During the entire history of professional reporting and transcription, few things have revolutionised the field like artificial intelligence (AI) and its applications in automatic speech recognition (ASR). They have changed methods and workflows across the world and transformed the way in which professionals talk about their craft.
In his column “Mice, Machines and Men” in Tiro 1/2023, our Scientific Adviser Carlo Eugeni categorises diamesic translation, including speech-to-text professions, into four groups according to the nature of the human-machine relationship within them: 1) fully human (e.g. pen shorthand), 2) computer-aided (e.g. digital machine stenography), 3) human-aided (e.g. ASR-based parliamentary and court reporting), and 4) fully automated (e.g. automatic live subtitling). This thought-provoking categorisation contains many valuable aspects for professional discussion.
The aspect of the above categorisation that is of most interest to me is “human-aided”and how it is used. The term depicts traditional reporting professions that nowadays use ASR tools to create a first draft for professional editors as the human aiding the AI, not the other way round. This idea poses important questions for speech-to-text professionals: what is the nature of the report, where is the essential work done, and who or what is the one aiding or assisting in it? Is there still a need for the human element in professional reporting and, if there is, what is it?
For me, professional reporting and transcription are, simply put, recontextualising a speech event in a written document so that it fulfils the needs of its target audiences. This is always a selective act and careful situational deliberation based on, for example, the strategy, principles and practices of the reporter or the reporting community. AI-based ASR software does not deliberate: it is built on algorithms, statistics and probabilities. It has no intelligence or understanding, at least not in the way that we traditionally perceive them. It might produce great results, but it does not truly know what it is doing.
Because of this, ASR output is not a report, even though it might sometimes resemble one—even a good one. It simply recognises sound-letter equivalencies and applies post-processing rules. This output only becomes a real report when it is confirmed and revised by a professional human reporter. It is he or she who verifies that the text reliably depicts the speech and that every detail can be understood correctly by the reader when disconnected from the original speech event with its spoken resources, such as prosody, tone, gestures, expressions and material settings. The reporter also considers what should be edited, added or deleted to give a complete picture of the reported event.
A reporter might not always do much to the ASR output quantitatively, but even a little amount is what really counts, qualitatively. The report gains its validity only after human intervention. Only then is it transformed from a mechanical probability exercise into a meaningful textual entity. This is why professional reporting is still deeply human.
This issue of Tiro includes many different perspectives on the human element in reporting and transcription. D’Arcy McPherson explores how Hansard Services at the Legislative Assembly of British Columbia, Canada, have balanced innovation with continuity by developing Parrot, an in-house ASR tool. He shows that through a carefully organised process, AI can be used as a collaborative partner rather than a threat to professionals, reducing manual labour while preserving trust and human judgment. Henk-Jan Eras revisits the theme of his previous Tiro article, discussing the increased use of AI-produced deep-fakes in politics, the threat of AI in replacing traditional jobs and the impact of AI on parliamentary reporting. He also outlines an ongoing pilot project in the reporting office of the Dutch Parliament, where the AI tools Whisper and ChatGPT are combined innovatively to produce an automatic draft for parliamentary reporters.
Introducing ASR in professional reporting is not merely a technical phenomenon. Ana Rita Pereira and Paulo Granja share the preliminary results of their study, showing that the use of ASR has had a considerable impact on how the reporters edit the reported speeches in the Portuguese parliament. Ana Luisa Reis, also from the Portuguese parliament, gives us an enlightening tour of the humorous slip-ups that the ASR makes because it does not understand the content of the speeches. In my article, I discuss the impact of ASR on parliamentary editing practices in the Finnish Parliament and examine the different changes that the AI suggests to the report.
The remarkable influence of AI-tools in reporting and transcription has not been restricted to parliamentary reporting. Yanou Van Gauwbergen analyses results from a study that explores Flemish university students’ perceptions of Flemish live subtitling of English lectures created by both manual respeaking and unedited ASR, illustrating the strengths and challenges of both techniques. Ji Tang presents how AI is rapidly changing the Chinese stenography industry. He claims, however, that stenography professions are not vanishing, but transforming, while still relying on human professionals for accuracy, confidentiality and nuanced interpretation.
Tiro is also a forum for dialogue and shared discoveries. Elena Davitti, Daniela Eichmeyer-Hell, Carlo Eugeni, Judith Platter, and Pilar Orero give us a conference report from the 16th International Conference on Live Subtitling and Speech-to-Text Interpreting, summarising the key findings of the various panels and focusing on the role of humans in professions that have been deeply affected by the arrival of AI. Johanna Mechler, Lilja Björk Stefansdottir, and Anton Karl Ingason have written a response to an earlier article by Kristján Sigurðsson (Tiro 1/2025), where he highlighted potential challenges in studying linguistic variation in written transcripts. In their response, the authors react to these concerns by explaining how they have met these challenges in their research.
The popular reporting term verbatim means “word-for-word”. In his regular column, Carlo Eugeni shines a light on the other, para- and extra-verbal side of reporting: how to report phenomena like tone of voice, gestures, expressions and context in a report that consists of words? In delving into this realm, he demonstrates yet another aspect of reporting that requires thoughtful consideration—and is therefore deeply human.
Eero Voutilainen is Tiro’s editor-in-chief.
