Dissertation Information for Victoria L. Rubin
- Victoria L. Rubin
- Library and Information Science
- Syracuse University (USA) (2006)
- Elizabeth DuRoss Liddy
- Susan Monica Bonzi
- Scott Richard Nicholson
- Jaklin Kornfilt
- Michael Anthony D'Eredita
- Barbara H Kwasnik
MPACT Status: Fully Complete
Title: Identifying certainty in texts
Abstract: This study empirically derives a framework for analyzing certainty about written propositions. CERTAINTY, or EPISTEMIC MODALITY, is a linguistic expression of an estimation of the likelihood that a particular state of affairs is, has been, or will be true. The study describes how explicitly marked certainty can be predictably and dependably identified in texts, and answers the following research questions: (1) How is explicit certainty expressed linguistically in English? (2) What are the patterns and frequencies of occurrence of explicit certainty markers? (3) How much inter-subjective agreement can be reached by readers when identifying and categorizing explicit certainty in texts?
A dataset of 2,343 sentences in 80 New York Times Service news reports and editorials was analyzed for occurrences of explicit certainty markers (e.g., definitely, somewhat ) based on a four-dimensional categorization model. The dimensions included: level ( absolute, high, moderate, low certainty, and uncertainty ), perspective ( the writer's point of view, reported direct participant's account, and reported expert's view ), focus ( opinions, emotions, or judgments and facts or events ), and time ( past, present, future, and irrelevant ). The intercoder reliability results are modest but show improvements with longer coder training and stricter codebook instructions.
The model's five-level typology is subdivided into forty-three syntactico-semantic classes. Central modal auxiliary verbs, gradable adjectives in their superlative degree, and adverbial intensifiers frequently express explicit certainty, while adjectival downtoners and adverbial value disjuncts rarely do so.
Explicitly-qualified statements occur at a significantly higher rate in editorials than in news reports (means of 0.94 and 0.7 certainty markers per sentence, respectively). Editorials have a high likelihood of starting and ending with an explicit certainty-qualified statement, while news reports tend to start with an implicitly certain statement, and have equal chances of ending with an implicitly or explicitly certainty-qualified statement.
The model can serve as an analytical framework for automated certainty identification, a novel type of analysis in Natural Language Processing (NLP). Such a framework allows tracking of certainty levels in texts over time and among text genres. This contribution potentially enhances automated search capabilities in NLP and other information access applications.
MPACT Scores for Victoria L. Rubin
A = 0
C = 1
A+C = 1
T = 0
G = 0
W = 0
TD = 0
TA = 0
calculated 2010-09-28 01:14:23
Advisors and Advisees Graph