Learning to Segment Actions from Visual and Language Instructions via Differentiable Weak Sequence Alignment | IEEE Conference Publication | IEEE Xplore