Text Algorithms in Economics

Abstract

This article provides an overview of the methods used for algorithmic text analysis in economics, with a focus on three key contributions. First, we introduce methods for representing documents as high-dimensional count vectors over vocabulary terms, for representing words as vectors, and for representing word sequences as embedding vectors. Second, we define four core empirical tasks that encompass most text-as-data research in economics and enumerate the various approaches that have been taken so far to accomplish these tasks. Finally, we flag limitations in the current literature, with a focus on the challenge of validating algorithmic output. Expected final online publication date for the Annual Review of Economics, Volume 15 is August 2023. Please see http://www.annualreviews.org/page/journal/pubdates for revised estimates.

Generative AI enhances individual creativity but reduces the collective diversity of novel content

“…Thus far, we have focused on the subjective evaluation of thirdparty readers; now, we turn to a more objective measure of the stories' content, to understand how generative AI affects the final stories produced. Using embeddings (26) obtained from OpenAI's embeddings application programming interface (API), we were able to compute the cosine similarity of the stories to all other stories within condition as well as the generative AI ideas (Fig. 4).…”

A Practical Guide to Conversation Research: How to Study What People Say to Each Other

“…This definition primarily serves to distinguish conversations from documents authored from a single perspective, including speeches, essays, newspaper and magazine articles, books, product reviews, legal documents, and social media posts. Although the "great bulk" of language use is conversational (Levinson, 2016), single-voice documents have been the dominant source material in applied text analysis, and many review articles in related fields have focused only on single-voice documents (e.g., Benoit, 2020;Berger et al, 2020;Boyd & Schwartz, 2021;Dehghani & Boyd, 2022;Gentzkow et al, 2019;Grimmer & Stewart, 2013;Hansen & Ash, 2023;Hirschberg & Manning, 2015;Jackson et al, 2022;Pennebaker et al, 2003). Many of the techniques developed for singlevoiced documents are also useful for studying conversations.…”

Remote Work across Jobs, Companies, and Space

“…Self-attention is the key idea that powers models such as BERT, RoBERTa, GPT, GPT-3, PALM, and, most famously, ChatGPT. For more details on these models, collectively called Transformers, see Ash and Hansen (2023).…”