Monografico_3_EN_Doxa_29

59.39% maintain that the coherence between statements is adequate. In terms of cohesion, 53.81% also gave a positive rating to the way in which the different words, sentences and paragraphs of the text, both syntactic and lexical-semantic, are related and joined together. The figure of 52.17% of those surveyed “totally agree” with the appropriateness of the news item, because it is well written (29.34%) and respects spelling and grammar rules (22.83%), thanks mainly to the use of short sentences. Journalists also gave a positive evaluation (51.09%) to the criteria of newsworthiness and accuracy, despite the fact that the information relates to the intention to vote in Spain in the general elections of 26 June 2016. Along the same lines, 68.97% of the professionals say they are “in total agreement” with the accuracy of the data regarding the news facts and the impartiality, neutrality and honesty with regard to reality. However, 87.23% think there is no diversity of views, so the quality of the sources is questionable (77.56%), since their origin is not reflected.

Finally, the journalists consider that the context and interpretation can be improved: 79.64% think it lacks variety of expressiveness regarding the facts narrated, so the information is considered to be monotonous (28.20%), lacking in rhythm (25.30%), poor in vocabulary (15.17%), and improvable in style (10.97%).

4. Discussion and conclusions

The present research carried out has allowed us to consider the initial objective of analysing the structure and organization chart of Narrativa Inteligencia Artificial, as well as studying its production processes and the quality of the journalistic texts it produces. The company was founded in 2015 and is the only company in the country that generates automated texts for the media. Since its founding, the startup has not stopped customizing its services to generate and increase its value, nor has it ceased responding to the demands of different companies and industries that deal with data overload, such as those of the financial sector, e-commerce, real estate, healthcare and telecommunications. In the media field, for example, it generates automated news for various Spanish media companies such as El Confidencial, Sport, El Periódico, El Español, El Independiente, 20 Minutos, El Heraldo, La Información, and MediaPro. Furthermore, it collaborates in the international field with the Associated Press agency in the collaborative writing of texts (Hypothesis 1).

The product and formats used by Narrativa Inteligencia Artificial reflect the cultural and technological change that the media industry is currently undergoing (Pavlik, 2015).

The company is contributing to the revitalization of journalism with the search for new communicative formulas and business models. This shows a panorama that is radically different from the trends of the traditional company in recent years when it comes to having new information products on the market.

The startup’s software, Gabriele, increases journalistic productivity by automatically, accurately and quickly generating around 20,000 journalistic texts on specific topics per week by indexing data from preset templates (Hypothesis 2). Specifically, the content generated is based on computational learning, or machine learning, through the input of large amounts of data that serve as a basis for writing or creating content. This situation allows the writer to move away from repetitive, routine tasks and develop those that are more creative, and that add value to journalistic work. However, in the production process the presence of the journalist is necessary, especially when selecting writings samples of interest that will serve as a model for later information.