On developing indicators with text analytics: exploring concept vectors applied to English and Chinese texts

This paper investigates how high-quality, vocabulary-based classifiers, useful for competitive intelligence, can be found for relatively small corpora of publicly available documents. Two corpora of recent annual reports are examined and compared, one in English and one in Chinese. The paper tests whether vocabularies can predict whether firms are relatively innovative or not, examining vocabularies of both content words and function words. We find that indeed the tested vocabularies do produce effective indicators or classifiers and, surprisingly, that function words are especially effective. The paper also provides extensive conceptual and theoretical background to frame the investigation in the context of an EMCUT problematic, that of mapping entities to classification schemes using information derived from text.