Algorithm 1 Weighted Frame Integration (WFI) Module

Require: Frame features $e_f=[f_1,\dots ,f_T]$, Text feature $e_x$ Ensure: Video feature $e_v$   1: $s\leftarrow \left[\right]$                                                                        ▹ Initialize similarity scores array   2: for  $i\leftarrow 1$  to  T  do   3:     $s_i\leftarrow \mathrm{dot}\_\mathrm{product}(f_i,e_x)$   4:     Append $s_i$ to s   5: end for   6: $a\leftarrow \mathrm{softmax}\left(s\right)$                                                ▹ Normalize scores to obtain weights   7: $e_v\leftarrow \mathrm{dot}\_\mathrm{product}(a,e_f)$                        ▹ Compute weighted sum of frame features   8: return  $e_v$