CN117633165A - Intelligent AI customer service dialogue guiding method - Google Patents
Intelligent AI customer service dialogue guiding method Download PDFInfo
- Publication number
- CN117633165A CN117633165A CN202311372840.6A CN202311372840A CN117633165A CN 117633165 A CN117633165 A CN 117633165A CN 202311372840 A CN202311372840 A CN 202311372840A CN 117633165 A CN117633165 A CN 117633165A
- Authority
- CN
- China
- Prior art keywords
- data
- commodity
- candidate
- data matrix
- matrix
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 32
- 239000011159 matrix material Substances 0.000 claims abstract description 152
- 239000013598 vector Substances 0.000 claims abstract description 54
- 238000012216 screening Methods 0.000 claims abstract description 6
- 238000012545 processing Methods 0.000 claims description 5
- 238000012163 sequencing technique Methods 0.000 claims description 2
- 230000000977 initiatory effect Effects 0.000 claims 6
- 238000000638 solvent extraction Methods 0.000 claims 1
- 230000006870 function Effects 0.000 description 3
- 238000002372 labelling Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
Classifications
-
- 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
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of product recommendation, in particular to an intelligent AI customer service dialogue guiding method. The method comprises the following steps: acquiring a question data matrix and an answer data matrix; calculating the relevance between the commodity vector corresponding to each candidate commodity and each row of data in each data matrix; according to the similarity of each line of data in the question data matrix and each line of data in the answer data matrix, obtaining the reply validity of each line of data in the answer data matrix to each line of data in the question text data, and further obtaining target data and non-target data; and determining the guiding rate of each candidate commodity according to the relevance corresponding to the target data and the relevance corresponding to the non-target data, and further screening the commodity to be recommended and recommending the commodity to the user. The invention realizes the guidance of the customer demands and improves the satisfaction degree of the users.
Description
Technical Field
The invention relates to the technical field of product recommendation, in particular to an intelligent AI customer service dialogue guiding method.
Background
With the rapid development of AI technology, intelligent customer service has been widely used in many business scenarios, such as online shopping. They can not only provide seven-day 24-hour per week services, but also can greatly reduce the operating costs of the enterprise. However, although the application of intelligent customer service is becoming widespread, there are still some problems in providing high quality customer service. For example, many intelligent customer service replies are too mechanical to understand the actual needs of the user well, resulting in a large number of customer complaints and negative effects on the enterprise brands, as well as financial losses. In addition, the existing AI customer service system generally performs dialogue management based on a predefined dialogue tree, and when the user does not know the specific name of the required commodity, the dialogue management method cannot help the user to perform effective commodity searching, and the matching is performed approximately through the problem label and the commodity, but the validity of the reply cannot be ensured, so that the satisfaction degree of the user on the product recommended by the intelligent AI customer service is lower.
Disclosure of Invention
In order to solve the problem that the satisfaction degree of a user for recommending products is low when the existing intelligent AI customer service recommends the products, the invention aims to provide an intelligent AI customer service dialogue guiding method, which adopts the following technical scheme:
the invention provides an intelligent AI customer service dialogue guiding method, which comprises the following steps:
acquiring question text data of a user and answer text data of a corresponding intelligent AI customer service; obtaining a data matrix based on the question text data and the answer text data, wherein the data matrix comprises a question data matrix and an answer data matrix;
acquiring commodity vectors corresponding to each candidate commodity, and calculating the relevance between the commodity vectors corresponding to each candidate commodity and each row of data in each data matrix; according to the similarity of each line of data in the questioning data matrix and each line of data in the answer data matrix, obtaining the reply validity of each line of data in the answer data matrix to each line of data in the questioning text data;
dividing data in a data matrix into target data and non-target data based on the reply validity; obtaining the recommendation rate of each candidate commodity according to the relevance corresponding to the target data; acquiring the non-recommendation rate of each candidate commodity according to the relevance corresponding to the non-target data;
determining a lead rate for each candidate good based on the recommended rate and the non-recommended rate for each candidate good; and screening commodities to be recommended based on all the guiding rates and recommending the commodities to the user.
