Computational Linguistics and Intelligent Text Processing: 8th International Conference, CICLing 2007, Mexico City, Mexico, February 18-24, 2007. Proceedings

Title
Preface
Organization
Table of Contents
Integration of Linguistic Resources for Verb Classification: FrameNet Frame, WordNet Verb and Suggested Upper Merged Ontology
	Introduction
	Background and Motivation
	Extension of FrameNet Verb Coverage
		Direct Retrieval – WordNet Synset
		WordNet Relation Links and Frame as Domain
		Affinity of Candidate Synsets with Domain Frame
		Linking FrameNet Frame with SUMO Concept
	Data Evaluation
		Evaluation Result
	Conclusion
	References
French EuroWordNet Lexical Database Improvements
	Introduction
	EuroWordNet: Presentation and Limits
	Improvement Made to the Relationships
		An Usable Database
		Update of the Semantic Relationships
	Inserting Definitions into EuroWordNet Thesaurus
		Wikipedia
		Definition Extraction in French Language
		General Process
		Results Analysis
	Conclusion
Building a Large-Scale Commonsense Knowledge Base by Converting an Existing One in a Different Language
	Introduction
	The Method
		Translating the English OpenMind Using a Commercial MT Software
		Translating the English ConceptNet with Heuristic Translation Rules
		Combining Two Translation Results
	Manual Evaluations of E-K Translated Concepts
	Concluding Remarks and Future Work
	References
Conquering Language: Using NLP on a Massive Scale to Build High Dimensional Language Models from the Web
	Introduction
	Estimating Language Presence
	Extracting Language Modeling Information
		Fetching Web Pages and Character Encoding
		Extracting Text from Web Pages
		Language Identification and Web Crawling
		Natural Language Processing
		Processing Output and Example
	Possible Applications
	Related Research and Conclusion
	References
On Heads and Coordination in Valence Acquisition
	Introduction
		Motivation and Outline
		The IPI PAN Corpus
		Poliqarp
	Distinguishing Syntactic and Semantic Heads
	Coordination
	XML Representation
	Extending the Poliqarp Query Language
		Simple Constructions
		Coordination
	Conclusion
Chinese Terminology Extraction Using Window-Based Contextual Information
	Introduction
	Related Work
	Algorithm Design
		The Preprocessing Module
		Automatic Term Extraction
		Terminology Verification
	Experiment and Discussion
		Performance of the Two Approaches
		The Hybrid Approach
	Conclusion
	References
Baby-Steps Towards Building a Spanglish Language Model
	What Is Spanglish?
	Linguistic Features of Spanglish
		Code-Switching
		Borrowing
		Code-Mixing
		Examples of Shallow Phenomena
	Language Models
	Data Collection
	Tools of the Trade
	Test Phase and Results
		SML Test
		UTI Test
	Final Remarks
	Current and Future Work
Latent Variable Models for Causal Knowledge Acquisition
	Introduction
	Related Work
	Statistical Models for Causal Knowledge Acquisition
		Model Structures
		Model Estimation
		Causality Detection
	Experiments
		Settings
		Results(1): The Effectiveness of Incorporating Dependencies Between Two Events into Causal Models
		Results(2): The Effectiveness of Class Labels
		Examples
	Conclusion
Finite-State Technology as a Programming Environment
	Introduction
	A Motivating Example
	An Alternative Implementation
	Comparison and Evaluation
	Discussion
Morphological Disambiguation of Turkish Text with Perceptron Algorithm
	Introduction
	Morphological Disambiguation
		Representation
		Problem Definition
		Methodology
	Baseline Trigram-Based Model
	Perceptron Algorithm
	Experiments
		Data Set
		Features
		Optimal Parameter and Feature Selection
		Results
	Conclusions
Part-of-Speech Tagging Using Word Probability Based on Category Patterns
	Introduction
	N-Gram POS-Tagging Models and Korean Language Characteristics
		Word N-Gram Is Not of Practical Use for Korean
		Morphotactic Constraints Within a Korean Word and Previous Alternatives
	Korean POS-Tagging Using Word Probability Based on Category Patterns
		Application of Category-Pattern-Based Model to Bayesian Models for POS-Tagging
		Parameter Training and POS Assigning
	Experimentation and Application
	Conclusions and Further Work
	References
Handling Conjunctions in Named Entities
	Introduction
	Problem Description
	Related Work
	Experimental Setup
		Corpus and Data Preparation
		The Tag Set
		Encoding
	The Algorithms
		Baseline
		Classifiers
	Results
		Evaluation Scheme
		Classification Results
	Analysis
		Conjunction Category Indicators
		Error Analysis
	Conclusions and Future Work
ANERsys: An Arabic Named Entity Recognition System Based on Maximum Entropy
	Introduction
	Named Entity Recognition in Arabic
	The Maximum Entropy Approach
	The Developed Resources
		ANERcorp$^13$: Two Corpora for Training and Test
		ANERgazet$^15$: Integrating Web-Based Gazetteers
	Experiments and Results
	Conclusions and Future Works
	References
Applying Machine Learning to Chinese Entity Detection and Tracking
	Introduction
	Related Work
	Mention Boundary Detection
		Character-Based Model Combined with Word Knowledge
		Boundary Detection with Conditional Random Fields
		Character-Based N-Gram Features and Wordlist-Based Features
		Head and Extent Combination
	Entity Attribute Identification
