top of page

Work in Progress

In this section, I present three articles: 

1. Reverse-engineering and Gradational Assemblage A Deep-learning Approach to Transnational TV series Adaptations (work in progress)

2. The influence of Film Techniques on Genre in Transnational TV Series Adaptations (work in progress)

3. Transnational TV Series Adaptations: What Artificial Intelligence Can Tell Us About Gender Inequality In France And The US (work published)

Paper #3:in progress 

Reverse-engineering and Gradational Assemblage

A Deep-learning Approach to Transnational TV series Adaptations


Based on a deep-learning approach to transnational TV series adaptations, we deconstruct, restructure, analyze, and compare cultural and cinematographic elements through data mining. Showcasing eight episodes of the American TV series Law & Order Criminal Intent and their equivalent French remake, Paris Enquêtes Criminelles, we seek recurring patterns of shot scales in conjunction with gender and characters' expressions of emotion to uncover and compare cultural representation and cinematographic trends in France and the US.


We rely on our AI toolkit, the Möbius Trip, to conduct our research. The Möbius Trip is a multimodal analysis engine based on machine learning techniques used to predict elements in audiovisual narration. The toolkit measures the screen time of male and female characters, identifies expressions of emotion[1], and recognizes shot scales[2].

Based on Lev Manovich's concept of reverse-engineering [3], we identify, isolate, and quantify each element. Relying on Franco Moretti's Distant Reading[4] model, we conceptualize different levels of reading of the audiovisual text by gradually zooming in on the data. We focus on shot scales between the French and American shows (figure 1). Next, we zoom in and analyze gender through shot scales. Lastly, we achieve in-depth reading by looking at the character's emotions through shot scales (Figure 2). 

We describe our approach as a Large-Scale Granularity Reading (figure 3). That is, the gradational assemblage of a wide range of elements with each other at a large scale to obtain comprehensive, precise, and complex recurring televisual patterns.


[1] Six universal expressions of emotion: Anger, Disgust, Fear, Happiness, Sadness, and Surprise

[2] Shot scales ranging from Big Close-up (BCU), Close-up (CU), Medium Close-up (MCU), Medium Long Shot (MLS), Long Shot (LS), and (Very Long Shot (VLS)


[3] Manovich, L. (2013). Visualizing Vertov. Russian Journal of Communication, 5(1), 44–55.


[4] Moretti, Franco. 2015. Distant reading. London: Verso.

Sample of the article:

Shot scales are a powerful tool to direct the audience's gaze on particular elements. Sergio Benini et al. define shot scale as, “The perceived distance of the camera from the subject of a filmed scene, namely shot scale, is a prominent formal feature of any film product, endowed with both stylistic and narrative functions” (Benini et al., 2019). It is one of the vital cinematographic features that regulate the relative size of characters’ faces, the relative proportion of the human figure to the background (Salt, 1992; Bowen and Thompson, 2013), arranging film content to emphasize an element (Carroll and Seeley, 2013)” (Rooney & Bálint, 2018) and directing the audience’s gaze (Cutting). While scholars agree about the importance of the shot scales as a fundamental expressive tool of the film language, they pass over the technical aspect of a shot scale. Annette Kuhn discusses the proportion of shot scale, “An informally agreed and widely accepted set of conventions which describe and define different framings of a film image, or apparent distances between camera and subject” (Kuhn, 2012 1321). Though there is a conventional usage of shot scales, “the terminology is quite elastic, and deals mainly with concepts” (Arijon, 2015).

It is agreed that shot scales are commonly divided into seven categories (or sometimes nine) depending on the model. (Benini, 2019, 1). Shot scales typically range from Very Long Shot (VLS), Long Shot (LS), Medium Long Shot (MLS), Medium Close-up (MCU), Close-up (CU), and Big Close-up (BCU). However, the terminology is not always consistent because BCU is also referred to as Extreme Close-up. Likewise, MLS is also called knee shot or American Shot. In addition, the convention often diverges, and the contrast between shot scale models can be pretty significant. Figure 1 highlights the difference between Barry Salt's scale in terms of proportion as well as in the terminology (e.g., MLS and Knee shot) and that of Daniel Arijon.

