CN115587869A - A financial product recommendation speech generation method and device - Google Patents

Abstract

The application provides a financial product recommended word generating method and device, relates to the field of artificial intelligence, can also be used in the financial field, and comprises: analyzing the financial product description text to be analyzed according to a rule metadata set corresponding to the financial product description text to be analyzed to obtain corresponding financial product information; determining the corresponding financial product recommendation degree according to the financial product information, a pre-constructed financial product recommendation degree generation model and the historical financial product purchase information of the customer to be recommended; and writing the financial product information and the financial product recommendation degree into a pre-constructed standard conversational template to obtain a corresponding financial product recommendation conversational sentence. The method and the device can acquire the financial product information in the financial product description text based on the color label, and generate the recommendation dialect of the financial product based on the financial product information.

Description

A financial product recommendation speech generation method and device

technical field

This application relates to the field of artificial intelligence and can be used in the financial field, specifically a method for generating financial product recommendation words.

Background technique

In the financial field, there are various types of financial product brochures. In order to understand the financial products to be recommended to customers, product managers need to read the corresponding financial product brochures. However, the text volume of financial product manuals is generally large, and key information that affects the recommendation process is easily overwhelmed.

At present, there are mainly two ways to extract the above-mentioned key information: first, manual extraction, which requires high labor costs; second, use the application program interface to directly operate the product manual file, and set the data extraction rules through the underlying program , to extract relevant information, but this method needs to write the underlying program according to the different types of product manuals when extracting information. Once a new type of product manual is launched, the relevant extraction code needs to be rewritten, and the technical threshold is high. In the absence of automatic extraction of key information in financial product brochures, it is even more impossible to automatically generate recommendation words corresponding to financial products.

Contents of the invention

Aiming at the problems in the prior art, this application provides a method and device for generating financial product recommendation words, which can obtain financial product information in financial product description text based on color labels, and generate financial product recommendation words based on the financial product information. surgery.

In order to solve the above technical problems, the application provides the following technical solutions:

In the first aspect, the present application provides a financial product recommendation speech generation method, including:

Analyzing the financial product description text to be parsed according to the rule metadata set corresponding to the financial product description text to be parsed to obtain corresponding financial product information;

Determine the corresponding financial product recommendation degree according to the financial product information, the pre-built financial product recommendation degree generation model and the financial product historical purchase information of the customer to be recommended;

Write the financial product information and the financial product recommendation degree into the pre-built standard speech template to obtain the corresponding financial product recommendation speech.

Further, the analyzing the financial product description text to be parsed according to the rule metadata set corresponding to the financial product description text to be parsed to obtain corresponding financial product information includes:

Reading the full amount of data in the description text of the financial product to be analyzed by using the API to obtain the corresponding target metadata set;

Carry out color labeling on the description text of the financial product to be analyzed according to the preset analysis keywords, and obtain the description text of the financial product containing the color label;

Traverse the target metadata set according to the rules in the rule metadata set, and read the financial product information specified by the rules; wherein, the rules are set based on the color labels.

Further, the determination of the corresponding financial product recommendation degree based on the financial product information, the pre-built financial product recommendation degree generation model and the financial product historical purchase information of the customer to be recommended includes:

Calculate the expected net value and expected net value yield of the financial product according to the financial product information;

Input the historical purchase information of the financial product, the financial product rate in the financial product information, the net value and the rate of return on the net value into the financial product recommendation generation model, and generate the financial product for the customer to be recommended. Product recommendation.

Further, the calculation of the expected net value and expected net value yield of the financial product according to the financial product information includes:

Obtaining the historical net value sequence of the financial product from the financial product information;

Calculating a corresponding historical equity yield sequence according to the historical equity sequence;

Using a time series analysis algorithm to fit the historical net worth sequence and the historical net worth return rate sequence to obtain the expected net worth and expected net worth return rate.

Further, inputting the historical purchase information of the financial product, the financial product rate in the financial product information, the net value and the rate of return on the net value into the financial product recommendation generation model, and generating the financial product for the financial product The financial product recommendation degree of the customer to be recommended, including:

Generate a corresponding recommended parameter feature set according to the historical purchase information of the financial product, the financial product rate in the financial product information, the net value and the net value yield;

Inputting the recommendation parameter feature set into the financial product recommendation generation model to generate the financial product recommendation degree of the financial product for the customer to be recommended.

Further, the step of constructing the financial product recommendation generation model includes:

Generate a historical feature set of recommended parameters corresponding to each customer based on the historical purchase information of each customer's financial product, the financial product rate in the financial product information, the net value and the net value yield;

According to the purchase results of each customer for the financial product, mark the positive and negative samples of each recommended parameter historical feature set;

The positive and negative sample marking results and the recommended parameter historical feature set are input into a distributed gradient boosting model for training to obtain the financial product recommendation generation model.

Further, the financial product recommendation speech generation method also includes:

Collecting the audio data when the financial product customer manager uses the financial product recommendation speech technique to recommend the corresponding financial product to the customer to be recommended, and generating corresponding audio conversion text;

Comparing the audio conversion text with the financial product recommendation speech, and performing early warning processing based on the comparison result.

Further, in the method for generating financial product recommendation speech, the comparison of the audio conversion text and the financial product recommendation speech, and performing early warning processing based on the comparison result includes:

The audio conversion text is input into the BERT model for semantic analysis to obtain the corresponding first sentence label;

Inputting the words recommended by the financial product into the BERT model for semantic analysis to obtain the corresponding second clause label;

Comparing the label of the first clause with the label of the second clause, if inconsistent, give an early warning.

