Text this: A new approach for novel object image captioning based on data key multitask conformer with representative mask.