I am a Marie Skłodowska-Curie postdoctoral fellow at Ca’ Foscari University of Venice working on my [https://pric.unive.it/projects/pythiaplus/home PythiaPlus] project.

My research uses machine learning to study the epigraphic cultures of the Mediterranean world, and I also work on the history and inscriptions of ancient Sicily.
Since obtaining my DPhil in Ancient History (University of Oxford) in 2021, I have held fellowships at the British School at Rome, Harvard's Centre for Hellenic Studies, and Google Cloud. In 2024 I will be returning to the UK for a Leverhulme EC Fellowship at the University of Nottingham.

I co-led the Pythia (2019) and Ithaca (2022) projects (the latter featured on the cover of the scientific journal ''Nature''), and am publishing the new inscriptions from Himera, Sicily. I live in Venice, but I'm often out on the Mediterranean trail.

"Computers aren't the thing, they're the thing that gets us to the thing" (J. MacMillan).

Find me on:
* thea {dot} sommerschield {at} gmail {dot} com
* [https://oxford.academia.edu/TheaSommerschield Academia.edu]
* [https://github.com/sommerschield GitHub]
* [https://scholar.google.it/citations?user=dUifv5oAAAAJ&hl=en Google Scholar]
* [https://twitter.com/TSommerschield Twitter]
* a Mediterranean island with good internet connection.

2020-06-09T17:37:20Z

TheaSommerschield:

<div id="mw-content-text" lang="en-GB" dir="ltr" class="mw-content-ltr">
<div class="mw-parser-output"><div id="toc" class="toc">
<div class="toctitle" lang="en-GB" dir="ltr">

<h2>Contents</h2>
</div>

<ul>
<li class="toclevel-1 tocsection-1"><a href="#Available"><span class="tocnumber">1</span> <span class="toctext">Available</span></a></li>
<li class="toclevel-1 tocsection-2"><a href="#Authors"><span class="tocnumber">2</span> <span class="toctext">Authors</span></a></li>
<li class="toclevel-1 tocsection-3"><a href="#Description"><span class="tocnumber">3</span> <span class="toctext">Description</span></a></li>
<li class="toclevel-1 tocsection-4"><a href="#References"><span class="tocnumber">4</span> <span class="toctext">References</span></a></li>
<li class="toclevel-1 tocsection-5"><a href="#Pythia_offline"><span class="tocnumber">5</span> <span class="toctext">Pythia offline</span></a></li>
<li class="toclevel-1 tocsection-6"><a href="#Media"><span class="tocnumber">65</span> <span class="toctext">Media</span></a></li>
<li class="toclevel-1 tocsection-7"><a href="#Contact"><span class="tocnumber">65</span> <span class="toctext">Contact</span></a></li>

</div>

<h2>
<span class="mw-headline" id="Available">Available</span>
</h2>

<ul>
<li>[https://colab.research.google.com/drive/16RfCpZLm0M6bf3eGIA7VUPclFdW8P8pZ PYTHIA: the deep learning model for the automatic restoration of Greek inscriptions]</li>
</ul>

<h2>
<span class="mw-headline" id="Authors">Authors</span>
</h2>

<ul>
<li>Thea Sommerschield, University of Oxford</li>
<li>Yannis Assael, DeepMind</li>
<li>Jonathan R. W. Prag, University of Oxford</li>
</ul>

<h2>
<span class="mw-headline" id="Description">Description</span>
</h2>

<p>
Ancient History relies on disciplines such as Epigraphy, the study of ancient inscribed texts, for evidence of the recorded past. However, these texts, "inscriptions", are often damaged over the centuries, and illegible parts of the text must be restored by specialists, known as epigraphists. This work presents a novel assistive method for providing text restorations using deep neural networks.
</p>

<p>
<b>Pythia</b> is the first ancient text restoration model that recovers missing characters from a damaged text input. Its architecture is carefully designed to handle long-term context information, and deal efficiently with missing or corrupted character and word representations.
</p>

<p>
To train it, we wrote a non-trivial pipeline to convert [[PHI Greek Inscriptions|PHI]], the largest digital corpus of ancient Greek inscriptions, to machine actionable text, which we call <i>PHI-ML</i>. On <i>PHI-ML</i>, <b>Pythia</b>'s predictions achieve a 30.1% character error rate, compared to the 57.3% of human epigraphists. Moreover, in 73.5% of cases the ground-truth sequence was among the Top-20 hypotheses of <b>Pythia</b>, which effectively demonstrates the impact of such an assistive method on the field of digital epigraphy, and sets the state-of-the-art in ancient text restoration.
</p>

