Two proofreaders, but no professional subtitlers, have been recruited for a pilot project aimed at adding automatically generated subtitles to TVM's 6 and 8pm news broadcasts.

The government recently announced that a pilot project introducing automated Maltese subtitles for news broadcasts is being launched as part of a wider initiative to strengthen the Maltese language's presence in the digital and artificial intelligence sphere, through a collaboration between TVM and the Maltese Language Centre.

As part of this pilot project, TVM's 6 and 8pm news bulletins will include automatically generated subtitles in Maltese. In between the two daily broadcasts, these generated subtitles will be refined and verified.

However, it seems that this process will be assisted by just two proofreaders, rather than by any of the numerous qualified subtitlers in the country. The proofreaders will step in and ameliorate the subtitles generated by the used speech-to-text AI.

This project forms part of SaqWI, a questions-and-answers body funded by Microsoft's AI for Good Lab that is being developed by Ċentru tal-Ilsien Malti (ĊIM), in collaboration with the University of Malta's Department of Artificial Intelligence.

These proofreaders, who are contracted by ĊIM on a contract for service, will go over and refine the generated subtitles in between the 6 and 8pm news broadcasts every day to improve the quality of subtitles provided by the speech-to-text tools being used.

Following this announcement, this newsroom contacted Dr Giselle Spiteri Miggiani, senior lecturer in the Department of Translation, Terminology, and Interpreting Studies at the University of Malta, to see whether professional subtitlers were engaged to participate in this pilot project.

She stated that neither she nor the UM Department of Translation, Terminology and Interpreting Studies at the University of Malta, nor its affiliated Inclusive Arts and Media Lab, were consulted. Spiteri Miggiani trains her students to become qualified subtitlers in accordance with media industry standards. In an interview with The Malta Independent on Sunday last month, she said that Malta must begin subtitling all its productions to abide by an EU legislation introduced last year. She had also mentioned the need for these subtitles to be carried out by professional subtitlers, which, through her work at the university, Malta now has the necessary human resources to add Maltese or English subtitles to all television productions.

She coordinates the Audiovisual Translation and Accessibility specialisation stream within the Master in Translation and Terminology Studies. Her academic research includes conducting reception studies to find out how to best incorporate inclusive subtitles for Maltese audiences; this includes research into local audiences' preferences towards Maltese subtitles.

On this pilot project, Spiteri Miggiani noted that "true subtitling requires adherence to technical and linguistic norms and conventions to ensure legibility, visual-aesthetic adequacy, appropriate display rates, reading speeds, readability, flow, and comprehension".

Essentially, while proofreaders ensure that quality writing standards are maintained, subtitlers have the expertise to see that subtitles fit appropriately on screen without obstructing the visual scene, are visible for an adequate amount of time but aren't too chunky, and allow the Deaf and hard of hearing to follow, on top of seeing that everything is written correctly and faithful to the source audio.

Without subtitling expertise, proofreaders can only oversee proper grammar and orthography and lack the skills to fix technical parameters that improve the quality of subtitles for viewers beyond having correct on-screen spelling, she said.

"Even in the case of 'same-language' captions, this involves linguistic adaptation, cultural localisation, textual additions for sensory accessibility, and inclusive language considerations - ideally informed by user-centred research and established Maltese subtitling guidelines," she said.

How the project came about

When contacted by this media house, the Culture Ministry said the project originated from ĊIM's own initiative to strengthen the use of Maltese in the digital and technical sphere, following its successful application for funding under Microsoft's Lingua scheme. This scheme is targeted for small languages to create open-source datasets that can help bolster corpora with clean data that can then be incorporated into LLMs to be used in AI technology.

The SaqWI body "is the creation of a localised dataset (that is based on Malta's history and heritage, its language and various aspects of culture) full of questions and answers that will be elicited from the large audio/audiovisual data owned by PBS, with which, ĊIM, as part of the same Culture Ministry has a working agreement to collaborate".

The ministry said that corpora/datasets must come in texts, and this project's technological element "will require ample training of the current Speech-to-Text tools available by Microsoft".

