Summarizing Differences between Text Distributions with Natural Language
How do two distributions of texts differ? Humans are slow at answering this, since discovering patterns might require tediously reading through hundreds of samples. We propose to automatically summarize the differences by "learning a natural language hypothesis": given two distributions D_0 and D_1, we search for a description that is more often true for D_1, e.g., "is military-related." To tackle this problem, we fine-tune GPT-3 to propose descriptions with the prompt: "[samples of D_0] + [samples of D_1] + the difference between them is _____". We then re-rank the descriptions by checking how often they hold on a larger set of samples with a learned verifier. On a benchmark of 54 real-world binary classification tasks, while GPT-3 Curie (13B) only generates a description similar to human annotation 7 of the time, the performance reaches 61 our best system using GPT-3 Davinci (175B) reaches 76 describe distribution shifts, debug dataset shortcuts, summarize unknown tasks, and label text clusters, and present analyses based on automatically generated descriptions.
READ FULL TEXT