NTT Communication Science Laboratories Open House 2013

Abstract

Reading picture books to children attracts the attention of educators as a way of aiding children’s language and cognitive development. We built a similarity book search system that we can easily find suitable books for our children from the many published picture books. We developed a morphological analyzer that is robust for hiragana and katakana words that are difficult to analyze correctly, and this made it possible to analyze picture books with great precision. Furthermore, taking the words in each stage of an infant’s development into consideration, we devised a high-speed similarity search technique by using a graph search algorithm. In future, we expect this approach to be applied to a system for recommending picture books and an educational program to assist children’s language development, and a supporting system capable of creating new picture books for each child.

Poster

Please click the thumbnail image to open the full-size PDF file.

Reference

H,Taira, S.Fujita, T. Kobayashi, “Distribution of the vocabulary which appears in picture-book texts," in Proc. of the 12th Annual Meeting of the Japanese Society of Baby Science (JSBS), p. 92, 2012.
H,Taira, S.Fujita, T. Kobayashi, “Analysis of the high frequency vocabulary in picture-book texts," in H24 Annual Meeting of Information Processing Society of Japan (IPSJ) Kansai Branch, F103, 2012. (in Japanese)
S.Fujita, H,Taira, T. Kobayashi, “A morphological analysis for texts of picture books," in Proc. of the 19th Annual Meeting of the Association for Natural Language Processing (NLP), A6-4, 2013.
K. Aoyama, K. Saito, H. Sawada, N. Ueda, “Fast approximate similarity search based on degree-reduced neighborhood graphs,”in Proc. 17th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD), pp. 1055-1063, 2011.

Presentor