Revised 5 Report on the Algorithmic Language

Transactions of the Japanese Society for Artificial Intelligence

Transactions of the Japanese Society for Artificial Intelligence

The a Everyday language computing (ELC) is a new computational paradigm that all people, from small children to aged persons, can access and use computing systems with his/her own everyday language. As a way to realize ELC, we proposed a framework of language-based operating system (LOS), and we are now working intensively to develop the fundamental part of it. In this paper, we report our status of research on LOS. One of the main components of LOS is the semiotic base, which is a database of linguistic knowledge based on Systemic Functional Linguistic Theory. We explain the architecture of the semiotic base and corpus experiments conducted to determine the contents of it. A client's secretary agent is a user interface of LOS which realizes everyday language communication between a user and the system. We explain a client model that realizes personalized communication and the plan module that manages the structure of dialogue. In LOS, all kinds of information processing are done through everyday language. Language protocol is a computer protocol based on everyday language, and language-based application is a kind of agent software that provides services through language. We explain the characteristics of language protocol, the structure of language-based application programming interface (LAPI), and text interpretation process of language-based applications. Finally, we discuss prospects of this project and mention future work that remains to be done. We argue that by processing meaning of language rather than processing numbers, we attempt to provide a more human-like computer system and an intelligent computational environment to all people.

The Journal of the Institute of Image Electronics Engineers of Japan, 2012

ており，約 3 割の低ビットレート化を実現できることを確認した． キーワード：ROI-JPEG, 顕著性マップ，画像圧縮，眼球運動計測，構造的類似性 <Summary> By allocating saliency maps to the region of interests (ROIs), ROI-JPEG has been demonstrated to generate JPEG images with high compression rates without degrading perceived quality. However, on putting ROI-JPEG to practical use, some issues need to be resolved: degradation of the resulted image caused by an improper saliency map and difficulty in implementing the code due to inability to use the existing JPEG libraries. This paper proposes a simplified implementation of ROI-JPEG by separating the part of the ROI coding to assure the perceived level of qualities: limit of detection, acceptability limit and to allow the use of an existing JPEG library to save the resulted image. Our experiments show that regardless of the structural similarity between maps of saliency and eye fixation distributions, the output images will receive no degrading perceived quality and about 20 % lower bitrate compared with the ordinary JPEG images by setting the minimum quality factor for the quantization matrices during the ROI coding to 45. Meanwhile, setting the minimum quality factor to 25 will generate about 30% lower bitrate with acceptable perceived quality of images.

Journal of Japan Society for Fuzzy Theory and Intelligent Informatics

Transactions of the Japanese Society for Artificial Intelligence

The task of inducing grammar structures has received a great deal of attention. The reasons why researchers have studied are different; to use grammar induction as the first stage in building large treebanks or to make up better language models. However, grammar induction has inherent computational complexity. To overcome it, some grammar induction algorithms add new production rules incrementally. They refine the grammar while keeping their computational complexity low. In this paper, we propose a new efficient grammar induction algorithm. Although our algorithm is similar to algorithms which learn a grammar incrementally, our algorithm uses the graphical EM algorithm instead of the Inside-Outside algorithm. We report results of learning experiments in terms of learning speeds. The results show that our algorithm learns a grammar in constant time regardless of the size of the grammar. Since our algorithm decreases syntactic ambiguities in each step, our algorithm reduces required time for learning. This constant-time learning considerably affects learning time for larger grammars. We also reports results of evaluation of criteria to choose nonterminals. Our algorithm refines a grammar based on a nonterminal in each step. Since there can be several criteria to decide which nonterminal is the best, we evaluate them by learning experiments.

Shizen gengo shori, 2005

2004

2005

Shisutemu Seigyo Jōhō Gakkai ronbunshi, 2003

Shizen gengo shori, 1995