What you can cram into a single vector: Probing sentence embeddings for linguistic properties
This work addresses the need for interpretability in sentence embeddings for NLP researchers, though it is incremental as it builds on existing evaluation methods.
The paper tackled the problem of understanding what linguistic information sentence embeddings capture by introducing 10 probing tasks to evaluate simple linguistic features, and applied them to three encoders trained in eight ways, revealing properties of the encoders and training methods.
Although much effort has recently been devoted to training high-quality sentence embeddings, we still have a poor understanding of what they are capturing. "Downstream" tasks, often based on sentence classification, are commonly used to evaluate the quality of sentence representations. The complexity of the tasks makes it however difficult to infer what kind of information is present in the representations. We introduce here 10 probing tasks designed to capture simple linguistic features of sentences, and we use them to study embeddings generated by three different encoders trained in eight distinct ways, uncovering intriguing properties of both encoders and training methods.