In the second aspect, the present application provides a financial product recommendation speech generation device, including:

A financial product information generating unit, configured to parse the financial product description text to be parsed according to the rule metadata set corresponding to the financial product description text to be parsed, and obtain corresponding financial product information;

The recommendation degree calculation unit is used to determine the corresponding financial product recommendation degree according to the financial product information, the pre-built financial product recommendation degree generation model and the financial product historical purchase information of the customer to be recommended;

The recommended speech unit is configured to write the financial product information and the financial product recommendation degree into a pre-built standard speech template to obtain the corresponding financial product recommended speech.

Further, the financial product information generation unit includes:

The target metadata set generation module is used to read the full amount of data in the description text of the financial product to be parsed by using the API to obtain the corresponding target metadata set;

A color label generation module, configured to color-label the description text of the financial product to be analyzed according to the preset analysis keywords, and obtain the description text of the financial product containing the color label;

The product information extraction module is configured to traverse the target metadata set according to the rules in the rule metadata set, and read the financial product information specified by the rules; wherein, the rules are set based on the color labels.

Further, the recommendation calculation unit includes:

A net value parameter calculation module, used to calculate the expected net value and expected net value yield of the financial product according to the financial product information;

The recommendation degree calculation module is used to input the historical purchase information of the financial product, the financial product rate in the financial product information, the net value and the net value return rate into the financial product recommendation degree generation model to generate the financial product The financial product recommendation degree for the customer to be recommended.

Further, the net worth parameter calculation module includes:

A historical net value sequence generating module, configured to obtain the historical net value sequence of the financial product from the financial product information;

A historical net worth yield sequence generation module, used to calculate a corresponding historical net worth yield sequence according to the historical net worth sequence;

The net worth parameter expectation module is used to use a time series analysis algorithm to fit the historical net worth sequence and the historical net worth return rate sequence to obtain the expected net worth and expected net worth return rate.

Further, the recommendation calculation module includes:

The parameter feature set generating submodule is used to generate a corresponding recommended parameter feature set according to the historical purchase information of the financial product, the financial product rate in the financial product information, the net value and the return rate of the net value;

The recommendation degree calculation sub-module is used to input the recommendation parameter feature set into the financial product recommendation degree generation model, and generate the financial product recommendation degree of the financial product for the customer to be recommended.

Further, the financial product recommendation speech generation device also includes:

A parameter historical feature set generation unit, used to generate a recommended parameter historical feature set corresponding to each customer according to the historical purchase information of financial products of each customer, the financial product rate in the financial product information, the net value and the net value yield;

The sample marking unit is used to mark positive and negative samples of each recommended parameter historical feature set according to the purchase result of each customer for the financial product;

A model training unit, configured to input the positive and negative sample marking results and the recommended parameter historical feature set into a distributed gradient boosting model for training to obtain the financial product recommendation degree generation model.

Further, the financial product recommendation speech generation device also includes:

The audio conversion unit is used to collect the audio data when the financial product account manager uses the financial product recommendation speech technique to recommend the corresponding financial product to the customer to be recommended, and generate corresponding audio conversion text;

The comparison and warning unit is used to compare the audio conversion text with the financial product recommendation speech, and perform warning processing based on the comparison result.

Further, the comparison warning unit includes:

The first sentence marking module is used to input the audio conversion text into the BERT model and carry out semantic analysis to obtain the corresponding first sentence label;

The second clause marking module is used to input the financial product recommendation speech technique into the BERT model for semantic analysis to obtain the corresponding second clause label;

The comparison and early warning module is used to compare the first clause label and the second clause label, and if they are inconsistent, give an early warning.

In a third aspect, the present application provides an electronic device including a memory, a processor, and a computer program stored on the memory and operable on the processor, and when the processor executes the program, the financial product recommendation speech generation is realized method steps.

In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method for generating financial product recommendation speech are realized.

In a fifth aspect, the present application provides a computer program product, including a computer program/instruction, and when the computer program/instruction is executed by a processor, the steps of the method for generating a financial product recommendation speech are implemented.

Aiming at the problems in the prior art, the financial product recommendation speech generation method and device provided by this application can obtain the financial product information in the financial product description text based on the color label, and generate the financial product recommendation speech based on the financial product information ; Combining the color label-based financial product information extraction method with the financial product recommendation generation model, it can analyze the financial product description text, and provide financial product customer managers with financial product recommendation words, and even target financial product customers Managers use financial product recommendation skills to warn customers of possible slips of the tongue when recommending financial products; thus guiding customers to adapt to and accept the net value transformation of financial products, reasonably assess their own risk tolerance, and improve customer investment experience.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. Those skilled in the art can also obtain other drawings based on these drawings without creative work.

Fig. 1 is one of the flowcharts of the financial product recommendation speech generation method in the embodiment of the present application;

Fig. 2 is the flowchart of obtaining the corresponding financial product information in the embodiment of the present application;

FIG. 3 is a flow chart of determining the corresponding financial product recommendation degree in the embodiment of the present application;

Fig. 4 is the flowchart of calculating the expected net value and expected net value yield of financial products in the embodiment of the present application;

Fig. 5 is a flow chart of generating the financial product recommendation degree of the financial product for the customer to be recommended in the embodiment of the present application;

FIG. 6 is a flow chart of building a financial product recommendation generation model in the embodiment of the present application;

Fig. 7 is the second flow chart of the financial product recommendation speech generation method in the embodiment of the present application;

FIG. 8 is a flowchart of early warning processing in the embodiment of the present application;

FIG. 9 is one of the structural diagrams of the financial product recommendation speech generation device in the embodiment of the present application;

FIG. 10 is a structural diagram of a financial product information generation unit in the embodiment of the present application;

FIG. 11 is a structural diagram of a recommendation calculation unit in an embodiment of the present application;

