.png){kind=icon}
.png){kind=icon}
Return
Using Coreference Resolution to Mitigate Measurement Error in Text Analysis
.png){kind=icon}
PRE
AI
DOI：10.1177/10944281251334777.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Content analysis has enabled organizational scholars to study constructs and relationships that were previously unattainable at scale. One particular area of focus has been on sentiment analysis, which scholars have implemented to examine myriad relationships pertinent to organizational research. This article addresses certain limitations in sentiment analysis. More specifically, we bring attention to the challenge of accurately attributing sentiment in text that mentions multiple firms. Whereas traditional methods often result in measurement error due to misattributing text to firms, we offer coreference resolution-a natural language processing technique that identifies and links expressions referring to the same entity-as a solution to this problem. Across two studies, we demonstrate the potential of this approach to reduce measurement error and enhance the veracity of text analyses. We conclude by offering avenues for theoretical and empirical advances in organizational research.
Keywords:
archival
data management
data cleaning
data
sampling
and modeling
machine learning and AI
types of research design
Journal
IF:
7.6
Papers:
883
Citations:
1.4W