The categories and labels vary depending on the exact proportion of a close-up or a long shot. This discrepancy can be explained by the loosening overtime of the early Hollywood industry's strict conventions (e.g., standardized production methods and framing guides) (Kuhn). This imprecision grew over time and has led to confusion between cinematographers, directors, and camera operators. As Monaco describes, "One person's close-up is another's "detail shot," and no Academy of film has (so far) sat in deep deliberation deciding the precise point at which a medium shot becomes a long shot or a long shot metamorphoses into an extreme long shot. Nevertheless, within limits, the concepts are valid" (Monaco & Lindroth, 2000, p197). The lack of convention of terms and concepts translates leads to confusion, as noted by Annette Kuhn, "For the contemporary cinematographer, director, and camera operator working together on a production, the solution to this imprecision is to agree before filming precisely how they wish to define each type of shot size, bearing in mind that *framing can be an important expressive tool." "A Dictionary of Film Studies." Apple Books. 1322. The ambiguity of the terms and conventions does not only impact the film industry but may also affect researchers. Not having the same language and sharing the same references makes the dialog among film scholars difficult and may lead to potential misunderstanding, inaccurate findings, and inexact interpretations. Such a hiatus highlights the need for a common reference. This need is further emphasized by the advent of new media, and the large-scale computational approach to film analysis calls for a defined shot scale framework and strict conventions. In this paper, we introduce an AI-based shot scale framework.

Paper #2: in progress

The Influence of Film Techniques on Genre in Transnational TV series Adapatation


Born from a collaboration between a humanities scholar and an artificial intelligence engineer, the Möbius Trip is a deep-learning multimodal toolkit that can predict TV series genre based on the combination of color, light, rhythm, and music. Following the tradition of formalist film theorists concerned with the technical elements of motion pictures, and Cultural Analytics, which focuses on computational, visualization, and big-data methods, we seek patterns in the film-making cultural tradition between the US and France. Our results indicate that Transnational TV series Adaptations, the transfer of a show into another culture, also implies a transfer in genre. Our goal is to offer a potent toolkit that can help TV series adaptors make more intentional decisions when adapting a show and make the adapted show more successful.

Paper #1: published 

Transnational TV Series Adaptations: What Artificial Intelligence Can Tell Us About Gender Inequality In France And The US


The present research analyzes the inequality of gender representation in transnational TV series. For this purpose, a content analysis was carried out on 18 episodes of the US crime show Law & Order: Criminal Intent and its French adaptation Paris Enquêtes Criminelles. To conduct this research, we used the artificial intelligence toolkit the Möbius Trip, which is equipped with a gender and emotion recognition feature and relies on big data. The main findings indicate that male characters overwhelmingly dominate the onscreen time equally in both the US and the French versions. The data also show that male characters are more emotionally expressive and that women tend to display a wider range of emotions. The French characters are slightly more emotionally expressive than their American counterparts. The data also suggest that male characters tend to display violent behavior and that female characters tend to be portrayed as a victim in both versions of the show. The emotions-related results show a trend, but the difference of emotions between male and female characters and between the French and American cultures remain fairly narrow.

In this paper, we have introduced our artificial intelligence software, the Möbius Trip, and we have presented our current results on gender representation in transnational TV series adaptations in France and in the US. According to the above results, both Law & Order: Criminal Intent and Paris Enquêtes Criminelles tend to show a negative characterization of females based on underrepresentation. Male characters overwhelmingly dominate the show. We also proved that the lack of representation of female characters is fairly similar in both versions of the show. We demonstrated that emotion is contingent on gender and that female characters show a wider range of emotions, but that male characters are more emotionally expressive. We also demonstrated that the French characters are more emotionally expressive and show a wider range of emotions than their American counterparts. Lastly, we have shown that male characters tend to display violent behavior and that female characters tend to be portrayed as victims. The emotions-related results showed a trend, but the difference of emotions between male and female characters and between the French and American cultures were fairly narrow. We will continue investigating other French and American TV series in order to confirm the trend we have revealed in this study.

To conclude, the Möbius Trip can be a potent tool to raise awareness of women’s discrimination in transnational TV series adaptation. In the future, we hope the Möbius Trip will help TV series directors check and balance their subjective opinions to make a more accurate decision based on real and objective data. We encourage TV series directors and broadcasters to define quantified progress objectives to improve the presence of women on their channels for more gender representation equity. In this way, our study could contribute to more awareness, social justice, and better cinematographic decisions.

Paper available here: 

bottom of page