Derwin Suhartono

Enhanced Latent Semantic Analysis by considering mistyped words in automated essay scoring

In this paper, we present an approach to consider the mistyped words which frequently occur in an essay. Not all the mistyped words are totally wrong. If they are caused by human errors, we should look further to handle it. Thus, enhanced version of Latent Semantic Analysis (LSA) is proposed. LSA is enhanced by calculating the value of mistyped words before constructing the term document matrix. Essays which are scored consists of 119 English essays of students writing assignment while several essays as gold standard are made by experts. All of the essays have been scored manually by the human grader. Enhanced LSA gives the closeness level of 0.242 to human judgment, while LSA indicates the closeness level of 0.244 to human judgment. Unfortunately, Enhanced LSA cannot outperform GLSA which is one of current methods. The experiment result indicates that Enhanced LSA by considering mistyped words has a better closeness value to human judgment compared with LSA

Personality prediction based on Twitter information in Bahasa Indonesia

The sheer usage of social media presents an opportunity for an automated analysis of a social media user based on his/her information, activities, or status updates. This opportunity is due to the abundant amount of information shared by the user. This fact is especially true for countries with high number of active social media users such as Indonesia. Extraction of information from social media can yield insightful results if done correctly. Recent studies have managed to leverage associations between language and personality and build a personality prediction system based on those associations. The current study attempts to build a personality prediction system based on a Twitter users information for Bahasa Indonesia, the native language of Indonesia. The personality prediction system is built on Support Vector Machine and XGBoost trained with 329 instances (users). Evaluation results using 10-fold cross validation shows that the system managed to reach highest average accuracy of 76.2310% with Support Vector Machine and 97.9962% with XGBoost.

A literature review of question answering system using Named Entity Recognition

Named Entity Recognition (NER) is well-known as the core component of question answering system. NER has traditionally been developed as a component for information extraction systems and current techniques are focused on this end use. This paper describes some approaches that are used by some research in this decade. We also do some analysis regarding the approach, resource (like corpora, language and so on), and result of the research. At the end of this paper, we conclude several issues regarding NER and find out that research in question answering system is still an interesting field to work on. Some resources are still able to be utilized for machine learning tools. Building NER for the origin country of the authors will be beneficial as well.

Optimization economic dispatch with transmission losses using by firefly algorithm

The major issue in power system is economic dispatch (ED) problem. The Economic dispatch problem is aim to distribute electrical power generators in some active electrical system to minimize the total of cost of generators. Here, the purpose of this paper is to achieve the combination of power output, maximum total of demand with the transmission losses, minimum the total cost of generating unit and design web based application to simplify the calculation of economic dispatch problem for user at PT. Indonesia Power. This paper introduces a solution to ED problem using Dr. Xin She Yang developed firefly algorithm. The result obtained by the proposed algorithm have been compared with the result from the data source and that results is more optimal using firefly algorithm than the results of the data source. Based on the result, it can be concluded that this web based application is very useful to user in calculating the economic dispatch problem with transmission losses at PT. Indonesia Power.

The effective noise removal techniques and illumination effect in face recognition using Gabor and Non-Negative Matrix Factorization

There is always noise inside the digital images. Noise is an unwanted component of the image. The existence of noise in a face image can degrade the accuracy of a face recognition. Therefore, we need a proper method that can cope noise or restore the quality of the image. The best method to overcome noise in the image is to use smoothing (filter). In this research, we discuss some techniques to overcome noise in face recognition task using Gabor and Non-Negative Matrix Factorization (NMF), as it is stated in the previous research that it still cannot handle images with noise yet. The noises discussed in this research consist of impulse noise (salt-and-pepper), additive noise (Gaussian) and multiplicative noise (speckle). The experiment was conducted by using two face databases; they were ORL and Extended Yale B. The result said that mean filter is the best coping technique for Gabor and NMF face recognition methods. We used K-Nearest Neighbors (KNN) as the classifier and it achieved 90.83% accuracy rate.

Predictive Text System for Bahasa with Frequency, n-gram, Probability Table and Syntactic using Grammar

Predictive text system is an alternative way to improve human communication, especially in matter of typing. Originally, predictive text system was intended for people who have flaws in verbal and motor. This system is aimed to all people who demands speed and accuracy in typing a document. There were many similar researches which develop this system that had their own strengths and weaknesses. This research attempts to develop the algorithm for predictive text system by combining four methods from previous researches and focus only in Bahasa (Indonesian language). The four methods consist of frequency, n-gram, probability table, and syntactic using grammar. Frequency method is used to rank words based on how many times the words were typed. Probability table is a table designed for storing data such as predefined phrases and trained data. N-gram is used to train data so that it is able to predict the next word based on previous word. And syntactic using grammar will predict the next word based on syntactic relationship between previous word and next word. By using this combination, user can reduce the keystroke up to 59% in which the average keystroke saving is about 50%.

PROTOTYPE DESIGN SYSTEM FOR MOBILE WEB USING RADIO FREQUENCY IDENTIFICATION AT PARKING BUILDING

Several factors cause people forget things. They are retrieval failure, interference failure to store and motivated forgetting. So, it is considered to have tools in order to prevent those. This happens almost everywhere, including when the drivers forgot where their car has been parked. Furthermore, the increase of parking fee is also considerable. The purpose of this research is to design a prototype system for mobile web as one of the helpful tools for car parking location reminder using Radio Frequency Identification (RFID) technology that implemented at the parking building. The research is conducted by direct observation of the current system, do a comparison of similar applications, design the architecture of the system, make the programs, create a mock-up and mobile web as a prototype. The result of this research is a prototype of parking location reminder using RFID-based on mobile web. The mobile web displays information menu such as the location and the parking information. Based on the research, it is concluded that the mobile web can be used as one of solutions for parking location reminder and allows users to view information such as the parking duration and the parking fee estimation.