Preferably, the obtaining the reply validity of each line of data in the answer data matrix to each line of data in the question text data according to the similarity between each line of data in the question data matrix and each line of data in the answer data matrix includes:
for the ith row of data in the answer data matrix and the jth row of data in the question data matrix:
the maximum value of the similarity between the commodity vectors corresponding to all the candidate commodities and the ith row of data in the answer data matrix is marked as a first maximum value, and the maximum value of the similarity between the commodity vectors corresponding to all the candidate commodities and the jth row of data in the question data matrix is marked as a second maximum value;
and obtaining the replying effectiveness of the ith row of data in the answer data matrix to the jth row of data in the question data matrix based on the similarity of the ith row of data in the answer data matrix and the jth row of data in the question data matrix and the first maximum value and the second maximum value.
Preferably, based on the similarity between the first maximum value, the second maximum value, and the ith row data in the answer data matrix and the jth row data in the question data matrix, obtaining the reply validity of the ith row data in the answer data matrix to the jth row data in the question data matrix includes:
acquiring the minimum value of the first maximum value and the second maximum value;
and determining the product of the minimum value and the similarity of the ith row data in the answer data matrix and the jth row data in the questioning data matrix as the replying validity of the ith row data in the answer data matrix to the jth row data in the questioning data matrix.
Preferably, the screening the goods to be recommended based on all the guiding rates and recommending the goods to the user includes:
sequencing all candidate commodities according to the sequence from the high guide rate to the low guide rate to obtain a candidate commodity sequence;
and taking the pre-preset number of candidate commodities in the candidate commodity sequence as commodities to be recommended, and recommending the commodities to be recommended to a user.
Preferably, the dividing the data in the data matrix into the target data and the non-target data based on the reply validity includes:
combining each row of data in the questioning data matrix with each row of data in the answer data matrix respectively to obtain at least two combined data;
based on reply effectiveness corresponding to each combination data, clustering all the combination data by adopting a k-means algorithm to obtain two clustering clusters, wherein the value of k is 2 when the k-means algorithm clusters;
and respectively calculating the average value of reply validity corresponding to all the combined data in each cluster, taking the combined data in the cluster with the largest average value as target data, and taking the reply validity corresponding to the combined data in the cluster with the smallest average value as non-target data.
Preferably, the obtaining the recommendation rate of each candidate commodity according to the relevance corresponding to the target data includes:
for any target data: the relevance between the commodity vector corresponding to each candidate commodity and the data in the questioning data matrix in the target data is marked as a first index, the relevance between the commodity vector corresponding to each candidate commodity and the data in the answer data matrix in the target data is marked as a second index, and the average value of the first index and the second index is used as the relevance value between the corresponding candidate commodity and the target data;
for any candidate commodity: and taking the sum of the association values of the candidate commodity and all the target data as the recommendation rate of the candidate commodity.
Preferably, the non-recommendation rate of each candidate commodity is obtained according to the relevance corresponding to the non-target data, including:
for any non-target data: the relevance of the commodity vector corresponding to each candidate commodity and the data in the questioning data matrix in the non-target data is marked as a third index, the relevance of the commodity vector corresponding to each candidate commodity and the data in the answer data matrix in the non-target data is marked as a fourth index, and the average value of the third index and the fourth index is used as the relevance value of the corresponding candidate commodity and the non-target data;
for any candidate commodity: and taking the opposite number of the sum value of the association values of the candidate commodity and all non-target data as the non-recommendation rate of the candidate commodity.
Preferably, the determining the guidance rate of each candidate commodity based on the recommendation rate and the non-recommendation rate of each candidate commodity includes:
for any candidate commodity: and taking the sum of the recommendation rate and the non-recommendation rate of the selected commodity as the guiding rate of the candidate commodity.
Preferably, the acquiring of the question data matrix and the answer data matrix includes:
inputting the question text data into a trained Bert language model, performing sentence breaking and vectorization processing on the question text data to obtain at least two multidimensional vectors with equal length, and combining the multidimensional vectors to obtain a question data matrix;
inputting the answer text data into a trained Bert language model, performing sentence breaking and vectorization processing on the answer text data to obtain at least two multidimensional vectors with equal length, and combining the multidimensional vectors to obtain an answer data matrix.
