This paper describes our followed methodology for the automatic extraction of disease mentions from tweets in Spanish as part of the SocialDisNER challenge within the 2022 Social Media Mining for Health Applications (SMM4H) Shared Task. We followed a Joint Learning ensemble architecture for the fine-tuning of top performing pre-trained language models in biomedical domain for Named Entity Recognition tasks. We used text generation techniques to augment training data. During practice phase of the challenge our approach showed results of 0.87 F1-Score.