The steel industry, as one of the fundamental base industries, plays a crucial role in meeting the deman‎ds of various sectors an‎d contributing to economic development. Key aspects such as forecasting deman‎d for steel products, improving quality, an‎d fostering innovation in processes an‎d raw materials are central to this industry. Moreover, the adoption of advanced technologies an‎d sustainable strategies aids in reducing environmental impacts an‎d enhancing efficiency. Relevant news about the steel industry is continuously published across various international news platforms. Given the diversity of products an‎d technological complexities within this sector, dynamic an‎d comprehensive analysis of global trends an‎d topics requires ongoing monitoring of credible news sources. In this study, a hybrid system leveraging Text Mining an‎d Machine Learning techniques has been developed to automatically summarize steel-related news from the perspective of key identified aspects over different time intervals. The main objective of the research is to achieve optimal performance in aspect-based summarization models with a focus on accuracy an‎d efficiency, by leveraging limited but domain-specific data, concentrating on the extraction an‎d clustering of relevant aspects from global steel-related news. To achieve this goal, English news articles related to the steel industry were collected from the reputable SteelOrbis website over a consecutive three-year period. Through a combination of topic modeling, clustering algorithm, an‎d literature review, 16 key aspects affecting the steel value chain were identified an‎d consolidated. The results indicate that the proposed model demonstrated average improvements of 5% in accuracy an‎d 61% in F-Measure for aspect classification. Furthermore, the aspect-based summarization models (Production-BART, Trading-BART, an‎d Environment-BART), compared to generic summarization, showed average improvements of 4.29% in ROUGE-1, 8.5% in ROUGE-2, 7.72% in ROUGE-L, an‎d 6.63% in ROUGE-3 using data categorized by primary aspects. This research, by innovatively integrating aspect-based analysis an‎d summarization, presents the first comprehensive effort to analyze steel industry news in this manner. At the end, by developing interactive dashboards, it enables practical utilization of the results in strategic an‎d long-term decision-making for steel industry managers.

صبا صارمي نيا

عليرضا بصيري , حسين خسروشاهي