Performance prediction model using text and quant data

I have a large data set of monthly reports in text format. The reports represent a summary of what has been happening in the division and what they expect to happen in the near term. There are historical quantitative metrics that cane be used as part of the model development process. I also have quantitative performance metrics that represent past performance for the groups and would like this data to be predicted for future performance. What I am seeking is a model that uses NLP, neural networks, some form of LSTM to identify key drivers of performance and make suitable predictions for future performance based on the historical data set. As we have an extensive number of groups, the model will have to be flexible to capture similar groups and build a model for them. The model also needs to be built using both the NLP data and the historical quantitative performance data upon which performance data will be predicted for the future.