---
title: "Using Word Embeddings for Identifying Emotions Relating to the"
description: "The paper asks whether body-part words that express emotion in Akkadian generally, as catalogued by Ulrike Steinert through… 23 references."
authors: ["Ellie Bennett", "Aleksi Sahala"]
year: 2023
venue: "Proceedings of the Ancient Language Processing Workshop"
source: "https://aclanthology.org/2023.alp-1.22/"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/bennett-2023-word-embeddings-identifying-emotions-relating/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/bennett-2023-word-embeddings-identifying-emotions-relating/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2023](https://md.research.elamur.ai/free/en/papers/2023/page/17/) / Using Word Embeddings for Identifying Emotions Relating to the Body in a Neo-Assyrian Corpus

# Using Word Embeddings for Identifying Emotions Relating to the Body in a Neo-Assyrian Corpus

**Authors:** Ellie Bennett and [Aleksi Sahala](https://md.research.elamur.ai/free/en/authors/aleksi-sahala/) · **Year:** 2023 · **Venue:** [Proceedings of the Ancient Language Processing Workshop](https://md.research.elamur.ai/free/en/venues/alp/) ([Workshop on Ancient Language Processing 2023](https://md.research.elamur.ai/free/en/venues/alp/2023/)) · **Publisher:** [INCOMA Ltd.](https://md.research.elamur.ai/free/en/publishers/incoma-ltd/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2023.alp-1.22/) · [PDF](https://aclanthology.org/2023.alp-1.22.pdf) · [OpenAlex](https://openalex.org/W6894142367)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary

The paper asks whether body-part words that express emotion in Akkadian generally, as catalogued by Ulrike Steinert through close reading and dictionaries, behave the same way in the narrower Neo-Assyrian period. The authors built a lemmatised Neo-Assyrian corpus from Oracc metadata, choosing generous inclusion criteria. They trained count-based word vectors on it with PMI-embeddings, a tool that combines PMI weighting with SVD and is designed for small corpora. For each of 22 body words, they took the ten nearest neighbours and looked them up in an emotion lexicon with 18 categories. They then read the actual attestations to see whether the flagged neighbours really carried emotional meaning.

Two words stood out: libbu ("interior/heart") and kabattu ("liver"). Kabattu clustered mainly with happiness and joy, with anger as a secondary field, mostly in royal inscriptions. Libbu's emotional neighbours pointed to fear and anger, which conflicts with traditional readings. The study shows that distributional methods can both confirm and challenge philological conclusions.

## Key points

- **Corpus:** after light deduplication, the Oracc-derived corpus has 7,969 texts, 1,014,890 tokens and 19,436 unique forms. Letters make up 33%, royal inscriptions 17%, transactions 15% and scholarly texts 14%.
- **PMI-embeddings settings:** 300 dimensions, shifted PMI with shift 7, a symmetric dynamic window of 3 words weighted by inverse distance, and Context Similarity Weighting with k=3. The defaults were tuned on a gold standard of word-similarity rankings from five Assyriologists.
- **Cosine ranges:** the average spread of cosine scores across each word's top ten neighbours was 0.16 (median 0.15). Libbu had the widest spread, which the authors read as broad usage. Kalītu ("kidney") had the narrowest, which suggests a specialised use.
- **Libbu's neighbours** include terror, pain and anger terms, but most of its neighbours are not emotional at all.
- **Kabattu's attestations:** 120 of its 176 occurrences are in royal inscriptions. All of its emotion-capable neighbours were confirmed as emotional in context, and three of the four relate to joy.

## Interesting ideas
- For corpora of about a million tokens, count-based PMI+SVD vectors are preferred over neural Word2vec/fastText, because neural methods are known to be unstable at that scale.
- Context Similarity Weighting downweights co-occurrences that come from repeated or near-duplicate passages, so formulaic text does not dominate the statistics.
- The spread between a word's first and tenth neighbour similarity serves as a simple, interpretable indicator of how broadly versus narrowly a word is used.
- The pipeline has two stages: embedding neighbours only generate candidates, and actual corpus attestations then verify them, preventing distributional similarity from being taken as fact.
- Shifted PMI deliberately keeps some statistically weak co-occurrences, because discarding all of them hurts sparse data.

