CN117634616A

CN117634616A - Content generation method, device, electronic equipment and medium

Info

Publication number: CN117634616A
Application number: CN202311693244.8A
Authority: CN
Inventors: 张树刚
Original assignee: Beijing QIYI Century Science and Technology Co Ltd
Current assignee: Beijing QIYI Century Science and Technology Co Ltd
Priority date: 2023-12-11
Filing date: 2023-12-11
Publication date: 2024-03-01

Abstract

The embodiment of the application provides a content generation method, a device, an electronic device and a medium, wherein the method is suitable for the electronic device with a content generation model, and the method comprises the following steps: acquiring a current guide text; determining a target field associated with the current guide text; acquiring background knowledge of the target field; updating the current guide text based on the background knowledge of the target field to obtain a target guide text; and inputting the target guide text into the content generation model to obtain reply content corresponding to the target guide text output by the content generation model. By adopting the embodiment of the application, the reply content with stronger specialty can be provided for the user.

Description

Content generation method, device, electronic equipment and medium

Technical Field

The present disclosure relates to the field of artificial intelligence technologies, and in particular, to a content generating method, device, electronic apparatus, and medium.

Background

With the continuous development of internet technology, the expectations for artificial intelligence (Artificial Intelligence, AI) content-generating products are also increasing. However, the current AI content-generating products generate reply content for the user according to the fixed correspondence of questions and answers. The content generation mode is mechanically dead, can only provide common reply content for users, cannot provide reply content with stronger professionals for users, and greatly influences the use experience of the users when the users use the AI content to generate products.

Disclosure of Invention

The embodiment of the application aims to provide a content generation method, a content generation device, electronic equipment and a medium, so as to provide reply content with stronger specialty for users. The specific technical scheme is as follows:

in a first aspect, an embodiment of the present application provides a content generation method, which is applicable to an electronic device deployed with a content generation model, and the method includes:

acquiring a current guide text;

determining a target field associated with the current guide text;

acquiring background knowledge of the target field;

updating the current guide text based on the background knowledge of the target field to obtain a target guide text;

and inputting the target guide text into the content generation model to obtain reply content corresponding to the target guide text output by the content generation model.

Optionally, the determining the target area associated with the current guide text includes:

in response to the domain determining instruction, a target domain associated with the current guide text is determined from at least one domain.

Carrying out semantic recognition on the current guide text to obtain a semantic recognition result;

and determining the target field associated with the current guide text based on the semantic recognition result.

Optionally, before updating the current guide text based on the background knowledge of the target field to obtain the target guide text, the method further includes:

acquiring history information, wherein the history information comprises a history guide text and reply content corresponding to the history guide text;

updating the current guide text based on the background knowledge of the target field to obtain a target guide text, including:

extracting a target abstract of the background knowledge;

and updating the current guide text based on the historical information and the target abstract to obtain a target guide text.

Optionally, the extracting the target abstract of the background knowledge includes:

splitting the background knowledge into a plurality of text fragments;

inputting the text fragments into a summary generation model to obtain a first summary corresponding to each text fragment;

summarizing the first abstracts corresponding to the text fragments to obtain a second abstract;

And inputting the second abstract into the abstract generating model to obtain the target abstract of the background knowledge.

splitting the background knowledge into a plurality of text fragments;

selecting one text segment from the plurality of text segments as a current text segment;

inputting a third abstract corresponding to a previous text segment and the current text segment into an abstract generation model to obtain a fourth abstract corresponding to the current text segment, wherein the third abstract is empty when the previous text segment does not exist;

and updating the current text segment to the last text segment, and returning to execute the step of selecting one text segment from the text segments as the current text segment until the text segments are all selected.

Optionally, in the background knowledge, the previous text segment is the above of the current text segment.

In a second aspect, an embodiment of the present application provides a content generating apparatus, the apparatus being adapted for an electronic device in which a content generating model is deployed, the apparatus including:

the first acquisition module is used for acquiring the current guide text;

The determining module is used for determining the target field to which the current guide text belongs;

the second acquisition module is used for acquiring background knowledge of the target field;

the updating module is used for updating the current guide text based on the background knowledge of the target field to obtain a target guide text;

and the obtaining module is used for inputting the target guide text into the content generation model and obtaining reply content corresponding to the target guide text output by the content generation model.

Optionally, the determining module is specifically configured to:

Optionally, the apparatus further includes a third acquisition module configured to:

before updating the current guide text based on background knowledge of the target field to obtain a target guide text, acquiring history information, wherein the history information comprises a history guide text and reply content corresponding to the history guide text;

The updating module comprises:

the extraction submodule is used for extracting the target abstract of the background knowledge;

and the updating sub-module is used for updating the current guide text based on the history information and the target abstract to obtain a target guide text.

