Data mining system using artificial intelligence

A computer-implemented method for analyzing documents filed by a plurality of third-party entities using a plurality of machine learning models to identify products, the computer-implemented method comprising:

• retrieving, by one or more processors, a document filed by a third-party entity of the plurality of third-party entities comprising metadata relating to the document and to the third-party entity and having a format unique to an industry with which the document is associated;
• extracting, using the one or more processors, media embedded in the document and generating a media file containing the extracted media and a text file containing extracted text;
• analyzing, using the one or more processors and using a computer-vision model that is a first machine learning (ML) model, the media file to identify and extract data from within the extracted media;
• processing, using the one or more processors and using a text extraction model that is a second ML model different from the first ML model, at least one of the text file or a text file generated from the text file to generate a plurality of text portions from the extracted text;
• generating, using the one or more processors and using an embedding model that is a third ML model different from the first and second ML models, a plurality of embeddings for the plurality of text portions;
• generating, using the one or more processors, a dataset comprising the plurality of text portions, the plurality of embeddings, and at least a portion of the metadata, wherein the dataset includes data compatible with natural language processing (NLP) and is configured to be queried, using NLP, to provide contextual data regarding the document and the third-party entity from the plurality of text portions;
• and
• processing the dataset to identify one or more product modifications to one or more products or to identify one or more new products based on information from the dataset.

A non-transitory computer readable medium for analyzing documents filed by a plurality of third-party entities using a plurality of machine learning models to identify products, the non-transitory computer readable medium comprising instructions stored thereon that, when executed by one or more processors, cause the one or more processors to perform operations comprising:

• retrieving multiple documents, each of the multiple documents comprising metadata and having a format unique to an industry with which the document is associated, wherein the multiple documents relate to multiple topics and the plurality of third-party entities;
• extracting media embedded in the multiple documents and generating a media file containing the extracted media and a text file containing extracted text;
• analyzing, using a computer-vision model that is a first machine learning (ML) model, the media file to identify and extract data from within the extracted media;
• processing, using a text extraction model that is a second ML model different from the first ML model, at least one of the text file or a text file generated from the text file to generate a plurality of text portions from the extracted text;
• generating, using an embedding model that is a third ML model different from the first and second ML models, a plurality of embeddings for the plurality of text portions;
• generating a dataset comprising the plurality of text portions, the plurality of embeddings, and at least a portion of the metadata, wherein the dataset includes data compatible with natural language processing (NLP) and is configured to be queried, using NLP, to provide contextual data regarding the multiple documents and the plurality of third-party entities from the plurality of text portions;
• and
• processing the dataset to identify one or more product modifications to one or more products or to identify one or more new products based on information from the dataset.