フリー問題

問題文

A model must relate a word near the start of a long document to a word near the end, where the distance between the two words varies greatly from case to case. Which mechanism handles that best?

選択肢

1. The mechanism that handles it best is hierarchical pooling in the model.
2. The mechanism that handles it best is convolutional filtering in the model.
3. The mechanism that handles it best is recurrent propagation in the model.
4. The mechanism that handles it best is attention weighting in the model.