No project description provided
Project description
reading-impact-model
Reading Impact Model for analyzing reading impact in online book reviews.
The Impact of Fiction
How does your favorite book make you feel?
If you’re an avid reader, this question might be hard to answer. Books can make us feel extatically happy or deeply sad. Books can inspire us, motivate us, or make us feel like we are a part of something that matters. As long as stories have existed, they have made people laugh, and they have made people cry.
In this research project, we are trying to measure the emotional impact of books and stories, by analyzing the kind of emotional responses that readers express in online reviews.
The Impact of Fiction is a project of the Huygens Institute and the KNAW Humanities Cluster, coordinated by Peter Boot (Huygens) and Marijn Koolen (KNAW Humanities Cluster).
Computational Analysis of Reading Impact
There are millions upon millions of reviews on the internet today. Because of the staggering number of online reviews available, computational analysis is the ideal tool for analyzing them. But emotional impact is not easy to detect computationally. That’s why we’ve created a list of words relating to features from literature (aspect-terms) and a list of words relating to literature’s emotional impact (impact-terms) and a set of rules formulated to measure impact in relation to specific aspects.
Here’s an example:
lovely | + | character | = | narrative engagement |
[impact-term] | + | [aspect-term] | = | type of impact |
By looking at examples from a large set of online reviews, we formulated more than 1300 rules of this kind, measuring different types of impact. For an explanation of our categories of impact, go to the “Explanation of impact” page.
Here’s another example: In 2020, researchers at the Huygens Institute completed a similar research project on Dutch online reviews. They found, among other things, that Harry Potter and the Half-Blood Prince scored exceptionally high on use of the word “magisch” (magical). Any human who knows Harry Potter could guess that this doesn’t mean an exceptionally high number of reviewers had a magical reading experience. Rather, in reference to Harry Potter, the word “magical” has little to do with emotional response because the plot is actually about magic.
There are lots of other fun and interesting things you could do if you had a clear sense of the emotional impact of books. For example, this previous study into Dutch reviews suggested that appreciation of a novel’s style is linked to reflection on that novel. On the other hand, narrative engagement and mentions of style or reflection are negatively correlated, meaning that books that are have a very engaging, often suspenseful narrative are less frequently described as having an affecting style or inviting reflection. Think of your favorite thriller: is this true in that case?
Installation and Usage
You can install the package via pip:
pip install reading-impact-model
Basic usage of the English language impact model:
from reading_impact_model.matchers.matcher import ImpactMatcher
matcher = ImpactMatcher(lang='en')
matcher.analyse_text('The book has beautiful writing.', doc_id='some_doc_id')
Which gives the following output:
[{'doc_id': 'some_doc_id',
'sentence_index': None,
'sentence': 'The book has beautiful writing.',
'reflection': 0,
'style': 1,
'attention': 0,
'humor': 0,
'surprise': 0,
'narrative': 0,
'negative': 0,
'positive': 1,
'match_index': 3,
'impact_term_type': 'term',
'impact_term': 'beautiful',
'impact_type': 'Style',
'match_lemma': 'beautiful',
'match_word': 'beautiful',
'condition_match_index': 4,
'condition_term': 'writing',
'condition_match_lemma': 'writing',
'condition_type': 'style',
'condition_match_word': 'writing'}]
There are different matchers that can incorporate syntax parsers to add POS and lemma information to word tokens, for improved rule matching.
E.g. the SpacyMatcher
accepts a Spacy parser (and requires you to have
installed spacy and an appropriate language model.)
from reading_impact_model.matchers.spacy_matcher import SpacyMatcher
import spacy
nlp = spacy.load('en_core_web_trf')
matcher = SpacyMatcher(parser=nlp)
Which matches the lemma sentence instead of the word sentences, which is not in the aspect dictionary.
[{'doc_id': 'some_doc_id',
'sentence_index': 0,
'sentence': 'The book contains some beautifully written sentences.',
'style': 1,
'surprise': 0,
'negative': 0,
'narrative': 0,
'humor': 0,
'attention': 0,
'reflection': 0,
'positive': 1,
'match_index': 4,
'impact_type': 'Style',
'impact_term': 'beautifully',
'match_word': 'beautifully',
'impact_term_type': 'term',
'match_lemma': 'beautifully',
'condition_type': 'style',
'condition_match_word': 'sentences',
'condition_match_index': 6,
'condition_match_lemma': 'sentence',
'condition_term': 'sentence'}]
Citing
If you use this package, please cite the following publications:
- Boot, P., & Koolen, M. (2020). Captivating, splendid or instructive? Assessing the impact of reading in online book reviews. Scientific Study of Literature, 10(1), 35-63. (pre-pub PDF)
- Koolen, M., Neugarten, J., & Boot, P. (2022). ‘This book makes me happy and sad and I love it’. A Rule-based Model for Extracting Reading Impact from English Book Reviews. Journal of Computational Literary Studies, 1(1).
Contributors
The reading-impact-model
package was developed by Marijn Koolen. The rule
set and dictionaries were created by Peter Boot, Julia Neugarten and Marijn
Koolen.
The following people are or have been involved in the Impact & Fiction project:
- Peter Boot
- Marijn Koolen
- Joris van Zundert
- Julia Neugarten
- Olivia Fialho
- Willem van Hage
- Ole Mussmann
- Carsten Schnober
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for reading_impact_model-1.0.3.tar.gz
Algorithm | Hash digest | |
---|---|---|
SHA256 | 550aacabe0d1be4dc61bda3735113f6d306083cf559199143dc63f0355b4c675 |
|
MD5 | 1bffbaecf91da017653b3bd264cdbd1b |
|
BLAKE2b-256 | 2e72f4c47e664670d357ffdba0a0e25ec6d5d55c33e96e68880e43992a24efa4 |
Hashes for reading_impact_model-1.0.3-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | b36b7139525ca0ff9f7dd248f9528082c366f35fcfcb9be1b63488e100d1eac0 |
|
MD5 | da38f2094fb054750a7bad05fcd6cdd0 |
|
BLAKE2b-256 | b01e783d4d0d5f9699a1197fcc3fd373993cffac7e33d54b1d9428f833f7b6c2 |