Extracting Investor Sentiment from Weblog Texts: A Knowledge-based Approach

This work presents a knowledge-based approach for extracting investor sentiment directly at high frequency from financial web logs based on domain expertise and linguistic knowledge and performs a semantic analysis that starts on the word and sentence level.

Abstract

Financial web logs contain a large amount of investor sentiments, i.e., expert assessments of financial instruments and market situations. These blogs provide potentially new and relevant information for investment managers. Since humans are not able to process and interpret the large amounts of available web information, an automated solution is required. We present a knowledge-based approach for extracting investor sentiment directly at high frequency. The approach performs a semantic analysis that starts on the word and sentence level. We employ ontology-guided and rule-based web information extraction based on domain expertise and linguistic knowledge. We evaluate our approach against standard machine learning approaches. A portfolio selection test using extracted sentiments provides evidence for the economic utility of investor sentiments from weblogs.