Preferably, the obtaining the commodity vector corresponding to each candidate commodity includes:
and spelling the names of the candidate commodities and the labels corresponding to the names into a sentence, inputting the spelled sentence into a trained Bert language model, and outputting the corresponding commodity vector.
The invention has at least the following beneficial effects:
firstly, acquiring a questioning data matrix and an answer data matrix, wherein the questioning data matrix can reflect the questioning information of a user, and the answer data matrix can reflect the answer information of intelligent AI customer service for questioning of the user; and then, according to the similarity between each line of data in the questioning data matrix and each line of data in the answering data matrix, obtaining the reply validity of each line of data in the answering data matrix for each line of data in the questioning text data, wherein the larger the reply validity is, the higher the repetition rate of the reply of the intelligent AI customer service and the questions asked by the user is, namely the more effective the reply of the intelligent AI customer service is, the relevance between the commodity vector corresponding to each candidate commodity and each line of data in each data matrix is calculated, the guiding rate of each candidate commodity is determined, the products to be recommended are further screened based on the guiding rate, the guiding of the customer demands is completed, and the satisfaction degree of the user is improved.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions and advantages of the prior art, the following description will briefly explain the drawings used in the embodiments or the description of the prior art, and it is obvious that the drawings in the following description are only some embodiments of the invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of an intelligent AI customer service session guiding method according to an embodiment of the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the present invention to achieve the preset purpose, the following detailed description is given below of an intelligent AI customer service dialogue guiding method according to the present invention with reference to the accompanying drawings and preferred embodiments.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following specifically describes a specific scheme of the intelligent AI customer service dialogue guiding method provided by the invention with reference to the accompanying drawings.
An embodiment of an intelligent AI customer service dialogue guiding method:
the specific scene aimed at by this embodiment is: the dialogue record of the user and the intelligent AI customer service is obtained in real time, the dialogue record comprises question text data of the user and answer text data of the intelligent AI customer service, and the guidance rate of each candidate commodity is determined by analyzing the question text data of the user and the answer text data of the intelligent AI customer service, so that commodities to be recommended are screened out of all candidate commodities according to the guidance rate, and the commodities to be recommended are recommended to the user.
The embodiment provides an intelligent AI customer service dialogue guiding method, as shown in fig. 1, which includes the following steps:
step S1, acquiring question text data of a user and answer text data of a corresponding intelligent AI customer service; and obtaining a data matrix based on the question text data and the answer text data, wherein the data matrix comprises a question data matrix and an answer data matrix.
Firstly, through an interface of an intelligent customer service platform, a dialogue record of a user and an AI customer service is collected in real time, wherein the dialogue record comprises question text data of the user and answer text data of the intelligent AI customer service corresponding to the question text data of the user.
After acquiring the question text data of the user and the answer text data of the corresponding intelligent AI customer service, in order to enable the computer to process the text language, the embodiment selects to use a pre-trained language model to perform vector conversion on the dialogue content, and because long text possibly exists in the dialogue process, the embodiment selects to perform fine tuning on the language model so that the pre-trained language model has a sentence breaking function, wherein the embodiment selects the Bert language model to perform sentence breaking and vectorization processing on single-time user question content or single-time customer service answer content.
When the Bert language model is fine-tuned, a sentence-breaking text data set needs to be constructed, wherein the sentence-breaking text data set is a text data set with sentence-breaking labels. This dataset contains articles and their sentence-break locations. For example: a BIO labeling method may be used, where B represents the beginning of a sentence, I represents the middle of a sentence, and O represents the outside of a sentence (i.e., not the beginning or middle of a sentence). The labeling process is performed by a person having an associated labeling experience, and the text dataset is selected from the public dataset. And then, the pretrained Bert language model is utilized, a data set with labels is used as the input of the Bert language model, and the difference between the prediction label and the real label of the model is optimized through cross entropy loss, so that the fine adjustment of the Bert language model is realized. The method for fine tuning the language model is well known and will not be described in detail here. The Bert language model is an existing model, and an implementer can adjust and select other pre-trained language models according to specific implementation scenes. Each sentence in the output of the bart model after fine adjustment can obtain a multidimensional vector with a fixed length, and then a section of text data can obtain a plurality of vectors with equal length and multiple dimensions. According to the embodiment, question text data are input into a trained Bert language model, sentence breaking and vectorization are carried out on the question text data, a plurality of multidimensional vectors with equal length are obtained, all the multidimensional vectors obtained at the moment are combined together to obtain a question data matrix, and each action of the question data matrix is a multidimensional vector; inputting the answer text data into a trained Bert language model, performing sentence breaking and vectorization on the answer text data to obtain a plurality of multidimensional vectors with equal length, combining all the multidimensional vectors obtained at the moment to obtain an answer data matrix, wherein each action of the answer data matrix is a multidimensional vector.
