Hello From Maochen.
Originated from  Shanghai, China.
A man seeking for unknown.
Reach me at contact maochen org
 

  Meta Inc.

  Microsoft

Querying Enterprise Knowledge Graph with Natural Language (2019)

Junyi Chai, Yonggang Deng, Maochen Guan, Yujie He, Bing Li and Rui Yan
in Proceedings of the CEUR Workshop 2019

Word Substitution in Short Answer Extraction: A WordNet-based Approach (2016)

Qingqing Cai, James Gung, Maochen Guan, Gerald Kurlandski and Adam Pease
in Proceedings of the Eighth Global WordNet Conference, Verginica Barbu Mititelu, Corina Forascu, Christiane Fellbaum, Piek Vossen (eds), pp 66-73

Patents

• Ontology Entity Type Detection from Tokenized Utterance - Serial # 16/370797 - 2019
• Training Utterance Generation for Knowledge Graph Based QA - Serial # 16/370759 - 2019
• Querying Knowledge Graph with Natural Language Input - Serial # 16/370721 - 2019

Amelia

Virtual agent for assisting customer answering domain related questions.
Constructing a hierarchical based complex Sentence Similarity computing system using distributed word representation.
Implemented an adapter layer for Stanford Parser to accept CoNLL-U format training data.
CRFs-based entity recognition.
Constructed dialog act recognizer by using LibLinear model with ~30% accuracy boosting on original dialogue scripts for 82 related topics.
Optimized PCFG parser model.

Feature Extraction Optimization for Multi-core Arch.

Optimize TweetEmotion.com.
Parallel feature extraction process.
Data level parallelism in training process.
Cell level and pair level parallelism in parsing process.
Non-blocking parallelism implemented by thread pool model and fork and join model.

Sentiment Classification System for Twitter

Supervised machine learning classification system.
Classify tweets extracted from twitter.com into 3 groups (Positive, Negative or Neutral).
Maximum entropy classifier for prediction.
Revised on Dec 2012, Backoff model added.
10-fold CV F1-Score: ~88%.