Web Scraping And Nlp For Price Prediction: The Hidden Signals In Hr Discourse

This study explores the intersection between human resources discourse and macroeconomic indicators through the use of web scraping, natural language processing (NLP), and machine learning techniques. By extracting job postings from Romanian platforms and correlating them with national economic indicators, we identify hidden signals embedded in hiring trends. We aimed to demonstrate how Human Resources (HR) data can anticipate shifts in prices, salaries, and economic performance in the age of AI, willing to uncover predictive signals in the given discourse. Using large-scale web scraping from major Romanian job platforms—eJobs and BestJobs—we compiled a dataset of IT and non-IT job advertisements across multiple industries. Textual and structural features from these listings were analyzed using NLP methods including Named Entity Recognition (NER), Latent Dirichlet Allocation (LDA), BERTopic for topic modeling, and BERT-based embeddings for semantic classification. Moreover, to contextualize these insights within broader economic trends, macroeconomic indicators such as the Harmonised Index of Consumer Prices (HICP) and industry-specific productivity data were integrated. Visualization techniques such as tSNE, PCA, and heatmaps were applied to capture correlations and trends across industries and timeframes. The results indicate strong associations between the linguistic features of job postings and key economic indicators. Notably, industries with higher semantic complexity in job descriptions tended to exhibit elevated labor costs and stronger consumer price growth. These findings suggest that HR discourse can serve as a reliable, early indicator of economic developments, offering valuable insights for both workforce planning and price prediction when they are combined with relevant macroeconomic data.