Optionally, the extracting submodule is specifically configured to:

splitting the background knowledge into a plurality of text fragments;

Optionally, the extracting submodule is specifically configured to:

splitting the background knowledge into a plurality of text fragments;

In a third aspect, an embodiment of the present application provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, where the processor, the communication interface, and the memory complete communication with each other through the communication bus;

a memory for storing a computer program;

and the processor is used for realizing any of the steps of the content generation method when executing the program stored in the memory.

In a fourth aspect, embodiments of the present application provide a computer-readable storage medium having a computer program stored therein, which when executed by a processor, implements any of the above-described content generation method steps.

The beneficial effects that this application embodiment provided:

in the technical scheme provided by the embodiment of the application, the current guide text is acquired, the target field associated with the current guide text is determined, the background knowledge of the target field is acquired, the current guide text is updated based on the background knowledge of the target field to obtain the target guide text, the target guide text is input into the content generation model, and the reply content corresponding to the target guide text output by the content generation model is obtained. Therefore, when the technical scheme provided by the embodiment of the application is used for generating the reply content for the user, not only the current guide text is considered, but also the target field associated with the current text is considered, and when the content generation model generates the reply content by the target field associated with the current text, the background knowledge of the target field is fully considered. Therefore, according to the technical scheme provided by the embodiment of the application, when the reply content is generated aiming at the current guide text, the background knowledge of the current guide text and the target field associated with the current guide text is considered, so that the reply content with stronger professionals can be generated for the user.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below.

Fig. 1 is a flow chart of a first content generating method according to an embodiment of the present application;

fig. 2 is a flow chart of a second content generating method according to an embodiment of the present application;

FIG. 3 is a schematic flow chart of a first method for extracting a background knowledge target abstract according to an embodiment of the disclosure;

FIG. 4 is a schematic flow chart of a second method for extracting a background knowledge target abstract according to an embodiment of the disclosure;

fig. 5 is an interaction schematic diagram of a content generating method according to an embodiment of the present application;

fig. 6 is a schematic structural diagram of a content generating device according to an embodiment of the present application;

fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

The technical solutions in the embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application.

With the continuous development of internet technology, the expectations for AI content generation products are also increasing. However, the current AI content-generating products generate reply content for the user according to the fixed correspondence of questions and answers. The content generation mode is mechanically dead, can only provide common reply content for users, cannot provide reply content with stronger professionals for users, and greatly influences the use experience of the users when the users use the AI content to generate products.

To solve the above problems, a user is provided with more specialized reply content. The embodiment of the application provides a content generation method which is applied to electronic equipment, such as a mobile phone, a tablet, a vehicle-mounted terminal and the like, in which a content generation model is deployed. The content generating model may be an existing large model after training, such as ChatGpt, or may be a large model obtained by training an initial model with training data, and the form of the content generating model is not specifically limited in this embodiment of the present application.

The content generating method provided by the embodiment of the present application is explained below with reference to specific embodiments.

As shown in fig. 1, fig. 1 is a flowchart of a content generating method according to an embodiment of the present application, where the content generating method includes steps S11 to S15:

step S11, acquiring the current guide text.

Step S12, determining the target field associated with the current guide text.

Step S13, background knowledge of the target field is acquired.

And step S14, updating the current guide text based on the background knowledge of the target field to obtain the target guide text.

And S15, inputting the target guide text into the content generation model to obtain reply content corresponding to the target guide text output by the content generation model.

In the above step S11, the current guidance text is the text generated by the guidance dialog. The current guide text may include a description of the task by the user and a requirement of the task by the user.

In the embodiment of the application, an application with a content generation function requirement, such as a social application, a shopping application and the like, can be installed on the electronic device. When the electronic equipment detects that a user opens an application with the requirement of a content generation function, and after the guide text is input in the application, the content generation model deployed on the electronic equipment is called to generate reply content corresponding to the current guide text for the user. The content generation model may be a natural language processing large model with a large number of parameters and complex structures. In one embodiment, the content generation model may be ChatGPT.

In one embodiment of the application, the electronic device prompts the user to enter the guide text when the user opens an application with a content generation functionality requirement. After the user inputs the guide text, the electronic device acquires the guide text input by the user as the current guide text.

In another embodiment of the present application, when a user opens an application with a content generation function requirement, the electronic device presents a plurality of preset guide texts to the user, and prompts the user to select one guide text. After the user selects the guide text, the electronic device obtains the guide text selected by the user as the current guide text.

In the step S12, the current guide text may be associated with a specific target area, such as a scenario question-answering area, a common sense area, etc. The electronic device may determine a domain to which the current guide text belongs as a target domain of the current guide text.

