Combining Denoising Autoencoders with Contrastive Learning to fine-tune Transformer Models