## For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

- **I1** (world): Cheap, inspectable PMI vectors could provide a transparent similarity signal when extending compiled lexical knowledge for rare verbs or nouns, where neural embeddings would be opaque.
- **I2** (learn): A similar penalty could be applied when aggregating silver data, so boilerplate sentences do not inflate counts used to induce repair rules or train parsers.
- **I3** (world): This measure could flag lemmas likely to be polysemous, which then need multiple WordNet/FrameNet senses rather than a single default.
- **I4** (learn): Any automatically proposed knowledge, such as a new verb-class or frame assignment, could be accepted only after checking concrete example sentences, which also yields explainable evidence.
- **I5** (learn): When thresholding agreement statistics on small silver sets, a softened cutoff may retain useful rare patterns instead of pruning them entirely.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [University of Helsinki](https://md.research.elamur.ai/free/en/universities/university-of-helsinki/) — University, Helsinki, Finland · 51 works · 5,318 citations

📚 Shelves (5): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [📚 Literature and Literary Theory](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/literature-and-literary-theory/) · 541 works · [🎭 Archeology](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/archeology/) · 64 works

🧭 Topics (1): [😊 Sentiment & emotion](https://md.research.elamur.ai/free/en/topics/sentiment-emotion/) · 1,035 works

## Cite

Cite as: Ellie Bennett and Aleksi Sahala. 2023. Using Word Embeddings for Identifying Emotions Relating to the Body in a Neo-Assyrian Corpus. Proceedings of the Ancient Language Processing Workshop.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{bennett2023using,
  title = {{Using Word Embeddings for Identifying Emotions Relating to the Body in a Neo-Assyrian Corpus}},
  author = {Ellie Bennett and Aleksi Sahala},
  booktitle = {Proceedings of the Ancient Language Processing Workshop},
  year = {2023},
  doi = {10.5281/zenodo.8263196},
  url = {https://doi.org/10.5281/zenodo.8263196}
}
```

- [← previous UPOS-DEPREL Mismatches: Detecting Annotation Errors and Improving UD Guidelines based on Linguistic Knowledge](https://md.research.elamur.ai/free/en/papers/yih-2023-upos-deprel-mismatches-detecting-annotation/)
- [next → Utilizing Suggestopedia Method in Teaching English Grammar](https://md.research.elamur.ai/free/en/papers/sa-diyah-2023-utilizing-suggestopedia-method-teaching-english/)
- [↑ back to papers of 2023](https://md.research.elamur.ai/free/en/papers/2023/page/17/)

## References

23 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Johannes Bach. 2022. Emotions and Assyrian Kingship. In Karen Sonik and Ulrike Steinert, editors, The Routledge Handbook of Emotions in the Ancient Near East, 1 edition, pages 329–358. Routledge, London. [↗](https://doi.org/10.4324/9780367822873-17)
2. Laura Battini. 2022. Sennacherib’s Sieges and Deportations Reliefs: How to Increase Emotions. Avar: An Interdisciplinary Journal of Life and Society in the Ancient Near East, 1(2):313–359. [↗](https://doi.org/10.33182/aijls.v1i2.2249)
3. Ellie Bennett. 2023. Age and Masculinities during the Neo-Assyrian Period. Journal of Cuneiform Studies, 75:123–154. Publisher: University of Chicago press. [↗](https://doi.org/10.1086/725222)