Fig. 12 is the structural diagram of the net value parameter calculation module in the embodiment of the present application;

Fig. 13 is a structural diagram of the recommendation calculation module in the embodiment of the present application;

Fig. 14 is the second structural diagram of the financial product recommendation speech generating device in the embodiment of the present application;

Fig. 15 is the third structural diagram of the financial product recommendation speech generating device in the embodiment of the present application;

Fig. 16 is a structural diagram of the comparison warning unit in the embodiment of the present application;

FIG. 17 is a schematic structural diagram of an electronic device in an embodiment of the present application;

FIG. 18 is a schematic flow diagram of a method for generating financial product recommendation words in the embodiment of the present application;

Fig. 19 is one of the schematic diagrams of model training in the embodiment of the present application;

FIG. 20 is the second schematic diagram of model training in the embodiment of the present application.

detailed description

The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all of them. Based on the embodiments in this application, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the scope of protection of this application.

It should be noted that the financial product recommendation speech generation method and device provided in this application can be used in the financial field, and can also be used in any field other than the financial field. The financial product recommendation speech generation method and device provided in this application The field of application is not limited.

The acquisition, storage, use, and processing of data in the technical solution of this application all comply with the relevant provisions of national laws and regulations.

In one embodiment, referring to Fig. 1, in order to obtain the financial product information in the financial product description text based on the color label, and generate a financial product recommendation speech based on the financial product information, this application provides a financial product recommendation speech generation method. The method firstly extracts the financial product information in the financial product description text based on the color label, and then generates a financial product recommendation speech based on the financial product information, so that the financial product account manager can use the recommended speech to recommend the customer (the financial potential buyers of the product) to recommend the financial product.

It should be noted that when generating financial product recommendation speech, the financial product recommendation speech generation method provided by this application is to combine the color labeling method with the financial product recommendation degree generation model, analyze the financial product description text, and Provide financial product recommendation speech skills for financial product account managers, and then use text comparison technology to address possible slips of the tongue (such as reading aloud) when financial product account managers use the financial product recommendation speech skills to recommend financial products to customers to be recommended error) put forward an early warning in order to guide customers to adapt to and accept the net value transformation of financial products, reasonably assess their own risk tolerance, and improve customer investment experience.

The financial product recommendation speech generation method provided by this application includes:

S101: Analyze the financial product description text to be analyzed according to the rule metadata set corresponding to the financial product description text to be analyzed, and obtain corresponding financial product information;

Specifically, in one embodiment, referring to FIG. 2 , step S101 includes: reading the full amount of data in the description text of the financial product to be analyzed by using an API to obtain a corresponding target metadata set (S201); Set the analytical keywords to color-label the financial product description text to be parsed to obtain the financial product description text containing color labels (S202); traverse the target metadata set according to the rules in the rule metadata set, and Read financial product information specified by the rules (S203); wherein, the rules are set based on the color labels.

The specific implementation is that the flow from step S201 to step S203 is as follows:

1. Use the application programming interface (such as Java POI) to read the full amount of data in the financial product description text (also called product manual) to be parsed to obtain the corresponding target metadata set. The target metadata set includes the full amount of data of the financial product description text.

2. For the product description text, set the color and font color of each cell in each table to obtain the financial product description text with color labels. Then, according to the color and font color of each cell above, the rule set is set for the product brochure template to obtain the rule metadata set. The rule metadata set at least records: what information is represented by the fonts of each color in the cells of each color.

3. According to business needs, select the specification template corresponding to the financial product specification text, and extract the rule metadata set according to the specification template.

3. Use the rule metadata set to model the rule set.

4. Use the rule set model to collect the information of the target metadata set.

5. The extracted financial product information is stored in storage media such as files and databases.

Specific examples are as follows:

In this example, the rules in the rules metadata set are as follows:

1) The orange background color of the table is used to determine the range of extracted information, which is the cell on the right side of the cell, where the font color "black" is fixed information, and the font color "green" is the entry item, and the format is "[table name/column name ]”

2) The background color of the table "dark blue" is used to determine the range of extracted information, which is the vertical range on the right side of the unit. The font color "white" in this table is fixed information, and the font color "yellow" is the entry item. The format is "[ Table Name]"

3) The table background color "light blue" is used to determine the range of extracted information, which is the entire column below the cell. The font color "black" in this table is fixed information, and the font color "orange" is the entry item. The format is "[ column name]"

Some parts of the product brochure template provided by a financial product issuing company are shown in Table 1 below:

Table 1

If the business scenario requires extraction of the fund name and Class A subscription fee information in the product brochure, the problem can be transformed into:

1) Extract the cell information on the right side of the "Fund Name" table, and store the relevant information in the JJZC field in the BASE_INFO table of the database.

2) Extract all the values under the "subscription amount (M)" column and the "rate" column in the right range of the "Class A subscription fee" table, and store the relevant information in the MONEY and RATE fields in the SGJE_INFO table of the database. Then business personnel can change the rule template in the following way by adjusting the background color of the form (see Table 2):

Table 2

When a financial product issuing company provides a new product specification, the steps to extract key information are:

For an example of the product specification, see Table 3:

table 3

The first step is to use Java POI to parse the uploaded product manual and get the target metadata set.

The second step is to find the corresponding template according to the product specification template type in the target metadata set, and extract the rule metadata set. An example rule metadata set is:

{

…………

Product brochure template: ...;

Rule 1:

{

Fixed information: fund manager;

Extraction range: right side;

Format: BASE_INFO/JJZC;

range append: []

};

Rule 2:

{

Fixed information: Class A subscription fee;

Extraction range: the vertical range on the right;

Format: SGJE_INFO;

other:[{

Fixed information: subscription amount (M);

Extract range: whole column;

Format: SGJE_INFO/MONEY;

range append: []

},{

Fixed information: rate;

Extract range: whole column;

Format: SGJE_INFO/RATE;

range append: []

};

…………

}

The third step is to traverse the rule metadata set, find rule 1, and extract information such as "fund manager", "right side", "BASE_INFO/JJZC" and so on.

