Generating a pseudo-code from a text summarization based on a convolutional neural network

A method, comprising:

• at a computing platform comprising at least one processor, and memory:
• translating, based on a neural machine translation model, a first document comprising text in a natural language different from English, to a second document comprising text in English;
• generating an attention-based convolutional neural network (CNN) for the second document;
• generating an intra attention net, of an encoder portion, comprising word attention nets outputs comprising features of one or more words of the encoder portion, and group attention nets comprising outputs of one or more group encoders;
• generating an inter attention net comprising outputs of the intra attention net of the encoder portion, and an intra attention net of a decoder portion;
• generating a first context vector for the encoder portion based on the intra attention net of the encoder portion;
• generating a second context vector for the decoder portion based on the intra attention net of the decoder portion;
• and
• determining an inter-attention score based on the first context vector and the second context vector;
• extracting, by applying the attention-based CNN, an abstractive summary of the second document;
• generating, based on a cascade image generation model applied to the abstractive summary, a flowchart;
• generating, based on the flowchart, a pseudo-code;
• and
• displaying, via an interactive graphical user interface, the flowchart, and the pseudo-code.

One or more non-transitory computer-readable media storing instructions that, when executed by a computing platform comprising at least one processor, and memory, cause the computing platform to:

• translate, based on a neural machine translation model, a first document comprising text in a natural language different from English, to a second document comprising text in English;
• generate an attention-based convolutional neural network (CNN) for the second document, wherein the attention-based CNN comprises an encoder portion and a decoder portion;
• generate a first context vector for the encoder portion based on a first weighted sum of first feature vectors for the encoder portion, wherein first weights are based on intra-attention scores for words in the encoder portion;
• and
• generate a second context vector for the decoder portion based on a second weighted sum of second feature vectors for the decoder portion, wherein second weights are based on intra-attention scores for words in the decoder portion;
• extract, by applying the attention-based CNN, an abstractive summary of the second document;
• generate, based on the abstractive summary, a flowchart;
• generate, based on a natural language generation model applied to the flowchart, a pseudo-code;
• and
• display, via an interactive graphical user interface, the flowchart and the pseudo-code;
• receive, via the graphical user interface, an indication of a change to one or more of the flowchart and the pseudo-code;
• and
• train, based on the indication, a machine learning model to perform one or more of the:
• translating the first document, generating the attention-based CNN, extracting the abstractive summary, generating the flowchart, and generating the pseudo-code.