Nature Language Processing
Overcoming Practical Issues of Deep Active Learning
Existing deep active learning algorithms achieve impressive sampling efficiency on natural language processing tasks. However, they exhibit several weaknesses in practice, including (a) inability to use uncertainty sampling with black-box models, (b) lack of robustness to noise in labeling, (c) lack of transparency. In response, we propose a transparent batch active sampling framework by estimating the error decay curves of multiple feature-defined subsets of the data.
We perform extensive experiments on four named entity recognition (NER) tasks and results show that our methods greatly alleviate these limitations without sacrificing too much sampling efficiency (Paper).
Distributional Inclusion Vector Embedding
We propose a novel word embedding method which preserves the distributional inclusion property in the sparse-bag-of-word (SBOW) feature. The embedding can be used to predict generality of words, detect the hypernym relation, and discover the topics from the raw text simultaneously. The extensive experiments show that the embedding effectively compresses the SBOW, and achieves new state-of-the-art performances on the unsupervised hypernym detection tasks (Paper, Code, Demo, Poster). We also show that DIVE could help us to do word sense induction more efficiently (Paper, Slides).
UMASS TAC 2016 system for relation extraction
TAC-KBP is one of the most challenging text-based information retrieval tasks. We integrate research which is done in UMASS IESL in the past year, including embedding linker, multilingual Universal Schema, and LSTM sentence embedding. We perform extensive error analysis and develop some novel techniques (such as using a search engine to reduce noise in training data) to tackle the problems (Paper).
Use Active Learning to Improve SGD
Inspired by active learning, we propose two alternatives to re-weight training samples based on lightweight estimates of sample uncertainty in stochastic gradient descent (SGD). Extensive experimental results on six datasets show that our methods reliably improve accuracy in various network architectures, including additional gains on top of other popular training techniques (Paper, Poster).
Student Modeling and Prerequisite Verification in Knowledge Tree
We extract answering logs of the exercises from Junyi Academy (http://www.junyiacademy.org/), an E-learning website similar to Khan Academy.
We use crowdsourcing and machine learning to discover relationships between exercises. Based on that, we will design a mechanism of adaptive test to improve learning experiences of Junyi academy (Paper, Presentation, Demo, Dataset).
Computer Vision and Multimedia
Active Sampling for estimating QoE model
We use Bayesian learning to model the non-linear relationships between quality of experience (QoE) and multiple factors.
Our experiment shows that active sampling can be used to reduce the number of samples collected from crowdsourcing for building such model (Paper).
Hierarchical Image Segmentation without Training
We proposed a general framework which applies classifiers with different complexity to discriminate segments in an image.
Our unsupervisedhierarchical segmentation results achieve similar or better performance in several standard benchmarks compared with the current state-of-the-art methods based on learning, and has been accepted to ACCV 2014 (Paper, Poster).
Decomposition of Multiple Foreground Co-segmentation
We proposed an efficient algorithm which decomposes the unsupervised Multiple Foreground Co-segmentation problem into three sub-problems: segmentation, matching and figure-ground classification.
Our method improves the accuracy of the state-of-the-art method by 13% in a standard benchmark, and has been accepted by CVIU (Paper).
Superpixel-Based Large Displacement Optical Flow
Video Object Extraction Using Saliency
We applied dictionary learning, visual and motion saliency to extract the foreground object from videos. This research has been accepted to TIP (Paper).
Reasoning 3-D Information from 2-D Images by Tracking
In a single surveillance video, we incorporate tracking data into the photo pop-up algorithm, and utilize RANSAC (random sample consensus) to reduce around 80% of tracking noise (PPT & Videos).