À Punt and MLLP (València Politècnica University): The final automated transcription project?
Automated transcription may be the solution that the Broadcast industry needs to achieve 100% accessibility to audiovisual content. But can the minimal latency and stability necessary be achieved while preserving the indispensable work of the transcriptors?
Vicente Fuster, head of the computer team of À Punt, recounts the last great development promoted by the Valencian Corporation: a transcription system for direct and deferred developed through an agreement with the Polytechnic University of Valencia - UPV.
The Law 7/2010 of March 31 (General Law of Audiovisual Communication that, by the way, it will be renewed in the near future), requires that the television audiovisual communication subtitulate at least the 75% of the programs, contributing additionally two hours a week of interpretation with sign language. We talk about a whole challenge. Although the accessibility of the deferred contents can be addressed if the necessary time is arranged, the live transcription continues to suppose difficulties To achieve the desired zero latency and a Minimum error margin. It is true that different human resources or digital tools are currently used. However, neither of the two worlds is perfect: transcriptors can offer a much more reliable collection, although they move away from real time. On the other hand, automation systems They will reduce latency, but they will fail loudly when capturing regional languages, such as Valencian, he gallego or the inhale.
The À Punt, led by Vicente Fuster and Paul Peiró Under the leadership of Higinio Añó Sanz (Director of the Department of Technology), has achieved the difficult balance between both worlds. Thanks to an agreement with the UPV, and after deep developments both at the level of engineering and lexical training with the media of the chain and the Valencian Academy of the Language (AVL), the one that possibly is the more solid system prepared to transcribe a Spanish co -official language. In it, a IA fast and efficient, a EMISSION INTEGRATION SYSTEM light and functional, and the Essential Transcript Team Labor, in charge of correcting the first result of the system and making rehabilled for Optimize the functioning of the tool.
Fuster deepens the keys to this solution that has also had a special involvement by Alfred Costa Folgado, director of the Anonymous Company of Community - SAMC; Empar Marco Estellés, previous director of the SAMC; Mar Iglesias, President of Function of the Valencian Corporation of Audiovisual Media - CVMA; Enrique Soriano Hernández, previous director of the Valencian Corporation of Audiovisual Media; and all the team of the Polytechnic University of Valencia, led by Alfons Juan Ciscar, UPV professor and director of the MLLP Research Group (Vrain Institute).
The context of à punt ... and Spanish televisions
The automated transcription and subtitulation initiative initiated by à Punt is based on the public service work that the corporation has to continue enhancing the Accessibility of its contents. The objective is clear: to bring each spectator of the Valencian Community closer through different accessibility tools, either Audio -description, subtitled or interpretation in Sign tongue.
The development is framed in a Complex global context, in which the big world agents such as Google or Microsoft They barely dedicate efforts to create Capable systems to perform the speech-to-text of languages such as Valencian, he gallego or the inhale. Similarly, given their global service strategy, these agents trust proposals cloud or hybrids; Solutions that either because the technology is not ready or due to the wide distances with the processing centers, do not reach the desired immediacy.
The needs of the project were on the table. A tool with some Great Percentages of successes in Valencian (adapting, in this way, to the lexical particularities that differentiate the language of Catalan) and that could integrate into the entire workflow of the Valencian chain. All with the aim of achieving subtitular in the less time possible audiovisual contents of Assisted by real -time computer form and based on intelligence artificial. The opportunity, as Fuster recounts, arrived from the hand of a working group of the Polytechnic University of Valencia.
The first steps
A permanently cuts the technological context by looking at future tools that could optimize their processes. However, as Fuster confesses, many times The technique is not prepared. There are developments (small initiatives from the academic world or alpha driven by technological) that address certain areas, but these are not usually approached to the minimums required by a television with the reach of A Punt.
Through several contacts, Fuster discovered that from the Polytechnic University of Valencia numerous research projects were being promoted, especially in the fields of the artificial intelligence, which could be perfectly applicable (prior adaptation) to the world television. With the support of the chain management, Fuster was able to boost a open day In à tip that served for different research groups They could expose their latest developments: Transcoding servers, feeling analysis, use of AI to get ahead of technical problems ... They put multiple lines of research on the table that could contribute new dimensions to the technical operation of television. However, a group stood out above all: MLLP, with his work on the Automated transcription.
After nine months of dialogue, procedures and administrative problems, à punt was able to sign a settlement with the UPV. In it, both institutions would commit to adapt the transcription system to the television operation, as well as continue rediscovering, developing and deepening the tool.
