Sentimental Analysis

Date: 23.05.02

Writer: 9tailwolf : doryeon514@gm.gist.ac.kr


Introduction

Ambiguity can cause a person to confuse. For example, address can be a number of location or can be a speech. To analyze the correct representation, determine ambiguity will be important.


Word Sense Disambiguation

Knowledge Based Method

Knowledge Based Method is a method that using meaning of vocabulary by defined in dictionary. There is a WordNet, BabelNet, ConceptNet, Freebase but WordNet is the most representative. WordNet is a Ontology that defined by the relationship(antonym, synonym, subterm, superterm) between words. By this ontology, there is a two algorithm, Dictionary-based, which is meaning inference method by Gloss and Graph based, which is a method by analyze the relationship between words. But both of algorithm has a disadvantage. Since there is no grammer information, It is hard to analyze meaning of sentence.

Supervised Based Learning

Supervised Based Learning is a algorithm that using labeling information to estimate meaning of sentence. Followings are the ways that can apply word sense disambiguation.

  • Naive Bayes Classifiers

                \(c = argmax_{c_{k}} P(c_{k}|\overrightarrow{x})\)
                \(= argmax_{c_{k}} \frac{P(\overrightarrow{x}|c_{k})P(c_{k})}{P(\overrightarrow{x})}\)
                \(= argmax_{c_{k}} P(\overrightarrow{x}|c_{k})P(c_{k})\)
                \(= argmax_{c_{k}} \Sigma_{i}^{n}[P(x_{i}|c_{k})P(c_{k})]\)

  • k-Nearest Neighbor Classifiers

               Find k-th nearest neighber and classify meaning of word.

  • Support Vector Machine

                By making Vector Equation to classify meaning of word.


Other way to analyze meaning of sentence

Semantic Role Analyze

Semantic Role is a way to analyze to find relationship between meaning of word. Following is a example of semantic role analyze.

  • Agent
  • Instrument
  • Patient
  • Experiencer
  • Benefactive
  • Source
  • Goal
  • Etc

And we can apply CRFs(Conditional Random Fields) or SVM(Support Vector Machine) algorithm.

Expression of Meaning

By using AMR(Abstract Meaning Representation) to make more meaning analyze model.