Exhibition Program

Science of Machine Learning

07

Search suitable for various viewpoints

- “Pitarie”: Picture book search with graph index based search -

Abstract

We propose a similarity search method for finding similar objects in a large-scale database. The search method is based on a graph index, where each vertex corresponds to a object and two vertices are connected by an edge when they satisfy a certain similarity condition. The graph index shows small-world behavior, that is, vertices can be reached from every other vertex by a small number of steps. Hence, searching the graph results in quick termination of the search process. Furthermore, since the graph index is constructed based on similarity between two objects, the search method is versatile and can be applied to wide variety of media such as text, images and audio. When applied to complex objects that are more than two media combined, such as picture books which consists of text and illustration, users can search from various viewpoints; users can find picture picture books that are not only similar in content but also similar in style of illustration.

References

  • [1] T. Hattori, T. Kobayashi, S. Fujita, Y. Okumura, K. Aoyama, “Pitarie: Picture Book Search with Interdisciplinary Approach,” NTT Technical Review, vol.14, no.7, pp. 1-8, 2016.
  • [2] K. Aoyama, K. Saito, H. Sawada, N. Ueda, “Fast approximate similarity search based on degree-reduced neighborhood graphs,” in Proc. The 17th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 1055-1063, 2011.

Poster

Photos

Contact

Takashi Hattori, Learning and Intelligent Systems Research Group, Innovative Communication Laboratory
Email: