About

I am an incoming computer science PhD student at University of Colorado, Boulder. I was previously a search engineer at Amazon Japan and its subsidiary A9.com. My Japanese name is “藤沼祥成”.

My main interest is on Natural Language Processing and Machine Learning. On the method side, I am intersted in unsupervised/semi-supervised/distant-supervised learning, or making use of unlabeled data because it requires less manual labor. On the application side, I am interested in cross-lingual processing on Japanese or other Asian languages. Japanese (or Chinese, Korean, etc.) is an unsegmented language and this fact makes it difficult to tokenize (word segmentation and POS-tagging) when compared to English.

You can find me on LinkedIn, Twitter, and Github.

My CV is here.

Publication

• Yoshinari Fujinuma, Alvin Grissom II: “Substring Frequency Features for Segmentation of Japanese Katakana Words with Unlabeled Corpora”, The 8th International Joint Conference on Natural Language Processing, Nov. 2017 (acceptance rate: 31%) pdf poster
• Yoshinari Fujinuma, Hikaru Yokono, Pascual Martinez-Gomez, Akiko Aizawa: “Distant-supervised Language Model for Detecting Emotional Upsurge on Twitter”, The 29th Pacific Asia Conference on Language, Information and Computation, Nov. 2015 pdf
• Yoshinari Fujinuma: “Detecting Japanese-English transliteration pairs from search query and clickthrough logs” Amazon Machine Learning Conference 2015 (Internal)

Preprints

• Yoshinari Fujinuma, Jordan Boyd-Graber, Michael J. Paul “Diagnosing Language Inconsistency in Cross-Lingual Word Embeddings” 2018 pdf
• Mozhi Zhang, Yoshinari Fujinuma, Jordan L. Boyd-Graber: “Exploiting Cross-Lingual Subword Similarities in Low-Resource Document Classification” arXiv 2018 pdf

Other Talks

• Probabilistic Graphical Models Chapter 4.1 - 4.4 (slides are in Japanese), July 2016
• IT業界における英語とプログラミングの関係性(English and Programming in Academia and Industry for Japanese Junior Students) at Tokyo Metropolitan University, April 2016 leaflet slides
• Query Understanding behind Product Search at NII, April 2016
• Query Understanding behind Product Search at Tohoku University, Feburary 2016
• 言語モデル入門(Introduction to Language Models, April 2014 Slides

OSS Projects

• Kuromoji FST: A finite-state transducer for efficient key-value (text -> integers) lookup.

My Personal Notes

• Checklists Before You Submit a Paper
• PCFG-related notes
• EM-related notes
• Dual decomposition-related notes
• SAGE-related notes
• CCA-related notes
• Vector/Matrix derivative-related notes
• Morphological prior paper -related notes
• LSTM and vanishing gradient-related notes
• Automatic differentiation-related notes
• Intro. to Geometry of Cross-Lingual Embeddings

Fun stuff

• Clustered Visualizaiton of Word Embeddings using Emojis ** a tweak of this tutorial
• Clustered Visualizaiton of EN-ES using Emojis

Email address

fujinumay at gmail dot com