# CERM: Context-aware Literature-based Discovery via Sentiment Analysis

arXiv: https://arxiv.org/abs/2402.01724 · HTML: https://mlsanity.com/paper/2402.01724/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 8 of its 33 references that are in the map.
- First posted on arXiv 27 Jan 2024.

## Builds on (most cited first, 8 of 8)

- [Parameter-Efficient Transfer Learning for NLP](https://mlsanity.com/paper/1902.00751/) (2019)
- [FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence](https://mlsanity.com/paper/2001.07685/) (2020)
- [BioBERT: a pre-trained biomedical language representation model for biomedical text mining](https://mlsanity.com/paper/1901.08746/) (2019)
- [Enriching Word Vectors with Subword Information](https://mlsanity.com/paper/1607.04606/) (2016)
- [Unsupervised Data Augmentation for Consistency Training](https://mlsanity.com/paper/1904.12848/) (2019)
- [Auxiliary Deep Generative Models](https://mlsanity.com/paper/1602.05473/) (2016)
- [SentiHood: Targeted Aspect Based Sentiment Analysis Dataset for Urban Neighbourhoods](https://mlsanity.com/paper/1610.03771/) (2016)
- [Efficient Fine-Tuning of BERT Models on the Edge](https://mlsanity.com/paper/2205.01541/) (2022)

## Built on by (most cited first, 1 of 1)

- [It is Simple Sometimes: A Study On Improving Aspect-Based Sentiment Analysis Performance](https://mlsanity.com/paper/2405.20703/) (2024)

## Abstract (arXiv)

Driven by the abundance of biomedical publications, we introduce a sentiment analysis task to understand food-health relationship. Prior attempts to incorporate health into recipe recommendation and analysis systems have primarily focused on ingredient nutritional components or utilized basic computational models trained on curated labeled data. Enhanced models that capture the inherent relationship between food ingredients and biomedical concepts can be more beneficial for food-related research, given the wealth of information in biomedical texts. Considering the costly data labeling process, these models should effectively utilize both labeled and unlabeled data. This paper introduces Entity Relationship Sentiment Analysis (ERSA), a new task that captures the sentiment of a text based on an entity pair. ERSA extends the widely studied Aspect Based Sentiment Analysis (ABSA) task. Specifically, our study concentrates on the ERSA task applied to biomedical texts, focusing on (entity-entity) pairs of biomedical and food concepts. ERSA poses a significant challenge compared to traditional sentiment analysis tasks, as sentence sentiment may not align with entity relationship sentiment. Additionally, we propose CERM, a semi-supervised architecture that combines different word embeddings to enhance the encoding of the ERSA task. Experimental results showcase the model's efficiency across diverse learning scenarios.