Data augmentation process to improve deep learning-based NER task in the automotive industry field

Searching and extracting information in a textual sequence drawn from scientific articles, queries on a search engine and posts in a discussion forum necessitate a process called Named Entity Recognition (NER). Nevertheless, the data available to achieve this process diverge depending on their nature and field of study. In this article, we look at the performance of Named Entity Recognition systems, and their complexity and ability to process data from different backgrounds. A comparative study between several state-of-the-art approaches, applied to different types of data (search engine queries and discussion forum posts) related to the automotive industry field, is proposed in order to select the approach that will best suit our instance. To do this, we shall rely on the results of metrics evaluating machine learning models such as precision, recall and F-score.