SQuASH System

A SQuASH System is a topic-focused multi-document summarization system that implements a SQuASH algorithm developed by the SQuASH Research Project.

• AKA: SQuASH Topic-focused Multi-Document Summarization System, SFU Question Answering Summary Handler, SQUASH Summarization System.
• Context:
  • It can be one of the first QA systems to report performance improvement through the application of Semantic Role Labeling.
  • It can process SQuASH System Document Sets containing 25-50 SQuASH system source documents per DUC specifications.
  • It can analyze SQuASH System Complex Questions requiring SQuASH system multi-faceted answers.
  • It can generate SQuASH System 250-Word Summaries as SQuASH system question responses.
  • It can employ SQuASH System Components including SQuASH system semantic parser, SQuASH system sentence selector, SQuASH system compressor, and SQuASH system post-editor.
  • ...
  • It can often utilize SQuASH System Linguistic Resources such as WordNet, PropBank, and SVM-based semantic role labelers.
  • It can often apply SQuASH System Processing Stages including SQuASH system preprocessing, SQuASH system core extraction, and SQuASH system postprocessing.
  • It can often integrate SQuASH System Innovations like semantic subgraph-based selection and automatic post-editing.
  • It can often achieve SQuASH System Performance measured by ROUGE scores and pyramid evaluations.
  • ...
  • It can range from being a DUC-2005 SQuASH System to being a DUC-2006 SQuASH System, depending on its SQuASH system version.
  • It can range from being a Baseline SQuASH System to being an Enhanced SQuASH System, depending on its SQuASH system feature completeness.
  • It can range from being a Research SQuASH System to being a Competition SQuASH System, depending on its SQuASH system deployment purpose.
  • It can range from being a Monolingual SQuASH System to being a Multilingual SQuASH System, depending on its SQuASH system language capability.
  • It can range from being a Domain-General SQuASH System to being a Domain-Specific SQuASH System, depending on its SQuASH system application domain.
  • It can range from being a Standalone SQuASH System to being an Integrated SQuASH System, depending on its SQuASH system architecture.
  • It can range from being a Fixed SQuASH System to being an Adaptive SQuASH System, depending on its SQuASH system learning capability.
  • It can range from being a Extractive-Only SQuASH System to being a Hybrid SQuASH System, depending on its SQuASH system generation approach.
  • ...
  • It can implement the SQuASH Algorithm for question-based summarization.
  • It can participate in Document Understanding Conferences for system evaluation.
  • It can be developed at Simon Fraser University by the Natural Language Lab.
  • It can contribute to Question Answering Research and Semantic Role Labeling Research.
  • It can influence subsequent QA systems and summarization systems.
  • ...
• Example(s):
  • DUC-2005 SQuASH System, described in (Melli et al., 2005).
  • DUC-2006 SQuASH System, described in (Melli et al., 2006).
  • SQuASH System Components, such as:
    • SQuASH Semantic Role Labeling Module, using SVM classifiers on PropBank annotations.
    • SQuASH Semantic Subgraph Selection Module, implementing graph-based sentence extraction.
    • SQuASH Sentence Compression Module, applying statistical compression techniques (Knight & Marcu, 2000).
    • SQuASH Automatic Post-Editing Module, performing coherence enhancement.
  • SQuASH System Configurations, such as:
    • SQuASH-Base Configuration, without post-editing module.
    • SQuASH-Full Configuration, with all modules enabled.
    • SQuASH-SRL Configuration, emphasizing semantic role labeling.
  • SQuASH System Applications, such as:
    • Biomedical SQuASH System, for medical document summarization (Shi et al., 2007).
    • News SQuASH System, for news article summarization.
  • ...
• Counter-Example(s):
  • Generic Summarization System, lacking question focus and semantic role labeling.
  • Single-Document System, processing only one document.
  • Keyword-Based System, using keyword matching without semantic analysis.
  • Pure Abstractive System, generating only new text without extraction.
  • Commercial QA System, designed for production rather than research.
• See: Simon Fraser University, Topic-focused Multi-Document Summarization System, Question Answer System, Semantic Role Labeling, Document Understanding Conference, SQuASH Algorithm, SQuASH Research Project, Multi-Document Summarization System, IBM Watson System.

---

---

References

2006

• (Melli et al., 2006) ⇒ Gabor Melli, Zhongmin Shi, Yang Wang, Yudong Liu, Anoop Sarkar and Fred Popowich. (2006.). “Description of SQUASH, the SFU Question Answering Summary Handler for the DUC-2006 Summarization Task.” In: Proceedings of DUC-2006.

2005

• (Melli et al., 2005) ⇒ Gabor Melli, Yang Wang, Yudong Liu, M. M. Kashani, Zhongmin Shi, Baohua Gu, Anoop Sarkar, and Fred Popowich. (2005). “Description of SQUASH, the SFU Question Answering Summary Handler for the DUC-2005 Summarization Task.” In: Proceedings of DUC-2005.

---