Claim Missing Document
Check
Articles

Found 5 Documents
Search
Journal : Journal of Software Engineering, Information and Communication Technology

Exploration of Spontaneous Speech Corpus Development in Urban Agriculture Instructional Videos Trisna Gelar; Aprianti Nanda
Journal of Software Engineering, Information and Communication Technology (SEICT) Vol 3, No 1: June 2022
Publisher : Universitas Pendidikan Indonesia (UPI)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.17509/seict.v3i1.44548

Abstract

Video transcription can be obtained automatically based on the original language translation of the video maker's speech, but the quality of the transcription depends on the quality of the audio signal and the natural voice of the speaker. In this study, Deep Speech is used to predict letters based on acoustic recognition without understanding language rules. The Common Voice multilingual corpus helps Deep Seech to transcribe Indonesian. However, this corpus does not accommodate the special topic of urban agriculture, so an additional corpus is needed to build acoustic and language models with the urban agriculture domain. A total of 15 popular videos with closed captions and nine E-Books with the theme of Horticulture (fruit, vegetables and medicinal plants) were curated. The video data were extracted into audio and transcription according to specifications as training data, while the agricultural text data were transformed into language models, which were used to predict recognition results. The evaluation results show that the number of epochs has an effect on improving the transcription performance. The language model score used during prediction improved WER performance as it interpreted words with agricultural terms. Another finding was that the model was unable to predict short words with informal varieties and located at the end of the sentence.
Klasifikasi Komentar Video Instruksional Populer Bertemakan Pekarangan Perkotaan menggunakan Auto-Keras Trisna Gelar; Aprianti Nanda Sari
Journal of Software Engineering, Information and Communication Technology (SEICT) Vol 1, No 1: December 2020
Publisher : Universitas Pendidikan Indonesia (UPI)

Show Abstract | Download Original | Original Source | Check in Google Scholar | Full PDF (414.445 KB) | DOI: 10.17509/seict.v1i1.29050

Abstract

Keterbatasan kompetensi menjadi halangan untuk memulai melakukan kegiatan pekarangan perkotaan. Mempraktikkan langkah-langkah pada video instruksional populer di Youtube dari individu maupun profesional dapat meningkatkan kompetensi diri. Namun, kualitas video instruksional(konten, audio dan visual) sangat bervariasi bergantung pada orang yang memproduksinya. Penonton secara langsung dapat berinteraksi dengan memberikan apresiasi (positif maupun negatif), tanggapan atau pertanyaan pada kolom komentar seputar topik yang dipresentasikan. Umpan balik tersebut digunakan untuk memperbaiki kualitas dari video seperti memberikan penjelasan mendalam untuk topik yang sering ditanyakan dan melanjutkan atau menghentikan video berdasarkan topik yang paling disukai atau sebaliknya. Pekerjaan klasifikasi komentar dapat diselesaikan dengan mudah menggunakan Auto-Keras karena proses pemilihan model, pencarian arsitektur neural-network dan evaluasi model terbaik dilakukan secara otomatis. Penelitian pada umumnya terdiri atas empat fase, yaitu (1) pengumpulan dataset, (2) text processing, (3) feature engineering, dan (4) pemodelan dan evaluasi. Pada penelitian ini telah terkumpul 5194 komentar berlabel(aspirasi, pertanyaan, dan pernyataan) dari 5 video instruksional populer bertemakan pekarangan kota yang dikurasi oleh penulis berdasarkan urutan views, likes dan dislikes tertinggi. Kualitas kalimat komentar diperbaiki pada fase persiapan melalui proses text cleaning, normalization, tokenization dan stemming. Pada proses normalization, kamus istilah pertanian menjadi informasi agar tidak tercampur dengan bahasa informal yang mirip. Kalimat komentar yang telah normal dikonversikan menjadi n-gram dan word embedding sebagai input auto-keras. Dari hasil pengujian evaluasi model, akurasi yang dihasilkan auto-keras dengan fitur word embedding mencapai 86.91% sedikit lebih baik dari akurasi fitur n-gram 86.33%.Keterbatasan kompetensi menjadi halangan untuk memulai melakukan kegiatan pekarangan perkotaan. Mempraktikkan langkah-langkah pada video instruksional populer di Youtube dari individu maupun profesional dapat meningkatkan kompetensi diri. Namun, kualitas video instruksional(konten, audio dan visual) sangat bervariasi bergantung pada orang yang memproduksinya. Penonton secara langsung dapat berinteraksi dengan memberikan apresiasi (positif maupun negatif), tanggapan atau pertanyaan pada kolom komentar seputar topik yang dipresentasikan. Umpan balik tersebut digunakan untuk memperbaiki kualitas dari video seperti memberikan penjelasan mendalam untuk topik yang sering ditanyakan dan melanjutkan atau menghentikan video berdasarkan topik yang paling disukai atau sebaliknya. Pekerjaan klasifikasi komentar dapat diselesaikan dengan mudah menggunakan Auto-Keras karena proses pemilihan model, pencarian arsitektur neural-network dan evaluasi model terbaik dilakukan secara otomatis. Penelitian pada umumnya terdiri atas empat fase, yaitu (1) pengumpulan dataset, (2) text processing, (3) feature engineering, dan (4) pemodelan dan evaluasi. Pada penelitian ini telah terkumpul 5194 komentar berlabel(aspirasi, pertanyaan, dan pernyataan) dari 5 video instruksional populer bertemakan pekarangan kota yang dikurasi oleh penulis berdasarkan urutan views, likes dan dislikes tertinggi. Kualitas kalimat komentar diperbaiki pada fase persiapan melalui proses text cleaning, normalization, tokenization dan stemming. Pada proses normalization, kamus istilah pertanian menjadi informasi agar tidak tercampur dengan bahasa informal yang mirip. Kalimat komentar yang telah normal dikonversikan menjadi n-gram dan word embedding sebagai input auto-keras. Dari hasil pengujian evaluasi model, akurasi yang dihasilkan auto-keras dengan fitur word embedding mencapai 86.91% sedikit lebih baik dari akurasi fitur n-gram 86.33%.
Geometry and Color Transformation Data Augmentation for YOLOV8 in Beverage Waste Detection Itikap, Sabar Muhamad; Abdurrahman, Muhammad Syahid; Soewono, Eddy Bambang; Gelar, Trisna
Journal of Software Engineering, Information and Communication Technology (SEICT) Vol 4, No 2: Desember 2023
Publisher : Universitas Pendidikan Indonesia (UPI)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.17509/seict.v4i2.64400

Abstract

In the bottle sorting process in real world, there are some beverage packaging waste that is deformed. Deformed objects can result in detection errors by an object detection system. Detection errors can also occur in attributes that share similar feature maps. Detection errors can be caused by models that are unable to generalize to the data. Several methods have been devised to prevent such issues, with data augmentation being one of them. To increase the variety of data, data enhancement techniques will be utilized. This research employs a data augmentation technique that concentrates on geometry transformations such as scaling and rotation, as well as color transformations such as hue, saturation, and brightness. Additionally, a combination of geometry and color transformations was conducted, resulting in a total of 39 experimental scenarios. This study demonstrates that data augmentation can affect the model's performance in terms of accuracy and the number of detection results. The combined method of scaling and rotation, which is applied to the original data, reveals the optimal experimental scenario with an accuracy of 88.4%.
The Investigation of Convolution Layer Structure on BERT-C-LSTM for Topic Classification of Indonesian News Headlines Fabillah, Dzakira; Auliarahmi, Rizka; Setiarini, Siti Dwi; Gelar, Trisna
Journal of Software Engineering, Information and Communication Technology (SEICT) Vol 4, No 2: Desember 2023
Publisher : Universitas Pendidikan Indonesia (UPI)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.17509/seict.v4i2.63742

Abstract

An efficient and accurate method for classifying news articles based on their topics is essential for various applications, such as personalized news recommendation systems and market research. Manual classification methods are tedious, prompting the use of deep learning techniques in this study to automate the process. The developed model, BERT-C-LSTM, combines BERT, the convolutional layer from CNN, and LSTM, leveraging their individual strengths. BERT excels at transforming text into context-dependent vector representations, The design of the classification model employs a blend of convolutional layers and LSTM, referred to as C-LSTM. The convolutional layer possesses the capability to extract salient elements, including keywords and phrases, from input data. On the other hand, the Long Short-Term Memory (LSTM) model exhibits the ability to comprehend the temporal context present in sequential data. This study aims to investigate the influence of the convolutional layer structure in BERT-C-LSTM on the classification of Indonesian news headline categorized into eight topics. The results indicate that there are no significant differences in accuracy between BERT-C-LSTM model architectures with a single convolutional layer and multiple parallel convolutional layers and the models using various filter sizes. Furthermore, the BERT-C-LSTM model achieves an accuracy that is not much different from the BERT-LSTM and BERT-CNN models, with accuracies reaching 92.6%, 92.1%, and 92.7%, respectively.
Morphological Grayscale Pre-processing to SAR Images for Reducing Noise in Ship Detection Based on YOLOv8 Pratidina, Caturiani; Safira, Decia; Gelar, Trisna; Permana, Heru; Suprihanto, Suprihanto; Syakrani, Nurjannah; Fauzi, Cholid
Journal of Software Engineering, Information and Communication Technology (SEICT) Vol 5, No 2: December 2024
Publisher : Universitas Pendidikan Indonesia (UPI)

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.17509/seict.v5i2.75970

Abstract

The development of a ship detection system using SAR pictures loaded with noise poses issues for pictures Intelligence (IMINT). The YOLOv8 model is utilized for ship identification. The preprocessing approaches entail employing a fusion of grayscale morphology techniques and image restoration using a harmonic mean filter and a bandpass. This technique is designed to assess the effect of noise reduction to enhance the accuracy of detecting objects in SAR images. The preprocessing technique is categorized into two methods: basic grayscale morphology (GM1-GM6) and a fusion of image restoration with grayscale morphology (GHB1-GHB6). The model's performance is assessed using mAP and IoU criteria. This research discovered that ship objects were not detected successfully in the presence of several types of noise. These failures were attributed to factors such as tiny ship size, low picture quality, and inadequate preprocessing techniques for noise handling. The findings indicate a substantial enhancement in ship detection, specifically in synthetic aperture radar (SAR) images affected by sidelobe noise. There were noticeable enhancements in the accuracy of images that underwent preprocessing using GHB5. GHB5 employs a combination of image restoration, closure, and erosion techniques.