Researchers Thea Sommerschield and Professor Jonathan Prag (Faculty of Classics, Oxford), alongside Yannis Assael (Google DeepMind and Oxford’s Department of Computer Science) explain how they have been working to restore damaged ancient Greek inscriptions using DeepMind AI.
Historians rely on different sources to reconstruct the thought, society and history of past civilisations. Many of these sources are text-based – whether written on scrolls or carved into stone, the preserved records of the past help shed light on ancient societies. Such sources include inscriptions, texts inscribed on a durable surface (such as stone, pottery or metal). Inscriptions are one of the main direct sources of new evidence from the ancient world, but the majority have suffered damage over the centuries, and parts of the text are illegible or lost (Figure 1). Restoring the missing or damaged text is one of the main undertakings of the discipline of Epigraphy. It is a complex and time consuming task, but ancient historians can estimate the likelihood of different possible solutions based on context clues in the inscription – such as grammatical and linguistic considerations, layout and shape, textual parallels, and historical context. Although complex, the restoration of these documents is necessary for a deeper understanding of civilisations past.
We have been using machine learning trained on these ancient inscribed texts to build a system that can furnish a more complete and systematically ranked list of possible restoration solutions, which we hope will augment historians’ understanding of a text.
Pythia, which takes its name from the woman who delivered the god Apollo's oracular responses at the Greek sanctuary of Delphi - is the first ancient text restoration model that recovers missing characters from a damaged text input using deep neural networks. Bringing together the disciplines of ancient history and deep learning, this work offers a fully automated aid to the text restoration task, providing ancient historians with multiple textual restorations, as well as the confidence level for each hypothesis.
Pythia takes a sequence of damaged text as input, and is trained to predict character sequences comprising hypothesised restorations of ancient Greek inscriptions (texts written in the Greek alphabet dating between the seventh century BCE and the fifth century CE). The architecture works at both the character- and word-level, thereby effectively handling long-term context information, and dealing efficiently with incomplete word representations (Figure 2). This makes it applicable to all disciplines dealing with ancient texts (philology, papyrology, codicology) and applies to any language (ancient or modern). To train Pythia, the largest digital corpus of ancient Greek inscriptions (PHI Greek Inscriptions) was converted to machine actionable text (called PHI-ML). On PHI-ML, PYTHIA’s predictions achieve a 30.1% character error rate, compared to the 57.3% of evaluated human epigraphists. Moreover, in 73.5% of cases the ground-truth sequence was among the Top-20 restoration hypotheses of Pythia, which effectively demonstrates the impact of this assistive method on the field of digital epigraphy, and sets the state-of-the-art in ancient text restoration.
The combination of machine learning and epigraphy has the potential to transform the study of ancient texts, and widen the scope of the historian’s work. For this reason, the Oxford and DeepMind teams collaborated to create an open-sourced online Python notebook, Pythia, and PHI-ML’s processing pipeline on GitHub. By so doing, we hope to aid future research and inspire further interdisciplinary work.
Read more about this work on the original DeepMind blog post, or the preprint article 'Restoring ancient text using deep learning: a case study on Greek epigraphy' on arXiv.
Analía Isabel Gerbaudo, Professor of Literary Theory and Didactics of Language and Literature at the National University of Litoral, Argentina, has been appointed as Global South Visiting Fellow at TORCH (The Oxford Research Centre in the Humanities).
The Global South visitor scheme, which sits in TORCH, is part of a wider aim to diversify the curriculum in Oxford’s humanities departments. The scheme is funded by the Andrew W. Mellon Foundation and is part of TORCH’s ‘Humanities & Identities’ series.
Dr Gerbaudo’s work tackles the difficult, charged and extremely topical question of the relationship between literary studies and political activism. The belief that practising criticism constitutes an important form of activism shapes the mission of the online literary journal that Gerbaudo founded in 2014, El Taco en la Brea, of which she remains the chief editor. It is also reflected in her most recent work as director of the independent publishing house Vera cartonera, linked to the ‘cartonero’ publishing movement which promotes literacy, diversity and social integration in Latin America by providing books at affordable prices. Dr Gerbaudo’s academic and professional work is produced outside the metropolitan context of Buenos Aires and is committed to geographical and social inclusion. She is currently researching the circulation of literary theory and Argentine literature during the last dictatorship and in the post-dictatorship period, especially with regard to questions of censorship, clandestine circulation and translation of texts, and underground reading practices. Gerbaudo is also an active translator and a committed mediator of Argentine literature and cultural politics in the global sphere.
Dr Gerbaudo said: ‘I would like to express my gratitude to those who have helped me develop and draw attention to a topic that I have been researching since 2006, the year I was admitted to the National Council of Scientific and Technical Research (CONICET). I study how Argentine literature and literary theory were taught in public universities in the years following the end of the military dictatorship. More specifically, I am concerned with the “fantasies of nano-intervention” that inspired the practices which my research has contributed to highlighting. Thanks again to all the people involved in the TORCH project for this opportunity.’
Dr Stefano-Maria Evangelista (English), Professor Laura Marcus (English) and Professor Benjamin Bollig (Medieval and Modern Languages) are sponsoring Dr Gerbaudo’s term at the University. While Dr Gerbaudo’s research has recently been published in the Journal of World Literature, her stimulating work remains very little known in Britain. This Global South Visiting Fellowship will provide her with a platform to share her work with English-speaking audiences both within Oxford and more widely, and to engage in ongoing debates in Oxford's Humanities faculties concerning disciplinary and interdisciplinary formations, literary translation, dissemination and the widening of the canon. Dr Gerbaudo’s expertise challenges the Eurocentric model of comparative literature by concentrating on Argentina as a key player in the global transmission of ideas and texts.
Dr Gerbaudo will be based at Oxford’s Faculty of Medieval and Modern Languages and Trinity College during Michaelmas Term 2019.
Throughout history, the experiences of women have largely been ignored, or understood only through representations in art, literature, music and other cultural mediums. This is true of the Belle Époque era (1870 – 1914), which is known for its idealised representations of women and femininity via traditional archetypes, such as the tranquillity of the homemaker, or the exoticism of the harem. A new conference run by TORCH (The Oxford Research Centre in the Humanities) aims to gain a deeper understanding of women during this era.
‘I wanted to put on an event as a way of fostering interdisciplinary discussion on women's lives,’ says conference organiser Dr Rhiannon Easterbrook, a Women in the Humanities Postdoctoral Writing Fellow at Oxford. ‘The two-day conference, co-organised with two DPhil students Sasha Rasmussen and Mara Gold, creates the space for dialogue on the representation and experiences of women at a transitional point in history from a global perspective: we have papers on women in India, Ireland, and the Caucasus, as well as France and Britain. Having been trained as a Classicist myself, it's been a fascinating journey learning how to engage with scholars from other disciplines. Additionally, how women create and preserve spaces for ourselves is still a contested subject today. Here we can look in-depth at how women's spaces relates to thorny topics, from sexuality to shopping, from work to idleness.’
The interest in understanding how women create and interact with space, from their homes to their offices, is particularly relevant to the Belle Époque era. During this time period, women were beginning to enter educational settings, workplaces, and new societal creations like department stores en masse. Their ability to navigate these new environments and roles for women, as well as how they experienced pleasure and desire in this different context, are of keen interest to academics. ‘The conference is about the experiences and representation of women during the Belle Époque, but with specific reference to how they inhabited or created spaces for themselves and how this allowed them to explore or suppress pleasure and desire,’ explains Dr Easterbrook.
As part of the conference, a concert will be held at the Holywell Music Room, featuring extracts from a new adaptation of Medea by Julia Pascal, a pioneering theatre director and playwright, and compositions by Cassandra White and Sarah Westwood.
‘In general, female composers and playwrights do not receive the same attention reserved for their male counterparts,’ says Dr Easterbrook. ‘We're delighted to share with members of the public this celebration of women's creativity, their meditations on pleasure and desire, and their drawing on the past to do so. This is a small corrective to a longstanding problem, which we hope will inspire our audience.’
‘Just as the women in our concert have forged connections between past and present and between their own creativity and that of their collaborators, we hope that both the public and our speakers will continue to make connections themselves, leading to fruitful collaboration, thoughtful scholarship, and a greater appreciation for creative pioneers.’
The ‘Women’s Spaces, Pleasure, and Desire in the Belle Époque’ conference will take place in Oxford on 3 – 4 June 2019. The concert is on 3 June 2019, 18:00 in the Holywell Music Room and is open to the public; register via Eventbrite.
The events are co-sponsored by Women in Humanities, St Hilda’s College and the Oxford Queer Studies Network.
Professor Katrin Kohl of Oxford's Faculty of Medieval and Modern Languages has written a letter to The Guardian calling on Ofqual to 'urgently adjust grade boundaries and implement proper quality control for Modern Foreign Languages (MFL) exams'. The letter has been signed by 150 university teachers, and The Guardian has also published a report on the issues raised. Here, Katrin Kohl gives further details about how the design and grading of exams are affecting MFL subjects and the pupils studying them.
Languages have long been considered ‘difficult’. The reasons are obvious – you can’t make progress without learning lots of vocabulary, you have to get your mind round illogical grammar rules and avoid getting discouraged by mistakes when applying them, and you project yourself publicly as an ignoramus every time you open your mouth to practise speaking. Moreover, words and rules are almost as quickly forgotten as they’re learned. Add to this the fact that English native speakers already know the most useful language in the world including the language of the internet and dominant pop culture, and it’s hardly surprising that foreign language learning in the UK is suffering.
There are many joys and rewards in learning languages, too – cognitive benefits, cultural enrichment, communicative empowerment, sense of adventure, creation of a new identity. Yet these require careful nurturing, patience and time. And time is in particularly short supply in crowded school timetables.
Powerful measures are needed if the difficulties are not to win the day. The most effective one is making the subject compulsory at school. In other European countries that’s normal. In England, that battle was lost in 2004 when the Labour government made languages optional at GCSE. Further nails were hammered into the languages coffin with the intensive promotion of STEM subjects as a career advantage, the abolition of the fourth AS subject from 2016, and the push towards fewer GCSEs with the reformed qualifications. Counter-measures by the government such as the EBacc and compulsory language teaching at primary level have not succeeded in reversing the trend.
There’s now widespread alarm at the rapid loss of language skills as schools reduce provision and universities close language departments. The All-Party Parliamentary Group on Modern Languages has demanded a Recovery Programme; the British Academy has issued a Call for Action together with the Royal Society, Academy of Medical Sciences and Royal Academy of Engineering; and the Arts and Humanities Research Council has invested some £16 million in research programmes designed to give languages a shot in the arm.
Meanwhile the spotlight is on the GCSE and A level exams in Modern Foreign Languages – are they fit for purpose? This is all the more critical in a context where other factors are impacting negatively on the subject. Yet schools report that it’s primarily the difficulty of the course and exams that is prompting learners to drop the subject. There are two interconnected issues here. One is ‘severe grading’. The other is the intrinsic difficulty of the exam papers, which in turn generates courses that are too demanding and makes for stressed teachers and learners. The exam regulator Ofqual is ultimately responsible for both issues since it oversees the work of the exam boards and maintains standards across subjects.
After some ten years of complaints from teachers, five years of support from the higher education subject community, and several consultations and research studies, Ofqual acknowledged last November that grading in MFL A levels is indeed, as teachers had claimed, ‘severe’ and that French, German and Spanish A levels are ‘of above average difficulty’. Yet Ofqual decided not to make an adjustment to the grades.
A consultation is now underway for a similar exercise with GCSEs in MFL. The decision expected in the autumn. So what about the impact of severe grading? Ofqual has been amassing statistical proof to show that there is no causal link with falling numbers. But can that possibly be the case? Which learner, parent or school will go for a subject that has statistically been proven even by the exam regulator to be ‘severely graded’ and thereby put the student’s university place at risk?
A key factor underlying excessive difficulty of the language exams for English learners is the presence of native and near-native speakers of the language in the exam cohort. This factor is unique to Modern Foreign Languages and it was partially addressed by Ofqual in 2017 with a small one-off adjustment to A level grading in French, German and Spanish. But what hasn’t yet been acknowledged is their effect on the exam papers.
This is significant, especially for smaller languages where the proportion of native speakers tends to be highest. Research commissioned by Ofqual showed that in the German A level sample, almost half the students gaining an A* were native-speakers, while at grade A, they made up almost a fourth. These are invisible to examiners, exam boards and Ofqual when it comes to scrutinising marks profiles. So even if the exam is far too difficult for non-native speakers, there will be enough marks gained at the top end to suggest the exam is working.
In fact an examiners’ report for the 2018 A level in German indicates that there may be insufficient awareness of difficulty as an issue. In the case of a reading comprehension question concerning a grammatically highly complex sentence with a word very unlikely to be familiar to an English learner, the examiner comments that the question ‘discriminated well. A few candidates answered this correctly and gained a mark’. The sample answer given in the report for this part of the exam is likely to be by a near-native speaker.
Learners, then, face a triple whammy – a rushed, stressful course that can’t possibly prepare them thoroughly for the exam at the end of it; a demoralising exam experience that makes them feel failures; and a grade that is below what they would get in another subject for equivalent performance.
So what’s to be done? There’s a window between now and Ofqual’s autumn decision for a change of direction. Ofqual needs to acknowledge the overwhelming evidence of anomalies in Modern Foreign Languages assessment – and act:
- Reopen the question of A level grading, and carry out the necessary adjustment to eliminate ‘severe grading’.
- Simplify the exam papers, and ensure that the exam boards start working with robust criteria for controlling the level of linguistic difficulty appropriately for non-native speakers.
- Gain better understanding of the impact of native and near-native speakers on exam papers, marking and grading, and make the necessary adjustments for all languages so non-native speakers are rewarded appropriately.
The subject community in schools and universities is keen to support this endeavour. If Ofqual does not address these matters now, language learning in the UK will face an inexorable further downward spiral caused by unrealistic expectations, exam difficulty, severe grading, irreversible loss of provision in schools and universities, and an intensifying teacher shortage.
You can read Ofqual’s response to the Guardian article and letter here.
Read Professor Kohl's letter to Ofqual, plus supporting documents on the Creative Multilingualism website.