In one embodiment of the present application, after acquiring the current guide text, the electronic device sends a domain determining instruction to the user, and prompts the user to select a target domain associated with the current guide text from the given at least one domain. After the user selects the target domain associated with the current guide text, the electronic device determines the target domain associated with the current guide text from at least one domain in response to the domain determination instruction.

For example, after a user opens a social application, the user starts to input a guide text, and after acquiring the current guide text, the electronic device sends a domain determining instruction to the user, and provides 3 domains of life domain, movie domain and scenario question-answering domain on the social application interface for the user to select. The user selects a life domain as the target domain with which the current guide text is associated. And the electronic equipment responds to the domain determining instruction of the user, and determines the life domain as the target domain associated with the current guide text.

According to the technical scheme provided by the embodiment of the application, the user selects the target field associated with the current guide text, so that the target field associated with the current guide text is determined to be more accurate, and a foundation is laid for generating more specialized reply content for the user.

In another embodiment, after the electronic device obtains the current guide text, the electronic device performs semantic recognition on the current text, matches the semantic recognition result with keywords in a plurality of fields based on the semantic recognition result, and determines the field with the highest matching degree as the target field associated with the current guide text.

For example, after a user opens an application, the entered guide text is: how the tomato fried eggs are cooked. The electronic equipment performs semantic recognition on the guide text to obtain two voice recognition results, namely tomato stir-frying and cooking, and matches the voice recognition results with keywords in multiple fields, and the matching degree with the life field is highest, so that the life field is determined as the target field associated with the current guide text.

According to the technical scheme provided by the embodiment of the application, the target field associated with the current guide text is determined based on the semantic recognition result by carrying out semantic recognition on the current guide text, so that the efficiency of determining the target field associated with the current guide text is higher, and the speed of generating the reply content is improved.

In the step S13, the electronic device obtains the background knowledge of the target domain after determining the target domain associated with the current guide text. Background knowledge may be text, pictures, video, audio, etc. types of content.

Taking the chat session application scenario as an example, the field may be a chat field, the current guide text may be a chat session, and the electronic device may prepare databases corresponding to the chat fields in advance, so as to obtain background knowledge of the target chat field after determining the target chat field to which the chat session belongs. For example, after determining that the target chat field to which the chat session belongs is a scenario question-answering field, the electronic device acquires background knowledge of the scenario question-answering field from a literature movie database prepared in advance.

After determining the target chat domain to which the chat session belongs, the electronic device may also retrieve background knowledge related to the target chat domain over the internet. The manner of obtaining background knowledge in the field of target chat is not particularly limited in the embodiments of the present application.

In the step S14, after the electronic device obtains the background knowledge of the target field, the electronic device may splice the background knowledge of the target field behind the current guide text, so as to update the current guide text.

The electronic device can also consider the background knowledge of the target field as a one-dimensional feature vector, consider the current guide text as a one-dimensional vector, and combine the two one-dimensional feature vectors into a two-dimensional feature vector, thereby realizing updating of the current guide text.

In the embodiment of the application, the target guide text is obtained by updating the current guide text based on the background knowledge of the target field, so that the target guide text contains the background knowledge of the target field, namely richer information, than the current guide text.

In the embodiment of the application, the electronic device may further update the current guide text based on other data to obtain the target guide text, so that the target guide text contains richer information. For example, the electronic device may also update the current guide text with the historical guide text and the reply content, the reply content type, the reply content style, and other data corresponding to the historical guide text. The embodiment of the present application does not specifically limit what data is used to update the current guide text.

In the step S15, the electronic device inputs the target guide text into the content generation model, and uses the content output by the content output model as the reply content corresponding to the target guide text, and displays the reply content corresponding to the target guide text to the user. The reply content may be text type, picture type, etc. The target guide text contains background knowledge related to the target field, and the target guide text is input into the content generation model, so that the content generation model can generate more specialized reply content for a user.

Based on the same inventive concept, the embodiment of the present application further provides a content generating method, as shown in fig. 2, fig. 2 is a schematic flow chart of a second content generating method provided in the embodiment of the present application, where the method includes steps S21 to S27, steps S21 to S23 and S27 are the same as steps S11 to S13 and S15, and steps S25 to S26 are a specific implementation manner of step S14.

Step S21, acquiring the current guide text.

Step S22, determining the target field associated with the current guide text.

Step S23, background knowledge of the target field is acquired.

The steps S21 to S23 are similar to the implementation process of the steps S11 to S13, and will not be described in detail here.

Step S24, history information is acquired.