4. Piotr Bojanowski, Edouard Grave, Armand Joulin, and Tomas Mikolov. 2017. [Enriching word vectors with subword information](https://md.research.elamur.ai/free/en/papers/bojanowski-2017-enriching-word-vectors-subword-information/). [↗](https://doi.org/10.1162/tacl_a_00051)
5. Dominik Bonatz. 2022. Fear and Terror in Assyrian Palace Reliefs. In Karen Sonik and Ulrike Steinert, editors, The Routledge Handbook of Emotions in the Ancient Near East, 1 edition, pages 525–543. Routledge, London. [↗](https://doi.org/10.4324/9780367822873-31)
6. Kenneth Ward Church and Patrick Hanks. 1990. [Word association norms, mutual information, and lexicography](https://md.research.elamur.ai/free/en/papers/church-1990-word-association-norms-mutual-information/). Computational Linguistics, 16(1):22–29. [↗](https://doi.org/10.3115/981623.981633)
7. Aleksi Sahala and Krister Lindén. 2020. Improving word association measures in repetitive corpora with context similarity weighting. In Proceedings of the 12th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management, IC3K 2020, Volume 1: KDIR, Budapest, Hungary, November 2-4, 2020. SCITEPRESS Science And Technology Publications. [find](https://scholar.google.com/scholar?q=Aleksi+Sahala+and+Krister+Lind%C3%A9n%2E+2020%2E+Improving+word+association+measures+in+repetitive+corpora+with+context+similarity+weighting%2E+In+Proceedings+of+the+12th+International+Joint+Conference+on+Knowle)
8. Aleksi Sahala and Saana Svärd. 2021. Language technology approach to “seeing” in Akkadian. In The Routledge Handbook of the Senses in the Ancient Near East, 1 edition, pages 561–575. Routledge, London. [↗](https://doi.org/10.4324/9780429280207-33)
9. Magnus Sahlgren. 2006. The word-space model: using distributional analysis to represent syntagmatic and paradigmatic relations between words in high-dimensional vector spaces. [find](https://scholar.google.com/scholar?q=Magnus+Sahlgren%2E+2006%2E+The+word%2Dspace+model%3A+using+distributional+analysis+to+represent+syntagmatic+and+paradigmatic+relations+between+words+in+high%2Ddimensional+vector+spaces%2E)
10. Hanspeter Schaudig. 2022. Anger and Hatred in Neo-Assyrian and Neo-Babylonian Royal Inscriptions. In Karen Sonik and Ulrike Steinert, editors, The Routledge Handbook of Emotions in the Ancient Near East, 1 edition, pages 631–647. Routledge, London. [↗](https://doi.org/10.4324/9780367822873-40)
11. Shih-Wei Hsu and Jaume Llop Raduà, editors. 2021. The Expression of Emotions in Ancient Egypt and Mesopotamia. BRILL. [find](https://scholar.google.com/scholar?q=Shih%2DWei+Hsu+and+Jaume+Llop+Radu%C3%A0%2C+editors%2E+2021%2E+The+Expression+of+Emotions+in+Ancient+Egypt+and+Mesopotamia%2E+BRILL%2E)
12. Karen Sonik. 2022a. Emotions and Body Language. The Expression of Emotions in Visual Art. In Karen Sonik and Ulrike Steinert, editors, The Routledge Handbook of Emotions in the Ancient Near East, 1 edition, pages 269–325. Routledge, London. [↗](https://doi.org/10.4324/9780367822873-15)
13. Jakob Jungmaier, Nora Kassner, and Benjamin Roth. 2020. Dirichlet-smoothed word embeddings for low-resource settings. In Proceedings of the 12th Language Resources and Evaluation Conference, pages 3560–3565, Marseille, France. European Language Resources Association. [↗](https://doi.org/10.48550/arxiv.2006.12414)
