Toshiko Aizono

Extracting word correspondences from bilingual corpora based on word co-occurrences information

A new method has been developed for extracting word correspondences from a bilingual corpus. First, the co-occurrence information for each word in both languages is extracted from the corpus. Then, the correlations between the co-occurrence features of the words are calculated pairwisely with the assistance of a basic word bilingual dictionary. Finally, the pairs of words with the highest correlations are output selectively. This method is applicable to rather small, unaligned corpora; it can extract correspondences between compound words as well as simple words. An experiment using bilingual patent-specification corpora achieved 28% recall and 76% precision; this demonstrates that the method effectively reduces the cost of bilingual dictionary augmenntation.

Corpus-dependent association thesauri for information retrieval

This paper presents a method for automatically generating an association thesaurus from a text corpus, and demonstrates its application to information retrieval. The thesaurus generation method consists of extracting terms and co-occurrence data from a corpus and analyzing the correlation between terms statistically. A new method for disambiguating the structure of compound nouns, which is a key component for term extraction, is also proposed. The automatically generated thesaurus is effectively used as a tool for exploring information. A thesaurus navigator having novel functions such as term clustering, thesaurus overview, and zooming-in is proposed.

Proposal and Evaluation of Policy Description for Information Lifecycle Management

Information lifecycle management (ILM) is attracting more attention as the costs of data retention and security and of regulatory compliance requirements are being increased by the explosive growth of information being handled and by increasingly stringent government regulations. The goal of ILM is to ensure that information is stored on the most appropriate medium providing the service level required at the phase of the informations lifecycle. This paper describes a method for describing and interpreting ILM policies in a way that information managers find easy to understand and that can be used to automate the ILM process

Parallel-Pathway Generator for Generative Adversarial Networks to Generate High-Resolution Natural Images

Generative Adversarial Networks (GANs) can learn various generative models such as probability distribution and images, while it is difficult to converge training. There are few successful methods for generating high-resolution images. In this paper, we propose the parallel-pathway generator network to generate high-resolution natural images. Our parallel network are constructed by parallelly stacked generators with different structure. To investigate the effect of our structure, we apply it to two image generation tasks: human-face image and road image which does not have square resolution. Results indicate that our method can generate high-resolution natural images with few parameter tuning.

Building a Conceptual Preference Model Based on Personal Purchase Records for Retail Service Improvement

Retailers need to understand customer preferences precisely in order to improve customer satisfaction. This research aims to correct retailers’ conceptual misunderstandings of preference based on actual purchase behavior. To this end, a novel conceptual preference model is introduced. This model can be defined by persons on the basis of their own knowledge and is updated utilizing actual data of purchase behavior. Typical patterns of misunderstanding are classified into three types: incorrect relationship, weak relationship, and unexpected relationship. A novel method has been developed for evaluating the model to extract these patterns of misunderstanding and validated by conducting an experiment comprised of a cafeteria analysis. The conceptual model includes 11 preference types defined on the basis of retailers’ knowledge and was evaluated utilizing ID-POS data. Results showed that the three patterns of misunderstanding could successfully be extracted. It was confirmed that this method is helpful for retailers to understand the appropriate preference model and that a model built by using the proposed method can help improve the business processes in a retail setting.

Forecasting Intermittent Demand with Generalized State-Space Model

We propose a method for forecasting intermittent demand with generalized state-space model using time series data. Specifically, we employ mixture of zero and Poisson distributions. To show the superiority of our method to the Croston, Log Croston and DECOMP models, we conducted a comparison analysis using actual data for a grocery store. The results of this analysis show the superiority of our method to the other models in highly intermittent demand cases.