This study introduces the Stakeholder Analyzer system for analyzing firm-related web forums, addressing the limitations of previous analyses that treated all participants as a homogeneous group. By applying stakeholder theory and systemic functional linguistic theory, the system identifies distinct participant groups with unique interests and opinions. Analysis of six Yahoo Finance forums over three months revealed that stakeholder-specific measures significantly predict firm stock returns, with adjusted R² values up to 0.83. The stakeholder models accurately predicted stock return directions 67% of the time and achieved a 17% return in simulated trading, outperforming aggregate-level analyses.