Webpage creation leveraging generative AI
Large language models (LLMs) are leveraged in order to dynamically generate webpages and to modify pre-existing webpages. The LLMs determine the intent of queries and modification requests and obtain relevant content using differently defined page generation strategies. Related apparatus, systems, techniques and articles are also described.
1 . A computer-implemented method comprising:
receiving a user-generated request to initiate forking of an existing webpage;
determining, using a large language model (LLM), an intent of the request;
generating a query based on the determined intent;
performing an Internet search to receive content responsive to the query; and
dynamically modifying and/or enriching, by the LLM and derived from the received content responsive to the query, the existing webpage to result in a modified webpage;
wherein: (i) forking maintains both the original webpage and the modified webpage available for subsequent access as separate webpages, each webpage having a distinct uniform resource locator (URL), (ii) the existing webpage and the modified webpage are presented in a browser interface that includes a content pane displaying webpage content and an AI copilot chat pane configured to receive natural language forking requests (iii) the dynamically modifying comprises parsing the original webpage content, creating a copy of the parsed original webpage content in a webpage data structure, and mixing the copy with LLM-generated content based on the Internet search to produce the modified webpage, and (iv) generating, by the LLM, a webpage layout for the modified webpage, the webpage layout defining content sections and content types for the modified webpage.
2 . The method of claim 1 further comprising:
determining, by the LLM and based on the modified webpage, at least one follow up question associated with the request;
supplementing the modified webpage based on the determined at least one follow up question associated with the request.
3 . The method of claim 2 , wherein the modified webpage is supplemented by performing an Internet search.
4 . The method of claim 2 , wherein the modified webpage is supplemented by further querying the LLM for complementary information to the modified webpage.
5 . The method of claim 1 further comprising:
receiving user-generated input changing content in the modified webpage.
6 . The method of claim 5 further comprising:
publishing the modified webpage.
7 . The method of claim 6 further comprising: embargoing the modified webpage prior for a pre-defined period of time, during which the modified webpage is not publicly available on the Internet and is subject to pre-publication analyses by one or more policy engines.
8 . The method of claim 7 further comprising:
performing, using at least one policy and during the embargo, computer-implemented analyses of the modified webpage to determine whether the modified webpage comprises unauthorized or prohibited content.
9 . The method of claim 8 further comprising:
deleting or redacting unauthorized or prohibited content prior to publication.
10 . The method of claim 8 further comprising:
preventing the modified webpage from being published if it is determined that the modified webpage comprises unauthorized or prohibited content.
11 . A system comprising:
at least one data processor; and
memory storing instructions which, when executed by the at least one data processor, result in operations comprising:
receiving a user-generated request to initiate forking of an existing webpage;
determining, using a large language model (LLM), an intent of the request;
generating a query based on the determined intent;
performing an Internet search to receive content responsive to the query; and
dynamically modifying and/or enriching, by the LLM and derived from the received content responsive to the query, the existing webpage to result in a modified webpage;
wherein: (i) forking maintains both the original webpage and the modified webpage available for subsequent access as separate webpages, each webpage having a distinct uniform resource locator (URL), (ii) the existing webpage and the modified webpage are presented in a browser interface that includes a content pane displaying webpage content and an AI copilot chat pane configured to receive natural language forking requests, (iii) the dynamically modifying comprises parsing the original webpage content, creating a copy of the parsed original webpage content in a webpage data structure, and mixing the copy with LLM-generated content based on the Internet search to produce the modified webpage, and (iv) generating, by the LLM, a webpage layout for the modified webpage, the webpage layout defining content sections and content types for the modified webpage.
12 . The system of claim 11 , wherein the operations further comprise:
determining, by the LLM and based on the modified webpage, at least one follow up question associated with the request;
supplementing the modified webpage based on the determined at least one follow up question associated with the request.
13 . The system of claim 12 , wherein the modified webpage is supplemented by performing an Internet search.
14 . The system of claim 12 , wherein the modified webpage is supplemented by further querying the LLM for complementary information to the modified webpage.
15 . The system of claim 11 , wherein the operations further comprise:
receiving user-generated input changing content in the modified webpage.
16 . The system of claim 15 , wherein the operations further comprise:
publishing the modified webpage.
17 . The system of claim 16 , wherein the operations further comprise:
embargoing the modified webpage prior for a pre-defined period of time, during which the modified webpage is not publicly available on the Internet and is subject to pre-publication analyses by one or more policy engines.
18 . The system of claim 17 , wherein the operations further comprise:
performing, using at least one policy and during the embargo, computer-implemented analyses of the modified webpage to determine whether the modified webpage comprises unauthorized or prohibited content.
19 . The system of claim 18 , wherein the operations further comprise:
deleting or redacting unauthorized or prohibited content prior to publication.
20 . The system of claim 18 , wherein the operations further comprise:
preventing the modified webpage from being published if it is determined that the modified webpage comprises unauthorized or prohibited content.
21 . The method of claim 1 , wherein the webpage layout comprises a page title, a content outline specifying a hierarchical arrangement of content sections, and a writing plan specifying content sources for each content section.
22 . The method of claim 1 , wherein the AI copilot chat pane is configured to display LLM-generated follow-up questions related to content in the content pane.
23 . The method of claim 1 , wherein the LLM determines the intent of the request by classifying the request into at least one of a plurality of intent categories comprising:
modifying the webpage, enriching the webpage, deleting content from the webpage, or rewriting the webpage.
24 . The method of claim 1 , further comprising: determining, by the LLM, that additional information beyond the received content is required to satisfy the request; and in response, generating one or more supplemental queries to obtain additional content from the Internet.
25 . The method of claim 1 , wherein the query specifies at least one of: (a) data sources from which to obtain content, or (b) content data types to obtain, the content data types comprising at least one of text, images, video, or audio.
26 . The method of claim 1 , wherein the dynamically modifying comprises the LLM generating HTML, CSS, or JavaScript code for rendering the modified webpage.
27 . The method of claim 1 , further comprising: receiving, via the AI copilot chat pane, a natural language prompt specifying desired modifications to the existing webpage; and wherein the user-generated request to initiate forking is based on the natural language prompt.
28 . The method of claim 1 , further comprising: searching pre-existing dynamically-generated webpages to identify content matching the query; and incorporating at least a portion of the identified matching content into the modified webpage.
29 . The method of claim 1 , wherein the modified webpage comprises multiple content modalities generated by the LLM, the content modalities comprising at least two of: text content, video content, image content, or map content.
30 . A computer-implemented method comprising:
receiving, via an AI copilot chat pane of a browser interface, a natural language prompt specifying desired modifications to an existing webpage displayed in a content pane of the browser interface;
determining, using a large language model (LLM), an intent of the natural language prompt, the intent comprising at least one of: modifying the existing webpage, enriching the existing webpage, deleting content from the existing webpage, or rewriting the existing webpage;
generating, by the LLM based on the determined intent, a contextualized query specifying at least one of data sources from which to obtain content or content data types to obtain;
performing an Internet search using the contextualized query to receive content responsive thereto;
generating, by the LLM and using webpage generation strategy instructions, a webpage layout comprising a content outline specifying a hierarchical arrangement of content sections and a writing plan specifying content sources for each content section;
parsing the existing webpage content and creating a copy of the parsed content in a webpage data structure;
dynamically generating, by the LLM, a modified webpage by mixing the copy of the parsed content with LLM-generated content derived from the received content responsive to the contextualized query and according to the webpage layout; and
storing the modified webpage with a distinct uniform resource locator (URL) while maintaining the existing webpage available for subsequent access with its original URL, such that both the existing webpage and the modified webpage are independently accessible.