In this embodiment of the present application, the history information includes a history guidance text and reply content corresponding to the history guidance text. The electronic device stores the history dialogue information during the dialogue with other users or with the content generation model, so that the electronic device can acquire the history information.

In one embodiment, the electronic device may extract, from the stored historical dialog information, all the historical guide texts within a preset time and the reply content corresponding to the historical guide texts as the historical information. The preset time may be set according to the need, for example, within half an hour, within one day, within one week, or the like.

For example, during a user and content generation model session, the electronic device extracts all the history guidance text and reply content corresponding to the history guidance text for half an hour as history information. For another example, during a user and content generation model session, the electronic device extracts all the historical guide text and the reply content corresponding to the historical guide text within one hour as the historical information.

In one embodiment, the electronic device may extract a preset number of history guidance texts and reply contents corresponding to the history guidance texts from the stored history dialogue information as the history information. The preset number may be set according to the need, for example, five, ten, twenty, or the like.

For example, in a user and content generation model dialogue, the electronic device extracts five pieces of history guidance text and reply content corresponding to the history guidance text as history information. For another example, during a user and content generation model session, the electronic device extracts ten pieces of history guidance text and reply content corresponding to the history guidance text as history information.

The execution sequence of the above step S21 and step S24 is not particularly limited in the embodiment of the present application.

And S25, extracting a target abstract of background knowledge.

In the embodiment of the application, the target abstract is the outline of the background knowledge, and the obtained simplified background knowledge is outlined. The electronic device may utilize a pre-trained summary extraction model to extract a target summary of background knowledge.

In the embodiment of the application, two schemes can be adopted to extract the target abstract of the background knowledge. The process of extracting the background abstract of background knowledge will be described in detail later.

And step S26, updating the current guide text based on the history information and the target abstract to obtain the target guide text.

In the embodiment of the application, after the electronic device acquires the history information and the target abstract, the history information and the target abstract can be spliced behind the current guide text, so that the current guide text is updated.

The electronic device can also consider the history information as a one-dimensional feature vector, consider the target abstract as a one-dimensional vector, consider the current guide text as a one-dimensional vector, and combine the three one-dimensional feature vectors into a three-dimensional feature vector, thereby realizing the update of the current guide text.

In the embodiment of the application, the target guide text is obtained by updating the current guide text based on the historical information and the target abstract, and the historical information fully considers the context information of the dialogue between the user and the content generation model, so that the reply content generated by the content generation model is smoother.

In the embodiment of the application, the current guide text can be updated by using all background knowledge of the target field to obtain the target guide text, so that the data volume of the target guide text is rich, and more accurate reply content can be obtained. The electronic device may also extract a summary of the obtained background knowledge, and update the current guide text with the summary of the background knowledge to obtain the target guide text. Thus, the calculated amount of the content generation model when generating the reply content is reduced, the efficiency of generating the reply content is improved, and the problem that the word number limitation possibly exists in the input of the content generation model can be effectively avoided. Specifically, whether the background knowledge needs to be extracted in a abstracted way or not can be set according to actual requirements.

Step S27, inputting the target guide text into the content generation model, and obtaining reply content corresponding to the target guide text output by the content generation model.

The implementation process of the step S27 is similar to that of the step S15, and will not be described in detail here.

Two schemes for extracting the objective abstract of background knowledge provided in the embodiments of the present application are specifically described below.

As shown in fig. 3, fig. 3 is a schematic flow chart of a first method for extracting a target abstract of background knowledge according to an embodiment of the application, including steps S31-S34:

step S31, splitting the background knowledge into a plurality of text fragments.

In the embodiment of the application, after the electronic device acquires the background knowledge, the background knowledge is split into a plurality of text fragments. When the background knowledge is text content, the electronic device can directly split the background knowledge into a plurality of text fragments. When the background knowledge is other types of content such as pictures, audio, video and the like, the electronic device needs to convert the background knowledge into text content first and then split the background knowledge converted into text content into a plurality of text fragments. When the electronic device splits the background knowledge into a plurality of text fragments, the background knowledge can be split into a plurality of text fragments with fixed lengths, and the background knowledge can also be split into a plurality of text fragments with different lengths. The embodiment of the present application is not particularly limited thereto.

And S32, inputting a plurality of text fragments into a abstract generating model to obtain a first abstract corresponding to each text fragment.

And step S33, summarizing the first summaries corresponding to the text fragments to obtain a second summary.

And step S34, inputting the second abstract into an abstract generating model to obtain a target abstract of background knowledge.

The electronic equipment inputs the text fragments into a summary generation model respectively to obtain a first summary corresponding to each text fragment, gathers the first summaries corresponding to the text fragments to obtain a second summary, and inputs the second summary into the summary generation model to obtain a target summary of background knowledge.

