Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, EMNLP 2011, 27-31 July 2011, John McIntyre Conference Centre, Edinburgh, UK, A meeting of SIGDAT, a Special Interest Group of the ACL

EMNLP 2011 · DBLP

The 2011 Conference on Empirical Methods in Natural Language Processing (EMNLP), held in Edinburgh, UK, served as a pivotal snapshot of a field in the midst of rapid expansion and structural transformation. As a primary venue for SIGDAT, the proceedings from this year reflect a community grappling with the challenges of growth, the necessity of reproducibility, and a shift toward more complex, data-driven modeling.

Structural Innovations and Challenges

Program Chairs Regina Barzilay and Mark Johnson identified two significant "growing pains" within the NLP community during this period. First was the issue of disciplinary fragmentation. As the field grew, the standard multi-track conference structure risked siloing sub-fields. To counter this, the 2011 program introduced additional plenary sessions, ensuring that research of broad interest reached the entire community rather than just specialists.

Second, the chairs addressed a critical concern regarding scientific integrity. They noted a rise in undisclosed double submissions and self-plagiarism, calling for the establishment of clearer community standards and improved information-sharing between major conferences to maintain the rigor of the peer-review process.

Methodological Shifts: Reproducibility and Joint Modeling

EMNLP 2011 marked a milestone in the movement toward "Open Science." For the first time, the conference allowed authors to submit up to 10MB of supplementary code and data. This policy change was a direct response to the need for reproducibility in empirical research, a standard that has since become a cornerstone of the field.

From a technical standpoint, the proceedings highlight a move away from traditional pipeline architectures. Instead, researchers began favoring joint models—such as those combining POS tagging and dependency parsing—to mitigate error propagation. Furthermore, there was a concerted effort to reduce the reliance on expensive labeled data, evidenced by a surge in research on unsupervised discovery of discourse relations and bootstrapping for semantic parsers.

The Diversification of Data and Domains

The datasets utilized in 2011 signaled a transition from clean, edited text (like the Wall Street Journal corpus) to the "noisy" reality of the modern web. Key trends included:

  • Social Media Analysis: A significant volume of research focused on the unique linguistic challenges of Twitter and microblogs, including named entity recognition and redundancy analysis in short-form text.
  • Multilingualism: The program featured extensive work on Arabic, Japanese, and Chinese, alongside cross-lingual projection techniques designed to port resources from high-resource to low-resource languages.
  • Knowledge Integration: Researchers increasingly integrated structured data from sources like WordNet and other large-scale knowledge bases to improve entity linking and semantic understanding.

Key Research Frontiers

Several sub-fields showed signs of maturation or significant evolution during this conference:

  • Sentiment and Opinion Mining: This area moved beyond simple polarity (positive/negative) toward complex structural mining and compositional matrix-space models.
  • Machine Translation (MT): MT research was heavily invested in reordering models and hierarchical phrase-based systems, pushing toward more syntactically-informed translation.
  • Scalability: As data sizes reached new heights, the introduction of "Approximate Scalable Bounded Space Sketches" demonstrated a growing need for computationally efficient NLP algorithms.

Conclusion and Community Growth

The metrics of EMNLP 2011 underscore the field's competitive nature and its expanding footprint. The conference received a record 628 submissions, ultimately accepting 149 papers for a selective 23.7% acceptance rate. Of these, 63% were presented as oral presentations and 36% as posters.

Supported by a program committee of 545 members, the 2011 proceedings capture a moment where NLP began to intersect more frequently with other disciplines—such as computer vision—and set the stage for the deep learning and open-data revolutions that would follow in the coming decade.

Summary updated

BibTeX
@inproceedings{2011proceedings,
  title = {{Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, EMNLP 2011, 27-31 July 2011, John McIntyre Conference Centre, Edinburgh, UK, A meeting of SIGDAT, a Special Interest Group of the ACL}},
  booktitle = {Conference on Empirical Methods in Natural Language Processing},
  year = {2011},
}

Loading related content...

Figure