Close this tab or window to return to previous window or tab
Andoni Aizpietia is part of the Human Speech and Language Technologies (HSLT) group at VICOMTECH-IK4, a group he joined in early 2013.
He holds an Engineering degree in Computer Science, obtained at the University of the Basque Country (UPV) in 2010. He then graduated with a Masters in Language Analysis and Processing (LAP), a degree offered by the IXA group at the UPV (2014).
He has participated in several European, National and Regional research projects, notably the Opener and Sumat FP7 projects, on several NLP topics, including core NLP technologies, sentiment analysis, automated quality estimation and satistical machine translation.
He is currently a member of the Machine Translation group within the HSLT department, involved in its core research and development activities.
Co-authored the Poster presentation ...
The reception of intralingual and interlingual automatic subtitling: an exploratory study within the HBB4ALL project
Co-authors: Anna Matamala, Andreu Oliver and Aitor Alvarez, being presented by Anna Matamala
HBB4ALL is a EC funded project (CIP-ICT-PSP-2013-5.1.) that builds on HbbTV, the European standard for broadcast and broadband multimedia converged services, and looks at how HBbTV technologies may be used to enhance access services such as subtitling. This poster presents the results of user testing on automatic subtitling carried out within the project. A first preliminary test with 56 students allowed to: (a) compare the comprehension of clips with automatic intralingual subtitling versus automatic interlingual subtitles in students with low English proficiency, (b) compare the comprehension reached by students with a low level of English using subtitles to that of students with a higher level using no subtitles, and most importantly (c) test the methodology. A second experiment aimed to determine if intralingual or interlingual automatic subtitling help to better understand news content. Three breaking news clips in English automatically subtitled in English and Spanish were used. Data from 30 volunteers exposed to the three conditions (no subtitles/intralingual/interlingual) were analysed. Comprehension was measured through a summarisation task and a questionnaire. An English proficiency control test as well as a demographics questionnaire were administered. The paper presents the results of both tests and discusses methodological issues.
The Hybrid Broadcast Television for All (HBB4ALL) project is a European Commission co-financed project within the Competitiveness and Innovation Framework Programme (CIP-ICT-PSP-2013-5.1.). The project builds on HbbTV, the European standard for broadcast and broadband multimedia converged services, and looks at how HbbTV technologies may be used to enhance access services such as subtitling, audio description or sign language on both the production and service sides.
The main objective of the HBB4ALL project is to test access services in various pilot implementations (from the definition to the operational phase) and gather implicit and explicit user feedback to assess the acceptance and the achievable quality of service in the various delivery scenarios. Specifically, pilot A researches a “Multi-platform subtitle workflow chain” and aims to advance HbbTV automatic multilingual subtitling functionalities. The pilot is built up on technology developed within other existing European projects, including technology for automatic subtitling generation and translation. This paper presents the results of the experiments carried out by testers within this specific pilot.
The experiments aimed to determine if either intralingual or interlingual automatic subtitling help to better understand news content. Experiments departed from the hypothesis that intralingual automatic subtitling would increase comprehension --as compared to a standard viewing with no subtitles-- but interlingual subtitling would not. The materials used were three two-minute English clips from the Reuters news service. The contents were automatically subtitled in English and automatically translated to Spanish. A post-edition process to correct mistakes was not included.
In the first experiment, 56 Political Science students were divided into two groups (lower level/higher level of English). This preliminary testing allowed us: (a) to compare the comprehension of English clips automatically subtitled in English (intralingual subtitling) and Spanish (interlingual) on students with low level of English, (b) to compare the comprehension reached by students with low level of English using subtitles to that of students with higher level of English without using subtitles, and most importantly, (c) to test the methodology, especially in terms of the best tools to be used when testing comprehension of audiovisual materials.
In the second experiment (main experiment), data from the exposure of 30 volunteers to the three conditions (no subtitles/intralingual/interlingual) were gathered and analysed. Comprehension was measured qualitatively through a summarisation task and quantitatively through a multiple-choice questionnaire, focusing on the key aspects of any breaking news story (who?, what?, when?, where?, how?, why?) and controlling for the information provided visually. An English proficiency control test as well as a demographics questionnaire were administered to the participants previous to the experiment.
The paper presents the rationale behind the experiments, the experimental design and the results. It also discusses methodological aspects such as the best tools to measure comprehension in audiovisual subtitled content.