---
title: "WordDecipher — paper by Chen and Liu (2024)"
description: "Current rewriting assistants offer several alternative phrasings or opaque fluency scores, but they do not help NNES see how… 29 references."
authors: ["Yuexi Chen", "Zhicheng Liu"]
year: 2024
venue: "arXiv:2404.07005"
source: "https://arxiv.org/abs/2404.07005"
doi: "10.48550/arxiv.2404.07005"
arxiv: "2404.07005"
licence: "CC BY 4.0 (arXiv)"
licence_url: "http://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/chen-2024-worddecipher-enhancing-digital-workspace-communication/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/chen-2024-worddecipher-enhancing-digital-workspace-communication/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2024](https://md.research.elamur.ai/free/en/papers/2024/page/11/) / WordDecipher: Enhancing Digital Workspace Communication with Explainable AI for Non-native English Speakers

# WordDecipher: Enhancing Digital Workspace Communication with Explainable AI for Non-native English Speakers

**Authors:** Yuexi Chen and [Zhicheng Liu](https://md.research.elamur.ai/free/en/authors/zhicheng-liu/) · **Year:** 2024 · **Venue:** arXiv:2404.07005 · **Licence:** [CC BY 4.0 (arXiv)](http://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/2404.07005) · [PDF](https://arxiv.org/pdf/2404.07005) · [DOI 10.48550/arxiv.2404.07005](https://doi.org/10.48550/arxiv.2404.07005) · [arXiv 2404.07005](https://arxiv.org/abs/2404.07005) · [OpenAlex](https://openalex.org/W4403413085)

**Citations:** 4 citations (OpenAlex, 2026-10-14)

## Summary
Non-native English speakers (NNES) often write workplace emails and chat messages that carry unintended social signals. They may sound too formal, too blunt or too distant, sometimes because they carry phrasing over from their first language. Current rewriting assistants offer several alternative phrasings or opaque fluency scores, but they do not help NNES see how those options differ. This short CHI'24 In2Writing workshop paper proposes WordDecipher, an explainable writing assistant with three components. An LLM first identifies the dominant social-intention dimensions in a draft, and pre-trained style embeddings then estimate how strongly each one is expressed. The user can correct these intensities either by editing the scores or by supplying a version written in their native language. A rewriter then asks the LLM for candidates that match the target intentions. Finally, a nuance explainer places the candidates in content and style embedding spaces so the user can see how they differ. The paper reports no user study. It illustrates the tool with one usage scenario, arguing that intent-aware, explainable rewriting can make NNES more confident in high-stakes text communication.

## Key points
- Motivation figures: more than two million international students study in English-speaking countries, and 29.8 million foreign-born workers make up 18.1% of the US civilian labor force.
- The pipeline has three parts: intention detector (LLM prompting plus style embeddings), user-guided rewriter, and nuance explainer.
- The example intention dimensions are bipolar, such as formal–informal, indirect–direct, distant–close, respectful–disrespectful and shy–bold. Each is scored on a 7-point scale.
- The tool builds on style embeddings from Wegmann et al., which are RoBERTa models fine-tuned with content controlled, and on LISA embeddings from Patel et al. LISA vectors are 768-dimensional, interpretable, and trained on LLM-annotated synthetic data.
- Rewrites can target paragraph, sentence or word level. Evaluation is limited to a scenario: a Chinese student emails a professor, with a second case of a worker who drafts in Chinese and then translates.

## Interesting ideas
- Two rewrites can be compared in separate content and style embedding spaces, which shows whether they differ in meaning or only in tone.
- Social intent can be represented as a few named, bipolar, graded dimensions that users can edit, rather than one tone label.
- The intent a writer really means can be inferred from their native-language draft and then used as the target for the English text. This rests on the observation that translation tends to distort how strongly social intentions come across.
- An LLM decides which dimensions matter for a given text, while a separate embedding-based measure assigns the actual numbers. This separates qualitative judgment from quantitative scoring.
- The system explains how parallel candidates differ instead of ranking them with an unexplained fluency score. This addresses the distrust NNES report toward such scores.

