nlp

nlp / SentimentByTerm / 0.1.3

README.md

Table of Contents

  1. Introduction
  2. Examples
  3. Credits

Introduction

This algorithm analyzes a document to find the approximate sentiment associated with each of a given set of terms. It does this by splitting the document into sentences and computing, for each provided term, the average sentiment of all sentences containing that term.

Input:

  • (Required): The document (String).
  • (Required): A list of terms (List<String>): Terms can either be provided by hand or extracted automatically with a tagging (for instance, /tags/AutoTagURL) or entity recognition algorithm (/StanfordNLP/NamedEntityRecognition).
  • (Optional): Coreferences in the form of a (Map<String,List<String>>), where each key is a member of the list of terms and the associated value is a list of terms that refer to the same entity. This ensures that all references to a given entity are accounted for, even if they don't all use the same term. This can be supplied either by hand or by previous application of an algorithm like /StanfordNLP/DeterministicCoreferenceResolution.

Note: Make sure that all strings in the term list and coreference map are lower case.

Output:

  • Sentiment associated with each individual word in the term list. Sentiments are scored between 0-4. (Very Negative, Negative, Neutral, Postive, Very Positive)

Examples

Example 1.

  • Parameter 1: Sample sentence.
  • Parameter 2: Term list.
  • Parameter 3: Mapping of conferences (aka. nicknames).
[
  "John Brown (Johnny to his friends) is amazing! Johnny is by far the best mechanic in the tri-state area. Bob Bozo is the worst.",
  ["john brown","bob"],
  {"john brown":["johnny"]}
]

Output:

{
    "john brown": 3.5,
    "bob": 1
}

Example 2.

  • Parameter 1: Sample sentence.
  • Parameter 2: Term list.
[
  "Samaritan will take over the world soon. Samaritan is compassionate. Samaritan is merciful. The Machine is good. Samaritan is evil.",
  ["samaritan", "machine"]
]

Output:

{
    "machine": 3,
    "samaritan": 2
}

Credits

For more information, please refer to http://nlp.stanford.edu/software/corenlp.shtml or Manning, Christopher D., Surdeanu, Mihai, Bauer, John, Finkel, Jenny, Bethard, Steven J., and McClosky, David. 2014. The Stanford CoreNLP Natural Language Processing Toolkit. In Proceedings of 52nd Annual Meeting of the Association for Computational Linguistics: System Demonstrations, pp. 55-60.