Network information integration method based on cue word and large language model

Abstract

The invention discloses a network information integration method based on cue words and a large language model, and mainly relates to the technical field of artificial intelligence large language model application. Comprising the following steps: S1, searching RSS feed sources needing to be integrated on a network; s2, all the collected RSS feed sources are stored in a configuration file; s3, writing cue words of the large language model, wherein the cue words are used for generating an article; s4, writing an information integration algorithm by using a programming language; according to the method, the big language model is guided to integrate the news disclosed on the network into an article through the cue word, so that the user can quickly and conveniently read the news.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence large language model application technology, and particularly relates to a network information integration method based on prompt words and a large language model. BACKGROUND

[0002] At present, the number of network information resources increases sharply. In the face of these large amounts of distributed and heterogeneous resources, users need to spend more time to obtain useful information. Although existing resource integration technologies have become increasingly perfect, it is still difficult to meet the actual needs of users. For example, when users read news through a news aggregation website or RSS, they need to view each one, which is time-consuming and laborious. SUMMARY

[0003] The present application aims to solve the problems in the prior art and provides a network information integration method based on prompt words and a large language model. The method guides the large language model to integrate publicly disclosed news on the network into an article, so that users can quickly and conveniently read the news.

[0004] To achieve the above-mentioned purpose, the present application realizes the following technical scheme:

[0005] A network information integration method based on prompt words and a large language model, comprising the following steps:

[0006] S1, searching for RSS feeds that need to be integrated on the network;

[0007] S2, saving all collected RSS feeds in a configuration file;

[0008] S3, writing prompt words for the large language model, which are used to generate articles;

[0009] S4, using a programming language to write an information integration algorithm, comprising the following steps:

[0010] S401, reading the configuration file;

[0011] S402, looping through the RSS feeds;

[0012] S403, reading the data of the RSS feeds;

[0013] S404, extracting the title in the feed content;

[0014] S405, combining the RSS feed content and the basic prompt words into new prompt words

[0015] S406, sending the new prompt words to the large language model;

[0016] S407, processing the reply of the large language model;

[0017] S408, send the large language model output to the text-to-speech model to generate an audio file;

[0018] S409, save the text output by the model in RSS format;

[0019] S410, save the audio file generated by the text-to-speech model;

[0020] S411, publish the generated RSS file and audio file.

[0021] Preferably, in step S3, the number of prompt words is determined according to the amount of content, one prompt word is used when the content is less, and multiple prompt words are used when the content is more.

[0022] Preferably, the prompt word when the content is less is: please write a briefing based on the following title, please output in paragraph format; the prompt word when the content is more includes a basic prompt word and a continue prompt word, the basic prompt word is: please write a briefing based on the following title, please output in paragraph format, please add a "continue output marker" at the end of the text when you need to continue output, and please add an "output end marker" at the end of the text if the output is complete; the continue prompt word is: please continue, please add a "continue output marker" at the end of the text when you need to continue output, and please add an "output end marker" at the end of the text if the output is complete.

[0023] Preferably, the continue output marker is 9009CONTINUE9009, and the output end marker is 8008COMPLETE8008.

[0024] Preferably, in step S404, an XML processing library is used to extract the title according to the XML tag.

[0025] Preferably, in step S407, when the new prompt word content is less and the model output is complete, the next step can be directly performed; when the content is more and the model output is not complete, first check whether the model output contains a "continue output marker", if the "continue output marker" appears, send the continue prompt word to the large language model, and repeat the above steps until the "output end marker" appears.

[0026] Compared with the prior art, the present application has the following advantages:

[0027] The present application collects news from the network through RSS at regular intervals, extracts the title from the collected RSS news, integrates the title and prompt word together, sends the article generated by the public large language model to the public text-to-speech model to generate audio, and stores the article and hard disk in the public cloud repository for users to read, so that users can quickly and conveniently read news. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 is a system framework diagram of the present application. DETAILED DESCRIPTION

[0029] The application will be further described below in connection with specific embodiments. It should be understood that these embodiments are only used to illustrate the application and not used to limit the scope of the application. Furthermore, it should be understood that after reading the content taught by the present application, those skilled in the art can make various modifications or changes to the present application, and these equivalent forms also fall within the scope defined by the present application.

[0030] Embodiment: As shown in the accompanying drawings Figure 1 The present application is a network information integration method based on prompt words and large language models. News on the network is collected through the RSS protocol, and the title is extracted and added to the prompt word input to the large language model. A integrated article is obtained through the large language model, and RSS is automatically generated, and the article can be automatically sent to the text-to-speech large model to generate an audio file for the reader to listen and read. Specifically, it includes the following steps:

[0031] S1, find the RSS feed sources that need to be integrated on the network.

