Information Sistem
Formal documentation regarding development goals, scientific data processing methodologies, and benefit contributions of the built system.
System Development Objectives
This system is developed with the main objective of presenting a transparent overview of the results of public sentiment analysis on government budget efficiency policies. By classifying public responses into positive, negative, and neutral polarities, this system helps simplify raw opinion data on Twitter (X) so that it is easy to read and understand as a material for evaluating operational savings policies of state apparatus.
Data Processing Methodology
Text Processing & Classification Flow
1. Text Preprocessing Stages (NLP)
Cleansing
Cleaning comment text from mentions, url links, RT symbols, hashtags, numbers, and special characters.
Case Folding
Converting all alphabet characters to lowercase uniformly to avoid word redundancy.
Stopword Removal
Filtering and removing Indonesian conjunctions or common words that do not have sentiment value.
Stemming
Returning affixed words to root words (lemmas) by removing prefixes & suffixes.
Tokenizing
Cutting the final clean sentence into an array of individual root words (tokens) for analysis.
Term weighting scheme to assign numerical values to each root word token based on its frequency of occurrence in a comment (Term Frequency) inversely proportional to its rarity in all database documents (Inverse Document Frequency).
Conditional probability-based classification method that calculates the probability weight of word occurrences in each sentiment class (positive, neutral, negative) with smoothing parameter optimization to determine the final estimated category of opinions.
Benefits of System Development
Theoretical & Practical Contributions
Theoretically (Academically)
Providing academic references and scientific contributions to the realm of Indonesian Natural Language Processing (NLP), especially in optimizing unstructured social media text data preprocessing and the application of Naive Bayes Classifier.
Practically (Social & Government)
Providing public opinion visualization and objective transparency media for government agencies to evaluate the success and real public acceptance of state budget efficiency policies.