Philippine Computing Journal

Includes bibliographical references.

Design, Implementation and Evaluation of a Scanned Filipino Text-to-Speech Device as a Reading Aid for the Blind and Visually Impaired -- Developing an Automated Pipeline for Information Extraction and Speech Classification from Floor Debate Records of the House of Representatives of the Philippines -- Automated Text Summarization of Research Papers Regarding the Effectiveness of Various Treatment Plans for Leukemia -- Extracting and Classifying Events from Social Media Posts forLife Story Generation -- Is There a Philippine English Vowel Space?.

[Article Title: Design, Implementation and Evaluation of a Scanned Filipino Text-to-Speech Device as a Reading Aid for the Blind and Visually Impaired/ Christian Paulo L. De Leon, Allan Adrian B. Gatchalian, Juan Miguel D. Ninobla, Nigel Roi D. Concepcion, David Joseph O. Catemprate and Ronald M. Pascual, p. 1-8]

Abstract: In this study, the authors provide new insights to the design, implementation and evaluation of a TTS system where the inputs are text images in Filipino. Many text-to-speech (TTS) systems have been studied and implemented in many languages. [8]Portable devices for Filipino TTS however have not been explored yet. With the aid of portable scanning device and Optical Character Recognition (OCR), our TTS system can generate synthetic Filipino speech from texts found on printed reading materials in Filipino.

https://pcj.csp.org.ph/index.php/pcj/issue/view/27/PCJ%20V13%20N1%20pp1-8%202018

[Article Title: Developing an Automated Pipeline for Information Extraction and Speech Classification from Floor Debate Records of The House of Representatives of the Philippines/ Carlos Alberto L. Arcenas, Miguel N. Galace, and RMarlene M. De Leon, Ph.D., p. 9-20]

Abstract: A lot can be learned about members of Congress in the way they speak and vote on bills during plenary proceedings. This paper tack-les the problem of automating the extraction of speech and vote data from legislative documents of the House of Representatives of the Philippines to create a classification model for determining votes based on given speeches. To accomplish this, a framework was conceptualized and developed to extract relevant data from structured documents. The framework was then applied to the official House website to gather documents which, in turn, coursed through automated text processing and data structuring. The resulting structured data is made accessible through an API, named the Floor reader API, to make the extracted information adaptable for further exploration. The resulting structured data was used to train a congressional vote classification model. Overall, 1,161Congressional Records and 728 House Journals were sourced for processing, resulting in the collection of 112,736 speeches across the debate of 752 bills, paired with 1,649 votes. The classification model developed was able to achieve performance measures in precision, recall, and F1-score of up to 85%, 84%, and 84% respectively

https://pcj.csp.org.ph/index.php/pcj/issue/view/27/PCJ%20V13%20N1%20pp9-20%202018

[Article Title: Automated Text Summarization of Research Papers Regarding the Effectiveness of Various Treatment Plans for Leukemia / Jan Apolline D. Estrella, Christian Philip L. Gelera, Christian S. Quinzon, Ethel C. Ong, and Edlen Mari M. Sanchez, p. 21-28]

Abstract: Text summarization involves the identification and extraction of sentences from long documents to produce shorter-length summaries that enable readers to have quick access to relevant information. In this paper, we describe our automated text summarizer that is capable of extracting essential information from research papers regarding various treatment plans for leukemia to aid users in reviewing online research publications more efficiently. The summarizer applies natural language processing and sentence scoring techniques to select parts of a paper, e.g. introduction, body, results, in bullet form. The generated summaries from three leukemia-related papers were then subjected for manual evaluation by experts and laymen. These summaries were found to consist of concise, informative and self-contained descriptions regarding the abstract and the introduction of a given article. However, these were also observed to have an excess of presented data, and a lack of background information, which an end-user is mostly likely required to know to fully comprehend the methods and findings of a study.

https://pcj.csp.org.ph/index.php/pcj/issue/view/27/PCJ%20V13%20N1%20pp21-28%202018

[Article Title: Extracting and Classifying Events from Social Media Posts for Life Story Generation / Robee Khyra Mae J. Te, Janica Mae M. Lam, Camille Alexis T.R. Saavedra, Alden Luc R. Hade, and Ethel Ong, p. 29-36]

Abstract: People use social media, in particular Facebook, to share stories about themselves and the things that interest them. However, a Facebook user’s posts by themselves cannot provide a concise narrative of events to tell a complete life story. Story generators can be designed to utilize these events extracted from posts that users share about themselves into a life story. Before this can be achieved, the story generator needs to be able to classify posts based on their textual content. Such techniques are already available in email, where it is possible to classify messages into categories. In this paper, we describe our approach in using automated classifiers to categorize a user’s posts, focusing on those that describe travelling, eating, dining and celebrating events. We then show the performance of our classifiers. Once events are identified, we extract details from these posts and store them into an event model. We end our paper with a short discussion of our story generation process that utilizes the topical and temporal relations that exist among the events.

https://pcj.csp.org.ph/index.php/pcj/issue/view/27/PCJ%20V13%20N1%20pp29-36%202018

[Article Title: Is There a Philippine English Vowel Space? / Nivea Jane Urdas, Jhon Rick Salvedia, Drew Dolot, John Meynard Sanchez, Jay Chester Usero, Ava Marie Villareal, and Dewi Retnoningsih, p. 37-49]

Abstract: Philippine languages commonly exhibit three or four-vowel systems consisting of an open mid-front unrounded [a], a high front unrounded [i], a mid-front unrounded [e], and one back vowel orthographically representing as [o] and [u]. This finding is in contrast with the five (5) orthographic characters <a, e, i, o, u> that are believed to exist in the Philippine vowel inventory. This study initially surveys the positions of the Philippine vowels in the vowel space diagrams using JPlot to first establish their locations during phonetic articulation. Informants are then asked to pronounce English vowels found between consonants /h/ and /d/ in environments (hVd environment where V stands for vowel) and from the phonological data gathered, the writers compared their phonetic articulations to those of Hillenbrand et al. (1995) and Peterson & Barney (1952). The phonetic output suggests that: (1) a Philippine English vowel space exists; (2) there seemed to be no significant difference between the Philippine English vowel space and that of the native speakers of American English in the study of Hillenbrand et al. (1995) and; (3) the suggested Philippine English vowel space is heavily influenced by the first language of Philippine language speakers.

https://pcj.csp.org.ph/index.php/pcj/issue/view/27/PCJ%20V13%20N1%20pp37-49%202018