For ease of understanding, the process of using scheme one to extract a target abstract of background knowledge is described below in connection with specific embodiments.

For example, the electronic device splits the background knowledge into 5 text segments with fixed lengths, namely, text segment 1-text segment 5, and inputs the text segment 1-text segment 5 into the abstract generation model respectively to obtain 5 first abstracts corresponding to the text segment 1-text segment 5: and summarizing the first abstract 1-5 to obtain a second abstract, and inputting the second abstract into an abstract generating model to obtain a target abstract of background knowledge.

By adopting the technical scheme of the embodiment of the application, when the target abstract of the background knowledge is extracted, the background knowledge is split into a plurality of text fragments, the first abstract is respectively extracted from the text fragments, the first abstracts are summarized to obtain the second abstract, the target abstract of the background knowledge point is obtained by utilizing the second abstract, and the target abstract of the background knowledge can be determined at a higher speed and with fewer calculation resources.

As shown in fig. 4, fig. 4 is a schematic flow chart of a second method for extracting a target abstract of background knowledge according to an embodiment of the application, including steps S41 to S44:

in step S41, the background knowledge is split into a plurality of text fragments.

This step is substantially similar to step S31 described above and will not be described in detail here.

Step S42, selecting a text segment from a plurality of text segments as a current text segment;

step S43, inputting a third abstract corresponding to the previous text segment and the current text segment into an abstract generating model to obtain a fourth abstract corresponding to the current text segment, and when the previous text segment does not exist, the third abstract is empty;

step S44, updating the current text segment to the last text segment, and returning to execute the step of selecting one text segment from the text segments as the current text segment until the text segments are all selected.

In the embodiment of the application, the electronic device first selects a text segment from a plurality of text segments as a current text segment. At this time, if the previous text segment does not exist, the third abstract corresponding to the previous text segment is empty, that is, the electronic device directly inputs the current text segment model into the abstract generation model to obtain the fourth abstract corresponding to the current text, and updates the current text segment to the previous text segment. At this time, the fourth abstract corresponding to the current text is updated to the third abstract of the last text segment. And then reselecting a text segment from the plurality of text segments as the current text segment. And re-executing the step of inputting the third abstract corresponding to the previous text segment and the current text segment into the abstract generating model to obtain a fourth abstract corresponding to the current text segment. And sequentially processing according to the rules until all fragments are processed to obtain a final fourth abstract, and determining the final fourth abstract as a target abstract of background knowledge.

In this embodiment of the present invention, the electronic device inputs the third abstract and the current text segment corresponding to the previous text segment into the abstract generating model, which may be that the third abstract and the current text segment corresponding to the previous text segment are combined to obtain a combined text segment, and the combined text segment is input into the abstract generating model, or that the third abstract and the current text segment corresponding to the previous text segment are input into the abstract generating model at the same time. Specifically, the method can be set according to actual requirements.

For ease of understanding, the process of using scheme two to extract the target abstract of background knowledge is described below in connection with specific embodiments.

For example, the electronic device splits the background knowledge into 5 text segments with fixed lengths, text segment 1-text segment 5, and selects text segment 1 first, at this time, there is no previous text segment, and then the third abstract corresponding to the previous text segment is empty, that is, the electronic device directly inputs text segment 1 into the abstract generation model, so as to obtain the fourth abstract corresponding to text segment 1. At this time, the text segment 1 is updated to be the last text segment, and the fourth abstract corresponding to the text segment 1 is updated to be the third abstract of the last text segment. And selecting the text segment 5 from the remaining 4 text segments as the current text segment, and inputting the third abstract of the text segment 1 and the text segment 5 into an abstract generating model by the electronic equipment to obtain a fourth abstract corresponding to the text segment 5. At this time, the text segment 5 is updated to be the last text segment, and the fourth abstract corresponding to the text segment 5 is updated to be the third abstract of the last text segment. And selecting the text segment 2 from the remaining 3 text segments as the current text segment, and inputting the third abstract of the text segment 5 and the text segment 2 into an abstract generating model by the electronic equipment to obtain a fourth abstract corresponding to the text segment 2. At this time, the text segment 2 is updated to the last text segment, and the fourth abstract corresponding to the text segment 2 is updated to the third abstract of the last text segment. And selecting the text segment 4 from the remaining 2 text segments as the current text segment, and inputting the third abstract of the text segment 2 and the text segment 4 into an abstract generating model by the electronic equipment to obtain a fourth abstract corresponding to the text segment 4. At this time, the text segment 4 is updated to be the last text segment, and the fourth abstract corresponding to the text segment 4 is updated to be the third abstract of the last text segment. And selecting the text segment 3 from the rest 1 text segments as the current text segment, and inputting the third abstract of the text segment 4 and the text segment 3 into an abstract generating model by the electronic equipment to obtain a fourth abstract corresponding to the text segment 3. At this time, all five text segments are selected, and the fourth abstract corresponding to the text segment 3 is taken as the target abstract of background knowledge.