## For Mova

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

- **I1** (back): The text → graph → text compressor could report meaning drift and register drift as two separate distances, so a round trip that keeps facts but changes politeness is flagged differently from one that loses facts.
- **I2** (frames): Speech-act frames in prag could carry explicit scalar attributes such as directness, formality and closeness, which downstream steps can inspect and adjust.
- **I3** (back): Translation through the meta-world could score the pragmatic dimensions of the source and the output and treat any mismatch as a translation error, even when the content is correct.
- **I4** (learn): LLM teacher tiers could propose which pragmatic labels apply, while a deterministic, inspectable scorer supplies the values, so silver data stays auditable.
- **I5** (answer): When several readings or answers are possible, Mova could show what separates them, such as which frame or entity changes, rather than giving only a confidence number.

*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 Maryland, College Park](https://md.research.elamur.ai/free/en/universities/university-of-maryland-college-park/) — University, College Park, United States · 132 works · 25,380 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

🧭 Topics (2): [English](https://md.research.elamur.ai/free/en/topics/english/) · 2,263 works · [🔬 Interpretability](https://md.research.elamur.ai/free/en/topics/interpretability/) · 130 works

## Cite

Cite as: Yuexi Chen and Zhicheng Liu. 2024. WordDecipher: Enhancing Digital Workspace Communication with Explainable AI for Non-native English Speakers. arXiv:2404.07005.

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

```bibtex
@misc{chen2024worddecipher,
  title = {{WordDecipher: Enhancing Digital Workspace Communication with Explainable AI for Non-native English Speakers}},
  author = {Yuexi Chen and Zhicheng Liu},
  year = {2024},
  doi = {10.48550/arxiv.2404.07005},
  eprint = {2404.07005},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2404.07005}
}
```

- [← previous Whisper Turns Stronger: Augmenting Wav2Vec 2.0 for Superior ASR in Low-Resource Languages](https://md.research.elamur.ai/free/en/papers/anidjar-2024-whisper-turns-stronger-augmenting-wav2vec/)
- [next → A Canonical Form for Flexible Multiword Expressions](https://md.research.elamur.ai/free/en/papers/odijk-2024-canonical-form-flexible-multiword-expressions/)
- [↑ back to papers of 2024](https://md.research.elamur.ai/free/en/papers/2024/page/11/)

