Special issue on noisy text analytics

摘要

Noise is an ever-present challenge in the field of document analysis. We face it at all levels and stages of the process. While human readers are adept at compensating for noisy inputs, machines still stumble, sometimes badly, when confronted by the consequences when noise is present. The goal of building robust systems that both tolerate noisy inputs on the one hand, and minimize the errors they pass on to downstream stages on the other, is an active area of research.