218 | 29, pp. 213-233 | doxa.comunicación

July-December of 2019

Algorithms and bots applied to journalism. The case of Narrativa Inteligencia Artificial: structure...

ISSN: 1696-019X / e-ISSN: 2386-3978

This trend toward this new way of writing journalistic texts is also being carried out parallel to the media. In this regard, there are now international companies offering automated news services. In Germany, 2txt, Aexea, AX Semantics, Retresco, Text-on and Textomatic stand out. In China, it is Tencent; and in the United States, Automated Insights, Bloomberg, Linguastat, and Narrative. France has Labsense and Syllabs; Israel has Articoolo; United Kingdom has Arria, and Sweden has United Robots. In Spain, one company that stands out is Prodigioso Volcán, which collaborates with Software Intelygenz to develop infographics and visual images through machine-learning comprehension processes. Another is the startup company Narrativa Inteligencia Artificial, which together with the University of Alcalá de Henares has created Gabriele, a software that writes approximately 20,000 articles weekly in real time for the news media. It is the only company in Spain that provides AI-generated news to different media such as Sport, 20 Minutos, El Periódico, El Confidencial, El Español, El Independiente, El Heraldo and La Información. However, its structure and organization chart, production processes and the quality of the texts it generates are unknown, so the object of this research is to respond to this unknown factor. It also works with some international companies such as The Social Audience, Property Finder, Intigral, Noon, and MSN (the collection of internet services offered by Microsoft). Consequently, the study of this entity can provide a more global vision.

2. Methodology

Design of the methodology for this research is based on a systematic review of the scientific literature, which is part of the secondary source research (Codina, 2017), and has allowed us to gain knowledge regarding the main contributions to the state of the question (Ramírez-Montoya & García-Peñalvo, 2018)

Furthermore, in order to carry out this research, the methodological triangulation of qualitative and quantitative techniques was chosen (Gaitán; Piñuel, 1998), which as Soler and Enríquez (2012) have pointed out, basically consists of contrasting information from different sources in order to obtain a sufficient contextualisation of the phenomena studied. First, a semi-structured in-depth interview was designed to address the most relevant issues regarding the object of study.

Four members of the company with complementary profiles were interviewed: David Llorente, founder of the company; Javier García; engineer in artificial intelligence; Alberto Moratilla; lead architect; and David Martínez de Lecea, director of operations. The interviews were conducted between January and May of 2019 and contained a series of questions related to the technical architecture of the project, the number of texts it produces weekly, the income it generates, and its value proposition, among other issues. Secondly, a participant observation was made during the process of creating an automated news item on unemployment in Spain in order to find out what role the journalist plays in the process. Marshall and Rossman (1989) define this qualitative technique as the systematic description of events, behaviour, and artifacts, the result of which must be a “written photograph” of the situation studied, according to Erlandson et al. (1993). This has been achieved using direct observation, informal interviews, and detailed field notes, as proposed by DeWalt & DeWalt (2002). The objective was to understand the functioning of its software, called Gabriele, and to gather additional information on its working methodology. The participant observation method took place during the month of May, 2019.

Finally, an online questionnaire was distributed to 145 journalists (N=145) to assess the quality of two news items produced by Gabriele. The company provided this information and one of the items is related to Apple’s quarterly results (Figure 1).