## References

29 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. 2024. Automatically Paraphrase your Academic Text - Writefull’s Paraphraser. [https://x.writefull.com/paraphraser](https://x.writefull.com/paraphraser). (Accessed on 02/09/2024). [↗](https://x.writefull.com/paraphraser)
2. 2024. Foreign-born workers were a record high 18.1 percent of the U.S. civilian labor force in 2022 : The Economics Daily: U.S. Bureau of Labor Statistics. [https://www.bls.gov/opub/ted/2023/foreign-born-workers-were-a-record-high-18-1-percent-of-the-u-s-civilian-labor-force-in-2022.htm](https://www.bls.gov/opub/ted/2023/foreign-born-workers-were-a-record-high-18-1-percent-of-the-u-s-civilian-labor-force-in-2022.htm). (Accessed on 02/12/2024). [↗](https://www.bls.gov/opub/ted/2023/foreign-born-workers-were-a-record-high-18-1-percent-of-the-u-s-civilian-labor-force-in-2022.htm)
3. 2024. International student - Wikipedia. [https://en.wikipedia.org/wiki/](https://en.wikipedia.org/wiki/) International_student. (Accessed on 02/12/2024). [↗](https://en.wikipedia.org/wiki/)
4. 2024. Ludwig • Find your English sentence. [https://ludwig.guru/](https://ludwig.guru/). (Accessed on 02/09/2024). [↗](https://ludwig.guru/)
5. 2024. Prompt engineering - OpenAI API. [https://platform.openai.com/docs/](https://platform.openai.com/docs/) guides/prompt-engineering. (Accessed on 02/13/2024). [↗](https://platform.openai.com/docs/)
6. Inc. 2023 Baydin. 2024. Respondable: Write Better Emails with Boomerang for Gmail. [https://www.boomeranggmail.com/respondable/](https://www.boomeranggmail.com/respondable/). (Accessed on 02/11/2024). [↗](https://www.boomeranggmail.com/respondable/)
7. Open AI. 2024. ChatGPT. [https://chat.openai.com](https://chat.openai.com). (Accessed on 02/11/2024). [↗](https://chat.openai.com)
8. Daniel Buschek, Martin Zürn, and Malin Eiband. 2021. [The impact of multiple parallel phrase suggestions on email input and composition behaviour of native and non-native english writers](https://md.research.elamur.ai/free/en/papers/buschek-2021-impact-multiple-parallel-phrase-suggestions/). In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems. 1–13.
9. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2018. [Bert: Pre-training of deep bidirectional transformers for language understanding](https://md.research.elamur.ai/free/en/papers/devlin-2019-bert-pre-training-deep-bidirectional/). arXiv preprint arXiv:1810.04805 (2018). [↗](https://arxiv.org/abs/1810.04805)
10. Wanyu Du, Zae Myung Kim, Vipul Raheja, Dhruv Kumar, and Dongyeop Kang. 2022. Read, revise, repeat: A system demonstration for human-in-the-loop iterative text revision. arXiv preprint arXiv:2204.03685 (2022). [↗](https://arxiv.org/abs/2204.03685)
11. Steven M Goodman, Erin Buehler, Patrick Clary, Andy Coenen, Aaron Donsbach, Tiffanie N Horne, Michal Lahav, Robert MacDonald, Rain Breaw Michaels, Ajit Narayanan, et al. 2022. LaMPost: Design and Evaluation of an AI-assisted Email Writing Prototype for Adults with Dyslexia. In Proceedings of the 24th International ACM SIGACCESS Conference on Computers and Accessibility. 1–18. [find](https://scholar.google.com/scholar?q=Steven+M+Goodman%2C+Erin+Buehler%2C+Patrick+Clary%2C+Andy+Coenen%2C+Aaron+Donsbach%2C+Tiffanie+N+Horne%2C+Michal+Lahav%2C+Robert+MacDonald%2C+Rain+Breaw+Michaels%2C+Ajit+Narayanan%2C+et+al%2E+2022%2E+LaMPost%3A+Design+and+Eval)
12. Grammarly. 2024. Introducing Grammarly’s New Tone Rewrite Suggestions | Grammarly. [https://www.grammarly.com/blog/tone-rewrite-suggestions/](https://www.grammarly.com/blog/tone-rewrite-suggestions/) (Accessed on 02/07/2024). [↗](https://www.grammarly.com/blog/tone-rewrite-suggestions/)
13. Grammarly Inc. 2024. Grammarly. [https://www.grammarly.com/](https://www.grammarly.com/) Accessed on 01/26/2024. [↗](https://www.grammarly.com/)
14. Takumi Ito, Tatsuki Kuribayashi, Masatoshi Hidaka, Jun Suzuki, and Kentaro Inui. 2020. Langsmith: An interactive academic text revision system. arXiv preprint arXiv:2010.04332 (2020). [↗](https://arxiv.org/abs/2010.04332)
