Данные на уровне составляющих

We study the problem of jointly aligning sen- tence constituents and predicting their sim- ilarities. While extensive sentence similar- ity data exists, manually generating reference alignments and labeling the similarities …...