POS Tagger With Hand-Crafted Features
This is a slight extension of a POS tagger, implemented with tensorflow. I adopted the parser to data with conllu format.
More …Graph-Based Dependency Parser
This parser extended a perceptron part-of-speech tagger into dependency parser. It works on any of universal dependency corpus in a conllu format.
More …POS Tagger With Elmo Embeddings
A part of speech tagger implemented in Python with TensorFlow using Universal Dependency corpus. Used pretrained Elmo Embeddings through TensorFlow Hub. The model includes Elmo Embeddings layer, Bidirectional LSTM layer, and a softmax dense layer. If an appropriate Elmo embedding is loaded, the tagger can be trained on any languages which have Universal Dependency corpus. Used Keras for incorporating some layers. Evaluated the model by F1 scores.
More …Skills Summary
Programming Languages:
- C
- C++
- JavaScript
- PHP
- Python
- Ruby
Work Experiences
Sierra Wireless | Software Test Engineer Coop | Jan 2018-Aug 2018
Implemented and executed automated tests for IoT devices’ user interface, including WiFi, cellular, LAN, and security. If some features failed a test, reported and filed the bug through JIRA. For testing port security and device connection, used Wireshark to monitor or investigate specific data flows.
Experimental Syntax Lab | Undergraduate Research Assistant | Oct 2016-Dec 2017
Responsible for experimenting a Japanese syntactic phenomenon “Null Object Construction.” Translated English sentences to corresponding Japanese sentences with the phenomenon and run the experiment implemented with PsychoPy2 with Japanese participants.