The Hub is hiring!

The Research Hub for Language in Forensic Evidence (School of Languages and Linguistics, University of Melbourne) has just advertised two five-year Level B research-only positions.

Please find information about the Research Fellow in Forensic Linguistics positions at the link below. The application deadline is Fri 1 December, 2023.

Read on for some background about the Research Fellow positions that might help you prepare your application. Check back here before you submit your application in case there are any updates. Questions? Email me at helen dot fraser at

Important note

These are Level B Research Fellow positions. As stated in the position information, the expectation is that candidates will already have a PhD in a relevant discipline. We can make an exception for a well-qualified applicant who has a nearly complete PhD in a relevant discipline and also has relevant professional experience – but these Research Fellow positions are not intended as an opportunity to undertake a postgraduate study.

About the Research Hub for Language in Forensic Evidence

The Hub was established in 2020, as a Strategic Initiative jointly funded by the Faculty of Arts and the School of Languages and Linguistics, in response to a number of problems that had been identified with the legal handling of recorded speech used as evidence in criminal trials.

The ten-minute video below gives a brief introduction to the issues the Hub seeks to address.

The Hub’s core questions

The short video above introduces the problem that originally sparked the research that eventually led to establishment of the Hub: use of police transcripts as ‘assistance’ to juries in understanding poor-quality forensic audio.

It also explains that the key problem is not the fact that the courts use police transcripts (though that is certainly not ideal). The key problem is the fact that all transcripts, whether by police, well-qualified experts, poorly-qualified experts, or anyone else, are evaluated by lawyers and judges, and handled in accordance with legal procedures which were developed with insufficient recognition of factors that affect the perception of forensic audio.

During the time this key problem was being uncovered, it also became clear that transcription is only one of multiple issues affecting the courts’ understanding of forensic audio. For example, many recordings feature languages other than English, which require not just transcription but translation. Transcripts in any language must represent not just what was said, but who said it (speaker attribution). Poor-quality audio is often ‘enhanced’ in hopes of improving its intelligibility. All of these aspects are currently handled in problematic ways. So the following four topics became, and remain, the core questions of the Hub’s research agenda and we have ongoing projects in each (along with some other topics of interest):

  • how to ensure reliable transcription of poor-quality English recordings
  • how to ensure reliable transcription and translation of forensic audio featuring languages other than English (note that the Hub does not currently work directly on forensic translation, but we do take an interest in ensuring that translators work from a reliable transcript, and represent speaker attribution reliably)
  • how to ensure transcripts reliably attribute utterances to particular speakers in the recording (note that this is different from speaker recognition, which seeks to help the court identify who the speakers are – the latter is an important topic of general interest to the Hub, but not a topic of current research focus)
  • how to ensure the audio itself is processed (or ‘enhanced’) in reliable ways that make it maximally easy for the court to hear the content (in conjunction with a reliable transcript).

In our short life so far, the Hub’s minuscule team (Director Prof Helen Fraser and Research Fellow Dr Debbie Loakes, both part-time) have made significant progress on each of these four topics. The Hub is now being extended and expanded to enable the work to be carried forward.

Important note: As might be obvious from the above, we aim to hire researchers who can understand, join with and contribute to the Hub’s existing research agenda, which is outlined briefly here. You application does not need to set out your personal proposal for how to solve the massively complex problems the Hub is addressing. See ‘What we are looking for’ below.

Two strands of research

For each of the Hub’s core questions, we have two broad strands of research, very closely intertwined. It will definitely be important for all members of the Hub to understand both strands, but we will provide opportunities for successful candidates to get up to speed from our previous work.

The ‘legal’ strand

The ‘legal’ strand seeks to help Australian law bring about appropriate change in its procedures for handling forensic audio. This is needed because, the way the law currently operates, merely having experts provide the court with a reliable transcript, or point out errors in an existing transcript, is not enough to ensure that the court gains a reliable impression of the content of forensic audio.

One important aspect of this work is finding effective ways to counter the misconceptions and unhelpful language ideologies that underlie current legal procedures.

An equally important aim of the legal strand is to ensure that linguistic science has a reliable understanding of how the relevant legal procedures really work – not just in principle but in practice. After all, in order to provide good recommendations, we need to understand the context in which those recommendations are to be implemented – but there are quite a few misconceptions among forensic transcription experts about how their evidence is used ‘behind the scenes’.

