Yinghao Li
Georgia Institute of Technology
22 Papers
28 Citations
Yinghao Li is an academic researcher from Georgia Institute of Technology. The author has contributed to research in topics: Computer science & Hidden Markov model. The author has an hindex of 4, co-authored 10 publications.
Chat about Author
Papers
Denoising Multi-Source Weak Supervision for Neural Text Classification.
TL;DR: A label denoiser is designed, which estimates the source reliability using a conditional soft attention mechanism and then reduces label noise by aggregating rule-annotated weak labels, which address the rule coverage issue.
•Posted Content
WRENCH: A Comprehensive Benchmark for Weak Supervision
TL;DR: Weak Supervision Benchmark (benchmark) as mentioned in this paper is a benchmark platform for evaluation of weak supervision methods, which consists of 22 real-world datasets for classification and sequence tagging; a range of real, synthetic, and procedurally generated weak supervision sources; and a modular, extensible framework for weak supervision evaluation.
37
STEAM: Self-Supervised Taxonomy Expansion with Mini-Paths
TL;DR: Zhang et al. as mentioned in this paper proposed a self-supervised taxonomy expansion model named STEAM, which leverages natural supervision in the existing taxonomy for expansion, and formulates a node attachment prediction task between anchor mini-paths and query terms.
32
BERTifying the Hidden Markov Model for Multi-Source Weakly Supervised Named Entity Recognition
Yinghao Li,Pranav Shetty,Lucas Liu,Chao Zhang,Le Song +4 more
- 01 Aug 2021
TL;DR: This paper proposed a conditional hidden Markov model (CHMM) which can effectively infer true labels from multi-source noisy labels in an unsupervised way, which can learn token-wise transition and emission probabilities from the BERT embeddings of the input tokens.
•Posted Content
Transformer-Based Neural Text Generation with Syntactic Guidance.
TL;DR: This work proposes to leverage the parallelism of Transformer to better incorporate parse trees and outperforms SOTA models both semantically and syntactically, improving the best baseline's BLEU score from 11.83 to 26.27.
17