Odor Descriptor Understanding through Prompting

TL;DR

This paper introduces two novel methods for generating odor descriptor embeddings that better capture olfactory meanings, outperforming existing approaches on a zero-shot NLP benchmark for odors.

Contribution

The paper proposes two new techniques to produce odor word embeddings aligned with olfactory meanings, surpassing prior state-of-the-art and standard prompting methods.

Findings

Generated embeddings outperform previous methods on the benchmark.

The methods improve alignment of embeddings with olfactory semantics.

Outperforms fine-tuning and prompting approaches.

Abstract

Embeddings from contemporary natural language processing (NLP) models are commonly used as numerical representations for words or sentences. However, odor descriptor words, like "leather" or "fruity", vary significantly between their commonplace usage and their olfactory usage, as a result traditional methods for generating these embeddings do not suffice. In this paper, we present two methods to generate embeddings for odor words that are more closely aligned with their olfactory meanings when compared to off-the-shelf embeddings. These generated embeddings outperform the previous state-of-the-art and contemporary fine-tuning/prompting methods on a pre-existing zero-shot odor-specific NLP benchmark.