Vasudev Lal


2022

pdf bib
Opinion-based Relational Pivoting for Cross-domain Aspect Term Extraction
Ayal Klein | Oren Pereg | Daniel Korat | Vasudev Lal | Moshe Wasserblat | Ido Dagan
Proceedings of the 12th Workshop on Computational Approaches to Subjectivity, Sentiment & Social Media Analysis

Domain adaptation methods often exploit domain-transferable input features, a.k.a. pivots. The task of Aspect and Opinion Term Extraction presents a special challenge for domain transfer: while opinion terms largely transfer across domains, aspects change drastically from one domain to another (e.g. from restaurants to laptops). In this paper, we investigate and establish empirically a prior conjecture, which suggests that the linguistic relations connecting opinion terms to their aspects transfer well across domains and therefore can be leveraged for cross-domain aspect term extraction. We present several analyses supporting this conjecture, via experiments with four linguistic dependency formalisms to represent relation patterns. Subsequently, we present an aspect term extraction method that drives models to consider opinion–aspect relations via explicit multitask objectives. This method provides significant performance gains, even on top of a prior state-of-the-art linguistically-informed model, which are shown in analysis to stem from the relational pivoting signal.

2021

pdf bib
InterpreT: An Interactive Visualization Tool for Interpreting Transformers
Vasudev Lal | Arden Ma | Estelle Aflalo | Phillip Howard | Ana Simoes | Daniel Korat | Oren Pereg | Gadi Singer | Moshe Wasserblat
Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: System Demonstrations

With the increasingly widespread use of Transformer-based models for NLU/NLP tasks, there is growing interest in understanding the inner workings of these models, why they are so effective at a wide range of tasks, and how they can be further tuned and improved. To contribute towards this goal of enhanced explainability and comprehension, we present InterpreT, an interactive visualization tool for interpreting Transformer-based models. In addition to providing various mechanisms for investigating general model behaviours, novel contributions made in InterpreT include the ability to track and visualize token embeddings through each layer of a Transformer, highlight distances between certain token embeddings through illustrative plots, and identify task-related functions of attention heads by using new metrics. InterpreT is a task agnostic tool, and its functionalities are demonstrated through the analysis of model behaviours for two disparate tasks: Aspect Based Sentiment Analysis (ABSA) and the Winograd Schema Challenge (WSC).