Thus far, the present embodiment obtains two data matrices, i.e., a question data matrix and an answer data matrix, respectively.
Step S2, commodity vectors corresponding to each candidate commodity are obtained, and the relevance between the commodity vectors corresponding to each candidate commodity and each row of data in each data matrix is calculated; and obtaining the reply validity of each line of data in the answer data matrix to each line of data in the question text data according to the similarity of each line of data in the question data matrix and each line of data in the answer data matrix.
In the guiding dialogue of the AI customer service, if the user continuously has similar sentences in the last reply and the next question, the answer content in the last reply has sentences meeting the needs of the user, but a plurality of recommended commodities possibly exist in the last reply, if the user is not asked in the next question, the guiding can be performed according to the commodities meeting the needs of the user as much as possible and not needed by the user, and further the intelligent guiding of the customer service dialogue is completed.
In this embodiment, the names of each candidate commodity and the corresponding label are pieced together to form a sentence, for example, a smart phone with a touch screen size of 6.1 and a weight of 200 g, and the pieced together is input into a trained Bert language model to output the corresponding commodity vector. Each candidate commodity can obtain a corresponding commodity vector. It should be noted that: the candidate commodities are all commodities in the database which are similar to the user query product.
Next, in this embodiment, the DTW distances between the commodity vector corresponding to each candidate commodity and each line of data in each data matrix are calculated, and the calculated negative correlation mapping value of the DTW distance is used as the corresponding correlation, where the negative correlation mapping value of the pearson correlation coefficient may be represented by an exponential function, for example: taking a value of an exponential function taking a natural constant as a base and taking the negative DTW distance as an index as a relevance; the inverse of the DTW distance may also be expressed as a correlation. The relevance between the commodity vector corresponding to each candidate commodity and each line of data in each data matrix is obtained, and for any candidate commodity, the relevance between the commodity vector corresponding to the candidate commodity and each line of data in the questioning data matrix is calculated; similarly, for any candidate commodity, a correlation is calculated between the corresponding commodity vector and each row of data in the answer data matrix.
If the text content of the question of the user and the answer text content of the intelligent AI customer service have similar sentences, the intelligent AI customer service reply is effective, so that the similarity of each line of data in the question data matrix and each line of data in the answer data matrix is calculated. Specifically, for any row of data in the questioning data matrix: respectively calculating cosine similarity of each line of data in the line of data and each line of data in the answer data matrix, and taking the obtained cosine similarity as similarity of each line of data in the line of data and each line of data in the answer data matrix; the larger the cosine similarity, the higher the similarity of the two corresponding lines of data. By adopting the method, the similarity between each row of data in the questioning data matrix and each row of data in the answer data matrix is obtained.
For the ith row of data in the answer data matrix and the jth row of data in the question data matrix: the maximum value of the similarity between the commodity vectors corresponding to all the candidate commodities and the ith row of data in the answer data matrix is marked as a first maximum value, and the maximum value of the similarity between the commodity vectors corresponding to all the candidate commodities and the jth row of data in the question data matrix is marked as a second maximum value; acquiring the minimum value of the first maximum value and the second maximum value; and determining the product of the minimum value and the similarity of the ith row data in the answer data matrix and the jth row data in the questioning data matrix as the replying validity of the ith row data in the answer data matrix to the jth row data in the questioning data matrix.
By adopting the method, the reply validity of each line of data in the reply data matrix to each line of data in the question data matrix can be obtained.
Step S3, dividing the data in the data matrix into target data and non-target data based on the reply validity; obtaining the recommendation rate of each candidate commodity according to the relevance corresponding to the target data; and obtaining the non-recommendation rate of each candidate commodity according to the relevance corresponding to the non-target data.