The fourth step is to traverse the rule to extract the target metadata set, find the data in the "right" cell of the "fund manager" table, that is, "a certain hybrid securities investment fund", and insert it into the JJZC field in the BASE_INFO table.

The fifth step is to continue to traverse rule 2, rule 3, etc. until all rules are traversed.

It can be seen from the above description that the financial product recommendation speech generation method provided by this application can analyze the financial product description text to be parsed according to the rule metadata set corresponding to the financial product description text to be parsed, and obtain corresponding financial product information.

S102: Determine the corresponding financial product recommendation degree according to the financial product information, the pre-built financial product recommendation degree generation model, and the financial product historical purchase information of the customer to be recommended;

Specifically, in one embodiment, referring to FIG. 3 , step S102 includes: calculating the expected net value and expected net value rate of return of the financial product according to the financial product information (S301); The financial product fee rate, the net value and the net value return rate in the financial product information are input into the financial product recommendation generation model to generate the financial product recommendation degree of the financial product for the customer to be recommended (S302).

Further, referring to FIG. 4 , step S301 includes: obtaining the historical net value sequence of the financial product from the financial product information (S401); calculating the corresponding historical net value yield sequence according to the historical net value sequence (S402); Using a time series analysis algorithm to fit the historical net worth sequence and the historical net worth rate of return sequence to obtain the expected net worth and expected net worth rate of return (S403).

During specific implementation, use statistical and time series analysis techniques to analyze product net value and income:

Step 1: Integrate the obtained net worth data {x _t , x _t-1 , x _t-2 ,…x _tn }; for example, a certain net worth wealth management company’s initial net worth is 1 on August 1, year ××××, and within a month There have been 3 changes in the net value, namely, the net value on August 5, ×××× was 1.002, the net value on August 12, ×××× was 1.006, and the net value on August 30, ×××× was 1.005, then the integration The acquired net worth data is {1, 1.002, 1.006, 1.005};

其中x_t为当前净值，x_t-n为初始净值，n为产品运营天数；根据步骤1的示例，如果投资者首发认购该净值型理财，那么××××年8月30日的净值收益率就是(1.005-1)×365/30＝6.08％；Step 2: Calculate the return on equity R _t ,

Among them, x _t is the current net value, x _tn is the initial net value, and n is the number of days of product operation; according to the example of step 1, if the investor first subscribes for the net value-based wealth management, then the net value return on August 30 of ×××× is (1.005-1)×365/30=6.08%;

Step 3: Draw net worth and net yield curves based on the net worth sequence and the calculated net worth yield sequence to reflect changes and discover trend laws.

Step 4: For sequences with regular trends, use time series analysis and forecasting techniques, such as LSTM and other model techniques, to predict the future net value of the product x _t+m = f ₁ (x _t ,x _t-1 ,…x _tn ) and net value Yield rate R _t+m = f ₂ (R _t , R _t-1 , R _t-2 , . . . , R _tn ).

It can be seen from the above description that the financial product recommendation speech generation method provided by this application can calculate the expected net value and expected net value return rate of the financial product according to the financial product information.

Further, referring to FIG. 5, step S302 includes: generating a corresponding recommended parameter feature set according to the historical purchase information of the financial product, the financial product rate in the financial product information, the net value and the rate of return on the net value (S501) Input the recommended parameter feature set into the financial product recommendation degree generation model, and generate the financial product recommendation degree of the financial product for the customer to be recommended (S502).

It should be noted that, in one embodiment, referring to FIG. 6 , the steps of constructing the financial product recommendation generation model include:

S601: Generate the historical feature set of recommended parameters corresponding to each customer according to the historical purchase information of financial products of each customer, the financial product rate in the financial product information, the net value and the net value rate of return; for example, integrate the net value data of the product (record is N), net value yield data (denoted as R), purchase cost data (denoted as C), and at the same time integrate the customer’s data on whether to hold other investment products while holding the product (denoted as Q), product mix adjustment (denoted as be recorded as T) and other data to form a modeling data feature set (recorded as X), then X={N, R, C, Q, T};

S602: According to the purchase results of each customer for the financial product, mark the positive and negative samples of each recommended parameter historical feature set; for example, identify the positive and negative samples according to whether the customer holds the product; The customer is identified as 1, otherwise it is identified as 0, as the positive and negative sample labels, recorded as Y;

S603: Input the positive and negative sample marking results and the recommended parameter historical feature set into a distributed gradient boosting model for training to obtain the financial product recommendation generation model. For example, modeling training is performed according to the modeling process of classic machine learning models (such as XGBoost, logistic regression, etc.); for example, model data (X, Y) is preprocessed first, and then the model is established and model training is performed, as shown in Figure 19 and Figure 19. 20 shown.

Then, the trained model is used to predict customers who do not own the product to obtain the recommendation potential of the product (that is, the recommendation probability predicted by the model), and then combine the experience of marketers to judge whether it is necessary to recommend the product to customers.

It can be seen from the above description that the financial product recommendation speech generation method provided by this application can input the historical purchase information of the financial product, the financial product rate in the financial product information, the net value and the rate of return on the net value into the financial product The product recommendation degree generation model generates the financial product recommendation degree of the financial product for the customer to be recommended.

S103: Write the financial product information and the financial product recommendation degree into a pre-built standard speech template to obtain a corresponding financial product recommendation speech.

