NLP Notes

Artik RTE 2 dataset'ini parse edebiliyoruz:

Yani elimizde her bir T-H (Text-Hypothesis) ikilisi icin:

(raw data)
THPair:
- id
- entailment: TRUE/FALSE
- task: QA/SUM/IE/IR
- text: "... ... ... ... .."
- hypothesis: ".. .. ..."

veya

(preprocessed data)
XTHPair:
-id
- entailment: TRUE/FALSE
- task: QA/SUM/IE/IR
- text:

• XSentence
• XSentence
• XSentence

- hypothesis:

• XSentence

-------------------

XSentence:
- serial (in other words index within the context)
- nodes:

• XNode (in other words, atomic tokens in the sentence)
  
• XNode
• XNode
• ...
• XNode

-------------------

XNode:
- id (unique within the sentence)
- word
- lemma
- pos
- relation:

• relation type
• parent XNode

Bazi yapilarin onunde 'X' var. Bu, o yapinin preprocessed dataset'ten olusturuldugunu gosteriyor.

Relation'daki 'parent node', cumleleri bir parse tree seklinde ifade edebilmemizi sagliyor.