This study explored the application of a machine learning algorithm to identify speech production areas using electrocorticographic mapping (ECM). Electrocortical stimulation mapping (ESM) was performed on 14 patients following the standard picture naming task to establish a ground truth. Machine learning was then used to translate the ECM signals to the ground truth. A deep learning network was used to address the challenge of separating speech production from speech perception signals caused from audio-feedback based error-correction from the patient listening to their own speech. Using a leave-one-patient-out procedure, the study was able to achieve a 0.91 ROC-AUC and 0.88 PR-AUC score. Variability in sensitivity and specificity are still present and limit the ability of applying universal thresholds on classifiers on a patient population. However, the balance between specificity and sensitivity can be improved by altering the threshold for patient specific ROC curves. This shows that brain activity recording has the information necessary for speech production mapping.