Multi-Domain Targeted Sentiment Analysis

Abstract

Targeted Sentiment Analysis (TSA) is a central task for generating insights from consumer reviews. Such content is extremely diverse, with sites like Amazon or Yelp containing reviews on products and businesses from many different domains. A real-world TSA system should gracefully handle that diversity. This can be achieved by a multi-domain model -one that is robust to the domain of the analyzed texts, and performs well on various domains. To address this scenario, we present a multi-domain TSA system based on augmenting a given training set with diverse weak labels from assorted domains. These are obtained through self-training on the YELP reviews corpus. Extensive experiments with our approach on three evaluation datasets across different domains demonstrate the effectiveness of our solution. We further analyze how restrictions imposed on the available labeled data affect the performance, and compare the proposed method to the costly alternative of manually gathering diverse TSA labeled data. Our results and analysis show that our approach is a promising step towards a practical domain-robust TSA system.

Comparative Analysis of Deep Natural Networks and Large Language Models for Aspect-Based Sentiment Analysis

“…In addition, it also beats the baseline accuracy for all the domains of DOTSA with 93.1% and 82.8% accuracy for hotel and restaurant domains respectively. Furthermore, PaLM yielded exceptional results for SemEval16/Restaurant Pontiki et al [49] and YASOToledo-Ronen et al [61] dataset also and outshined the performance of DeBERTa as well as the baseline accuracy. Particularly, PaLM yielded 93.5% and 98.1% accuracy for SemEval16/Restaurant and YASO datasets respectively.…”

“…YASO is a targeted sentiment analysis evaluation dataset for open-domain reviews [45], [61]. It is a crowd-sourced dataset containing more than 2,000 English user comments extracted from Yelp 7 , Amazon [24], Stanford Sentiment Treebank (SST) [56], and OPINOSIS [19].…”

“…The YASO dataset is not limited to any particular review domain, thus providing a broader perspective for open-domain targeted sentiment analysis (TSA). However, some major domains were identified by the authors in [61], as follows: 400 sentences were categorized as restaurants, 412 as electronics, 161 as hotels, 144 as automotive, 500 as movies, and the remaining 596) sentences were labeled as others. Whereas, the domain label is not publicly available with the dataset.…”

Interpreting Sentiment Composition with Latent Semantic Tree

“…That is, it decomposes the classification of a sentence into a hierarchical tree structure explicitly showing how the polarity of the sentence come from the composition of its subconstituents. Early works are mainly based on manual rules and semantic lexicon that is constructed either manually (Wilson et al, 2005;Kennedy and Inkpen, 2006) or automatically (Dong et al, 2015;Toledo-Ronen et al, 2018). Nowadays, represented via different forms of tree, sentiment composition is often learned explicitly or implicitly in the endto-end learning manner of neural network models.…”