Some key readings for the legal strand

The ‘linguistics’ strand

The ’linguistics’ strand seeks to build the research base needed to ensure that all audio used as evidence in criminal trials is accompanied by a reliable transcript, produced via accountable, evidence-based methods, with reliable translation and speaker attribution where needed – and provided to jurors in ways that assist them to reach a reliable interpretation of the content of the audio evidence.

Some key commitments of the linguistics strand

End-to-end perspective

This commitment follows from all that has been said so far: it is not enough for an expert witness to gain a reliable impression of the audio content – though that is an essential first step. What really matters is whether court as a whole, especially the jury, gains a reliable impression of the audio content.

Interdisciplinary approach

One of the biggest barriers to progress with forensic transcription is the widespread assumption that it is a matter purely for acoustic phonetics or speech science. Certainly these disciplines are highly relevant, but other branches of linguistics (many of which use transcription in various ways), and indeed other disciplines (such as law, forensic science, cognitive science) can contribute valuable insights in relation to research methodology, law-linguistics communication, and other aspects.

Not just ‘applied’ research

Achieving the Hub’s core goals is not just a matter of applying existing findings of linguistic science. Forensic transcription is a very different undertaking from transcription in ‘normal’ branches of linguistic science.

In fact, just being an expert in any branch of linguistic science (including acoustic phonetics) does not guarantee aptitude for the practical task of forensic transcription. This is partly because forensic audio is often of very poor quality. A more important factors however is that forensic transcription has a very specific purpose (determining what words were spoken) which is quite different from the purpose of most research transcripts, where the words are already known and the issue is how to represent them in relation to some research question. The good news is that expertise in forensic transcription does not necessarily require personal aptitude in transcribing forensic audio. Rather it requires knowing how to harness and extend relevant knowledge and skill to ensure a reliable transcript is produced.

Needs a team, not just an individual

In fact the Hub is committed to the concept (which is now becoming more widespread) that providing a demonstrably reliable transcript needs a team, not just an individual. Of course an individual can prepare a transcript. However the harder task is not preparing the transcript, but evaluating it. In ‘normal’ linguistics or phonetics setting, evaluating a transcript (e.g. by a student) is simply a matter of comparing its content against a known ‘right answer’ (or ‘ground truth’) – but that is the one thing we can’t do in forensics, where the whole point is that the content is unknown, and even the context is disputed. How do you evaluate a transcript without knowing the right answer? There are many possible answers to investigate – but one thing we know for sure is that leaving evaluation to the courts is not a viable method.

The Hub approach is to have the work done by accredited practitioners following an evidence-based method designed and managed by experts, and able to be updated in a controlled manner as new research results become available. Notably, this is how responsible forensic sciences in non-language disciplines are handled.

Theoretical considerations

One reason forensic transcription seems like such a hard problem is that it asks us to do something rather different from the traditional goals of research in speech perception and related fields. Much academic study in these fields aims to model what experiment participants do as they listen to speech under various conditions, and to explain why they make the errors they do. In forensics we need to develop understanding that lets us become better ourselves (as individuals and teams) at avoiding transcription errors even under difficult conditions. Fortunately recent developments in Bayesian theory and predictive processing theory are congenial to this aim.

Some key readings for linguistics strand

It’s an ambitious agenda – can you help?

There’s no doubt the Hub has set itself some big questions to investigate! We have made significant progress (some but not all represented in our publications, listed at Key references from the Hub). In particular we are fortunate to have had some brilliant assistance from the Language Testing Research Centre, helping us develop tests for transcribers wanting to work within our method, and from Melbourne Data Analytics Platform, helping us develop a platform to enable us to carry out experiments rapidly and effiiciently.

But we still have a lot of work to do in coming years, and we are going to need a committed team to help us take it forward. If you take inspiration from our ambitious agenda, combining genuine real-world impact with genuine advancement of academic knowledge across a range of disciplines – do consider applying to become one of our five-year Research Fellows. We can’t wait to get started with our new team.

What we are looking for

Important: We certainly need researchers with existing knowledge in their own specialised area. But above all we are looking for candidates who are ready to learn, explore, and contribute to development of new knowledge. One of the key commitments of the Hub, as noted, is that forensics is not just a matter of ‘applying’ existing knowledge from other branches of linguistic science. Forensic linguistics offers productive challenges to existing knowledge that can help us improve our general scientific understanding of the nature of spoken language and how it works.

As a reminder

The application deadline is Fri 1 December, 2023.

Information about the Research Fellow in Forensic Linguistics positions is at this link – please read the requirements carefully:

Find more information about the Hub on the Hub blog ( or at If you have questions don’t hesitate to email me: helen dot fraser at I’m keen to make contact with great potential applicants!