FIAT/IFTA World Conference 2024

Considering the growing popularity of podcasts due to their inherent convenience for consumption in areas such as deepening knowledge and entertainment, and taking the example of the Rai Play Sound platform, where original podcasts can be found, in addition to the radio offering of new and/or repertoire programs, we will analyse, through concrete examples, the methods and verification processes implemented by Rai’s Rights Management Department when rights clearances are requested in order to use archive material in new in-house or third parties podcast productions.

We will examine the different types of contracts, from the oldest ones, which only provided for radio and television rights, to the newest ones where exploitation within podcasts is already provided for in the contractual clauses, explaining what risks we take and why. Rights Management Department follows Rai’s Legal Department’s guidelines , but its decisions are also the result of comparisons and evaluations made for each individual case based on various factors: whether internal use or commercialization of the podcast is envisaged; the nature of the repertoire requested (movies, entertainment, investigations, sports, etc.); whether the podcast consists of the full retransmission of a program already broadcasted or whether the material must be inserted into a new work.

We will also discuss cases where the Fair Use principle can be applied for the free use of archives (for example, for purposes of criticism, comment, information, teaching, instruction, or research).

During the presentation, we will show some specific and potentially critical cases, discussing how we behave in the presence of material protected by copyright and agreements between Rai and collecting societies.

This session will explore how the BFI approached its largest ever rights clearance project, researching over 30,000 titles and clearing rights for 13,000 film and TV works for a new streaming service in UK libraries, BFI Replay, as part of the recently completed Heritage 2022 programme. I'll show how we devised our approach to compliance & risk across the multiple layers of rights and the different categories of rights holders and works, and the tools, processes and policies we developed to support this work and the results we achieved. I'll share some of the main obstacles we faced, like the information gap, and how we engaged with rights holders and consider some of the longer term challenges that can be created by project driven rights work.

INA is a French public company whose first mission is to preserve and develop a vast collection of audiovisual and audio archives from French public radios and TV channels.

Due to the quantity of rights managed and the large number of rightsholders in the collection of INA’s archives, exploiting it involves taking risks.

In order to limit risk exposure, INA has developed a politic of collective agreement conclusion with CMOs (for the authors rights) and unions (for the performers rights). In the same time, individual agreements are concluded / negotiated with third-party property rights (rightsholders non-members of CMOs, coproducers, organizers of sports events / shows, multiple rightsholders of contents previously incorporated in INA’s archives (photo, phonograms, literary works, etc.).

INA’s Legal Department also have to deal with rights of the personality, image, freedom of expression, private life… and others (with the IA, RGPD).

Our purpose is to facilitate the work of our operational teams with clear guidelines on each subject. A legal team at the INA’s Legal Department is dedicated to the legal analysis of our contents with the perspective of an exploitation to manage the risk. INA has launched a project to create a SI which enables to combine the legal data on the contents and the royalties data in order to simplify the decision making. However, human intervention is still necessary to appreciate the risk and the opportunity of an exploitation.

We will present results of combining the latest research in automatic speech recognition (ASR) with European high performance computing (HPC) and large quantities of raw audiovisual data contained in national radio and television archives. The aim of the work was two-fold, firstly, to advance ASR by building models on large public data collections and secondly, to harness the large audio-visual media archives for large-scale qualitative and quantitative media research by generating an automatic indexation based on all spoken content that is decoded by ASR. Only the largest global companies can have access both to the latest ASR development, huge computing resources and huge audio collections, but their commercial interests do not treat all languages equally.

In Europe, most languages are spoken in small countries which, however, have advanced radio and television archives containing millions of hours of broadcasted media content. The latest publicly funded HPC initiatives have also opened researchers an access to unprecedented computational resources. By utilizing the computing and archives it is possible for researchers to develop and publish large pre-trained speech models for many languages without depending on the commercial interests of the large global companies. The large speech models can be pre-trained in a self-supervised fashion which can benefit also from untranscribed and uncategorized audio collections. When openly published, these models make it then easy and quick to develop speech technology applications, such as accurate recognizers for ASR and speech, speaker and audio characteristics for these languages by fine-tuning the models using a feasible amount of transcribed target data.

In the past years, Large Language Models (LLMs) have been increasingly developed and used for various Natural Language Processing (NLP) tasks. As an example, LLMs can be employed for content classification without a need for manual human annotation or exhaustive training datasets. However, such LLMs are associated with high computational cost during inference, preventing their wide adoption in the audiovisual industry, such as large scale audiovisual collections.

In this context, we want to automatically classify broadcast news into topics. Topic classification can be useful for audiovisual content retrieval or monitoring, but remains a challenging problem, particularly due to the difficult task of segmenting continuous feed into homogenous extracts. To solve this issue, we propose a framework applied on French TV and radio news where we transcribe a dataset of 11.7k hours, broadcasted in 2023 on 21 channels with a State-of-the-Art transcription model. A LLM is used in few-shot conversation mode to obtain a topic classification on those transcriptions. We define a topic classification scheme based on the IPTC categories.

Using the generated LLM annotations, we explore the finetuning of a specialized smaller classification model. To evaluate the performances of these models, and to estimate the subjectivity of the topic categorization task, we construct and annotate a test set of 03h44m. We demonstrate that while LLM's inference costs makes them prohibitive for large scale analysis of audiovisual collections, they can be used to generate synthetic datasets used to train less complex models (students) associated to much smaller inference time and better classification performances.

Finally, using an automatic gender classification tool, we compute the speaking time per gender depending on the topic, to determine if some subjects are predominantly reserved to specific genders. We show that women are notably under-represented in subjects such as sports and politics.

The transformative potential of applying artificial intelligence in the context of audiovisual archiving has been demonstrated across a spectrum of use cases related to search & exploration, preservation, artistic expression and big-data analysis. This presentation explores research findings regarding the integration of AI solutions specifically in humanities and social sciences research.  

The emergence of Large Language Models (LLMs) and Large Multimedia Models (LMMs) poses new opportunities and questions in this area. For instance, LLMs could support the content annotation tasks for researchers working on framing analysis over large amounts of textual data. But more investigation is required to determine how such systems could be fine tuned and integrated into research workflows to deliver satisfactory and trustworthy outcomes. The goal of this research was to develop a roadmap that guides the development and use of LLMs and LMMs in SSH research. It (i) Identifies potential SSH research areas (such as framing analysis) that could benefit from the use of LMMs / LLMs, (ii) Identify limitations and risks of these techniques, (iii) provide recommendations on how LLMs and LMMs could be designed and fine tuned to fit SSH scholar needs.

We will also reflect on how researchers, archives, and the broader cultural sector can contribute to promoting more responsible AI practices in society and raising "AI Literacy." This entails exploring practical ways in which cultural institutions can actively engage with the ethical and societal implications of AI, such as curating exhibitions that examine AI's impact on human lives, hosting workshops to educate the public about AI technologies and their implications, and collaborating with researchers to develop resources that foster critical thinking about AI in diverse communities.