By Dr. Stella Bullo
When I first heard developers talk about tokenisation and lemmatisation, I thought those concepts sounded strangely familiar. They did not seem like exotic new tech terms. I had heard and used them for years. My work with corpus linguistics had taught me how to slice language into parts, trace its roots, and see how meaning shifts with form. For me, this was not jargon. It was a familiar landscape, only now I was walking it with a different compass. Little did I know that my former tech-scared self would end up using these concepts in a fully technical context every day, in front of a console I once treated like a dangerous animal.
I spent two decades in academia, teaching and researching across the discipline of linguistics: from semantics, morphology and pragmatics to critical discourse analysis, corpus linguistics, conceptual metaphor theory and intercultural communication. Along the way, I wrote books on clichés and evaluative language, examined sentiment in marketing and health communication, analysed fear and resilience during the first months of the COVID-19 pandemic, and explored how language shapes public understanding. These were never just theoretical pursuits. In hindsight, they were the foundations for any technology that hopes to process language as people truly use it.
For me, language was never only a subject to study. It was the lens through which I explored human experience. I examined how we narrate pain, express desire, negotiate identity, and endure suffering. I came to understand how words can carry urgency, intimacy, and resistance, and how a single metaphor can open the door to empathy or close it entirely. In technology, those same skills mark the difference between a tool that merely processes words and one that genuinely understands the people behind them.
Then I began building a tool I wished I had during my own years of medical dismissal as an endometriosis patient. At first it was a metaphor tagger, designed as a research aid. Over time it became something more: Explain My Pain, a web application that helps patients describe their pain in ways clinicians can hear. It grew from both a research and a personal need into a tool that could bridge a gap that had long felt unbridgeable.
I did not set out to become a developer. I set out to solve a problem. To do it, I taught myself Python, front-end development, APIs, and, at some point, I realised that I was already halfway through learning natural language processing. I discovered that the processes I had honed as a linguist, such as identifying patterns, mapping meaning and interpreting context, were the same skills that could train a machine to listen more like a human.
NLP appealed to me not only as a technical challenge but also as a new canvas for everything I knew about language. Tokenisation became more than splitting text. It became a way to honour each fragment of someone’s story. Lemmas were not just dictionary forms but the underlying shapes of meaning, stripped of surface distractions. Sentiment analysis was not simply a number on a scale but a way of tracing hope, fear, or determination through the turns of a sentence.
The work demanded precision and empathy. It was code, but it was also culture, pragmatics, idiomatic turns, and discourse markers that revealed more than they seemed to. My experience told me that no algorithm could truly understand language without these layers. Machines could be taught to parse what was said. Linguistics could help them sense why it mattered.
Now I work at an unusual intersection. I am a linguist, a developer, a writer, a researcher, a patient, and a builder. I can design taxonomies of figurative language, parse sentiment in pandemic discourse, map the way metaphor carries fear or resolve across cultures, and turn those insights into working software.
I am interested in the impact linguistics can have in the current world where tech plays a pivotal role. For me, NLP has become more than a set of tools. It is the place where my knowledge of language and my skills in technology belong together. It feels like home. I want to shape tools that do not just process data but understand it in the context of human lives. I want to work where data and human experience meet, where the stakes are high, and where understanding someone better can be just a click away and change what happens next.
This reflection is linked to my app Explain My Pain, which grew out of my academic research project The Language of Endometriosis.
Explore the Live app.
You can find my academic publications on my website and on Google Scholar.