A successful development in just one year
After nine months, the agreement was ready to be signed in February 2020. However, the COVID-19 arrived in Spain. The pandemic delayed the signature and application of the alliance, displacing the implementation of the initiative to October 2020. This agreement, signed for two years With another two years extension options, it has a series of objectives. A year later, many of them have already been fulfilled. Among them, the integration and system implementation: At the end of October the system began operating in all live emissions, and since mid -November, this operation has spread to the Deferred programs.
The adaptation of the UPV solution, despite the talent of both teams, has not been a Simple work. They have automated numerous processes, they have interconnected all issuance and production systems with the automated transcription system and Integrated the tool With the API for direct Fingertext of Anglatecnic (à tip provider). Similarly, the system has been adapted to the needs of each program. As Fuster recalls, "it is not the same to make an informative in which a normalized language is used and a script is available, than to make a live with the complexity that you may be talking in Valencian and respond in Spanish." The system, flexible and constantly evolving, continues to multiply its reach with the passing of the weeks.
Improving the day -to -day workers
One of the keys to the automated transcription project is that it was never designed thinking about replace The team of transcriptores From à punt. On the contrary, as it emphasizes Carpenter, the objective was to improve their performance, making the transcriptive person work "with less pressure and achieving a better result."
In this way, transcriptors perform closely related functions With the work they did previously: Correct the transcription proposed by the system, adapt elements to make them compatible with the subtitling system, supervise the transcription processes of the deferred contents or perform the rehabilled: live phrases that replicates what was commented by the protagonists of various formats. The system, being trained to understand the voices of the transcriptors and when they do their locutions in an isolated and controlled space, is able to respond with greater fidelity In certain contrexts.
Fuster highlights that the transcription equipment has Expressed your satisfaction for this methodology. In addition, he considers that using the previous model was not sustainable to meet one of the goals of à punt: that the 100% of its programming is subtitled. "We could reach this goal if we hired a hundred specialized transcriptors, but the money we have is what it is and we cannot do it. What we have to do is increase the productivity of the processes. That is why we have applied this system," says the head of computer science of à Punt.
Success and latency
Today, the results of the proposal are very promising. The automatic transcription success rate in the global program of à Punt in Valencian 80% (against him 55% of the success of Google systems), amount that rises to more than one 90% (against 68% of Google) of success In the informative of the chain.
Vicente Fuster and UPV team work for Improve these percentages, although they recognize the merit of the results after just one year working actively on the project.
With regard to the latency of the system, Fuster stands out that it is set in 0, 8 seconds, A figure that is described as an achievement. The computer manager justifies this milestone in the decision to function on premises with completely local servers. These servers, by the way, are able to attend "without problem" up to two direct at the same time.
A promising future
The automated transcription solution has barely given its First steps. The agreement signed between à Punt and UPV includes new developments that will continue to provide valuable television resources. Among these developments, we find the application of a new artificial intelligence that will allow a Label and cataloging media to favor the work of the area of documentation; and the generation of a label tree Based on transcripts so that the journalist can easily locate the most appropriate piece to complement his report or news.
Similarly, the à punt team is also working on the creation of an avatar that allows to do Sign tongue based on it speech-to-text of the transcription system; and in the automated generation of subtitles in Valencian, Spanish or English based on the original audio in one of these three languages. This last application, designed to "enrich educational content", will take your first steps over 2022.
Beyond these projects in the medium term, a new improvement is contemplated that will reach beginning of next year: The authorization of Subtitulation systems (supported by the solution of Automated transcription) integrated into the live signal emission what À Punt offers through your web page and mobile applications.
Will the system reach other televisions?
Today, numerous television public have contacted the team of À Punt to be interested in the automated transcription solution. Once they discover their performanceThey are as surprised as interested in knowing more details, in order to value their inclusion in their respective workflows.
Today, Fuster does not know what the future of this tool will be Fuera de to punt. Can it be marketed by television itself? The university working group You can spread it and achieve an economic revenue from it? The Information Technology of à Punt breathes relieved to know that "fortunately" That decision It does not correspond to it. However, he acknowledges that he would feel “proud” that the system will use on other televisions, since it would endorse the intense work done by both your team and the UPV.
Similarly, Fuster considers that the success of this tool throws A clear message About him regional television value how À Punt: “In this context, in which there are sectors that proclaim that investing on autonomic television is to throw money in the trash, projects such as this show that our media have all their meaning; thanks to regional television, we are doing projects that will never be carried out in many other private initiatives”.
A report by Sergio Julián Gómez
Did you like this article?
Subscribe to our NEWSLETTER and you won't miss a thing.
