The present embodiment then combines each line of data in the question data matrix with each line of data in the answer data matrix, respectively, to obtain a plurality of combined data, for example: for the ith row of data in the answer data matrix and the jth row of data in the question data matrix: and splicing the ith row of data in the answer data matrix and the jth row of data in the question data matrix together to obtain a new row of data, wherein the row of data is combined data. Any line of data in the answer data matrix and any line of data in the question data matrix can form one combined data, so that a plurality of combined data are obtained.
Based on reply effectiveness corresponding to each combination data, clustering all the combination data by adopting a k-means algorithm to obtain two clustering clusters, wherein the value of k is 2 when the k-means algorithm clusters; the k-means algorithm is prior art and will not be described in detail here. The average value of reply validity corresponding to all the combined data in each cluster is calculated, the combined data in the cluster with the largest average value is used as target data, the reply validity corresponding to the combined data in the cluster with the smallest average value is used as non-target data, namely all the combined data are classified, and the target data and the non-target data are respectively obtained.
For any target data: the target data is formed by combining one row of data in the questioning data matrix and one row of data in the answering data matrix, so that the relevance between the commodity vector corresponding to each candidate commodity and the data in the questioning data matrix in the target data is marked as a first index, the relevance between the commodity vector corresponding to each candidate commodity and the data in the answering data matrix in the target data is marked as a second index, and the average value of the first index and the second index is used as the relevance value between the corresponding candidate commodity and the target data.
By adopting the method, the association value of each candidate commodity and each target data can be obtained, and one candidate commodity and each target data have a corresponding association value. For any candidate commodity: and taking the sum of the association values of the candidate commodity and all the target data as the recommendation rate of the candidate commodity.
For any non-target data: the non-target data is formed by combining one row of data in the questioning data matrix and one row of data in the answer data matrix, so that the relevance between the commodity vector corresponding to each candidate commodity and the data in the questioning data matrix in the non-target data is marked as a third index, the relevance between the commodity vector corresponding to each candidate commodity and the data in the answer data matrix in the non-target data is marked as a fourth index, and the average value of the third index and the fourth index is used as the relevance value between the corresponding candidate commodity and the non-target data.
By adopting the method, the association value of each candidate commodity and each non-target data can be obtained, and one candidate commodity and each non-target data have a corresponding association value. For any candidate commodity: and taking the opposite number of the sum value of the association values of the candidate commodity and all non-target data as the non-recommendation rate of the candidate commodity.
Step S4, determining the guidance rate of each candidate commodity based on the recommendation rate and the non-recommendation rate of each candidate commodity; and screening commodities to be recommended based on all the guiding rates and recommending the commodities to the user.
The present embodiment has obtained the recommended rate and the non-recommended rate of each candidate commodity, and next the present embodiment will determine the guidance rate of each candidate commodity from the recommended rate and the non-recommended rate of each candidate commodity. Specifically, for any candidate commodity: and taking the sum of the recommendation rate and the non-recommendation rate of the selected commodity as the guiding rate of the candidate commodity. By adopting the method, the guidance rate of each candidate commodity is obtained.
The greater the guidance rate of the candidate good, the more likely the corresponding candidate good is to be of interest to the user, i.e., the more should be recommended to the user. Based on the above, in this embodiment, all candidate commodities are ordered according to the order of the guidance rate from large to small to obtain a candidate commodity sequence; and taking the first preset number of candidate commodities in the candidate commodity sequence as commodities to be recommended, namely taking the preset number of candidate commodities with the maximum guiding rate as commodities to be recommended, and further recommending all the commodities to be recommended to the user in the embodiment. The preset number in this embodiment is that in a specific application, the implementer may set according to a specific situation.
So far, by adopting the method provided by the embodiment, commodity recommendation is carried out on the user, and user demand guidance is completed.
