Using generative AI models for content searching and generation of confabulated search results
Methods and systems provide content searching and retrieval using generative artificial intelligence (AI) Models. The system is configured to receive a user search for content, media or item listings. The user search is provided to a generative AI based search sub-system and to a traditional search sub-system. A first search result listing is generated by the generative AI based subsystem, and a second search result listing is generated by the traditional search sub-system. The first search result listing and the second search result listing are aggregated together and provided for display to a user client device.
1 . A method for content searching and retrieval performed by a processing engine hosted on a computer or multiple computers comprising operations of:
receiving a search query for a search of real items, the search query received from an input from a client device;
generating a first listing of items by utilizing the search query with a search module not based on generative artificial intelligence (AI);
providing, for display within a graphical user interface of the client device, a first listing of items in response to the search query within a first response time;
generating a second listing of items utilizing one or more generative Al models by:
generating one or more query embeddings based on the received search query;
determining a set of exemplar embeddings by utilizing a similarity between the generated one or more query embeddings and one or more exemplar embeddings from pre-existing search queries;
identifying candidate real item embeddings from an embeddings vector database comprising real item embeddings of real items based on a comparison between the real item embeddings and the set of exemplar embeddings;
generating, by one or more generative Al models with the search query, a set of confabulated item embeddings;
determining a set of real item embeddings for the search query utilizing similarity scores between the candidate real item embeddings and the set of confabulated item embeddings; and
generating the second listing of items by utilizing real items corresponding to the determined set of real item embeddings; and
providing for display, within the graphical user interface of the client device, the second listing of items by populating the second listing of items with the displayed first listing of items within a second response time occurring after the first response time.
2 . The method of claim 1 , further comprising the operations of determining the set of real item embeddings by:
generating a similarity score from a comparison between a real item embedding and one or more confabulated item embeddings from the set of confabulated item embeddings; and
selecting the real item embedding for the second listing of items based on the similarity score satisfying a similarity threshold value.
3 . The method of claim 1 , wherein the first listing of items comprises item textual descriptions and one or more images associated with the item.
4 . The method of claim 1 , wherein a real item from the real items comprises a digital content items.
5 . The method of claim 1 , further comprising the operations of:
generating, by the one or more generative Al models with the search query, a confabulated item representing the search query, wherein the confabulated item comprises artificial content; and
generating a confabulated item embedding for the set of confabulated item embeddings from the confabulated item.
6 . The method of claim 1 , further comprising the operations of populating the second listing of items with the displayed first listing of items by merging the second listing of items within the displayed first listing of items.
7 . The method of claim 1 , further comprising the operations of introducing the one or more query embeddings in a feature pool for machine learning model inferencing.
8 . The method of claim 1 , further comprising the operations of providing for display, within the graphical user interface of the client device, the second listing of items comprising an item description.
9 . A system comprising one or more processors configured to perform by a processing engine hosted on a computer or multiple computers, operations of:
receiving a search query for a search of real items, the search query received from an input from a client device;
generating a first listing of items by utilizing the search query with a search module not based on generative artificial intelligence (AI);
providing, for display within a graphical user interface of the client device, a first listing of items in response to the search query within a first response time;
generating a second listing of items utilizing one or more generative Al models by:
generating one or more query embeddings based on the received search query;
determining a set of exemplar embeddings by utilizing a similarity between the generated one or more query embeddings and one or more exemplar embeddings from pre-existing search queries;
identifying candidate real item embeddings from an embeddings vector database comprising real item embeddings of real items based on a comparison between the real item embeddings and the set of exemplar embeddings;
generating, by one or more generative Al models with the search query, a set of confabulated item embeddings;
determining a set of real item embeddings for the search query utilizing similarity scores between the set ofcandidate real item embeddings and the set of confabulated item embeddings; and
generating the second listing of items by utilizing real items corresponding to the determined set of real item embeddings; and
providing for display, within the graphical user interface of the client device, the second listing of items by populating the second listing of items with the displayed first listing of items within a second response time occurring after the first response time.
10 . The system of claim 9 , further comprising the operations of determining the set of real item embeddings by:
generating a similarity score from a comparison between a real item embedding and one or more confabulated item embeddings from the set of confabulated item embeddings; and
selecting the real item embedding for the second listing of items based on the similarity score satisfying a similarity threshold value.
11 . The system of claim 9 ,, wherein the first listing of items comprises item textual descriptions and one or more images associated with an item.
12 . The system of claim 9 , further comprising the operations of:
generating, by the one or more generative Al models with the search query, a confabulated item representing the search query, wherein the confabulated item comprises artificial content; and
generating a confabulated item embedding for the set of confabulated item embeddings from the confabulated item.
13 . The system of claim 9 , further comprising the operations of providing for display, within the graphical user interface of the client device, the second listing of items comprising an item description.
14 . Non-transitory computer storage that stores executable program instructions that, when executed by one or more computing devices, configure the one or more computing devices to perform operations comprising:
receiving a search query for a search of real items, the search query received from an input from a client device;
generating a first listing of items by utilizing the search query with a search module not based generative artificial intelligence (AI);
providing, for display within a graphical user interface of the client device, a first listing of item in response to the search query within a first response time;
generating a second listing of items utilizing one or more generative Al models by:
generating one or more query embeddings based on the received search query;
determining a set of exemplar embeddings by utilizing a similarity between the generated one or more query embeddings and one or more exemplar embeddings from pre-existing search queries;
identifying candidate real item embeddings from an embeddings vector database comprising real item embeddings of real items based on a comparison between the real item embeddings and the set of exemplar embeddings;
generating, by one or more generative AI models with the search query, a set of confabulated item embeddings;
determining a set of real item embeddings for the search query utilizing similarity scores between the candidate real item embeddings and the set of confabulated item embeddings; and
generating the second listing of items by utilizing real items corresponding to the determined set of real item embeddings; and
providing for display, within the graphical user interface of the client device, the second listing of items by populating the second listing of items with the displayed first listing of items within a second response time occurring after the first response time.
15 . The non-transitory computer storage of claim 14 , further comprising the operations of determining the set of real item embeddings by:
generating a similarity score from a comparison between a real item embedding and one or more confabulated item embeddings from the set of confabulated item embeddings; and
selecting the real item embedding for the second listing of items based on the similarity score satisfying a similarity threshold value.
16 . The non-transitory computer storage of claim 14 , wherein the first listing of items comprises item textual descriptions and one or more images associated with an item.
17 . The non-transitory computer storage of claim 14 , further comprising the operations of:
generating, by the one or more generative Al models with the search query, a confabulated item representing the search query, wherein the confabulated item comprises artificial content; and
generating a confabulated item embedding for the set of confabulated item embeddings from the confabulated item.
18 . The non-transitory computer storage of claim 14 , further comprising the operations of populating the second listing of items with the displayed first listing of items by merging the second listing of items within the displayed first listing of items.
19 . The non-transitory computer storage of claim 14 , further comprising the operations of providing for display, within the graphical user interface of the client device, the second listing of items comprising an item description.
20 . The non-transitory computer storage of claim 14 , further comprising the operations of providing for display, within the graphical user interface of the client device, the second listing of items comprising an item description and an image.