2018/09/04 by Carlos-Emiliano González-Gallardo, Malek Hajjem, González-Gallardo, Carlos-Emiliano +5
Computer Science · #Natural Language Processing Techniques #Handwritten Text Recognition Techniques
paper · pdf · doi:10.48550/arxiv.1809.00994
In this paper we propose a new approach to evaluate the informativeness of\ntranscriptions coming from Automatic Speech Recognition systems. This approach,\nbased in the notion of informativeness, is focused on the framework of\nAutomatic Text Summarization performed over these transcriptions. At a first\nglance we estimate the informative content of the various automatic\ntranscriptions, then we explore the capacity of Automatic Text Summarization to\novercome the informative loss. To do this we use an automatic summary\nevaluation protocol without reference (based on the informative content), which\ncomputes the divergence between probability distributions of different textual\nrepresentations: manual and automatic transcriptions and their summaries. After\na set of evaluations this analysis allowed us to judge both the quality of the\ntranscriptions in terms of informativeness and to assess the ability of\nautomatic text summarization to compensate the problems raised during the\ntranscription phase.\n