When the scheme II is adopted to extract the target abstract of the background knowledge, the background knowledge is split into a plurality of text fragments, and the abstract is extracted from the background knowledge in a recursion mode, so that the target abstract of the background knowledge is more accurate.

In one embodiment, in background knowledge, the previous text segment is above the current text segment. I.e. to extract a summary of the text segment in the context order of the text segment in the background knowledge.

For example, the electronic device splits the background knowledge into 5 text segments with fixed lengths, namely a text segment 1-a text segment 5, firstly selects the text segment 1, inputs the text segment 1 into a abstract generation model to obtain a third abstract corresponding to the text segment 1, merges the third abstract of the text segment 1 with a text segment 2, inputs the abstract generation model to obtain a third abstract corresponding to the text segment 2, merges the third abstract corresponding to the text segment 2 with a text segment 3, inputs the abstract generation model to obtain a third abstract corresponding to the text segment 3, merges the third abstract corresponding to the text segment 3 with a text segment 4, and inputs the abstract generation model to obtain a third abstract corresponding to the text segment 4; merging the third abstract corresponding to the text segment 4 with the text segment 5, and inputting an abstract generating model to obtain a third abstract corresponding to the text segment 5; at this time, all five text segments are selected, and the third abstract corresponding to the text segment 5 is taken as the target abstract of background knowledge.

Furthermore, when the scheme II is adopted to extract the target abstract of the background knowledge, the background knowledge is split into a plurality of text fragments, the abstract is extracted from the background knowledge in a recursion mode, and the abstract is extracted according to the context sequence of the text fragments in the background knowledge, so that the obtained abstract is more coherent with the context of the background knowledge, and the target abstract of the background knowledge is also more accurate.

In one embodiment of the application, after the device obtains the reply content corresponding to the target guide text, the electronic device obtains a plurality of session records from a dialogue process between the user and the content generation model. And inputting the plurality of session records into a abstract generation model to obtain abstracts corresponding to the session records of multiple days.

In the embodiment of the application, the abstract of the session record is directly generated by using the abstract generation model, the formed abstract text may be rough, and some entity words, time words or phrases may have wrong marks or missed marks. Aiming at the problem, the electronic device can output prompt information to prompt a user to correct the session abstract, and the correction content can comprise: correcting the illness, correcting the mislabeled words in the conversation abstract, supplementing the missed labeled words in the conversation abstract, and the like.

In the embodiment of the application, in order to clearly reflect the occurrence time of the abstract of the session record, a time tag may be added to the abstract. Specifically, the electronic device counts the generation time of each obtained session record, takes the generation time of most session records as the time tag of the session group, and adds the time tag into the abstract corresponding to the session record.

After the electronic equipment corrects the conversation abstract and adds the time tag, the abstract corresponding to the conversation record can be sent to the terminal of the user in a mail mode and the like so as to be referred by the user. In the embodiment of the application, the preset terminal can be electronic equipment such as a mobile phone and a tablet of the target user.

In one embodiment, after the electronic device corrects the session digest and adds the time tag, the digest corresponding to the session record may be stored in a database or a local memory of the electronic device, so that the user may conveniently retrieve the session digest in the future.

In one embodiment, after the electronic device corrects the session digest and adds the time tag, the digest corresponding to the session record may be sent to the terminal of the user by means of mail, and at the same time, the digest corresponding to the session record is stored in the database or the local memory of the electronic device.

According to the technical scheme provided by the embodiment of the application, the conversation record in the chat session of the user and the content generation model is generated into the concise abstract, and the abstract is sent to the preset terminal and/or stored, so that the user can conveniently inquire and retrieve the abstract of the conversation record.

In the embodiment of the application, when receiving a query request for a chat session sent by a user, the electronic device determines target information matched with the query request from a summary database corresponding to a stored session record according to the query request, and outputs the target information to a user terminal for the user to review.

According to the technical scheme provided by the embodiment of the application, the brief abstract is generated from the session record in the chat session between the user and the content generation model, so that when the user views the chat session, the lengthy chat session can be replaced by the abstract presenting mode, the content presenting is more concise and visual, and the user experience is improved. In addition, the user can also quickly locate the chat session based on the session abstract, inquire the required chat content, and greatly improve the user experience.

