Question

我之前在我的应用程序中使用了 stanford-parser-2.0.4-models.jar 。现在我想将我的应用程序移植到 stanford-corenlp-3.2.0-models.jar 。我在我的应用程序中使用 edu.stanford.nlp.trees.EnglishGrammaticalRelations.PURPOSE_CLAUSE_MODIFIER 和 edu.stanford.nlp.trees.EnglishGrammaticalRelations.COMPLEMENTIZER 来识别目的子句修饰符来自语义图边缘的和 complementizer 关系，但不幸的是我在最新版本的stanford-corenlp-3.2.0-models.jar中看不到它们。有人可以建议我怎样才能使用新罐子来解释一下这可能是在新罐子里避免这些关系背后的原因。

Answer 1

我可以在stanford-corenlp-3.2.0-sources.jar中找到这些细节。作为其中的一部分，他们删除了这些关系，并将其视为现有关系的特殊情况。从源代码中找到以下评论

The "purpose clause modifier" grammatical relation has been discontinued
It is now just seen as a special case of an advcl.  A purpose clause
modifier of a VP is a clause headed by "(in order) to" specifying a
purpose.  Note: at present we only recognize ones that have
"in order to" or are fronted.  Otherwise we can't use our surface representations to
distinguish these from xcomp's. We can also recognize "to" clauses
introduced by "be VBN".
<p/>
Example: <br/>
"He talked to the president in order to secure the account" &rarr;
<code>purpcl</code>(talked, secure)


The "complementizer" grammatical relation is a discontinued grammatical relation. A
A complementizer of a clausal complement was the word introducing it.
It only matched "that" or "whether". We've now merged this in with "mark" which plays a        similar
role with other clausal modifiers.
<p/>
<p/>
Example: <br/>
"He says that you like to swim" &rarr;
<code>complm</code>(like, that)

在新的stanford-corenlp-3.2.0-models.jar中缺少关系

1 个答案: