Detecting machine text

A computer-implemented method for detecting machine text, the method comprising:

• creating a corresponding set of historical features for historical text lines within a block of historical text, in response to receiving the block of historical text;
• training a machine-learning model to cluster the sets of historical features into clusters of historical features based on similarities within the sets of historical features;
• identifying one of the clusters of historical features as corresponding to human text;
• classifying, each historical text line that corresponds to the identified cluster of historical features as human text, and each historical text line that corresponds to another one of the clusters of historical features as machine text;
• creating a corresponding set of features for text lines within a block of text generated by either of a first human and a second human, in response to receiving the block of text;
• applying the trained machine-learning model to cluster the sets of features into clusters of features based on similarities within the sets of features;
• identifying one of the clusters of features as corresponding to human text;
• classifying, each text line that is within the block of text and which corresponds to the identified cluster of features as human text generated by either of the first human and the second human, and each text line that is interleaved with the classified human text within the block of text and which corresponds to another one of the clusters of features as machine text inserted by either of the first human and the second human;
• and
• applying, human text analysis to each text line classified as human text generated by either of the first human and the second human and machine text analysis to each text line classified as machine text inserted by either of the first human and the second human.

A computer program product, comprising a non-transitory computer-readable medium having a computer-readable program code embodied therein to be executed by one or more processors, the program code including instructions to:

• create a corresponding set of historical features for historical text lines within a block of historical text, in response to receiving the block of historical text;
• train a machine-learning model to cluster the sets of historical features into clusters of historical features based on similarities within the sets of historical features;
• identify one of the clusters of historical features as corresponding to human text;
• classify, each historical text line that corresponds to the identified cluster of historical features as human text, and each historical text line that corresponds to another one of the clusters of historical features as machine text;
• create a corresponding set of features for text lines within a block of text generated by either of a first human and a second human, in response to receiving the block of text;
• apply the trained machine-learning model to cluster the sets of features into clusters of features based on similarities within the sets of features;
• identify one of the clusters of features as corresponding to human text;
• classify, each text line that is within the block of text and which corresponds to the identified cluster of features as human text generated by either of the first human and the second human, and each text line that is interleaved with the classified human text within the block of text and which corresponds to another one of the clusters of features as machine text inserted by either of the first human and the second human;
• and
• apply, human text analysis to each text line classified as human text generated by either of the first human and the second human and machine text analysis to each text line classified as machine text inserted by either of the first human and the second human.