Dessi Puji Lestari

Emotion recognition on Indonesian television talk shows

As interaction between human and computer continues to develop to the most natural form possible, it becomes more and more urgent to incorporate emotion in the equation. The field continues to develop, yet exploration of the subject in Indonesian is still very lacking. This paper presents the first study of emotion recognition in Indonesian, including the construction of the first emotionally colored speech corpus in the language, and the building of an emotion classifier through an optimized machine learning process. We construct our corpus using television talk show recordings in various topics of discussion, yielding colorful emotional utterances. In our machine learning experiment, we employ the support vector machine (SVM) algorithm with feature selection and parameter optimization to ensure the best resulting model possible. Evaluation of the experiment result shows recognition accuracy of 68.31% at best.

Construction and analysis of Indonesian Emotional Speech Corpus

In this paper we present Indonesian Emotional Speech Corpus (IDESC), the first ever corpus in Indonesian that contains various emotion contents. As interaction between human and computer makes its way to the most natural form possible, it becomes more and more urgent to incorporate emotion in the equation. However, in Indonesian, this aspect is yet to be explored. The acquisition of an emotion corpus serves as a foundation in further research regarding the subject. In constructing IDESC, we aim at natural and real emotion that is applicable to human-computer interaction. The corpus consists of three episodes of Indonesian talk show in different genres: politics, humanity, and entertainment. Each episode is carefully segmented and labeled based on its emotion content, resulting in 2179 segments worth 1 hour, 34 minutes, and 49.7 seconds of speech. The corpus is still in its early stage of development, yielding exciting possibilities of future works.

Filled Pause Detection in Indonesian Spontaneous Speech

Detecting filled pause in spontaneous speech recognition is very important since most of the speech is spontaneous and the most frequent phenomenon in Indonesian spontaneous speech is filled pause. This paper discusses the detection of filled pauses in spontaneous speech of Indonesian by utilizing acoustic features of the speech signal. The detection was conducted by employing statistical method using Naive Bayes, Classification Tree, and Multilayer Perceptron algorithm. To build the model, speech data were collected from an entertainment program. Word parts in the data were labeled and its features were extracted. These include the formant and pitch stability, energy-drop, and duration. Half an hour of sentences contains 295 filled pause and 2082 non-filled pause words were employed as training data. Using 25 sentences as testing data, Naive Bayes gave best detection correctness, 74.35 % on a closed data set and 71.43 % on an open data set.

Spatial data model for corporate based on Google Maps platform

Spatial data wrapped and processed by an application known as Geographic Information Systems (Geographical Information System / GIS). In desktop based GIS, the spatial information services only occurs when a variety of basic data has been loaded into the applications database. The real challenge for corporate is how to create a GIS that smart, secure, and relatively low cost. It is a real difficulty for a corporate to build such system, because they have to supply the basic data, their specific data, and define the relation. The development of information technology has provide ready to access, worldwide scale, web based mapping system, like Google Maps. Google Maps already provides the basic spatial information. By using Google Maps as basic and background information layer, corporates can focus on their specific spatial data layer on top of it. The main study of this research is to provide a way to create smart, secure, and relatively low cost geographic information services for corporate by utilizing basic spatial data provided by Google Maps. The paper highlight our first step in creating the system by proposing a new model of corporate spatial data management based on Googgle Maps platform.

Automatic Extraction Phonetically Rich and Balanced Verses for Speaker-Dependent Quranic Speech Recognition System

This paper discussed how to collect phonetically rich and balanced verses as speech corpus for quranic recognition system. The Quranic phonology was analyzed based on the qira’a of ‘Asim in the riwaya of Hafs to transform arabic text of Holy Quran into alphabetical symbols that represent all possible sounds (QScript) when Holy Quran is read. The entire verses of Holy Quran were checked to select verses-set which met the criteria of a phonetically rich and balanced corpus. The selected verses contained 180 verses of 6236 whole verses in Quran. Statistical phonemes distribution similarity of selected verses was 0.9998 compared to phonemes distiribution in whole Quran. To determine the effect of using this corpus, early development speaker-dependent Quranic recognition system based on CMU Sphinx was developed. MFCC was used as feature extraction. The system used HMM with 3-emitting-states based on tri-phone. For language model, the system used N-gram with word as a basis. The system was trained using recitation from 3 speakers and obtained a recognition accuracy of 97.47 %.

