Build Software to Generate Thematically Relevant & Unique Music Album Titles

Hello, I am looking for a machine learning/neural networks specialist familiar with similarity learning to build a software system capable of generating unique data (in this case new music album titles) that is thematically relevant and grammatically correct. Ideally the software would (1) analyse an extensive and existing dataset (several thousand album titles) (2) detect patterns in the data (i.e. the arrangement of nouns-adjectives-verbs and their relation to each other at the basic level + thematic arrangement) (3) output unique & new data (new album titles) in a similar theme and format and in a way that makes sense grammatically Also ideal would be software that allows for the output to be varied, e.g. "30 Relaxing Melodies to Unwind" could be one result, as would "Best of Summer 2018 Relaxation Collection". An option for changing variables such as the number of tracks, season, year, etc. would be helpful. Do get in contact if you are interested, or if you think there is a better way of achieving the desired goal.