<p>
To aid further research in the field we created a freely accessible [https://colab.research.google.com/drive/16RfCpZLm0M6bf3eGIA7VUPclFdW8P8pZ online interactive Python notebook], where researchers can query one of our models to get text restorations and visualise the attention weights.
</p>

<p>
Furthermore, the PHI-ML dataset and processing pipeline have been published on [https://github.com/sommerschield/ancient-text-restoration GitHub], so that any researcher may regenerate PHI-ML and train new models offline. Follow the instructions below under the section ''"Pythia offline"''.
</p>

<h2>
<span class="mw-headline" id="References">References</span>
</h2>

<p>
Assael, Y., Sommerschield, T., Prag, J. “Restoring Ancient Text Using Deep Learning: A Case Study on Greek Epigraphy.” In
''Empirical Methods in Natural Language Processing (EMNLP - IJCNLP 2019)''. Association for Computational Linguistics. 6369-6376.
</p>

<p>
Available at:
<ul>
<li>[https://www.aclweb.org/anthology/D19-1668/ ACL anthology]</li>
<li>[https://arxiv.org/abs/1910.06262 arXiv preprint]</li>
</ul>
</p>

<p>
When using any of this project's source code, please cite:

<pre>
@inproceedings{assael2019restoring,
title={Restoring ancient text using deep learning: a case study on {Greek} epigraphy},
author={Assael, Yannis and Sommerschield, Thea and Prag, Jonathan},
booktitle={Empirical Methods in Natural Language Processing},
pages={6369--6376},
year={2019}
}</pre>
</p>

<h2>
<span class="mw-headline" id="Pythia-offline">Pythia offline</span>
</h2>

<p>
The following snippets provide references for regenerating PHI-ML and training new models offline.
</p>

<p>
'''Dependencies'''
<pre>
pip install -r requirements.txt && \
python -m nltk.downloader punkt
</pre>

'''PHI-ML dataset generation'''
<pre>
# Download PHI (this will take a while)
python -c 'import pythia.data.phi_download; pythia.data.phi_download.main()'

# Process and generate PHI-ML
python -c 'import pythia.data.phi_process; pythia.data.phi_process.main()'
</pre>

Preprocessed PHI-ML uploaded by @Holger.Danske800: [https://drive.google.com/drive/folders/1PQaWYmB02Sc2OC9yokajcsw65wIcLxGD link]

'''Training'''
<pre>
python -c 'import pythia.train; pythia.train.main()'
</pre>

'''Evaluation'''
<pre>
python -c 'import pythia.test; pythia.test.main()' --load_checkpoint="your_model_path/"
</pre>

'''Docker execution'''
<pre>
./build.sh
./run.sh <GPU_ID> python -c 'import pythia.train; pythia.train.main()'
</pre>
</p>

<h2>
<span class="mw-headline" id="Media">Media</span>
</h2>

<p>
Our research has been featured in blog posts by [https://deepmind.com/research/publications/Restoring-ancient-text-using-deep-learning-a-case-study-on-Greek-epigraphy DeepMind] and the [http://www.ox.ac.uk/news/arts-blog/restoring-ancient-greek-inscriptions-using-ai-deep-learning University of Oxford], and has been published in articles on [https://www.thetimes.co.uk/article/oracle-of-ai-solves-classic-conundrums-v6pdtsps0 The Times], [https://www.newscientist.com/article/2220438-deepmind-ai-beats-humans-at-deciphering-damaged-ancient-greek-tablets/ NewScientist], [https://www.repubblica.it/tecnologia/2019/10/24/news/non_solo_go_l_ai_di_google_batte_anche_gli_archeologi-239393979/ La Repubblica], [https://www.kathimerini.gr/1047963/article/epikairothta/episthmh/py8ia-o-gonos-ths-google-kai-ellhna-ereynhth-poy-diavazei-misokatestrammenes-arxaies-epigrafes Ekathimerini], [https://techcrunch.com/2019/10/18/ai-is-helping-scholars-restore-ancient-greek-texts-on-stone-tablets/ TechCrunch] and others.
</p>
<p>
It has also been presented at [https://www.emnlp-ijcnlp2019.org/ EMNLP - IJCNLP 2019] in Hong Kong, at the British School at Rome, the Oxford Epigraphy Workshop, the OIKOS National Research School in Classical Studies, during lectures at Venice Ca' Foscari and Rome La Sapienza, and most recently at the [https://www.youtube.com/watch?v=nKSfzHYmLtQ&t=26s Digital Classicist London Seminar].
</p>

<p>
Pythia is a partner project of [[epigraphy.info]].

