Skip to content
This repository has been archived by the owner on Dec 12, 2018. It is now read-only.

Latest commit

 

History

History
18 lines (13 loc) · 2.22 KB

README_ner.md

File metadata and controls

18 lines (13 loc) · 2.22 KB
Raw

How to Get Named Entities from the Stanford Named Entity Recognizer (NER) via this Apache Thrift Server

How to Interact with the Methods and Data Structures

The core return type here is a data structure called NamedEntity which has four members:

  • entity: A string containing the actual named entity itself, potentially a multi-word expression if that's what Stanford NER recognized.
  • tag: A string containing the tag assigned to this named entity (PERSON, LOCATION, etc.). Should always be upper-case.
  • startOffset: All named entities exist in some sentence. This integer represents the starting character offset of this named entity in its sentence.
  • endOffset: Like startOffset, only tells you the character offset of the last character of the named entity in its sentence.
  • sentence_num: An integer referring to the index (starting from 0) into the original list of sentences provided to Stanford NER, showing in which sentence this named entity occurred.

In order to get these NamedEntity objects, you have three choices, depending on what kind of data you'd like to recognize named entities in. The return type for ALL of these is a Java ArrayList/Python list containing NamedEntity objects corresponding to entities recognized across the ENTIRETY of your text, no matter how many sentences, parse trees, etc. were passed in. If you'd like to recognize named entities in:

  • arbitrary (potentially several sentences worth of), untokenized, un-parsed, un-tagged text, and you're cool with CoreNLP handling all of those tasks for you, call get_entities_from_text(text). text is a Java String/Python str or unicode.
  • one sentence worth of tokens (the output from some sentence and then word tokenizer), call get_entities_from_tokens(tokens), where tokens is a Java List<String>/Python list containing str/unicode. Since Stanford NER requires either parse trees or POS-tagged text, the Stanford Parser will be called.
  • one or more parse trees in Stanford Parser's "oneline" output format, call get_entities_from_trees(trees), where trees is a Java List<String>/Python list containing str/unicode.