Conference article

This paper analyzes the tone (including polarity and semantic orientations) in a corpus of financial reports in Spanish. Specifically, we look at the Letter to Shareholders section of the Annual Reports, which concentrates on an analysis of the financial performance, corporate strategies and other aspects relevant to investors. We use a semantic analysis tool developed for Spanish narratives, based on lexicons and phrase-structure information. We divide the corpus in four subgroups based on the net earnings figure as a benchmark, to identify differences in tone between profit firms and loss firms. This paper confirms that Spanish financial narratives suffers a communicative bias towards positive terms (Pollyanna effect). Additionally, we provide a gold standard of financial narratives, based on a random selection of 1% of the sentences of the corpus of Letters. We run a first evaluation of three different sentiment analysis tools (Azure, Watson and FinT-esp) compared to the GS and observe that tone analysis in the financial narratives domain breaks with classical sentiment analysis (based on subjective feelings, value judgments, emotions). Financial narratives tone is linked to measurable facts and figures (financial results) and investors’ expectations about the future performance of the firm.

Opinion & Sentiment Analysis, Financial narratives Spanish