Hiroki Nakayama
Researcher, Programmer at TIS Inc.
Tokyo, Japan
Summary
🤩
Rockstar🎓
Top SchoolHiroki Nakayama is a Tokyo-based researcher and programmer with 11 years of experience specializing in natural language processing and machine learning. At TIS Inc. he conducts applied NLP research while contributing significant open-source work—authoring and improving tools for embeddings, sequence labeling evaluation, and BiLSTM-CRF models for NER and POS tagging. His GitHub contributions show hands-on expertise from implementing Skip-gram variants and ELMo-backed NER pipelines to core evaluation metrics used in sequence labeling benchmarks. He also experiments across modalities, having built a webcam-based screen-hiding proof of concept that blends classical computer vision and small CNN models. Comfortable moving between research, engineering, and production code, he combines rigorous M.Eng. NLP training with pragmatic tool-building that benefits both researchers and practitioners.
11 years of coding experience
Master of Engineering (M.Eng.), Natural Language Processing, Master of Engineering (M.Eng.), Natural Language Processing at The University of Electro-Communications
Japanese, Chinese
Stackoverflow
685reputation
94kreached
6answers
3questions
Github Skills (26)
opencv10
gensim10
bilstm10
python10
sequence-labeling10
word-embeddings10
machine-learning10
mask-rcnn10
keras10
deep-learning10
tensorflow10
natural-language-processing10
computer-vision10
faster-rcnn10
nlp10
Programming languages (4)
Github contributions (5)
Hironsan/anago
Jun 2017 - Apr 2021
Bidirectional LSTM-CRF and ELMo for Named-Entity Recognition, Part-of-Speech Tagging and so on.
Role in this project:
Back-end Developer
Back-end DeveloperContributions:5 releases, 294 commits, 18 PRs in 3 years 9 months
Contributions summary:Hiroki primarily focused on enhancing the project's data loading and processing capabilities. Contributions included the implementation of a reader for a specific dataset (KWDLC-1.0), and the creation of data loading functions for the CoNLL2003 dataset, enabling the model to ingest and utilize various data sources. The user also designed a model using a Bidirectional LSTM, implementing the most basic model to load and process the model to perform named entity recognition, showing expertise in implementing components of an NLP project.
part-of-speechentitycrfspeech-taggingnatural-language-processing
Hironsan/awesome-embedding-models
Dec 2016 - Apr 2019
A curated list of awesome embedding models tutorials, projects and communities.
Role in this project:ML Engineer
ML EngineerContributions:32 commits, 5 PRs, 27 pushes in 2 years 4 months
Contributions summary:Hiroki primarily contributed to implementing and refining word embedding models. Their work includes adding example codes for Skip-gram and Skip-gram with negative sampling, along with utility functions. The user also integrated analogy evaluation and updated the training process to utilize the text8 corpus. Furthermore, the user added a Jupyter notebook to interactively explore the Skip-gram model.
nlpword-embeddingsmachine-learningcommunitiessentence-embeddings
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.