Specifically, in one embodiment, referring to FIG. 7 , the method for generating financial product recommendation words further includes:

S701: Collect audio data when the financial product account manager recommends the corresponding financial product to the customer to be recommended using the financial product recommendation speech technique, and generate corresponding audio conversion text;

S702: Compare the audio conversion text with the financial product recommendation speech, and perform early warning processing based on the comparison result.

Further, referring to FIG. 8, step S702 includes: inputting the audio conversion text into the BERT model for semantic analysis to obtain the corresponding first sentence label (S801); inputting the financial product recommendation speech into the BERT model Perform semantic analysis to obtain the corresponding second clause label (S802); compare the first clause label and the second clause label, and give an early warning if they are inconsistent (S803).

It can be understood that, referring to FIG. 18 , the above steps S701 to S702 can be specifically executed as follows:

Step 1: Set a standard speech template.

Step 2: The financial product account manager enters the financial product code on the front end of the webpage;

Step 3: The background generates a parameter list according to the financial product code, for example {"pdid":xxxxx,"jjlx":xxxxx,"jjzc":xxxxx,……………}, read the corresponding value in the list, Combined with the standard speech template, a financial product recommendation speech is formed, which is transmitted to the front-end page and provided to the financial product account manager;

Step 4: The financial product account manager reads aloud according to the words recommended by the financial products. The front-end collects the voice through the high-speed camera, and uses ASR technology to convert the acquired voice into structured data; at the same time, it also performs structured processing on the words recommended by the financial products.

Step 5: Build the BERT text model for the structured data corresponding to the words spoken by the financial product account manager in step 4 and the structured data corresponding to the words recommended by the financial product, and use short sentence sliding windows to cut them according to different semantics points, label the clause labels respectively.

Therefore, at the tag level, it is possible to compare whether the audio conversion text is consistent with the words recommended by financial products, and if not, an early warning message can be issued.

Finally, for example, the language of financial product recommendation can be:

"What you are about to buy is a ×××× fund, the manager is ×××× fund company, the product code is ××××, and the issuance method is public offering. The net value data of this product is ××××, and the net value return It is ××××, after analysis, this product will maintain a steady growth trend, and the matching degree with your investment tendency is ××××, so we recommend you to buy this product. Do you know and agree with the above situation?"

It can be seen from the above description that the financial product recommendation speech generation method provided by this application can obtain the financial product information in the financial product description text based on the color label, and generate the financial product recommendation speech based on the financial product information; Combining the financial product information extraction method with the financial product recommendation generation model, it can analyze the financial product description text, provide financial product recommendation words for financial product account managers, and even use financial product recommendation for financial product account managers Huashu gives early warning to customers of slips of the tongue that may occur when recommending financial products; thereby guiding customers to adapt and accept the transformation of financial products to net value, reasonably assess their own risk tolerance, and improve customer investment experience.

Based on the same inventive concept, the embodiment of the present application also provides an apparatus for generating financial product recommendation words, which can be used to implement the methods described in the above embodiments, as described in the following embodiments. Since the problem-solving principle of the financial product recommendation speech generation device is similar to the financial product recommendation speech generation method, the implementation of the financial product recommendation speech generation device can refer to the implementation of the software performance benchmark determination method, and the repetition will not be repeated. As used below, the term "unit" or "module" may be a combination of software and/or hardware that realizes a predetermined function. Although the systems described in the following embodiments are preferably implemented in software, implementations in hardware, or a combination of software and hardware are also possible and contemplated.

In one embodiment, referring to FIG. 9 , in order to obtain the financial product information in the financial product description text based on the color label, and generate a financial product recommendation speech based on the financial product information, this application provides a financial product recommendation speech generation The device includes: a financial product information generation unit 901 , a recommendation degree calculation unit 902 and a recommended speaking unit 903 .

Financial product information generation unit 901, configured to analyze the financial product description text to be analyzed according to the rule metadata set corresponding to the financial product description text to be analyzed, and obtain corresponding financial product information;

The recommendation degree calculation unit 902 is configured to determine the corresponding financial product recommendation degree according to the financial product information, the pre-built financial product recommendation degree generation model and the financial product historical purchase information of the customer to be recommended;

The recommendation speech unit 903 is configured to write the financial product information and the financial product recommendation degree into a pre-built standard speech template to obtain the corresponding financial product recommendation speech.

In one embodiment, referring to FIG. 10 , the financial product information generation unit 901 includes: a target metadata set generation module 1001 , a color label generation module 1002 and a product information extraction module 1003 .

The target metadata set generation module 1001 is used to read the full amount of data in the description text of the financial product to be analyzed by using the API to obtain the corresponding target metadata set;

A color label generation module 1002, configured to color-label the explanatory text of the financial product to be analyzed according to the preset analysis keywords, to obtain the explanatory text of the financial product containing the color label;

The product information extraction module 1003 is configured to traverse the target metadata set according to the rules in the rule metadata set, and read the financial product information specified by the rules; wherein, the rules are set based on the color labels.

In one embodiment, referring to FIG. 11 , the recommendation calculation unit 902 includes: a net worth parameter calculation module 1101 and a recommendation calculation module 1102 .

A net value parameter calculation module 1101, configured to calculate the expected net value and expected net value yield of the financial product according to the financial product information;

The recommendation degree calculation module 1102 is configured to input the historical purchase information of the financial product, the financial product rate in the financial product information, the net value and the rate of return on the net value into the financial product recommendation generation model to generate the financial product The financial product recommendation degree of the product for the customer to be recommended.

In one embodiment, referring to FIG. 12 , the equity parameter calculation module 1101 includes: a historical equity sequence generation module 1201 , a historical equity return sequence generation module 1202 and an equity parameter expectation module 1203 .