Handling arbitrary polygon query based on the boolean overlay on a geographical information system

Current Geographical Information System (GIS) has been applied the classical point-in-polygon algorithm such as the ray casting algorithm and winding algorithm to conduct spatial analysis. These algorithms have linear complexity with the number of points available in the map. In this paper, we introduce an efficient algorithm to handle arbitrary polygonal queries on a GIS based on the boolean overlay. The algorithm is able to find whether a point is inside, outside, or on the boundary of a given polygon by utilizing a very simple boolean overlay mechanisms in a GIS. Experimental results show that the algorithm works very accurate with 100% accuracy and works very fast.

Shared-hidden-layer Deep Neural Network for Under-resourced Language the Content

Training speech recognizer with under-resourced language data still proves difficult. Indonesian language is considered under-resourced because the lack of a standard speech corpus, text corpus, and dictionary. In this research, the efficacy of augmenting limited Indonesian speech training data with highly-resourced-language training data, such as English, to train Indonesian speech recognizer was analyzed. The training was performed in form of shared-hidden-layer deep-neural-network (SHL-DNN) training. An SHL-DNN has language-independent hidden layers and can be pre-trained and trained using multilingual training data without any difference with a monolingual deep neural network. The SHL-DNN using Indonesian and English speech training data proved effective for decreasing word error rate (WER) in decoding Indonesian dictated-speech by achieving 3.82% absolute decrease compared to a monolingual Indonesian hidden Markov model using Gaussian mixture model emission (GMM-HMM). The case was confirmed when the SHL-DNN was also employed to decode Indonesian spontaneous-speech by achieving 4.19% absolute WER decrease.

Positive and negative emotions recognition from speech signal using acoustic and lexical features

This paper aims to present the work on emotion recognition in Indonesian spoken language. For this effort, we construct Indonesian Emotional Corpus (IDEC). In constructing the corpus, we aim at natural emotional occurrences. We choose television talk shows due to its natural emotional content. We extract two types of features from IDEC: acoustic and lexical features. We employ Support Vector Machine (SVM) and Random Forest algorithm to model the emotions utilizing these two types of features. Experiment result shows an average F-measure of 0.772 at best for 2 emotion classes based on valence dimension. We hope to continue this work and move towards more precise emotion recognizer by collecting more data, incorporating other features, and defining more complex emotion classes.

Transcriber: An Android application that automates the transcription of interviews in Indonesian

In this paper, Transcriber that can be used to automatically transcribe interviews in Indonesian using speech-to-text and speaker diarization technology is described. The main feature of the software is generating interview transcription automatically and providing an option if grouping by group of speakers is required. Transcriber is designed to work in two modes that give users the freedom to provide recording file input or perform live recording. Automatic Speech Recognition (ASR) used in Transcriber was developed by utilizing KALDI and the ASR model developed in the previous research, while the speaker diarization was developed with LIUM Speaker Diarization and successfully optimized for Indonesian with DER 35.10%. These technologies are integrated into a whole system that has client-server architecture and is capable of transcribing interviews for 1.10 times the duration of input recordings excluding connection latency. Transcriber has also managed to gain positive feedback on the usability testing.

Designing an arisan mobile application for novice users using user-centered design approach

A lot of smartphone users are still novice and have difficulty in operating tasks using their smartphone, including running applications, for example running an online Arisan application. Arisan is a traditional social gathering that is common in Indonesia. Members of arisan contribute to and take turns at winning, and most of the arisan users are novice users. In this paper, we present the results of our observations to determine problems faced by the novice users when using an online arisan application. The user-centered design approach is employed to analyze the problems and to design the new online arisan application prototype. The prototype is designed based on the usability goal and the user experience defined for the novice users. We conduct a usability testing to evaluate the prototype. The testing results show that the prototype fulfills the usability goal and the user experience.