In correct nouns , dates, facts and figures

[Ashbajawed]

Hello , hope you are fine. I am working on text generation module of Grover and have successfully trained Grover's mega network on common crawl data with different parameters. Its generating results better then before but still I am facing following issues,

1. Sometimes it changes the person's name or gender entirely
2. sometime makes spelling mistakes in it.
3. It changes the dates and time mention in original news articles to something else

Can I solve these issues by more training or hyper parameter tuning? or it is issue not at all? Is there any other way I can make my model from making spelling mistakes and generating incorrect facts ad figures