Enhancing text classification performance by preprocessing misspelled words in Indonesian language

Abstract

Supervised learning using shallow machine learning methods is still a popular method in processing text, despite the rapidly advancing sector of unsupervised methodologies using deep learning. Supervised text classification for application user feedback sentiments in Indonesian Language is one of the applications which is quite popular in both the research community and industry. However, due to the nature of shallow machine learning approaches, various text preprocessing techniques are required to clean the input data. This research aims to implement and evaluate the role of Levenshtein distance algorithm in detecting and preprocessing misspelled words in Indonesian language, before the text data is then used to train a user feedback sentiment classification model using multinomial Naïve Bayes. This research experimented with various evaluation scenarios, and found that preprocessing misspelled words in Indonesian language using the Levenshtein distance algorithm could be useful and showed a promising 8.2% increase on the accuracy of the model's ability to classify user feedback text according to their sentiments.

Improving sentiment reviews classification performance using support vector machine-fuzzy matching algorithm

“…Fuzzy matching (FM) can be used to measure the inequality between two strings, where text matches can be found, even with misspelled and different words [3]. Measurement of text similarity can also apply Euclidean distance similarity (EDS) [7]. Principal component analysis (PCA) is a conversion technique that allows to reduce the size of a data set that includes a large number of interrelated dimensionalities, so that the current data can be expressed with a smaller number of variables [8].…”

“…The tokenizing technique helps to reduce the dimensions of the problem in a group of texts. By cutting a group of sentences into chunks of words makes the analysis process into a simpler form [1], [7], [9]- [11]. Capitalization variability in this dataset can cause problems during classification and degrade performance.…”

“…The technique of converting capital letters to lowercase is the most common method of dealing with problems in text data. The lowercasing technique helps avoid different variations of the same word as determined by the case [5], [7], [9]- [15]. The existence of the stop word in a group of text is quite a lot, but this word has no meaning to be analyzed.…”

“…Stemming technique [1], [5]- [7], [9]- [12] has the advantage of being able to change words into basic forms (removing affixes). This step can reduce the analysis process on unnecessary text.…”

“…According to Romadhon and Kurniawan [17] apply kNN and produce 75% accuracy. To simplify the dimensions of the analysed text, an algorithm is used FM [3], PCA [8], [24], [25], EDS [7].…”