Currently, we do a combination of Max, Mean pooling at the sentence feature level to get PIM score. This may not be an effective option, we should consider the count of sentences close to single mind sentence and/or any other scoring metric to make the score robust.
Currently, we do a combination of Max, Mean pooling at the sentence feature level to get PIM score. This may not be an effective option, we should consider the count of sentences close to single mind sentence and/or any other scoring metric to make the score robust.