Paper Reading #1: picoTrans: An Icon-driven User Interface for Machine Translation on Mobile Devices

Introduction:  

Authors: Wei Song, Kumiko Tanaka-Ishii, Andrew Finch, and Eiichiro Sumita

Reference information: This paper was a collaboration between the University of Tokyo Graduate School of Science and Technology and the NICT Language Translation Group in Japan.

Author bios:  
Wei Song is a doctoral candidate in the Graduate School of Information Science and Technology at the University of Tokyo; his bachelor's degree is from the Beijing University of Aeronautics and Astronautics, and his master's is from the University of Tokyo.
Kumiko Tanaka-Ishii was a professor at the University of Tokyo that moved to Kyushu University in April 2012. She specializes in language processing, mathematical modeling of language, language software, and computational semiotics. She previously worked at the National Institute of Advanced Industrial Science and Technology in Japan.
Andrew Finch and Eiichiro Sumita have worked on several translation programs and systems.

Finding related works is difficult due to the specificity of the interface; fwe

Summary: The authors of this paper created an interface based on a picture book method of communication. With this interface, the user taps appropriate icons to create sentences; once the icons are selected, a sentence is created in the user's language. If the sentence is not what the user intended, they are able to alter it until it's correct. The sentence is then translated to the target language. To double-check translation correctness, users can back-translate the sentence in the target language. In this research, only icons about traveling were used (e.g. "I want to go" and "the restaurant").

Related work: 
A Picture Language Translator - Hiroshi Migiyama, Taiichiro Hayashi, Nikolay Mirenkov
A Translation Method from Natural Language Specifications into Formal Specifications Using Contextual Dependencies - Yasunoria Ishihara, Hiroyuki Seki, Tadao Kasami

Portable computer based language translation - Davis, S.

Study and Correlation Analysis of Linguistic, Perceptual, and Automatic Machine Translation Evaluations - Mireia Farrus, Marta R. Costa-jussa, Maja Popovic

An Advanced Review of Hybrid Machine Translation - H.W. Xuan, W. Li, G.Y. Tang

The language translation interface: A perspective from the users - Dominique Estival

Icon Design Study in Computer Interface - Rushan Yan

The difficulty in finding relevant works demonstrates the novelty of the project. Few people have used an icon interface with machine translation, apparently.

Evaluation: To evaluate the usefulness of the project, the creators found 100 sentences involving travel and determined whether equivalent sentences could be created. This resulted in a quantitative, unbiased evaluation (e.g. "For 49 of the 74 sentences that we were able to cover with our system (66% of them), the system proposed the correct source sentence to the user..."). They also analyzed the number of key press actions between this method and typing out sentences; this system was found to use 57% of the number of key presses as the text-entry method. This was a systemic evaluation, showing that the entire system works. However, the authors do state that this prototypical system is incomplete and that further development could make it more accurate.

Discussion:  I feel as though this system is useful, but only for specific people in specific circumstances. Translation of spoken words seems more useful and much quicker than using icons, and typing out the sentences seems more efficient as far as time goes. The fewer key presses would give a translation tool to people unaccustomed to or unskilled in using typical interfaces. The work is novel, combining two previously used translation interfaces ("picture books" and straight text translation). The evaluation seemed incomplete; while the number of key presses is one measure of efficiency, I feel like time taken while constructing sentences should have been investigated.