Firstly, acquiring a questioning data matrix and an answer data matrix, wherein the questioning data matrix can reflect the questioning information of a user, and the answer data matrix can reflect the answer information of intelligent AI customer service for questioning of the user; and then according to the similarity between each line of data in the questioning data matrix and each line of data in the answering data matrix, obtaining the reply validity of each line of data in the answering data matrix for each line of data in the questioning text data, wherein the larger the reply validity is, the higher the repetition rate of the reply of the intelligent AI customer service and the questions asked by the user is, namely the more effective the reply of the intelligent AI customer service is, the relevance between the commodity vector corresponding to each candidate commodity and each line of data in each data matrix is calculated, the guiding rate of each candidate commodity is determined, the products to be recommended are further screened based on the guiding rate, the guiding of the customer demands is completed, and the satisfaction degree of the user is improved.
It should be noted that: the foregoing description of the preferred embodiments of the present invention is not intended to be limiting, but rather, any modifications, equivalents, improvements, etc. that fall within the principles of the present invention are intended to be included within the scope of the present invention.
Claims (10)
1. An intelligent AI customer service dialogue guiding method is characterized by comprising the following steps:
acquiring question text data of a user and answer text data of a corresponding intelligent AI customer service; obtaining a data matrix based on the question text data and the answer text data, wherein the data matrix comprises a question data matrix and an answer data matrix;
acquiring commodity vectors corresponding to each candidate commodity, and calculating the relevance between the commodity vectors corresponding to each candidate commodity and each row of data in each data matrix; according to the similarity of each line of data in the questioning data matrix and each line of data in the answer data matrix, obtaining the reply validity of each line of data in the answer data matrix to each line of data in the questioning text data;
dividing data in a data matrix into target data and non-target data based on the reply validity; obtaining the recommendation rate of each candidate commodity according to the relevance corresponding to the target data; acquiring the non-recommendation rate of each candidate commodity according to the relevance corresponding to the non-target data;
determining a lead rate for each candidate good based on the recommended rate and the non-recommended rate for each candidate good; and screening commodities to be recommended based on all the guiding rates and recommending the commodities to the user.
2. The intelligent AI customer service dialogue guidance method according to claim 1, wherein the obtaining the reply validity of each line of data in the reply data matrix to each line of data in the question text data according to the similarity of each line of data in the question data matrix and each line of data in the reply data matrix comprises:
for the ith row of data in the answer data matrix and the jth row of data in the question data matrix:
the maximum value of the similarity between the commodity vectors corresponding to all the candidate commodities and the ith row of data in the answer data matrix is marked as a first maximum value, and the maximum value of the similarity between the commodity vectors corresponding to all the candidate commodities and the jth row of data in the question data matrix is marked as a second maximum value;
and obtaining the replying effectiveness of the ith row of data in the answer data matrix to the jth row of data in the question data matrix based on the similarity of the ith row of data in the answer data matrix and the jth row of data in the question data matrix and the first maximum value and the second maximum value.
3. The intelligent AI customer service dialogue guidance method of claim 2, wherein obtaining validity of the ith row of data in the answer data matrix for the reply of the jth row of data in the question data matrix based on the first maximum value, the second maximum value, and similarity of the ith row of data in the answer data matrix and the jth row of data in the question data matrix comprises:
acquiring the minimum value of the first maximum value and the second maximum value;
and determining the product of the minimum value and the similarity of the ith row data in the answer data matrix and the jth row data in the questioning data matrix as the replying validity of the ith row data in the answer data matrix to the jth row data in the questioning data matrix.
4. The intelligent AI customer service dialogue guidance method according to claim 1, wherein the screening of goods to be recommended and recommending to a user based on all the guidance rates comprises:
sequencing all candidate commodities according to the sequence from the high guide rate to the low guide rate to obtain a candidate commodity sequence;
and taking the pre-preset number of candidate commodities in the candidate commodity sequence as commodities to be recommended, and recommending the commodities to be recommended to a user.
5. The intelligent AI customer service dialogue initiation method of claim 1, wherein the partitioning of data in a data matrix into targeted data and non-targeted data based on the reply validity comprises:
combining each row of data in the questioning data matrix with each row of data in the answer data matrix respectively to obtain at least two combined data;
based on reply effectiveness corresponding to each combination data, clustering all the combination data by adopting a k-means algorithm to obtain two clustering clusters, wherein the value of k is 2 when the k-means algorithm clusters;
and respectively calculating the average value of reply validity corresponding to all the combined data in each cluster, taking the combined data in the cluster with the largest average value as target data, and taking the reply validity corresponding to the combined data in the cluster with the smallest average value as non-target data.
