representation-learning

A Unified Speaker Adaptation Approach for ASR

A unified speaker adaptation approach consisting of feature adaptation and model adaptation for ASR.

UXLA: A Robust Unsupervised Data Augmentation Framework for Zero-Resouce Cross-Lingual NLP

We propose UXLA, a novel data augmentation framework for self-supervised learning in zero-resource transfer learning scenarios.

Evaluating Pronominal Anaphora in Machine Translation: An Evaluation Measure and a Test-suite

An extensive, targeted dataset that can be used as a test suite for pronoun translation, covering multiple source languages and different pronoun errors drawn from real system translations, for English

Discourse-informed Sen2Vec

CON-S2V: A Generic Framework for Incorporating

Recurrent Neural Models for Fine-grained Opinion Analysis

Publications Pengfei Liu, Shafiq Joty, Helen Meng. Fine-grained Opinion Mining with Recurrent Neural Networks and Word Embeddings. In Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP-2015), Lisbon, Portugal, 2015. @InProceedings{liu-joty-meng-emnlp-15, author = {Liu, Pengfei and Joty, Shafiq and Meng, Helen}, title = {Fine-grained Opinion Mining with Recurrent Neural Networks and Word Embeddings}, booktitle = {Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing}, year = {2015}, address = {Lisbon, Portugal}, series = {EMNLP’15}, pages = {1433–1443}, url = {http://aclweb.

SegBot: A Generic Neural Text Segmentation Model with Pointer Network

Online Demo Figure 1 shows the model architecture of SegBot. For EDU segmentation, the units in the input $ U0 \ to \ U8 $ are words in a sentence. Formally, given an input sequence $ U = (U_1, U_2, … , U_N) $ of length $N$, we get its distributed representations $ X = (x_1, x_2, … , x_N $ by looking up the corresponding embedding matrix, where $x_n \in R^k$ is the representation for the unit $U_n$ with $K$ being the dimensions.