
| US20120265635A1 (en) | Social network recommendation polling | |
| WO2016044608A1 (en) | System and method for a semi-lit market | |
| US11793296B2 (en) | Beauty product creation platform | |
| US20150341304A1 (en) | Personalized sharing at the user level | |
| CN115358554A (en) | Demand scheme recommendation method and device for computing power network | |
| KR20230138237A (en) | Method and system for operating sns-based re-commerce platform using hashtag | |
| JP6959044B2 (en) | Recording server, recording method and program | |
| Delawari | Online-business in Afghanistan, current trend and challenges ahead: A conceptual study | |
| CN111523914A (en) | User satisfaction evaluation method, device and system and data display platform | |
| US11227300B2 (en) | Computer-network-based referral service functions and user interfaces | |
| KR20190097615A (en) | apparatus and method for determining manufacture quantitybased on data related to design | |
| US20240119481A1 (en) | Collaboration management system and method | |
| CN115587238A (en) | Product brand marketing big data service method | |
| Laesser et al. | Tourism in the new economy | |
| Barry et al. | Factors affecting the intention to use mobile commerce in Malaysia: An integration of TAM And IS success model | |
| CN111222934A (en) | Information issuing method, device and computer system |
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 |