15. Takumi Ito, Naomi Yamashita, Tatsuki Kuribayashi, Masatoshi Hidaka, Jun Suzuki, Ge Gao, Jack Jamieson, and Kentaro Inui. 2023. Use of an AI-powered Rewriting Support Software in Context with Other Tools: A Study of Non-Native English Speakers. In Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology. 1–13. [find](https://scholar.google.com/scholar?q=Takumi+Ito%2C+Naomi+Yamashita%2C+Tatsuki+Kuribayashi%2C+Masatoshi+Hidaka%2C+Jun+Suzuki%2C+Ge+Gao%2C+Jack+Jamieson%2C+and+Kentaro+Inui%2E+2023%2E+Use+of+an+AI%2Dpowered+Rewriting+Support+Software+in+Context+with+Other+Too)
16. Yewon Kim, Mina Lee, Donghwi Kim, and Sung-Ju Lee. 2023. [Towards Explainable AI Writing Assistants for Non-native English Speakers](https://md.research.elamur.ai/free/en/papers/kim-2023-towards-explainable-ai-writing-assistants/). arXiv preprint arXiv:2304.02625 (2023). [↗](https://arxiv.org/abs/2304.02625)
17. Max Kreminski and Michael Mateas. 2021. Reflective Creators.. In ICCC. 309–318. [find](https://scholar.google.com/scholar?q=Max+Kreminski+and+Michael+Mateas%2E+2021%2E+Reflective+Creators%2E%2E+In+ICCC%2E+309%E2%80%93318%2E)
18. Hajin Lim, Dan Cosley, and Susan R Fussell. 2022. Understanding Cross-lingual Pragmatic Misunderstandings in Email Communication. Proceedings of the ACM on Human-Computer Interaction 6, CSCW1 (2022), 1–32. [↗](https://doi.org/10.1145/3512976)
19. Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. [Roberta: A robustly optimized bert pretraining approach](https://md.research.elamur.ai/free/en/papers/liu-2019-roberta-robustly-optimized-bert-pretraining/). arXiv preprint arXiv:1907.11692 (2019). [↗](https://arxiv.org/abs/1907.11692)
20. Slack Technologies LLC. 2024. Where work happens | Slack. [https://slack.com/](https://slack.com/). (Accessed on 02/12/2024). [↗](https://slack.com/)
21. Manas Mhasakar, Shikhar Sharma, Apurv Mehra, Utkarsh Venaik, Ujjwal Singhal, Dhruv Kumar, and Kashish Mittal. 2024. Comuniqa: Exploring Large Language Models for improving speaking skills. arXiv preprint arXiv:2401.15595 (2024). [↗](https://arxiv.org/abs/2401.15595)
22. Juri Opitz and Anette Frank. 2022. SBERT studies meaning representations: Decomposing sentence embeddings into explainable semantic features. In Proceedings of the 2nd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics and the 12th International Joint Conference on Natural Language Processing. 625–638. [find](https://scholar.google.com/scholar?q=Juri+Opitz+and+Anette+Frank%2E+2022%2E+SBERT+studies+meaning+representations%3A+Decomposing+sentence+embeddings+into+explainable+semantic+features%2E+In+Proceedings+of+the+2nd+Conference+of+the+Asia%2DPacific+C)
23. Ajay Patel, Delip Rao, and Chris Callison-Burch. 2023. Learning Interpretable Style Embeddings via Prompting LLMs. arXiv preprint arXiv:2305.12696v2 (2023). [↗](https://arxiv.org/abs/2305.12696)
24. QuillBot. 2024. Paraphrasing Tool - QuillBot AI. [https://quillbot.com/](https://quillbot.com/) (Accessed on 02/07/2024). [↗](https://quillbot.com/)
25. Nils Reimers and Iryna Gurevych. 2019. [Sentence-bert: Sentence embeddings using siamese bert-networks](https://md.research.elamur.ai/free/en/papers/reimers-2019-sentence-bert-sentence-embeddings-siamese/). arXiv preprint arXiv:1908.10084 (2019). [↗](https://arxiv.org/abs/1908.10084)
26. Anna Wegmann and Dong Nguyen. 2021. Does it capture STEL? A modular, similarity-based linguistic style evaluation framework. arXiv preprint arXiv:2109.04817 (2021). [↗](https://arxiv.org/abs/2109.04817)
27. Anna Wegmann, Marijn Schraagen, and Dong Nguyen. 2022. Same author or just same topic? towards content-independent style representations. arXiv preprint arXiv:2204.04907 (2022). [↗](https://arxiv.org/abs/2204.04907)
28. Wordtune. 2024. Wordtune: Free AI Writing Assistant | Write Better Today. [https://www.wordtune.com/](https://www.wordtune.com/) (Accessed on 02/07/2024). [↗](https://www.wordtune.com/)
29. Ann Yuan, Andy Coenen, Emily Reif, and Daphne Ippolito. 2022. Wordcraft: story writing with large language models. In 27th International Conference on Intelligent User Interfaces. 841–852. *Figure 2: ChatGPT-generated email for Yinuo’s inquiry on research opportunities in Prof. Miller’s lab* [find](https://scholar.google.com/scholar?q=Ann+Yuan%2C+Andy+Coenen%2C+Emily+Reif%2C+and+Daphne+Ippolito%2E+2022%2E+Wordcraft%3A+story+writing+with+large+language+models%2E+In+27th+International+Conference+on+Intelligent+User+Interfaces%2E+841%E2%80%93852%2E+%2AFigure+2%3A+)

*Updated 1 November 2026*