Dissecting Span Identification Tasks with Performance Prediction
This work provides guidance for model selection in NLP span identification tasks, but it is incremental as it builds on existing performance prediction methods.
The paper tackled the problem of understanding how properties of span identification tasks influence their difficulty and model performance, by analyzing tasks via performance prediction and identifying key task properties that inform predictions, with results showing that span frequency is important for LSTMs and CRFs help with infrequent spans and non-distinctive boundaries.
Span identification (in short, span ID) tasks such as chunking, NER, or code-switching detection, ask models to identify and classify relevant spans in a text. Despite being a staple of NLP, and sharing a common structure, there is little insight on how these tasks' properties influence their difficulty, and thus little guidance on what model families work well on span ID tasks, and why. We analyze span ID tasks via performance prediction, estimating how well neural architectures do on different tasks. Our contributions are: (a) we identify key properties of span ID tasks that can inform performance prediction; (b) we carry out a large-scale experiment on English data, building a model to predict performance for unseen span ID tasks that can support architecture choices; (c), we investigate the parameters of the meta model, yielding new insights on how model and task properties interact to affect span ID performance. We find, e.g., that span frequency is especially important for LSTMs, and that CRFs help when spans are infrequent and boundaries non-distinctive.