A historical net value sequence generation module 1201, configured to obtain the historical net value sequence of the financial product from the financial product information;

A historical equity yield sequence generating module 1202, configured to calculate a corresponding historical equity yield sequence according to the historical equity sequence;

The net worth parameter prediction module 1203 is used to use time series analysis algorithm to fit the historical net worth sequence and the historical net worth rate of return sequence to obtain the expected net worth and expected net worth rate of return.

In one embodiment, referring to FIG. 13 , the recommendation degree calculation module 1102 includes: a parameter feature set generation submodule 1301 and a recommendation degree calculation submodule 1302 .

The parameter feature set generation submodule 1301 is used to generate a corresponding recommended parameter feature set according to the historical purchase information of the financial product, the financial product rate in the financial product information, the net value and the rate of return on the net value;

The recommendation degree calculation sub-module 1302 is configured to input the recommendation parameter feature set into the financial product recommendation degree generation model, and generate the financial product recommendation degree of the financial product for the customer to be recommended.

In one embodiment, referring to FIG. 14 , the device for generating words for financial product recommendation further includes: a parameter history feature set generation unit 1401 , a sample labeling unit 1402 and a model training unit 1403 .

The parameter historical feature set generating unit 1401 is used to generate the recommended parameter historical feature set corresponding to each customer according to the historical purchase information of financial products of each customer, the financial product rate in the financial product information, the net value and the net value yield;

A sample marking unit 1402, configured to mark positive and negative samples of each recommended parameter historical feature set according to the purchase result of each customer for the financial product;

The model training unit 1403 is configured to input the positive and negative sample marking results and the recommended parameter historical feature set into a distributed gradient boosting model for training to obtain the financial product recommendation degree generation model.

In one embodiment, referring to FIG. 15 , the device for generating words for financial product recommendation further includes: an audio conversion unit 1501 and a comparison and warning unit 1502 .

The audio conversion unit 1501 is configured to collect audio data when the financial product account manager recommends the corresponding financial product to the customer to be recommended using the financial product recommendation speech technique, and generate corresponding audio conversion text;

The comparison and warning unit 1502 is configured to compare the audio conversion text with the financial product recommendation words, and perform warning processing based on the comparison result.

In one embodiment, referring to FIG. 16 , the comparison and warning unit 1502 includes: a first sentence marking module 1601, a second sentence marking module 1602 and a comparison and warning module 1603.

The first sentence marking module 1601 is used to input the audio conversion text into the BERT model for semantic analysis to obtain the corresponding first sentence label;

The second clause marking module 1602 is configured to input the financial product recommendation speech into the BERT model for semantic analysis to obtain the corresponding second clause label;

The comparison and early warning module 1603 is configured to compare the first sentence tag with the second sentence tag, and if they are inconsistent, give a warning.

From the perspective of hardware, in order to obtain the financial product information in the financial product description text based on the color label, and generate a financial product recommendation language based on the financial product information, this application provides a method for implementing the financial product recommendation language An embodiment of an electronic device with all or part of the content in the method for generating a technique, the electronic device specifically includes the following content:

A processor (Processor), a memory (Memory), a communication interface (Communications Interface) and a bus; wherein, the processor, the memory, and the communication interface complete mutual communication through the bus; the communication interface is used to realize the Information transmission between the financial product recommendation speech generating device and related equipment such as core business systems, user terminals, and related databases; the logic controller can be a desktop computer, tablet computer, and mobile terminal, and this embodiment is not limited thereto. In this embodiment, the logic controller can be implemented with reference to the embodiments of the method for generating financial product recommendation words and the embodiment of the device for generating financial product recommendation words in the embodiments, the contents of which are incorporated herein and repeated I won't repeat them here.

It can be understood that the user terminal may include smart phones, tablet electronic devices, network set-top boxes, portable computers, desktop computers, personal digital assistants (PDAs), vehicle-mounted devices, smart wearable devices, and the like. Wherein, the smart wearable device may include smart glasses, smart watches, smart bracelets and the like.

In practical applications, part of the method for generating financial product recommendation words may be executed on the side of the electronic device as described above, or all operations may be completed in the client device. Specifically, the selection may be made according to the processing capability of the client device, the limitation of the user usage scenario, and the like. This application is not limited to this. If all operations are completed in the client device, the client device may further include a processor.

The above-mentioned client device may have a communication module (that is, a communication unit), which can communicate with a remote server to realize data transmission with the server. The server may include a server on the side of the task scheduling center, and may also include a server of an intermediate platform in other implementation scenarios, such as a server of a third-party server platform that has a communication link with the server of the task scheduling center. The server may include a single computer device, or a server cluster composed of multiple servers, or a server structure of a distributed device.

FIG. 17 is a schematic block diagram of a system configuration of an electronic device 9600 according to an embodiment of the present application. As shown in FIG. 17 , the electronic device 9600 may include a central processing unit 9100 and a memory 9140 ; the memory 9140 is coupled to the central processing unit 9100 . It should be noted that this FIG. 17 is exemplary; other types of structures may also be used to supplement or replace this structure, so as to realize telecommunication functions or other functions.

In an embodiment, the function of the financial product recommendation speech generation method may be integrated into the central processing unit 9100 . Wherein, the central processing unit 9100 may be configured to perform the following control:

S101: Analyze the financial product description text to be analyzed according to the rule metadata set corresponding to the financial product description text to be analyzed, and obtain corresponding financial product information;

S102: Determine the corresponding financial product recommendation degree according to the financial product information, the pre-built financial product recommendation degree generation model, and the financial product historical purchase information of the customer to be recommended;

S103: Write the financial product information and the financial product recommendation degree into a pre-built standard speech template to obtain a corresponding financial product recommendation speech.