14. Karen Sonik. 2022b. Emotions and Body Language. The Expression of Emotions in Visual Art. In Karen Sonik and Ulrike Steinert, editors, The Routledge Handbook of Emotions in the Ancient Near East, 1 edition, pages 269–325. Routledge, London. [↗](https://doi.org/10.4324/9780367822873-15)
15. Omer Levy, Yoav Goldberg, and Ido Dagan. 2015. [Improving distributional similarity with lessons learned from word embeddings](https://md.research.elamur.ai/free/en/papers/levy-2015-improving-distributional-similarity-lessons-learned/). Transactions of the Association for Computational Linguistics, 3:211–225. [↗](https://doi.org/10.1162/tacl_a_00134)
16. Mikko Luukko. 2021. Expressions of Joy and Happiness in Neo-Assyrian. In Shih-Wei Hsu and Jaume Llop Raduà, editors, The Expression of Emotions in Ancient Egypt and Mesopotamia, pages 255–282. BRILL. [↗](https://doi.org/10.1163/9789004430761_013)
17. Karen Sonik and Ulrike Steinert, editors. 2022. The Routledge Handbook of Emotions in the Ancient Near East, 1 edition. Routledge, London. [find](https://scholar.google.com/scholar?q=Karen+Sonik+and+Ulrike+Steinert%2C+editors%2E+2022%2E+The+Routledge+Handbook+of+Emotions+in+the+Ancient+Near+East%2C+1+edition%2E+Routledge%2C+London%2E)
18. Ulrike Steinert. 2021. Pounding Hearts and Burning Livers: The Sentimental Body in Mesopotamian Medicine and Literature. In Shih-Wei Hsu and Jaume Llop Raduà, editors, The Expression of Emotions in Ancient Egypt and Mesopotamia, pages 410–469. BRILL, Leiden. [↗](https://doi.org/10.1163/9789004430761_018)
19. Ulrike Steinert. 2022. Emotion and the Body: Embodiment, Conceptual Metaphor, and Linguistic Encoding of Emotions in Akkadian. In The Routledge Handbook of Emotions in the Ancient Near East, 1 edition, pages 51–87. Routledge, London. [find](https://scholar.google.com/scholar?q=Ulrike+Steinert%2E+2022%2E+Emotion+and+the+Body%3A+Embodiment%2C+Conceptual+Metaphor%2C+and+Linguistic+Encoding+of+Emotions+in+Akkadian%2E+In+The+Routledge+Handbook+of+Emotions+in+the+Ancient+Near+East%2C+1+edition)
20. Saana Svärd, Tero Alstola, Heidi Jauhiainen, Aleksi Sahala, and Krister Lindén. 2020. Fear in Akkadian Texts: New Digital Perspectives on Lexical Semantics. In S-W Hsu and J. Llop-Raduà, editors, The Expression of Emotions in Ancient Egypt and Mesopotamia, number 116 in Culture and History of the Ancient Near East, pages 470–502. Brill, Leiden. [↗](https://doi.org/10.1163/9789004430761_019)
21. Peter D Turney and Patrick Pantel. 2010. [From frequency to meaning: Vector space models of semantics](https://md.research.elamur.ai/free/en/papers/turney-2010-frequency-meaning-vector-space-models/). Journal of artificial intelligence research, 37:141–188. [↗](https://doi.org/10.1613/jair.2934)
22. Jonathan Valk. 2016. “They Enjoy Syrup and Ghee at Tables of Silver and Gold”: Infant Loss in Ancient Mesopotamia. Journal of the Economic and Social History of the Orient, 59(5):695–749. [find](https://scholar.google.com/scholar?q=Jonathan+Valk%2E+2016%2E+%E2%80%9CThey+Enjoy+Syrup+and+Ghee+at+Tables+of+Silver+and+Gold%E2%80%9D%3A+Infant+Loss+in+Ancient+Mesopotamia%2E+Journal+of+the+Economic+and+Social+History+of+the+Orient%2C+59%285%29%3A695%E2%80%93749%2E)
23. Janine Wende. 2022. Akkadian Emotion Terms. In Karen Sonik and Ulrike Steinert, editors, The Routledge Handbook of Emotions in the Ancient Near East, 1 edition, pages 150–170. Routledge, London. [↗](https://doi.org/10.4324/9780367822873-9)

*Updated 2 November 2026*