To what extent does content selection affect surface realization in the context of headline generation?

Headline generation is a task where the most important information of news article condensed and embodied into single short sentence. This normally addressed by summarization techniques, ideally combining extractive abstractive methods together with sentence compression or fusion techniques. Although Natural Language Generation (NLG) techniques have not been directly exploited for headline generation, they may provide better mechanisms than to paraphrase text. Therefore, this paper analyzes evaluates effectiveness NLG generating headlines. In NLG, both content selection surface realization are equally important—there no point in text without knowing topic. Considering premise, we therefore take HanaNLG—a hybrid approach—as basis, analyze effect generated when different strategies integrated at macroplanning stage. The experiments conducted show that, despite using any sophisticated method, proposed approach provided following benefits: i) it coherent, linguistically structured headline; ii) obtained results on standard datasets (i.e., DUC 2003 2004) that were comparable several competitive systems, terms and, iii) headlines whole (PLM-HanaNLG) preferred human assessors compared those best performing system 2003.