We create custom functions to handle different types of excel files, then use classification and NER to identify which function and with what parameters to use it to get the context from the excel files to create the answer.

We use NER and classification to classify concept like guides, rapports, products, then extract properties like prices, dates, features, description, titles etc. We do the same with user queries, and filter the dataset in order to retrieve the complete answer.

We use NER to extract information like dates, total, prices, quantity etc, then we classify the query to identify what documents and how they must be filtered and sorted. We create custom NER / classification models.