View on GitHub

三澤賢祐(みつざわ けんすけ)

Kensuke Mitsuzawa

日本語に切り替えるためには、下のボタンを押してください

At the date of February 20th, 2019


Abstract

Data Scientist/Data engineer

I'm specialized in NLP(Natural Language Processing) with my linguistics background.

For work experience, I have experience of data analytics from log data for the social game, machine learning techniques for NLP solutions, such as text summarization, document clustering/classification, information retrieval etc.

Not only data analysis, but also I have experience of system development for data analytics. For that, I'm able to consider hardware, DB specification, deployment, and operations with using Jenkins, Docker etc.

Plus, I have +2 years work experience in a semi-international work environment(50% Japanese members, 50% international members) In that environment, I'm able to communicate with other members with English.

I also had experience of "data-driven" PR planning of corporation. As one example, I published a paper with data-analysis.

For research perspective, I'm interested in information extraction/information retrieval from the text. This is because I have worked for 3.5 years on projects to extract business useful information from the text which is posted to an opinion-platform called "Fuman Kaitori"[1] so that I feel technical limitations of existing NLP methods.

To this day, I had research experience in Auto-Error-Correction of English text, document classification for folk narratives and Rule-based translation of Sign language from Japanese text.

I use Python mainly.

[1] This is served only in Japan. Please refer this paper to know "Fuman Kaitori"

Resumes

English(Europass) 履歴書

Academic background

Educations

Funding

  • NAIST Creative and International Competitiveness Project (CICP2012). 780,000 JPY, Project Leader (2012.06-2013.03)

Job history

Insight Tech Ltd. February 2015 - November 2018

Experience at Insight Tech Ltd.

Drecom Inc. April 2014 - January 2015

Experience at Drecom Inc.

Skills

Natural Languages

  • Japanese(Native)
  • English(Daily conversation) 2 years experience to work with international members in English
  • Persian(Elementary proficiency)
  • French(Elementary proficiency) 4th grade in "Test in Practical French Proficiency" (2016/12/16)
  • Italian(Elementary proficiency) 4th grade in "Test in Practical Italian Proficiency" (2018/03)

OS

  • MacOS: using as main environment for system development, machine learning model development or internet browsing
  • Linux: using as production server of analysis-sytem or using as training/evaluating server of machine learning models. I'm able to install OS from scratch and to make environemnt as administrator server.
  • Windows: minimum usage if it's needed.

Programming Languages

Python
Python 85%

Possible: Packaging, Web app(Flask, Django), natural language processing, machine learning(scikit-learn etc.), deep learning(chainer, tensorflow), Data analysis script, Data visualization, Jupyter notebook, Code optimization(Cython)

R
R 60%

Possible: data analysis script, reporting system(Rmd), Webapp(shiny), Data visualization, machine learning

Scala
Scala 30%

Possible: natural language processing

Javascript
Javascript 40%

Possible: Data visualization(D3.js, C3.js), Jquery, Vue.JS, Build this homepage ;)

Bash
Bash 60%

Possible: batch script

C/C++
C/C++ 15%

Possible: implementation of mathmatical operation for basic algorithm

Databases

MySQL, Postgresql, Redis, SQLite3, MongoDB, ElasticSearch

Cloud computing

AWS lambda, AWS Batch, API Gateway, S3, Kinesis stream

NLP/Machine learning

  • NLP: morphological analysis, syntactic parsing, predicate-argument analysis, text classification, text similarity, topic models, named entity recognition(+ wikification), feature selection, language modeling
  • Machine learning: KNN, Clustering, Regression, Classifier(Naive Bayse, Perceptron, SVM), Topic models(LDA, LSI), Sequential labeling(HMM, CRF), Neural network(FFNN, Auto-encoder, LSTM, CNN)
  • CMS tools

  • Wordpress: Admin of enterprise web page, Edit web page design, SEO
  • Else tools

    • Development tools: Pycharm, vim, git, Jenkins, travis-ci, Docker, Apache
    • Business tools: MS office(Excel, Powerpoint, Word), Google office tools(Spreadsheet, Slides, Documentation), Atlassian confluence
    • Communication tools: Slack

