RAG
metadata
artificial intelligence
RAG: the hybrid approach to text generation that combines search and generation techniques via input queries to locate relevant information within a database.
The client, a leading insurance group, was faced with the need to quickly retrieve information contained in large volumes of documents, some of which were very old, including current company documentation and all its historical documents, including policies, contracts and the like.
To meet this need, Gruppo SCAI adopted an innovative approach in the field of artificial intelligence, known as Retrieval-Augmented Generation. Thanks to RAG, it was possible to index the retrieved documents, extract information from them and collect it in a structured database, facilitating consultation by users.
This was achieved by exploiting the enormous text-generative abilities of public language models, without the need to re-train them, in the context of confidentiality of business domain documents that must remain inaccessible to public LLMs.
The insurance group used Artificial Intelligence to extract metadata from company resumes. This information was then fed into a structured database facilitating the production of resumes in a standard format using the chatbot’s generative abilities. The resumes were subsequently indexed and entered into the confidential corporate knowledge base, where the RAG approach was applied. Thanks to this process, the HR department achieved remarkable results in terms of searching for, consulting and comparing profiles.
RAG systems were therefore implemented and adapted to retrieve sensitive (guaranteeing confidentiality) and relevant information based on user preferences or the specific context of the communication, without generating any generic or repetitive responses, but providing more specific and detailed outputs. In particular, the search could be carried out both through the selection of keywords and by exploiting natural language, i.e. through complete sentences, proposed as they use the ‘similarity’ of the concept and not just the single term.
Thanks to the implementation of the RAG approach, it was possible to significantly improve both the information retrieval process, resulting in faster processing of related tasks, and its management within the insurance group, through more effective retrieval and more comprehensive and satisfactory responses even in complex or ambiguous contexts.
Thanks to access to external sources of information, the answers obtained today are richer and more comprehensive; the user seeking information has a better understanding of the subject matter.
expertise
