In this work, we propose a method that combines two popular research areas by injecting linguistic structures into pre-trained language models in the parameter-efficient fine-tuning (PEFT) setting. In our approach, parallel adapter modules encoding different linguistic structures are combined using a novel Mixture-of-Linguistic-Experts architecture, where Gumbel-Softmax gates are used to determine the importance of these modules at each layer of the model. To reduce the number of parameters, we first train the model for a fixed small number of steps before pruning the experts based on their important scores. Our experiment results with three different pre-trained models show that our approach can outperform state-of-the-art PEFT methods with a comparable number of parameters. In addition, we provide additional analysis to examine the experts selected by each model at each layer to provide insights for future studies.
The standard approach for neural topic modeling uses a variational autoencoder (VAE) framework that jointly minimizes the KL divergence between the estimated posterior and prior, in addition to the reconstruction loss. Since neural topic models are trained by recreating individual input documents, they do not explicitly capture the coherence between words on the corpus level. In this work, we propose a novel diversity-aware coherence loss that encourages the model to learn corpus-level coherence scores while maintaining high diversity between topics. Experimental results on multiple datasets show that our method significantly improves the performance of neural topic models without requiring any pretraining or additional parameters.
The multi-head self-attention mechanism of the transformer model has been thoroughly investigated recently. In one vein of study, researchers are interested in understanding why and how transformers work. In another vein, researchers propose new attention augmentation methods to make transformers more accurate, efficient and interpretable. In this paper, we combine these two lines of research in a human-in-the-loop pipeline to first discover important task-specific attention patterns. Then those patterns are injected, not only to smaller models, but also to the original model. The benefits of our pipeline and discovered patterns are demonstrated in two case studies with extractive summarization and topic segmentation. After discovering interpretable patterns in BERT-based models fine-tuned for the two downstream tasks, experiments indicate that when we inject the patterns into attention heads, the models show considerable improvements in accuracy and efficiency.
Discourse processing is a suite of Natural Language Processing (NLP) tasks to uncover linguistic structures from texts at several levels, which can support many downstream applications. This involves identifying the topic structure, the coherence structure, the coreference structure, and the conversation structure for conversational discourse. Taken together, these structures can inform text summarization, machine translation, essay scoring, sentiment analysis, information extraction, question answering, and thread recovery. The tutorial starts with an overview of basic concepts in discourse analysis – monologue vs. conversation, synchronous vs. asynchronous conversation, and key linguistic structures in discourse analysis. We also give an overview of linguistic structures and corresponding discourse analysis tasks that discourse researchers are generally interested in, as well as key applications on which these discourse structures have an impact.
We explore the use of natural language processing and machine learning for detecting evidence of Parkinson’s disease from transcribed speech of subjects who are describing everyday tasks. Experiments reveal the difficulty of treating this as a binary classification task, and a multi-class approach yields superior results. We also show that these models can be used to predict cognitive abilities across all subjects.
We investigate if writers with dementia can be automatically distinguished from those without by analyzing linguistic markers in written text, in the form of blog posts. We have built a corpus of several thousand blog posts, some by people with dementia and others by people with loved ones with dementia. We use this dataset to train and test several machine learning methods, and achieve prediction performance at a level far above the baseline.
We explore a novel computational approach for analyzing member participation in small group social sequences. Using a complex state representation combining information about dialogue act types, sentiment expression, and participant roles, we explore which sequence states are associated with high levels of member participation. Using a Markov Rewards framework, we associate particular states with immediate positive and negative rewards, and employ a Value Iteration algorithm to calculate the expected value of all states. In our findings, we focus on discourse states belonging to team leaders and project managers which are either very likely or very unlikely to lead to participation from the rest of the group members.