Let's take for example the first sentence:
The white-haired old man was sitting in his favorite chair, holding a thick book and rubbing his tired eyes.
If using StanfordCoreNLP object, we could make use of the following features.
- split corpus docs in sentences and tokens
- foreach token, we get:
- word
- POS
- NE
We can also get the parse tree of the sentence:
(ROOT (S (NP (DT The) (JJ white-haired) (JJ old) (NN man)) (VP (VBD was) (VP (VP (VBG sitting) (PP (IN in) (NP (PRP$ his) (JJ favorite) (NN chair)))) (, ,) (VP (VBG holding) (NP (DT a) (JJ thick) (NN book))) (CC and) (VP (VBG rubbing) (NP (PRP$ his) (JJ tired) (NNS eyes))))) (. .)))... and the dependency graph of the sentence:
-> sitting-VBG (root)
-> man-NN (nsubj)
-> The-DT (det)
-> white-haired-JJ (amod)
-> old-JJ (amod)
-> was-VBD (aux)
-> chair-NN (prep_in)
-> his-PRP$ (poss)
-> favorite-JJ (amod)
-> holding-VBG (conj_and)
-> man-NN (nsubj)
-> book-NN (dobj)
-> a-DT (det)
-> thick-JJ (amod)
-> rubbing-VBG (conj_and)
-> man-NN (nsubj)
-> eyes-NNS (dobj)
-> his-PRP$ (poss)
-> tired-JJ (amod)... and the coreference link graph:
{
1=CHAIN1-["The white-haired old man" in sentence 1, "his" in sentence 1, "his" in sentence 1, "his" in sentence 2, "he" in sentence 2, "His" in sentence 3, "him" in sentence 6, "I" in sentence 8, "I" in sentence 9, "he" in sentence 9, "you" in sentence 14],
2=CHAIN2-["his favorite chair" in sentence 1],
5=CHAIN5-["his tired eyes" in sentence 1, "His eyes" in sentence 3, "my eyes" in sentence 36],
9=CHAIN9-["his nineteen-year-old granddaughter , Valerie ," in sentence 2, "his nineteen-year-old granddaughter" in sentence 2, "Valerie" in sentence 2, "her" in sentence 3, "Valerie" in sentence 12, "I" in sentence 14, "I" in sentence 14, "Valerie" in sentence 15, "She" in sentence 16, "Valerie" in sentence 18, "I" in sentence 19, "Valerie 's" in sentence 23, "You" in sentence 23, "your" in sentence 24, "Valerie" in sentence 40, "Valerie" in sentence 45, "Valerie" in sentence 51, "I" in sentence 54, "Valerie" in sentence 56, "you" in sentence 57, "your" in sentence 57, "Valerie" in sentence 58, "her" in sentence 58, "Valerie" in sentence 61, "she" in sentence 61],
}Assume you have a sentence (just like the one above). In order to obtain the dependency graph, you just have to do the following:
String sentences = "The white-haired old man was sitting in his favorite chair, holding a thick book and rubbing his tired eyes."
Preprocessor preprocessor = new Preprocessor();
preprocessor.stanfordPreprocess(sentences);
depGraph = preprocessor.dependencyGraph;If you just want to obtain the dependecy graph of the hole corpus, use:
Preprocessor preprocessor = new Preprocessor();
preprocessor.stanfordPreprocess(null);
depGraph = preprocessor.dependencyGraph;Each node of the dependency graph has its lemValue and posTag fields filled after preprocessing.