Miki Ueno

2-Scene Comic Creating System Based on the Distribution of Picture State Transition

Understanding picture by computer has become one of the most important topics in computer science. However, there are few researches have been reported about human like picture understanding by computer. The main reason of difficulty is that lots of picture expressions contain more lyric aspect than natural language expressions. Comic is one of the best target of picture understanding researches because pictures in comics express simply and clearly story, therefore we can presume that pictures in comics have strong universality. Picture understanding is defined as understanding situations and estimating transition between current scene and next scene. In this paper, The novel method which generates pictures using prepared picture parts and image objects databases is proposed. We also show the 2-scene comics creating system using user inputs picture and propose the representation of picture state transition.

Novel Chatterbot System Utilizing Web Information

Recently, the use of various chatterbots has been proposed to simulate conversation with human users. Several chatterbots can talk with users very well without a high-level contextual understanding. However, it may be difficult for chatterbots to reply to specific and interesting sentences because chatterbots lack intelligence. To solve this problem, we propose a novel chatterbot that can directly use Web information. We carried out computational experiments by applying the proposed chatterbot to “2channel” (2ch) and “Twitter”.

Computational Interpretation of Comic Scenes

Understanding intellectual products such as comics and picture books is one of the important topics in the field of artificial intelligence. Hence, stepwise analysis of a comic story, i.e., features of a part of the image, information features, features relating to continuous scene etc., by human and by a combination of several classifiers was pursued. As the first step in this direction, several classifiers for comics are constructed in this study by utilizing a convolutional neural network, and the results of classification by a human annotator and by a computational method are compared.

Novel chatterbot system utilizing Web information for estimating current user interests

Recently, the use of various chatterbots has been proposed to simulate conversation with human users. Several chatterbots can talk with users very well without a high-level contextual understanding. However, it may be difficult for chatterbots to reply to specific and interesting sentences because chatterbots lack intelligence. To solve this problem, we have proposed a novel chatterbot that can directly use Web information. In this paper, we propose a novel method of estimating current user interests to make conversation natural.

Artificial Curation for Creating Learners Manual based on Data Semantics and User Personality

Curation services contribute to create manual web pages for learners and teachers especially for e-Learning. However, it is difficult to determine the quality of web pages for certain learning purpose. In this paper, we propose the novel method of generating learners’ manuals automatically based on semantic information of web pages and users personality. As an example, we implemented the application based on the proposed method and it was applied to the process of learning Git, which is one of popular version control systems. From our experiments, we discuss the relation between semantic of web pages and user personalities.

Classification of Two Comic Books based on Convolutional Neural Networks

Unphotographic images are the powerful representations described various situations. Thus, understanding intellectual products such as comics and picture books is one of the important topics in the field of artificial intelligence. Hence, stepwise analysis of a comic story, i.e., features of a part of the image, information features, features relating to continuous scene etc., was pursued. Especially, the length and each scene of four-scene comics are limited so as to ensure a clear interpretation of the contents. In this study, as the first step in this direction, the problem to classify two four-scene comics by the same artists were focused as the example. Several classifiers were constructed by utilizing a Convolutional Neural Network(CNN), and the results of classification by a human annotator and by a computational method were compared. From these experiments, we have clearly shown that CNN is efficient way to classify unphotographic gray scaled images and found that characteristic features of images to classify incorrectly.

The Convolutional Neural Network Model Based on an Evolutionary Approach For Interactive Picture Book

Creating interactive picture books based on human “Kansei” is one of the most interesting and difficult issues in the artificial intelligence field. We have proposed a novel interactive picture book based on Pictgent (Picture Information Shared Conversation Agent) and CASOOK (Creative Animating Sketchbook). Since our system accepts human sketches instead of natural languages, a high degree of sketch recognition accuracy is required. Recently, convolutional neural networks (CNNs) have been applied to various image- recognition tasks successfully. We have also adopted a CNN model for the sketch recognition of the proposed interactive picture book. However, it takes a considerable effort to tune the hyperparameters of a CNN. In this paper, we propose a novel parameter tuning method for CNNs using an evolutionary approach. The effectiveness of the proposed method is confirmed by a computer simulation that uses, as an example, a scribble-object recognition problem for the interactive picture book.

Estimation of structure of four-scene comics by convolutional neural networks

The computational interpretation of comics is one of the important topics being studied in the field of artificial intelligence and image recognition. There are a lot of challenging tasks to undertake in order to interpret comics, i.e., recognize objects in gray-scaled drawing image, extract emotional information of scenes, and define models of continuous scenes by considering the structure of comics. In this paper, we focused on four scene comics and their transition. Four-scene comics have a structure which originated in four-part of Chinese-poetry so creators clearly draw the semantic distance between each scene. It is very important for expressing the interesting and lyrical aspects of comics. To detect the transition of scenes, convolutional neural networks(CNNs) are constructed and computer experiments were carried out. The results suggest that CNN is able to detect the transition of scenes and that the features of each scene are quite different.

Classification of doll image dataset based on human experts and computational methods: A comparative analysis

In this paper, we analyze the important features for the difficult dataset to classify, comparing the process between human experts and computational method in order to built useful application on specific domain utilizing only computational techniques at practical level. To achieve this, “Doll” images dataset were constructed and the features were described in detail. After that, two types of experiments were carried out between human beginner participants and computational method. As a results, we obtained typical features of this dataset and discussed for applying the method to similar domains. In addition, the results suggest that it is efficient for classifying such a dataset to combine human experts knowledge and computational way.

A study on the efficiency of creating stories by the use of templates

The efficiency of creating stories by writers is very important for publishing a plentiful supply of products in the market. Although there are only a handful of matured writers, little research has been reported regarding a computational approach to support the creation of stories. In order to help amateur writers with creating well organized stories, we propose a newly developed system based on the four templates. In this paper, we constructed a system referring to the previous study and presented the outline of the novel system by utilizing templates. To confirm the efficiency of templates, user experiments were performed. From the experiments, we found that the backgrounds of characters are the most significant ingredient for the creation of attractive stories. Finally, we discuss the requirements for developing our system.