Introduction
The 2002 Conference on Empirical Methods in Natural Language Processing (EMNLP) represented a pivotal moment in the history of computational linguistics. As documented in the proceedings' preface, this period marked the transition of empirical methods—characterized by data-driven and statistical approaches—from a specialized niche into the academic mainstream. The conference organizers faced the dual challenge of managing unprecedented institutional growth while attempting to preserve the intellectual risk-taking that defined the field's early years.
Managing Explosive Growth
In 2002, EMNLP experienced a 100% increase in submissions compared to the previous year, receiving 142 papers. This volume was comparable to the flagship ACL (Association for Computational Linguistics) conferences of the mid-1990s, signaling that EMNLP had evolved from a small workshop into a primary venue for the community.
To accommodate this surge, the organizers implemented several structural changes:
- Parallel Sessions: For the first time, the conference moved away from a single-track format to handle the increased number of accepted papers.
- Reviewer Workload: The "reviewer tax" became a significant concern, with a panel of 53 reviewers often evaluating more than 10 papers each on short notice.
- Extended Schedule: The conference duration and session lengths were adjusted to maintain a standard 25-minute presentation slot per paper.
The Methodology of Selection
The program chairs adopted a rigorous organizational methodology to ensure the quality of the proceedings. Utilizing a double-blind review process, the committee focused heavily on the "gray zone"—those borderline papers where numerical scores were ambiguous. In these instances, the chairs prioritized the qualitative prose of reviewer comments over raw metrics, seeking to identify work with genuine substance rather than just high averages.
Ultimately, the conference achieved a competitive acceptance rate of 24.6% (35 papers) for full presentation, with a total "printed" acceptance rate of 28.9% when including reserve papers.
The Hunt for the "Next Big Thing"
A recurring theme in the 2002 preface is the search for radical innovation. The organizers explicitly sought papers that deviated from the mainstream, hoping to discover the next foundational breakthrough on par with the landmark works of the late 1980s and 1990s, such as Kenneth Church’s work on tagging or Michael Collins’ lexicalized parsers.
However, the results of this search were sobering. Despite a deliberate call for non-traditional submissions, only one paper out of 142 was flagged by a reviewer as having the potential to be a "next big thing." This observation highlights a perennial challenge in maturing scientific fields: as a specific paradigm (in this case, empirical methods) becomes successful and institutionalized, truly radical innovation becomes increasingly difficult to identify and reward within the standard peer-review framework.
Conclusion
EMNLP 2002 was more than just a successful academic gathering; it was a benchmark for the maturity of the SIGDAT community. While the chairs candidly acknowledged the subjective nature of the selection process and the strain on the reviewing pool, the conference solidified the empirical movement's dominance. It left the community with a lingering question that remains relevant today: how can a rapidly growing field maintain its capacity for disruptive, field-defining research while managing the logistical demands of its own success?
Summary updated