<p>The digitized world has made information an asset, and organizations face the challenge of identifying which information must be secured. Information classification is the cornerstone process for distinguishing confidential from non-confidential information and defining different sensitive levels. However, implementing information classification in organizations encounters several difficulties due to its manual approach. In order to overcome real-world barriers, scientific research has evaluated the application of Natural Language Processing (NLP) to automate the process. Current studies are primarily based on supervised learning approaches, which demand significant effort to annotate sensitive labels and lack flexibility for changes in the information classification scheme. So, this study proposes a new NLP model using textual features to estimate the information value. The model assesses document value from two perspectives: (i) personal information associated with laws and regulations and (ii) confidential information related to the organizational context. The model applies information extraction and topic modeling to extract document features and a regression model to estimate information value. We designed an experiment considering six experimental scenarios comprising distinct information classification schemes. We assessed five regression classifiers, and Decision Three demonstrated the most consistent classification metrics, achieving an accuracy higher than 0.8 in all scenarios.</p>