Natural Language Processing based Automatic Ingestion of Automotive Data

In order to develop automated processes involving the elaboration of textual data, Natural Language Processing (NLP) provides many different types of tasks. One of them is Named Entity Recognition (NER), which aims to identify and classify certain entities inside the text. Pretrained NER models can offer excellent performances on generic data, but they could be not as effective on specialized data, such as company data, as they show unique features and patterns. In order to overcome this obstacle, a possible solution could be to create a specialized model, trained specifically on the company dataset. The goal of this thesis is to apply different models and methodologies in order to develop a Machine Learning model able to identify key data points of cars (such as make, model and transmission type) using the version names of the vehicles. The work will be devided in 3 main phases: analysis of the dataset and the logs obtained from the precedent model the company already implemented, selection and training of the models, evaluation of performances on unknown data.