US 12,271,694 B1
Machine learning-based automated narrative text scoring including emotion arc characterization
Swapna Somasundaran, Plainsboro, NJ (US); Xianyang Chen, Princeton, NJ (US); and Michael Flor, Lawrence Township, NJ (US)
Assigned to Educational Testing Service, Princeton, NJ (US)
Filed by Educational Testing Service, Princeton, NJ (US)
Filed on Apr. 23, 2021, as Appl. No. 17/238,368.
Claims priority of provisional application 63/014,975, filed on Apr. 24, 2020.
Int. Cl. G06F 40/284 (2020.01); G06F 40/30 (2020.01); G06N 20/00 (2019.01)
CPC G06F 40/284 (2020.01) [G06F 40/30 (2020.01); G06N 20/00 (2019.01)] 21 Claims
OG exemplary drawing
 
1. A computer-implemented method for characterizing quality of a narrative comprising:
receiving data comprising a narrative text;
determining that the narrative text meets a minimum length threshold;
preprocessing words within the narrative text, wherein the preprocessing comprises:
tokenizing the words;
tagging part-of-speech of the words; and
dependency parsing the words;
extracting a plurality of events from the preprocessed words, the extracting a plurality of events using, in parallel, two or more different extraction techniques;
aggregating the extracted events;
generating a waveform based on the aggregated extracted events that characterizes a plurality of emotional arcs within the narrative text;
extracting a plurality of waveform elements from the waveform, wherein the waveform elements comprise a maximum peak value, a number of peaks, and a highest positive slope value; and
scoring a narrative quality of the narrative text based on the extracted plurality of waveform elements and using a machine learning model trained to correlate emotional arc waveforms with narrative quality scores.