It can be seen from the above description that the financial product recommendation speech generation method provided by this application can obtain the financial product information in the financial product description text based on the color label, and generate the financial product recommendation speech based on the financial product information; Combining the financial product information extraction method with the financial product recommendation generation model, it can analyze the financial product description text, provide financial product recommendation words for financial product account managers, and even use financial product recommendation for financial product account managers Huashu gives early warning to customers of slips of the tongue that may occur when recommending financial products; thereby guiding customers to adapt and accept the transformation of financial products to net value, reasonably assess their own risk tolerance, and improve customer investment experience.

In another embodiment, the financial product recommendation speech generation device may be configured separately from the central processing unit 9100, for example, the financial product recommendation speech generation device of the data composite transmission device may be configured as a chip connected to the central processing unit 9100, through the central processing unit 9100 The processor is controlled to realize the function of the financial product recommendation speech generation method.

As shown in FIG. 17 , the electronic device 9600 may further include: a communication module 9110 , an input unit 9120 , an audio processor 9130 , a display 9160 , and a power supply 9170 . It should be noted that the electronic device 9600 does not necessarily include all the components shown in FIG. 17; in addition, the electronic device 9600 may also include components not shown in FIG. 17, and reference may be made to the prior art.

As shown in FIG. 17 , the central processing unit 9100 is also sometimes referred to as a controller or operating control, and may include a microprocessor or other processor devices and/or logic devices. The central processing unit 9100 receives input and controls various components of the electronic device 9600. The operation of the component.

Wherein, the memory 9140 may be, for example, one or more of a cache, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory, or other suitable devices. The above-mentioned failure-related information may be stored, and a program for executing the related information may also be stored. And the central processing unit 9100 can execute the program stored in the memory 9140 to realize information storage or processing.

The input unit 9120 provides input to the central processing unit 9100 . The input unit 9120 is, for example, a button or a touch input device. The power supply 9170 is used to provide power to the electronic device 9600 . The display 9160 is used to display display objects such as images and characters. The display can be, for example, an LCD display, but is not limited thereto.

The memory 9140 may be a solid state memory, for example, a read only memory (ROM), a random access memory (RAM), a SIM card, and the like. There can also be memory that retains information even when power is off, can be selectively erased and is provided with more data, an example of which is sometimes called EPROM or the like. Memory 9140 may also be some other type of device. The memory 9140 includes buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 may include an application/function storage part 9142 for storing application programs and function programs or procedures for executing operations of the electronic device 9600 through the central processing unit 9100 .

The memory 9140 may also include a data storage unit 9143 for storing data such as contacts, digital data, pictures, sounds and/or any other data used by the electronic device. The driver storage part 9144 of the memory 9140 may include various drivers of the electronic device for communication functions and/or for executing other functions of the electronic device (such as messaging applications, address book applications, etc.).

The communication module 9110 is a transmitter/receiver 9110 that transmits and receives signals via an antenna 9111 . The communication module (transmitter/receiver) 9110 is coupled to the central processing unit 9100 to provide input signals and receive output signals, which can be the same as that of a conventional mobile communication terminal.

Based on different communication technologies, multiple communication modules 9110, such as a cellular network module, a Bluetooth module and/or a wireless local area network module, may be provided in the same electronic device. The communication module (transmitter/receiver) 9110 is also coupled to a speaker 9131 and a microphone 9132 via an audio processor 9130 to provide audio output via the speaker 9131 and receive audio input from the microphone 9132, thereby implementing common telecommunication functions. Audio processor 9130 may include any suitable buffers, decoders, amplifiers, etc. In addition, the audio processor 9130 is also coupled to the central processing unit 9100, so that the microphone 9132 can be used to record on the machine, and the speaker 9131 can be used to play the sound stored on the machine.

The embodiments of the present application also provide a computer-readable storage medium capable of implementing all the steps in the method for generating financial product recommendation words in the above-mentioned embodiments in which the executive body is the server or the client, and the computer-readable storage medium stores the A computer program, when the computer program is executed by a processor, implements all the steps of the method for generating financial product recommendation words in the above embodiments in which the execution subject is a server or a client, for example, when the processor executes the computer program, the following is implemented step:

S101: Analyze the financial product description text to be analyzed according to the rule metadata set corresponding to the financial product description text to be analyzed, and obtain corresponding financial product information;

S102: Determine the corresponding financial product recommendation degree according to the financial product information, the pre-built financial product recommendation degree generation model, and the financial product historical purchase information of the customer to be recommended;

S103: Write the financial product information and the financial product recommendation degree into a pre-built standard speech template to obtain a corresponding financial product recommendation speech.

It can be seen from the above description that the financial product recommendation speech generation method provided by this application can obtain the financial product information in the financial product description text based on the color label, and generate the financial product recommendation speech based on the financial product information; Combining the financial product information extraction method with the financial product recommendation generation model, it can analyze the financial product description text, provide financial product recommendation words for financial product account managers, and even use financial product recommendation for financial product account managers Huashu gives early warning to customers of slips of the tongue that may occur when recommending financial products; thereby guiding customers to adapt and accept the transformation of financial products to net value, reasonably assess their own risk tolerance, and improve customer investment experience.

Those skilled in the art should understand that the embodiments of the present invention may be provided as methods, apparatuses, or computer program products. Accordingly, the present invention can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (apparatus), and computer program products according to embodiments of the invention. It should be understood that each procedure and/or block in the flowchart and/or block diagram, and a combination of procedures and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions may be provided to a general purpose computer, special purpose computer, embedded processor, or processor of other programmable data processing equipment to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing equipment produce a An apparatus for realizing the functions specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to operate in a specific manner, such that the instructions stored in the computer-readable memory produce an article of manufacture comprising instruction means, the instructions The device realizes the function specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.

