By text summarization, large texts or documents can be concise while preserving the context of the main document. This research work is focused on summarizing online product reviews in Bengali Language. By using online user feedback, future buyers will be able to take decisions whether to buy the product or not. However, it is very difficult for a buyer to decide by reading manually all the reviews. Machine learning techniques are available nowadays for text summarization which is a promising solution for extracting information from those user reviews. There are many well-known summarization tools for English language but in Bengali, there are very few and insufficient tools for text summarization. In this paper, a model has been for abstract Bangla text summarization on online product reviews using a Recurrent Neural Network(RNN). Long ShortTerm Memory (LSTM) and Sequence-to-Sequence (Seq2Seq) based RNN has been applied here. Experimented results underscore that the training loss is reduced to 0.0034 and able to generate a frequent predictive summary from original texts or documents. This research work also includes a Word Mover’s Distance(WMD) method which checks the similarity between human and machine summary. It is observed that, the applied WMD in this work for text similarity model has been able to achieve better performance than the Jaccard method.