Over the past few decades, text production and designing intelligent systems to make a text comprehensible to a machine and fabricate new text have been a complex challenge for artificial intelligence researchers. In addition to being fast, such systems should bring into play the methods of language modeling, syntactic or grammatical analysis, and semantic and conceptual analysis of a set of texts efficiently. In this regard, several algorithms based on neural networks have been proposed by researchers. The long short-term memory deep learning algorithm is investigated in this study as a system with the ability to memorize long sequences and learn long-term dependencies. Since the combination of several powerful models can lead to the development of a stronger model with higher accuracy and performance, we decided to present a hybrid algorithm including three models based on long short-term memory (bidirectional long short-term memory network, long short-term memory network, deep long short-term memory network) and stacking ensemble learning algorithm. The components of the hybrid algorithm are taught in parallel to improve the performance of text generation. The tests performed on all 3 models and the proposed algorithm prove that the obtained structure can reduce the production error and achieve higher accuracy

long short-term memory algorithm, text generation, incremental framework, the ensemble learning algorithm