For easy understanding, a method for generating content provided in the embodiments of the present application will be specifically described below with reference to fig. 5 by taking a chat application scenario as an example. Fig. 5 is an interaction schematic diagram of a content generating method according to an embodiment of the present application, which relates to a user side, a large model side and a background database side.

For the background database side, the method specifically comprises the following steps:

in step S51, background information is acquired.

In the embodiments of the present application, the background information is the background information mentioned above. The electronic device may obtain background information for the target area from a background repository.

Step S52, selecting a background data compression method.

In the embodiment of the present application, the background data compression is the abstract for extracting background knowledge mentioned above. The electronic device may optionally select the background data compression method provided in steps S53 and S54 to compress the background data. After step S52 is performed, the electronic device may perform only step S53, may perform only step S54, or may perform both step S53 and step S54. The selection can be specifically performed according to actual requirements.

Step S53, dividing the long text into segments with fixed sizes, extracting abstracts from each segment through a large model, and extracting total abstracts through the large model after summarizing all abstracts.

In the embodiment of the application, the background information is split into a plurality of text fragments, the text fragments are extracted respectively to be summarized to obtain a total abstract, the total abstract is utilized to obtain the target abstract of the background knowledge, and the target abstract of the background knowledge can be determined at a high speed and with less calculation resources.

In step S54, the long text is divided into segments with a fixed size, the abstract of the first segment is extracted through the large model, the abstract is combined with the second text segment, the combined content is extracted through the large model, the abstract is combined with the third segment, and the abstract is sequentially processed according to the above rule until all the segments are processed to obtain the final abstract.

In the embodiment of the application, the electronic equipment splits the background knowledge into a plurality of text fragments, extracts the abstract of the background knowledge in a recursion mode, and the target abstract of the background knowledge is obtained more accurately.

Step S55, generating a background knowledge abstract and feeding the background knowledge abstract to the large model.

In embodiments of the present application, a large model may be generated for the content mentioned above. The electronic equipment can obtain the background knowledge abstract by utilizing the two background information compression methods, and feeds the obtained background knowledge to the large model.

For the user side, the method specifically comprises the following steps:

in step S56, a problem is posed.

In the embodiment of the application, the user proposes a problem based on chat function software installed on the electronic equipment. I.e. the user enters the guide text.

Step S57, extracting the contextual chat record from the chat session of the user.

In the embodiment of the present application, the contextual chat record may be the history information mentioned above. The electronic device may extract a session of a preset time period from a chat session of a user with the content generation model or a chat session of a user with another user as a contextual chat record, or may extract a preset number of sessions from the chat session as contextual chat records.

Step S58, the problems and the contexts are combined and fed into the large model.

In the embodiment of the present application, the large model may generate a model for the above-mentioned content, and the electronic device inputs the questions posed by the user and the contextual chat records extracted from the chat session into the large model, and further, the large model may output the answers to the questions posed by the user.

In this embodiment of the present application, the electronic device may execute the steps S51 to S55 first, then execute the steps S56 to S58 first, or execute the steps S56 to S58 first, then execute the steps S51 to S55, or execute the steps S51 to S55 and the steps S56 to S58 simultaneously, which is not limited in the execution order of the steps S51 to S58.

Step S510, updating the context information of the chat session of the user, and if the context information is too long, extracting the abstract and storing the abstract.

In this embodiment of the present application, after the electronic device completes step S59, that is, after the large model generates the reply content for the question posed by the user, the chat session of the user starts to be updated, and the question posed by the user and the reply content generated by the large model are used as the context information for the next question posed by the user. When the context information of the chat session after updating is overlong, the abstract of the context information is extracted, so that the problem of limiting the input word number of the large model when the context is input into the large model later is avoided.

Step S511, after the chat is finished, the chat summary is extracted through the large model, sent to the user through the mail, and stored in the database.

In the embodiment of the application, when the current chat session is ended, the electronic equipment generates the abstract of the session record through the abstract generation model on the large model, sends the abstract to the user, and stores the session abstract into the database, so that the subsequent retrieval is convenient.

For the large model side, the method specifically comprises the following steps:

step S59, generating an answer according to the question.

In the embodiment of the application, the electronic device can splice the background knowledge abstract and the questions proposed by the user into the target guide text, input the target guide text into the large model, and the large model can output answers to the questions proposed by the user. The electronic device can splice the background knowledge abstract, the questions posed by the user and the context extracted from the chat session into a target guide text, input the target guide text into the large model, and then output the answers of the questions posed by the user by the large model. The input of the large model can be set according to specific actual requirements.

