Skip to content

NLP analysis of sensory comments for 'Vera' hops, to identify common descriptors and unique flavor signatures

License

Notifications You must be signed in to change notification settings

maggie-reiter/hops-sensory-analysis

Repository files navigation

Hops sensory analysis for 'Vera' variety registration

Overview

Contains sensory data and analysis code for new hop variety registration. NLP analysis of tasting panel comments characterizes aroma and flavor in dried hops and finished beer.

Dataset

File: Vera Sensory Comments.csv

268 sensory comments from professional tasting events:

  • Dried Hops: 181 comments
  • Beer: 87 comments

Tasting events

  • Hopsource 2022-2024 (Washington and Idaho)
  • Craft Brewers Conference (CBC) 2022, 2024
  • Hop Research Council (HRC) Winter Meeting 2024 (multiple breweries)
  • Arbeiter Brewing Company Experimental Hop Rub 2023

Data structure

Event,Sample,Comments
Hopsource 2022 - Washington,Dried Hops,Complex and clean
Hopsource 2022 - Washington,Dried Hops,Vinous grape
...

Analysis workflow

1. Data cleaning (1_clean_data.ipynb)

Preprocess raw sensory comments for text analysis

Steps:

  • Acronym expansion (GO/O/G/ONG → "onion garlic", DMTS → "dimethyl trisulfide", etc.)
  • Text standardization and cleaning
  • Data reshaping to JSON format
  • Summary statistics on comment length distribution

2. Keywords and phrases analysis (2_keywords_phrases.ipynb)

Extract and analyze key sensory descriptors and linguistic patterns

Steps:

  • N-gram analysis: Most frequent words, word pairs, and three-word phrases
  • Part-of-speech tagging: Extraction of adjectives and adverbs
  • Categorical analysis: Prevalence of different flavor categories
  • Off-flavor analysis: Quantification of onion/garlic descriptors

Findings

Word frequency

Most frequent words, bigrams, and trigrams across comments. Numbers in parentheses are raw counts.

Dried Hops Beer
Top word fruit (29) fruit (23)
High‑frequency words tropical (27), citrus (25), peach (16), garlic (15), onion (15), mango (14), pineapple (13), fruity (11), bright (11) hop (15), citrus (11), sweet (10), pleasant (10), bitterness (10), tropical (9)
High-frequency bigrams onion garlic (11), stone fruit (7), passion fruit (4) stone fruit (5), hop flavor (5), sweet fruit (4)
High-frequency trigrams tropical stone fruit (3), slight onion garlic (3)
Top adjectives tropical (25), bright (11), sweet (9) sweet (10), tropical (9) clean (8)

Flavor prevalence

Prevalence of flavor expressions based on grouped keyword patterns (e.g., "tropical" captures tropical, pineapple, mango, passion, etc.).

Term Dried hops n Dried hops % Beer n Beer %
Tropical fruits 48 27% 21 24%
Citrus fruits 45 25% 18 22%
Fruit 39 22% 21 24%
Stone fruits 23 13% 9 10%
Onion garlic 18 10% 5 6%
Melons 8 4% 5 6%
Berries 7 4% 0 0%

Publication

Title
Journal of Plant Registrations, 2025

Analysis supports new hop variety registration by profiling sensory attributes and demonstrating distinct flavor profile and market potential.

About

NLP analysis of sensory comments for 'Vera' hops, to identify common descriptors and unique flavor signatures

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published