some confusions about convert_to_head.py

[leileilin]

Hello, I have a new question about convert to heads.py file, in which some span and clusters will be deleted. Is this the case as follows? In those cases "A" and "A & B" are different spans with the same head word, "A". In our implementation such cases were simply discarded from the training set, because they were few and we were able to perform well, even though we couldn't predict any of such cases during inference. like u said in #2 thanks.

[vdobrovolskii]

I am sorry, I don't quite understand your question. Are you asking if what I said in issue #2 is true? Yes, it is

[leileilin]

I am sorry, I don't quite understand your question. Are you asking if what I said in issue #2 is true? Yes, it is

I mean the deleted span and clusters in convert_to_head.py is the same as the above example i proposed?

[vdobrovolskii]

What example are you referring to?

[leileilin]

In those cases "A" and "A & B" are different spans with the same head word, "A". In our implementation such cases were simply discarded from the training set, because they were few and we were able to perform well, even though we couldn't predict any of such cases during inference.

In those cases "A" and "A & B" are different spans with the same head word, "A". In our implementation such cases were simply discarded from the training set, because they were few and we were able to perform well, even though we couldn't predict any of such cases during inference. it is the example you refered to in #2 .

[vdobrovolskii]

But what do you want me to say? You are quoting me to myself, do I need to additionally confirm what I said is still true? Or what? 😆

[leileilin]

But what do you want me to say? You are quoting me to myself, do I need to additionally confirm what I said is still true? Or what? 😆

what i want to say is the deleted span and clusters in convert_to_head.py is the same as the above example i proposed? i know the example is true but the relationship between them is not certain for me. 😆

[vdobrovolskii]

You mean, is there any difference between deleting a span and deleting a cluster? Or you are talking about the variables deleted_spans and deleted_clusters in convert_to_heads.py?

[leileilin]

You mean, is there any difference between deleting a span and deleting a cluster? Or you are talking about the variables deleted_spans and deleted_clusters in convert_to_heads.py?

I mean, I don't know the basis for deleting span and clusters in convert_to_head.py? is the same as the example above?

[leileilin]

You mean, is there any difference between deleting a span and deleting a cluster? Or you are talking about the variables deleted_spans and deleted_clusters in convert_to_heads.py?

I mean, I don't know the basis for deleting span and clusters in convert_to_head.py? is the same as the example above?

simplly to say , according to what to delete span and cluster in convert_to_head.py?

[vdobrovolskii]

Yes, according to that you would delete spans. And then all the singleton clusters also get deleted.

[leileilin]

Yes, according to that you would delete spans. And then all the singleton clusters also get deleted.

in other words, if "A" and "A & B" are different spans with the same head word, "A". so "A & B" will be deleted. right ?

[vdobrovolskii]

yes

[leileilin]

yes

If I don't delete it, can I train with such a dataset?

[vdobrovolskii]

You can, but it will be suboptimal (you will have the same head point to differen spans in span predictor module). And you will need to merge unrelated clusters.