<h2>
<span class="mw-headline" id="Contact">Contact</span>
</h2>

<ul>
<li>[http://thea.sommerschield@classics.ox.ac.uk thea.sommerschield@classics.ox.ac.uk]</li>
</ul>

</div>

Pythia

2020-06-09T17:19:58Z

TheaSommerschield: Created page with "<div id="mw-content-text" lang="en-GB" dir="ltr" class="mw-content-ltr"> <div class="mw-parser-output"><div id="toc" class="toc"> <div class="toctitle" lang="en-GB" dir="ltr"..."

<div id="mw-content-text" lang="en-GB" dir="ltr" class="mw-content-ltr">
<div class="mw-parser-output"><div id="toc" class="toc">

<div class="toctitle" lang="en-GB" dir="ltr">

<h2>Contents</h2>
</div>

<ul>
<li class="toclevel-1 tocsection-1"><a href="#Available"><span class="tocnumber">1</span> <span class="toctext">Available</span></a></li>
<li class="toclevel-1 tocsection-2"><a href="#Authors"><span class="tocnumber">2</span> <span class="toctext">Authors</span></a></li>
<li class="toclevel-1 tocsection-3"><a href="#Description"><span class="tocnumber">3</span> <span class="toctext">Description</span></a></li>
<li class="toclevel-1 tocsection-4"><a href="#References"><span class="tocnumber">4</span> <span class="toctext">References</span></a></li>
<li class="toclevel-1 tocsection-5"><a href="#Pythia_offline"><span class="tocnumber">5</span> <span class="toctext">Pythia offline</span></a></li>
<li class="toclevel-1 tocsection-6"><a href="#Media"><span class="tocnumber">65</span> <span class="toctext">Media</span></a></li>
<li class="toclevel-1 tocsection-7"><a href="#Contact"><span class="tocnumber">65</span> <span class="toctext">Contact</span></a></li>

</div>

<h2>
<span class="mw-headline" id="Available">Available</span>
</h2>

<ul>
<li>[https://colab.research.google.com/drive/16RfCpZLm0M6bf3eGIA7VUPclFdW8P8pZ PYTHIA: the deep learning model for the automatic restoration of Greek inscriptions]</li>
</ul>

<h2>
<span class="mw-headline" id="Authors">Authors</span>
</h2>

<ul>
<li>Thea Sommerschield, University of Oxford</li>
<li>Yannis Assael, DeepMind</li>
<li>Jonathan R. W. Prag, University of Oxford</li>
</ul>

<h2>
<span class="mw-headline" id="Description">Description</span>
</h2>

<p>
Ancient History relies on disciplines such as Epigraphy, the study of ancient inscribed texts, for evidence of the recorded past. However, these texts, "inscriptions", are often damaged over the centuries, and illegible parts of the text must be restored by specialists, known as epigraphists. This work presents a novel assistive method for providing text restorations using deep neural networks.
</p>

<p>
<b>Pythia</b> is the first ancient text restoration model that recovers missing characters from a damaged text input. Its architecture is carefully designed to handle long-term context information, and deal efficiently with missing or corrupted character and word representations.
</p>

<p>
To train it, we wrote a non-trivial pipeline to convert [[PHI Greek Inscriptions|PHI]], the largest digital corpus of ancient Greek inscriptions, to machine actionable text, which we call <i>PHI-ML</i>. On <i>PHI-ML</i>, <b>Pythia</b>'s predictions achieve a 30.1% character error rate, compared to the 57.3% of human epigraphists. Moreover, in 73.5% of cases the ground-truth sequence was among the Top-20 hypotheses of <b>Pythia</b>, which effectively demonstrates the impact of such an assistive method on the field of digital epigraphy, and sets the state-of-the-art in ancient text restoration.
</p>

<p>
To aid further research in the field we created a freely accessible [https://colab.research.google.com/drive/16RfCpZLm0M6bf3eGIA7VUPclFdW8P8pZ online interactive Python notebook], where researchers can query one of our models to get text restorations and visualise the attention weights.
</p>

<p>
Furthermore, the PHI-ML dataset and processing pipeline have been published on [https://github.com/sommerschield/ancient-text-restoration GitHub], so that any researcher may regenerate PHI-ML and train new models offline. Follow the instructions below under the section ''"Pythia offline"''.
</p>

<h2>
<span class="mw-headline" id="References">References</span>
</h2>

<p>
Assael, Y., Sommerschield, T., Prag, J. “Restoring Ancient Text Using Deep Learning: A Case Study on Greek Epigraphy.” In
''Empirical Methods in Natural Language Processing (EMNLP - IJCNLP 2019)''. Association for Computational Linguistics. 6369-6376.
</p>

<p>
Available at:
<ul>
<li>[https://www.aclweb.org/anthology/D19-1668/ ACL anthology]</li>
<li>[https://arxiv.org/abs/1910.06262 arXiv preprint]</li>
</ul>
</p>

<p>
When using any of this project's source code, please cite:

<pre>
@inproceedings{assael2019restoring,
title={Restoring ancient text using deep learning: a case study on {Greek} epigraphy},
author={Assael, Yannis and Sommerschield, Thea and Prag, Jonathan},
booktitle={Empirical Methods in Natural Language Processing},
pages={6369--6376},
year={2019}
}</pre>
</p>

<h2>
<span class="mw-headline" id="Pythia-offline">Pythia offline</span>
</h2>

<p>
The following snippets provide references for regenerating PHI-ML and training new models offline.
</p>

<p>
'''Dependencies'''
<pre>
pip install -r requirements.txt && \
python -m nltk.downloader punkt
</pre>

'''PHI-ML dataset generation'''
<pre>
# Download PHI (this will take a while)
python -c 'import pythia.data.phi_download; pythia.data.phi_download.main()'

# Process and generate PHI-ML
python -c 'import pythia.data.phi_process; pythia.data.phi_process.main()'
</pre>

Preprocessed PHI-ML uploaded by @Holger.Danske800: [https://drive.google.com/drive/folders/1PQaWYmB02Sc2OC9yokajcsw65wIcLxGD link]

'''Training'''
<pre>
python -c 'import pythia.train; pythia.train.main()'
</pre>

'''Evaluation'''
<pre>
python -c 'import pythia.test; pythia.test.main()' --load_checkpoint="your_model_path/"
</pre>

'''Docker execution'''
<pre>
./build.sh
./run.sh <GPU_ID> python -c 'import pythia.train; pythia.train.main()'
</pre>
</p>

<h2>
<span class="mw-headline" id="Media">Media</span>
</h2>

<p>
Our research has been featured in blog posts by [https://deepmind.com/research/publications/Restoring-ancient-text-using-deep-learning-a-case-study-on-Greek-epigraphy DeepMind] and the [http://www.ox.ac.uk/news/arts-blog/restoring-ancient-greek-inscriptions-using-ai-deep-learning University of Oxford], and has been published in articles on [https://www.thetimes.co.uk/article/oracle-of-ai-solves-classic-conundrums-v6pdtsps0 The Times], [https://www.newscientist.com/article/2220438-deepmind-ai-beats-humans-at-deciphering-damaged-ancient-greek-tablets/ NewScientist], [https://www.repubblica.it/tecnologia/2019/10/24/news/non_solo_go_l_ai_di_google_batte_anche_gli_archeologi-239393979/ La Repubblica], [https://www.kathimerini.gr/1047963/article/epikairothta/episthmh/py8ia-o-gonos-ths-google-kai-ellhna-ereynhth-poy-diavazei-misokatestrammenes-arxaies-epigrafes Ekathimerini], [https://techcrunch.com/2019/10/18/ai-is-helping-scholars-restore-ancient-greek-texts-on-stone-tablets/ TechCrunch] and others.
</p>
<p>
It has also been presented at [https://www.emnlp-ijcnlp2019.org/ EMNLP - IJCNLP 2019] in Hong Kong, at the British School at Rome, the Oxford Epigraphy Workshop, the OIKOS National Research School in Classical Studies, during lectures at Venice Ca' Foscari and Rome La Sapienza, and most recently at the [https://www.youtube.com/watch?v=nKSfzHYmLtQ&t=26s Digital Classicist London Seminar].
</p>

<p>
Pythia is a partner project of [[epigraphy.info]].

<h2>
<span class="mw-headline" id="Contact">Contact</span>
</h2>

<ul>
<li>[http://thea.sommerschield@classics.ox.ac.uk thea.sommerschield@classics.ox.ac.uk]</li>
</ul>

</div>