    Publications

    Google citation

    International

    Reviewed

    • Kazuhiro Akiyama, Kensuke Mitsuzawa, Kazuya Narita, Tadahiko Kumamoto, Akiyo Nadamoto. Clause-level Negative-opinion Analysis for Classifying Reviews on Multiple Domains. International Conference on Information Integration and Web-based Applications & Services(iiWAS2018), pp.xxx-xxx, 2018
    • Kensuke Mitsuzawa, Maito Tauchi, Mathieu Domoulin, Masanori Nakashima and Tomoya Mizumoto. FKC Corpus: a Japanese Corpus from New Opinion Survey Service. In proceedings of the Novel Incentives for Collecting Data and Annotation from People: types, implementation, tasking requirements, workflow and results, pp.11-18, Portorož, Slovenia, May 2016. [URL]

    Workshop

    • Ippei Yoshimoto, Tomoya Kose, Kensuke Mitsuzawa, Keisuke Sakaguchi, Tomoya Mizumoto, Yuta Hayashibe, Mamoru Komachi, Yuji Matsumoto. NAIST at 2013 CoNLL Shared Task Grammatical Error Correction. In Proceedings of the Seventeenth Conference on Computational Natural Language Learning: Shared Task, pp.26-33, August 2013. [URL]

    Japan domestic conference

    No reviewed

    • 秋山 和寛 (甲南大), 三澤 賢祐 (Insight Tech), 成田 和弥 (Insight Tech),熊本 忠彦 (千葉工大),灘本 明代 (甲南大, "CRFを用いた複数ドメインの消費者投稿文におけるネガティブ感情分類", 第11回Webとデータベースに関するフォーラム(WebDB Forum 2018), 信学技報, vol. 118, no. 213, DE2018-18, pp. 55-60, 2018. (学生奨励賞受賞)
    • 三澤 賢佑, 成田和弥 (Insight Tech/JST), 伊藤友博 (Insight Tech), 柴田知秀, 河原大輔, 黒橋禎夫 (京大/JST). 意見分析に適した意見タグ獲得改善への取り組み. 言語処理学会第24回年次大会 発表論文集, pp.572-575, March 2018.
    • 秋山 和寛 (甲南大), 三澤 賢祐 (Insight Tech), 成田 和弥 (Insight Tech),熊本 忠彦 (千葉工大),灘本 明代 (甲南大). CRFを用いたレビューにおける節単位毎の感情推定. DEIM2018, March 2018. [URL]
    • 三澤 賢佑, 成田和弥, 田内真惟人, 中島正成, 黒橋禎夫. 定量調査のための意見調査コーパス構築への取り組み. 言語処理学会第23回年次大会 発表論文集, pp.1014-1017, March 2017. [URL]
    • 成田和弥, 田内真惟人, 三澤賢祐, 中島正成. 社内データに基づくイノベータ人財のピックアップ. 言語処理学会第23回年次大会 発表論文集, pp.628-631, March 2017. [URL]
    • 三澤賢祐, 田内真惟人, Mathieu Domoulin, 中島正成, 水本智也. ネガティブ評判情報に特化したコーパスの構築と分析. 言語処理学会第22回年次大会 発表論文集, pp.501-504, March 2016. [URL]
    • 三澤賢祐, 田内真惟人, Mathieu Domoulin, 中島正成, 水本智也. 意見投稿プラットフォームにおける意見クラスタリングの試み. 言語処理学会第22回年次大会 発表論文集, pp.1037-1040, March 2016. [URL]
    • 池田可奈子(首都大), 三澤賢祐. Twitterを利用した日本語感情表現辞書の自動構築, NLP若手の会 第10回シンポジウム, ポスター発表, 2015
    • 三澤賢祐, 松本裕治. 異言語資源を利用したモチーフラベルの自動推定, 言語処理学会第20回年次大会発表論文集, pp.213 - 216, June 2014. [URL]
    • 三澤賢祐, 酒井啓道, 吉川友也, 水本智也, 松本裕治. 格構造に注目した日本語-日本語手話の並び替えと述語項構造に注目した語義曖昧性解消. 第27回人工知能学会全国大会論文集, pp.1-4, June 2013. [URL]

    Activities

    Software - Python package

    Talks(Jp only)

    Interests

    • Kendo: A good friend to enjoy my life

    • Hokushin Ittō-Ryū Hyōhō: Great and wonderful martial arts of Japanese sword. Way to be Samurai.

    • Travel: the world is huge indeed. These are countries I've visited so far.


      visited 12 states (5.33%)
      Create your own visited map of The World


    Contacts

    email

    kensuke.mit xx gmail.com

    xx = at_mark

    Others