6. The intelligent AI customer service dialogue guidance method according to claim 5, wherein obtaining the recommendation rate of each candidate commodity according to the correlation corresponding to the target data comprises:
for any target data: the relevance between the commodity vector corresponding to each candidate commodity and the data in the questioning data matrix in the target data is marked as a first index, the relevance between the commodity vector corresponding to each candidate commodity and the data in the answer data matrix in the target data is marked as a second index, and the average value of the first index and the second index is used as the relevance value between the corresponding candidate commodity and the target data;
for any candidate commodity: and taking the sum of the association values of the candidate commodity and all the target data as the recommendation rate of the candidate commodity.
7. The intelligent AI customer service dialogue guidance method of claim 5, wherein obtaining the non-recommendation rate for each candidate commodity according to the relevance of the non-target data comprises:
for any non-target data: the relevance of the commodity vector corresponding to each candidate commodity and the data in the questioning data matrix in the non-target data is marked as a third index, the relevance of the commodity vector corresponding to each candidate commodity and the data in the answer data matrix in the non-target data is marked as a fourth index, and the average value of the third index and the fourth index is used as the relevance value of the corresponding candidate commodity and the non-target data;
for any candidate commodity: and taking the opposite number of the sum value of the association values of the candidate commodity and all non-target data as the non-recommendation rate of the candidate commodity.
8. The intelligent AI customer service dialogue initiation method of claim 1, wherein determining the initiation rate for each candidate item based on the initiation rate and the non-initiation rate for each candidate item comprises:
for any candidate commodity: and taking the sum of the recommendation rate and the non-recommendation rate of the selected commodity as the guiding rate of the candidate commodity.
9. The intelligent AI customer service dialogue initiation method of claim 1, wherein the acquisition of the questioning data matrix and the answer data matrix comprises:
inputting the question text data into a trained Bert language model, performing sentence breaking and vectorization processing on the question text data to obtain at least two multidimensional vectors with equal length, and combining the multidimensional vectors to obtain a question data matrix;
inputting the answer text data into a trained Bert language model, performing sentence breaking and vectorization processing on the answer text data to obtain at least two multidimensional vectors with equal length, and combining the multidimensional vectors to obtain an answer data matrix.
10. The intelligent AI customer service dialogue guiding method as claimed in claim 1, wherein the acquiring the commodity vector corresponding to each candidate commodity comprises:
and spelling the names of the candidate commodities and the labels corresponding to the names into a sentence, inputting the spelled sentence into a trained Bert language model, and outputting the corresponding commodity vector.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311372840.6A CN117633165B (en) | 2023-10-20 | 2023-10-20 | Intelligent AI customer service dialogue guiding method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311372840.6A CN117633165B (en) | 2023-10-20 | 2023-10-20 | Intelligent AI customer service dialogue guiding method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117633165A true CN117633165A (en) | 2024-03-01 |
| CN117633165B CN117633165B (en) | 2024-05-31 |
Family
ID=90029434
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311372840.6A Active CN117633165B (en) | 2023-10-20 | 2023-10-20 | Intelligent AI customer service dialogue guiding method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117633165B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2011070422A (en) * | 2009-09-25 | 2011-04-07 | Dainippon Printing Co Ltd | Goods recommending device |
| CN109271592A (en) * | 2018-09-30 | 2019-01-25 | 东软集团股份有限公司 | A kind of item recommendation method, device, storage medium and program product |
| WO2020073534A1 (en) * | 2018-10-12 | 2020-04-16 | 平安科技（深圳）有限公司 | Pushing method and apparatus based on re-clustering, and computer device and storage medium |
| US20210191962A1 (en) * | 2020-05-27 | 2021-06-24 | Beijing Baidu Netcom Science Technology Co., Ltd. | Question answering method and language model training method, apparatus, device, and storage medium |