Based on the same inventive concept, the embodiments of the present application further provide a content generating apparatus, as shown in fig. 6, which is suitable for an electronic device deployed with a content generating model, and the apparatus includes:

A first obtaining module 61, configured to obtain a current guide text;

a determining module 62, configured to determine a target domain to which the current guide text belongs;

a second obtaining module 63, configured to obtain background knowledge of the target area;

an updating module 64, configured to update the current guide text based on background knowledge of the target field to obtain a target guide text;

and the obtaining module 65 is configured to input the target guide text to the content generation model, and obtain reply content corresponding to the target guide text output by the content generation model.

Alternatively, the determining module 62 may be specifically configured to:

Optionally, the apparatus may further include a third acquisition module configured to:

The update module 64 may include:

Optionally, the extracting submodule may be specifically configured to:

splitting the background knowledge into a plurality of text fragments;

Optionally, the extracting submodule may be specifically configured to:

splitting the background knowledge into a plurality of text fragments;

The embodiment of the present application further provides an electronic device, as shown in fig. 7, including a processor 71, a communication interface 72, a memory 73, and a communication bus 74, where the processor 71, the communication interface 72, and the memory 73 perform communication with each other through the communication bus 74,

a memory 73 for storing a computer program;

the processor 71 is configured to implement any of the content generation methods described above when executing the program stored in the memory 73.

The communication bus mentioned by the above electronic device may be a peripheral component interconnect standard (Peripheral Component Interconnect, abbreviated as PCI) bus or an extended industry standard architecture (Extended Industry Standard Architecture, abbreviated as EISA) bus, or the like. The communication bus may be classified as an address bus, a data bus, a control bus, or the like. For ease of illustration, the figures are shown with only one bold line, but not with only one bus or one type of bus.

The communication interface is used for communication between the terminal and other devices.

The memory may include random access memory (Random Access Memory, RAM) or non-volatile memory (non-volatile memory), such as at least one disk memory. Optionally, the memory may also be at least one memory device located remotely from the aforementioned processor.

The processor may be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU for short), a network processor (Network Processor, NP for short), etc.; but also digital signal processors (Digital Signal Processor, DSP for short), application specific integrated circuits (Application Specific Integrated Circuit, ASIC for short), field-programmable gate arrays (Field-Programmable Gate Array, FPGA for short) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

In yet another embodiment provided herein, there is also provided a computer readable storage medium having stored therein a computer program which, when executed by a processor, implements the content generating method of any of the above embodiments.

In a further embodiment provided herein, there is also provided a computer program product containing instructions that, when run on a computer, cause the computer to perform the content generation method of any of the above embodiments.

In the above embodiments, it may be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, produces a flow or function in accordance with embodiments of the present application, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions may be stored in or transmitted from one computer-readable storage medium to another, for example, by wired (e.g., coaxial cable, optical fiber, digital Subscriber Line (DSL)), or wireless (e.g., infrared, wireless, microwave, etc.). The computer readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that contains an integration of one or more available media. The usable medium may be a magnetic medium (e.g., floppy Disk, hard Disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid State Disk (SSD)), etc.

It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

In this specification, each embodiment is described in a related manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for the apparatus embodiments, the electronic device embodiments, the computer storage medium embodiments, and the computer program product embodiments, the description is relatively simple, as relevant to the method embodiments being referred to in the section of the description of the method embodiments.

The foregoing description is only of the preferred embodiments of the present application and is not intended to limit the scope of the present application. Any modifications, equivalent substitutions, improvements, etc. that are within the spirit and principles of the present application are intended to be included within the scope of the present application.

Claims

1. A content generation method, the method being suitable for an electronic device in which a content generation model is deployed, the method comprising:

acquiring a current guide text;

determining a target field associated with the current guide text;

acquiring background knowledge of the target field;

2. The method of claim 1, wherein the determining the target area with which the current guide text is associated comprises:

3. The method of claim 1, wherein the determining the target area with which the current guide text is associated comprises:

4. The method of claim 1, further comprising, prior to said updating the current guide text based on background knowledge of the target area to obtain target guide text:

extracting a target abstract of the background knowledge;

5. The method of claim 4, wherein the extracting the target digest of the background knowledge comprises:

splitting the background knowledge into a plurality of text fragments;

6. The method of claim 4, wherein the extracting the target digest of the background knowledge comprises:

splitting the background knowledge into a plurality of text fragments;

7. The method of claim 6, wherein in the background knowledge the last text segment is the previous to the current text segment.

8. A content generation apparatus, the apparatus being adapted for an electronic device in which a content generation model is deployed, the apparatus comprising:

The first acquisition module is used for acquiring the current guide text;

9. The electronic equipment is characterized by comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus;

a memory for storing a computer program;

a processor for carrying out the method steps of any one of claims 1-7 when executing a program stored on a memory.

10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored therein a computer program which, when executed by a processor, implements the method steps of any of claims 1-7.