昇思25天学习打卡营第17天|LSTM+CRF

BIOE label: 

B start of an entity; O background; I other parts of an entity

We first compute a certain score.

def compute_score(emissions, tags, seq_ends, mask, trans, start_trans, end_trans):
    seq_length, batch_size = tag.shape
    mask = mask.astype(emissions.dtype)
    score = start_trans[tags[0]]
    score += emissions[0, mnp.arange(batch_size),tags[0]]
    for i in range(1, seq_length):
        score += trans[tags[i-1], tags[i]] * mask[i]
        score  += emissions[i, mnp.arange(batch_size), tags[i]] * mask[i]
    last_tags = tags[seq_ends, mnp.arange(batch_size)]
    score += end_trans[last_tags]
    return score

how to understand the score? 

Just two thing:

 1. When we consider a input seq: x = {x1, x2, x3...} and 

a label y = {y1, y2, y3 ...} correspondingly, there exists a transportation probablity 

where score(x, y) can represent the probablity  from x to generate y.

2. Okay, so we need to acculumate probablity of x_i to y_i, and y_(i-1) to y_i since it is important to generate a reasonable next label after one has been generated.

so the former probablity is called emission probablity , and the next is called transportation probablity , and we can define a kind of score by:

 

Next we define  a concept called normalizer., which represents the denominator of the formula below: