Impact case study database

Dr Hill, Prof Pagel and Dr Garret’s research supported the availability of Tibetan resources online and improved the learning of Modern and Classical Tibetan in Tibet and around the world. It made learning Tibetan more relevant to younger generations and helped ensuring that the cultural heritage of Tibet is preserved in the digital domain and handed on to future generations. To achieve this, the SOAS team, particularly Hill, engaged both with multinational technology companies, such as Google and Microsoft, and with Tibetan preservation and learning-focused organizations, which target the Tibetan diaspora, students of Tibetan, translators, religious communities and communities interested in learning and preserving Tibetan.

Multinational technological companies such as Microsoft have also faced challenges in digitizing Tibetan, thus limiting their ability to provide comprehensive services to Tibetan audiences and learners on a par with other languages. From 2018, Hill engaged with Microsoft and helped to solve the problem by providing a larger corpus of digitized Tibetan texts, which enabled Microsoft developers to train language models covering a larger vocabulary and offering more linguistically accurate predictions in the Tibetan version of their SwiftKey predictive keyboard for smartphones. The Natural Language Processing (NLP) Technical Lead at Microsoft indicated how fundamental Hill’s support was to develop the ability to process Tibetan: ‘in a language like Tibetan, where words are not delimited by spaces, we were restricted to offering predictions of syllables instead of words, which is an extremely poor experience for the user. Dr. Hill provided us with a Tibetan tokenizer . . . that has enabled us to move from syllable predictions to word predictions, putting our Tibetan offer on a par, in terms of quality, with other languages . . . Dr. Hill’s word breaker did not just save us time; we simply wouldn’t have been able to build a comparable one’ [5.1].

The Tibetan Swiftkey keyboard had 1,128 active monthly users as of July 2020 [5.8]. Microsoft reported that this resulted in a net increase in the quality of Microsoft prediction engine for Tibetan (an average 27% increase in typing speed, as measured on 6600 Tibetan users). Qualitative feedback was also extremely positive, both from users and from partners pre-installing the Tibetan keyboard on smartphones in China [5.1]. A civil servant from the Huangnan Tibetan Autonomous Prefecture, Qinghai Province, China stated: ‘I love using the [Tibetan keyboard] auto-completion when type on my phone . . . the slide typing is great, it’s the first slide typing keyboard for Tibetan in the world, ever’. A pastoralist from Mgo mang town, Hainan Tibetan Autonomous Prefecture, Qinghai Province, China described how ‘recently I found out about the SwiftKey, it is a better keyboard simply because it has auto-completion function. I have using it for a few months now and I am satisfied with it. Hope more people will use it so that we will have more words predictions as auto-completion suggestions’ [5.2].

For similar challenges, Hill’s support was also fundamental to Google’s efforts to improve its product offers – including Translate, Chrome, Maps, YouTube, the Google Assistant – in Tibetan language. A Google Speech Analytical Linguist explained how, from March 2016, Hill ‘provided consultation, advice and insight into the Tibetan language and script which has greatly benefited our work on Tibetan. His tokenization method for segmenting text into words has been a great source of inspiration, and his comments and help with our spelling-to-pronunciation mapping rules have significantly improved their quality and accuracy . . . [his research project] has significantly advanced our understanding of the complexities of the Tibetan language and script, and greatly impacted research and development of natural language processing solutions for the language . . . to better support our Tibetan-speaking users’ [5.3].

Esukhia is an NGO dedicated to the professional development of Tibetan Language Resources for the use of scholars, translators and Buddhist practitioners both in Asia and the West. Esukhia largely teaches students through online courses around the world. It also provides face-to-face courses in Dharamsala (India), and short programmes in Ladakh and Goa (India). More than 2000 students enrolled with Esukhia from 2012. Hill worked closely with Esukhia from May 2014, providing the foundation for Esukhia’s own research and development of Tibetan language NLP. Esukhia considered that ‘simply put, without [Hill’s] research and his support, we would not have [been] able to create these resources’ [5.4]. More specifically, Hill provided the technology and tools that underlay many important projects by Esukhia, including the planned definitive edition of the Tibetan Buddhist Canon – a key publication for academic studies, traditional monastic contexts as well as for large-scale translation projects, such as 84000: Translating the Words of the Buddha. Esukhia also graded texts and created learning resources for beginning readers – including Tibetan-speaking children and foreigners wanting to learn Tibetan as well as more professional and innovative Tibetan language textbooks. Esukhia indicated that the beginner-level textbook had already reached nearly 1,000 users across 63 countries by July 2020 [5.4], filling a gap in availability of learning resources for language learners in UK and USA. Esukhia also reported that feedback from independent users had been ‘overwhelmingly positive in helping new students gain a foothold in a language many have trouble learning’. Users expressed their gratitude for these new learning resources as compared to others they have tried: one user described how they now ‘enjoy learning Tibetan!’; another commented how easier it was compared to previous attempts, describing it as an ‘exciting way to learn which WORKS!!!’ [5.4].

The Buddhist Digital Resource Center (BDRC) is a Boston-based non-profit organisation dedicated to seeking out, preserving, organizing and disseminating Buddhist literature through the digitisation and preservation of Tibetan texts in a free online library used by scholars, translators and representatives of the tradition, especially Tibetan lamas and monks. The online library has over 5,000 active monthly users from over 15 countries – over half of the hits coming from Asia (mainly China, India, Nepal and Japan) where the Tibetan diaspora lives. From July 2013, Hill worked with BDRC to ensure that their internal search engine could search through Tibetan text, improving user experience and making a more sophisticated database. BDRC stated that its ‘technical team has embraced the breakthroughs introduced by Dr. Hill and have been exploring ways to build upon his momentous contribution to this key resource in Tibetan Studies’, including planning for the development of new improvements such as search ranking, named entity recognition, automated line alignment between multiple editions of a scripture, and lexicography tools, thus enriching the work of scholars and translators worldwide. BDRC was also ‘proud to have amplified [Hill’s] innovations to a global audience’ [5.5].

Nettle is the only open-source, open-access comprehensive learning resource developed for online delivery of instruction in Classical Tibetan language. It was developed by Professor Frances Garrett, University of Toronto based upon the tokenizer developed by Hill. Garrett said that without the research undertaken by SOAS, Nettle would have been different and not as sophisticated as it is [5.6]. Nettle has 2 main user groups, Toronto-based students and Classical Tibetan language learners around the world. From 2016 when Nettle was developed and introduced, allowing for more flexibility through online classes, approximately 25 to 40 University of Toronto-based students took Tibetan each year – compared to around 5 students in previous years. Student feedback was positive, with class evaluations ranging from average to high. Nettle can also be used by anyone online, although it does not track online, non-student users, Garret received many emails from around the world expressing gratitude for the online resources. Several came from India and Nepal, but also from Europe [5.6].

In May 2019, Hill was contacted by Christian Steinert, a software developer working for the Foundation for the Preservation of the Mahayana Tradition (FPMT). Steinert developed a Tibetan-English dictionary application for mobile phones; he explained that he very much benefitted from Hill’s research, with which he had enhanced the app substantially. After the app was revised thanks to Hill's research, the userbase of the dictionary almost tripled and in 2019 the app was used regularly by over 4,500 users all around the world. User feedback was also overwhelming positive: ‘I now use this as my primary dictionary, replacing both Monlam and The Tibetan Translation Tool . . . The inclusion of the Verb Lexicon is very helpful, as verb tense is not included in complete form anywhere else. Thank you for this dictionary’; ‘The results are much more detailed than the other dictionaries I’ve used. It even shows you tenses and their correct spellings’. Hill’s research also helped Steinert as an FPMT tutor providing non-academic online courses about Buddhist philosophy [5.7].