These computer program instructions can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby The instructions provide steps for implementing the functions specified in the flow chart or blocks of the flowchart and/or the block or blocks of the block diagrams.

In the present invention, specific examples have been applied to explain the principles and implementation methods of the present invention. The description of the above examples is only used to help understand the method of the present invention and its core idea; meanwhile, for those of ordinary skill in the art, according to this The idea of the invention will have changes in the specific implementation and scope of application. To sum up, the contents of this specification should not be construed as limiting the present invention.

Claims (12)

1. A method for generating a financial product recommendation statement, comprising:

analyzing the financial product description text to be analyzed according to a rule metadata set corresponding to the financial product description text to be analyzed to obtain corresponding financial product information;

determining corresponding financial product recommendation degrees according to the financial product information, a pre-constructed financial product recommendation degree generation model and the historical financial product purchasing information of the customer to be recommended;

and writing the financial product information and the financial product recommendation degree into a pre-constructed standard conversational template to obtain a corresponding financial product recommendation conversational sentence.

2. The method according to claim 1, wherein the parsing the description text of the financial product to be parsed according to the rule metadata set corresponding to the description text of the financial product to be parsed to obtain the corresponding financial product information comprises:

reading the full data in the financial product description text to be analyzed by using an application program interface to obtain a corresponding target metadata set;

according to preset analysis keywords, carrying out color labeling on the financial product description text to be analyzed to obtain a financial product description text containing a color label;

traversing the target metadata set according to rules in the rule metadata set, and reading financial product information specified by the rules; wherein the rule is set based on the color label.

3. The method of claim 1, wherein the determining the recommendation level of the financial product according to the information of the financial product, a pre-established model for generating the recommendation level of the financial product, and historical purchase information of the financial product of the customer to be recommended comprises:

calculating an expected net worth and an expected net worth rate of return of the financial product according to the financial product information;

inputting the historical purchase information of the financial products, the financial product rate in the financial product information, the net value and the net value rate of return into the financial product recommendation degree generation model, and generating the financial product recommendation degree of the financial products for the customer to be recommended.

4. The method of claim 3, wherein the calculating the expected net worth and expected net worth profitability of the financial product based on the financial product information comprises:

obtaining a historical net worth sequence of the financial product from the financial product information;

calculating a corresponding historical net worth rate of return sequence according to the historical net worth sequence;

and fitting the historical net worth sequence and the historical net worth rate of return sequence by using a time sequence analysis algorithm to obtain the expected net worth and the expected net worth rate of return.

5. The method of claim 3, wherein the step of inputting the historical purchase information of financial products, the rates of financial products in the financial product information, the net worth and the net worth rate of return into the model for generating recommendations for financial products of the financial products for the customer to be recommended comprises:

generating a corresponding recommendation parameter characteristic set according to the historical financial product purchase information, the financial product rate in the financial product information, the net value and the net value earning rate;

and inputting the recommendation parameter feature set into the financial product recommendation degree generation model to generate the financial product recommendation degree of the financial product for the customer to be recommended.

6. The method of claim 3, wherein the step of constructing the model for generating a recommendation level of a financial product comprises:

generating a recommendation parameter historical characteristic set corresponding to each customer according to historical financial product purchase information of each customer, financial product rates in the financial product information, net worth and net worth rate of return;

carrying out positive and negative sample marking on each recommended parameter historical characteristic set according to the purchase result of each customer for the financial product;

and inputting the positive and negative sample marking results and the recommendation parameter historical characteristic set into a distributed gradient lifting model for training to obtain the financial product recommendation degree generation model.

7. The financial product recommendation statement generating method according to claim 1, further comprising:

collecting audio data when a financial product customer manager uses the financial product recommendation dialect to recommend a corresponding financial product to the customer to be recommended, and generating a corresponding audio conversion text;

and comparing the audio conversion text with the financial product recommendation language, and performing early warning processing based on a comparison result.

8. The method of claim 7, wherein the comparing the audio conversion text with the financial product recommendation language and performing an early warning process based on the comparison result comprises:

inputting the audio conversion text into a BERT model for semantic analysis to obtain a corresponding first clause label;

inputting the financial product recommended dialect into the BERT model for semantic analysis to obtain a corresponding second sentence division label;

and comparing the first sentence dividing label with the second sentence dividing label, and if the first sentence dividing label is inconsistent with the second sentence dividing label, carrying out early warning.

9. A financial product recommendation utterance generation apparatus, comprising:

the financial product information generating unit is used for analyzing the financial product description text to be analyzed according to the rule metadata set corresponding to the financial product description text to be analyzed to obtain corresponding financial product information;

the recommendation degree calculation unit is used for determining the corresponding financial product recommendation degree according to the financial product information, a pre-constructed financial product recommendation degree generation model and the historical financial product purchase information of the customer to be recommended;

and the recommendation dialect unit is used for writing the financial product information and the financial product recommendation degree into a pre-constructed standard dialect template to obtain a corresponding financial product recommendation dialect.

10. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the method of generating financial product recommendations of any one of claims 1 to 8 when executing the program.

11. A computer-readable storage medium having stored thereon a computer program, wherein the computer program, when executed by a processor, implements the steps of the financial product recommended dialogues generation method of any one of claims 1 to 8.

12. A computer program product comprising computer program/instructions, characterized in that the computer program/instructions, when executed by a processor, implement the steps of the financial product recommendation dialog generation method of any of claims 1 to 8.

Images