| WO2021249528A1 (en) * | 2020-06-12 | 2021-12-16 | 支付宝(杭州)信息技术有限公司 | Intelligent dialogue method and apparatus and electronic device |
| CN114548092A (en) * | 2022-02-24 | 2022-05-27 | 广州华多网络科技有限公司 | Customer service session scheduling method and device, equipment, medium and product thereof |
| CN114626909A (en) * | 2022-02-15 | 2022-06-14 | 广州伊智信息科技有限公司 | Commodity recommendation method, system, equipment and medium |
| CN114896517A (en) * | 2022-05-26 | 2022-08-12 | 西安建筑科技大学 | Commodity recommendation method, system, equipment and storage medium |
| JP7240065B1 (en) * | 2021-10-13 | 2023-03-15 | 泰 六川 | Liquor information processing device, program and storage medium |
-
2023
- 2023-10-20 CN CN202311372840.6A patent/CN117633165B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2011070422A (en) * | 2009-09-25 | 2011-04-07 | Dainippon Printing Co Ltd | Goods recommending device |
| CN109271592A (en) * | 2018-09-30 | 2019-01-25 | 东软集团股份有限公司 | A kind of item recommendation method, device, storage medium and program product |
| WO2020073534A1 (en) * | 2018-10-12 | 2020-04-16 | 平安科技（深圳）有限公司 | Pushing method and apparatus based on re-clustering, and computer device and storage medium |
| US20210191962A1 (en) * | 2020-05-27 | 2021-06-24 | Beijing Baidu Netcom Science Technology Co., Ltd. | Question answering method and language model training method, apparatus, device, and storage medium |
| WO2021249528A1 (en) * | 2020-06-12 | 2021-12-16 | 支付宝(杭州)信息技术有限公司 | Intelligent dialogue method and apparatus and electronic device |
| JP7240065B1 (en) * | 2021-10-13 | 2023-03-15 | 泰 六川 | Liquor information processing device, program and storage medium |
| CN114626909A (en) * | 2022-02-15 | 2022-06-14 | 广州伊智信息科技有限公司 | Commodity recommendation method, system, equipment and medium |
| CN114548092A (en) * | 2022-02-24 | 2022-05-27 | 广州华多网络科技有限公司 | Customer service session scheduling method and device, equipment, medium and product thereof |
| CN114896517A (en) * | 2022-05-26 | 2022-08-12 | 西安建筑科技大学 | Commodity recommendation method, system, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117633165B (en) | 2024-05-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11334635B2 (en) | Domain specific natural language understanding of customer intent in self-help | |
| CN110162693B (en) | Information recommendation method and server | |
| CN108304439B (en) | Semantic model optimization method and device, intelligent device and storage medium | |
| US8983971B2 (en) | Method, apparatus, and system for mobile search | |
| US11429405B2 (en) | Method and apparatus for providing personalized self-help experience | |
| CA3114298C (en) | Recommendation method and system and method and system for improving a machine learning system | |
| CN114266443A (en) | Data evaluation method and device, electronic equipment and storage medium | |
| CN115062732A (en) | Resource sharing cooperation recommendation method and system based on big data user tag information | |
| CN114881712B (en) | Intelligent advertisement putting method, device, equipment and storage medium | |
| CN112749330A (en) | Information pushing method and device, computer equipment and storage medium | |
| CN112818091A (en) | Object query method, device, medium and equipment based on keyword extraction | |
| CN111209351A (en) | Object relation prediction method and device, object recommendation method and device, electronic equipment and medium | |
| CN113792134B (en) | User service method and system based on digital twin technology | |
| CN113656699B (en) | User feature vector determining method, related equipment and medium | |
| CN112330387B (en) | Virtual broker applied to house watching software | |
| CN117633165B (en) | Intelligent AI customer service dialogue guiding method | |
| CN116957128A (en) | Service index prediction method, device, equipment and storage medium | |
| CN115293818A (en) | Advertisement putting and selecting method and device, equipment and medium thereof | |
| CN113010664B (en) | Data processing method and device and computer equipment | |
| CN114330519A (en) | Data determination method and device, electronic equipment and storage medium | |
| CN112035740A (en) | Project use duration prediction method, device, equipment and storage medium | |
| CN113781076B (en) | Prompting method, device, equipment and readable storage medium | |
| CN117609444A (en) | Searching question-answering method based on large model | |
| CN118193692A (en) | Intelligent customer service interaction method and system | |
| CN117272048A (en) | Data set processing method and device based on machine learning platform |
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 |