For this, "PBS will be generously providing the study with ample audio/audiovisual data". In doing so, it will provide Microsoft's speech-to-text technology with a robust application. This software will be trained on the SaqWI dataset and then be further trained on the 6pm news items later this year.

Hence, the generated subtitles are part and parcel with this project, "meant to build the SaqWI dataset from audio/audiovisual data to textual data automatically", the ministry stated.

What impact does the absence of professional subtitlers have?

Spiteri Miggiani's 2024 academic research has highlighted the importance of these technical parameters. For instance, with reading speeds, that is, "the subtitling display rate that impacts the viewers' reading speed and is typically measured in number of characters per second" she had written that "establishing the length and duration of subtitles is crucial to ensure a readable, understandable, and stress-free experience while ensuring that the viewers' attention to the images is not significantly compromised".

This same paper also notes that "viewers who are not accustomed to subtitles in a specific language are likely to require more time to read and understand them". With Maltese subtitles not being standard or common in local media at the time of writing, this is one aspect that a qualified subtitler could factor for, and adjust, something that a proofreader may not have the knowledge or skills to do themselves.

Another point she wrote in this academic paper on Inclusive subtitles for all: Developing practices and guidelines for accessible media content in newly experimenting territories, was that the technical specifics and parameters adopted in subtitles "are bound to influence the audience's experience" - for better or worse.

She also told this newsroom that "AI becomes truly powerful only when combined with deep human expertise in the disciplines concerned".

"While utilising AI technologies for intralingual live captions is extremely advantageous for the speed and simultaneous synchronisation required, human editing, adaptation, and revision remain crucial," she said.

The academic stressed that any process featuring the creation of subtitles for television broadcasts must include qualified/professional subtitlers.

"The transition from AI-generated transcription to quality subtitles - particularly in the editing window between the 6 and 8pm news bulletins - requires professional subtitlers," Spiteri Miggiani said. "This is not a task that can be improvised; the 'human-in-the-loop' is essential to ensure technical and linguistic adequacy."

She stated that "achieving high-quality standards is not merely a technical goal; it is an act of respect that values the time and needs of the Maltese audience".

"Failure to meet these norms and conventions can easily result in distraction, a lack of understanding, frustration, and unnecessary cognitive overload, rather than achieving the intended functional outcomes and audience response, especially when implementing a new practice locally," Spiteri Miggiani said. "To be truly useful, accessibility features must meet the specific needs of the d/Deaf and hard-of-hearing communities, as well as those who simply benefit from an enhanced viewing experience."

She believes that this pilot initiative should serve as an opportunity to implement these norms and local guidelines, particularly during the subtitle refinement stage, "by engaging specialists trained specifically in Audiovisual Translation (AVT) and Accessibility" - like those emerging from her UM Department. She said that this pool of trained experts, who hold postgraduate degrees in Translation (specialisation in AVT), often also hold Maltese language degrees from the same Faculty.

"It is vital to value our local graduates and specialists who have trained within our national university system - specifically in languages, translation, film, and the humanities at the Faculty of Arts - while consulting and coordinating with academics, researchers, and subject matter experts specialised in these fields," she said.

Why weren't subtitling experts incorporated into this initiative?

Spokespersons for the ministry told this newsroom that "the aim is to strengthen Mt-Mt technology", that is technology that promotes Maltese-to-Maltese communication.

While proofreaders' involvement was deemed "necessary throughout the whole project", the ministry made no reference to professional subtitlers in its responses to this newsroom.

It gave no indication that the two proofreaders engaged for this project are also qualified subtitlers.

The ministry added that "dubbing, voiceovers and/or translations would not be employed" in this two-in-one project.

"Rather, upon further strengthening the Mt-Mt speech-to-text technology, the idea is to take this technology further to eventually strengthen automated Eng/EU27 languages-Mt translations (and vice-versa). In the long run, we aim to see Mt-LSM and Mt-Braille technologies," said the Culture ministry.

Spiteri Miggiani concluded that she looks forward to seeing "this much-needed and important pilot project open the way for further opportunities in the media sector that combine AI technologies with human language, translation, and accessibility expertise".