[0032] S2, save all collected RSS feeds in a configuration file.

[0033] S3, write prompt words for the large language model, which are used to generate articles.

[0034] When the content is less, a large language model prompt word can be used, such as: Please write a brief report based on the following title, please output in paragraph format.

[0035] When the content is more and longer, multiple large language model prompt words can be used, such as:

[0036] Basic prompt word: Please write a brief report based on the following title, please output in paragraph format, please add "9009CONTINUE9009" at the end of the text when you need to continue output, and add "8008COMPLETE8008" at the end of the text if the output is complete.

[0037] Continue prompt word: Please continue, please add "9009CONTINUE9009" at the end of the text when you need to continue output, and add "8008COMPLETE8008" at the end of the text if the output is complete.

[0038] S4, use a programming language to write an information integration algorithm, including the following steps:

[0039] S401, read the configuration file;

[0040] S402, loop through the RSS feeds;

[0041] S403, read the data of the RSS feed;

[0042] S404, extract the title in the feed content;

[0043] Use the XML processing library (such as Python's Beautiful Soup) to extract the title according to the XML tag.

[0044] S405, combine the RSS feed content and the basic prompt word into a new prompt word

[0045] S406, send the new prompt word to the large language model;

[0046] Use the library recommended by each public large language model to send the prompt word to the large language model.

[0047] S407, process the reply of the large language model;

[0048] When the content of the new prompt word is less and the model output is complete, the next step can be directly performed;

[0049] When the content is more and the model output is not complete,

[0050] 1. Check if the model output contains a continue output marker such as "9009CONTINUE9009";

[0051] 2. When the continue output marker appears, send a continue prompt word to the large language model;

[0052] 3. Repeat steps 1-2 until the end marker appears, such as "8008COMPLETE8008".

[0053] S408, send the large language model output to the text-to-speech model to generate an audio file;

[0054] S409, save the text output by the model in RSS format;

[0055] S410, save the audio file generated by the text-to-speech model;

[0056] S411, publish the generated RSS file and audio file.

Claims

1. A network information integration method based on a prompt word and a large language model, characterized in that, The method comprises the steps of: S1, searching for RSS feeds that need to be integrated on the network; S2, saving all collected RSS feeds in a configuration file; S3, writing prompt words for a large language model, which are used to generate articles; S4, writing an information integration algorithm using a programming language, comprising the steps of: S401, reading the configuration file; S402, looping through the RSS feeds; S403, reading the data of the RSS feeds; S404, extracting the title from the feed content; S405, combining the RSS feed content and the basic prompt words into new prompt words S406, sending the new prompt words to the large language model; S407, processing the reply of the large language model; S408, sending the output of the large language model to a text-to-speech model to generate an audio file; S409, saving the text output by the model in RSS format; S410, saving the audio file generated by the text-to-speech model; S411, publishing the generated RSS file and audio file.

2. The network information integration method based on the prompt word and the large language model according to claim 1, characterized in that, In step S3, the number of prompt words is determined according to the amount of content. When the content is less, one prompt word is used; when the content is more, multiple prompt words are used.

3. The network information integration method based on the prompt word and the large language model according to claim 2, characterized in that, When the content is less, the prompt word is: Please write a briefing based on the following title, please output in paragraph format; When the content is more, the prompt word includes a basic prompt word and a continue prompt word. The basic prompt word is: Please write a briefing based on the following title, please output in paragraph format, please add "continue output marker" at the end of the text when you need to continue output, and please add "output end marker" at the end of the text if the output is complete. The continue prompt word is: Please continue, please add "continue output marker" at the end of the text when you need to continue output, and please add "output end marker" at the end of the text if the output is complete.

4. The network information integration method based on the prompt word and the large language model according to claim 3, characterized in that, The continue output marker is 9009CONTINUE9009, and the output end marker is 8008COMPLETE8008.

5. The network information integration method based on the prompt word and the large language model according to claim 1, characterized in that, In step S404, the title is extracted using an XML processing library according to the XML tag.

6. The network information integration method based on the prompt word and the large language model according to claim 1, characterized in that, In step S407, when the content of the new prompt word is less and the model output is complete, the next step can be performed directly; when the content is more and the model output is not complete, first check whether the "continue output marker" is contained in the model output. When the "continue output marker" appears, send the continue prompt word to the large language model, and repeat the above steps until the "output end marker" appears.

Images