	Coreference Resolution
	Experimental Results
		Mention Boundary Detection Results
		Entity Attribute Identification Results
		Entity Tracking Results
	Conclusion
	References
Evaluation of an Automatic Extension of Temporal Expression Treatment to Catalan
	Introduction
	History of TERSEO Extensions
	Extension to Catalan
	Evaluation of the New Extension
		Corpus Development
		Results
	Conclusions
A Generalized Approach to Word Segmentation Using Maximum Length Descending Frequency and Entropy Rate
	Introduction
	Related Work
	Proposed Method
	A Walk-Through Example
	Evaluation and Experimental Results
	Conclusion and Future Work
	References
Tagging Sentence Boundaries in Biomedical Literature
	Introduction
	Methods
		Special SBD Issues in Biomedical Literature
		Rule-Based Approach for Biomedical Literature SBD
	Results
	Discussion
	References
Probabilistic Classifications with TBL
	Introduction
	Transformation Based Learning
	Probability Estimation with TBL Classifiers
		The Proposed Method
		Equivalence Class Partitioning
		Smoothing
		Related Work
	Experiments
		Cross Entropy and Perplexity
		Rejection Curve
		Active Learning
	Conclusions
The Non-associativity of Polarized Tree-Based Grammars
	Introduction
	Existing Polarity Systems
		XMG Colors
		PUGs
	General Polarity Systems
	Conclusion
Dependency Analysis of Clauses Using Parse Tree Kernels
	Introduction
	Problem Setting
		Clausal Dependency Identification
		Parse Tree Kernels
	Clause Dependency Analysis with Parse Tree Kernels
		Dependency Relation in Korean Clauses
		Clause Representation
		Multi-class Classification
	Experiments
	Conclusions
Unsupervised Method for Parsing Coordinated Base Noun Phrases
	Introduction
	Motivation
	Previous Work
		Syntactic Parsing
		Web Counts
	Approach and Statistical Modeling
	The Models
	Experimental Setup and Results
		Web Counts
		Evaluation Measures
	Conclusion
Text Categorization for Improved Priors of Word Meaning
	Introduction
	Finding Predominant Senses
	Creating the Domain Corpora
		The GigaWord Corpus
		The Classifier
		The Domain Corpora
		Domain Rankings
	Experiments and Evaluation
		Hand-Labelled Versus Automatically Classified
		Senseval
		Domain Salient Words
	Discussion and Future Research
Case-Sensitivity of Classifiers for WSD: Complex Systems Disambiguate Tough Words Better
	Introduction
	Prediction Factors
		Case Factors
		System Factors
	Optimal Ensembling Method
	Evaluation
		Test Setting
		Base System Complexity vs Tough / Easy Words
		Base System Complexity vs Best Optimal Ensembles
	Discussion
	Conclusions and Future Work
	References
Word Clustering for Collocation-Based Word Sense Disambiguation
	Introduction
	Related Work
		The Yarowsky Algorithm
		Word Clustering
	Extending the Collocation List
	Experiment
		Data Set
		Experimental Setup
		Experiment Results
	Discussion of Results
		Relationship Between F-Measure with Word-Class and Corpus
		Error Analysis
	Conclusion and the Future Work
	References
Lexical Constellations and the Structure of Meaning: A Prototype Application to WSD
	Introduction
	Meaning by Constellation and WSD
	DFA and WSD
	The Algorithm in Action
	Some Considerations and Conclusions
	References
Rule-Based Protein Term Identification with Help from Automatic Species Tagging
	Introduction
	Related Work
	Data and Ontology
	Hybrid Approachs to TI
		Assigning Potential CM Identifiers to Protein Mentions
		Term Disambiguation
		Results
	Conclusions
Unsupervised Discrimination of Person Names in Web Contexts
	Introduction
	Lexical Features
	Second Order Context Representation
	Cluster Stopping
	Experimental Data
	Evaluation
	Experimental Results
	Discussion and Conclusions
Learning for Semantic Parsing
	Introduction
	Sample Applications and Their MRLs
	Systems for Learning Semantic Parsers
		Scissor
		Wasp
		Krisp
	Experimental Evaluation
	Future Research
	Conclusions
The Usefulness of Conceptual Representation for the Identification of Semantic Variability Expressions
	Introduction
	Conceptual Representation Space with Latent Semantic Analysis
		WordNet Domains
		WordNet Alignment with SUMO
		Latent Semantic Analysis
	A Walk-Through Example
	Evaluation
		Microsoft Paraphrase Corpus
		Evaluation Measures
		Results
		Discussion
	Conclusions
Characterizing Humour: An Exploration of Features in Humorous Texts
	Introduction
		Related Work
	Datasets for Computational Humour
		One-Liners
		Humorous News Articles
	Automatic Humour Recognition
		Negative Datasets
		Text Classification
		Classification Results
	Characteristics of Verbal Humour
	Human Centeredness
	Polarity Orientation
	Discussion and Conclusions
Representing Emotions with Linguistic Acuity
	Introduction
	Related Work
	Emotions in Soap Opera
		Types of Emotional Cues
	Annotating Text with Emotions
	Linguistic Analysis
		Words with Explicit Emotional Senses
		Disambiguating Arguments and Expressions
		Reassessing the Emotional Weight of a Given Expression
		Strengthening Vague Expressions into Marked Ones
	Encoding Emotions with Parsing
	Discussion and Conclusion
An Evaluation of UNL Usability for High Quality Multilingualization and Projections for a Future UNL++ Language
	Introduction
	Embedding a Comparative Task-Related Evaluation of UNL in a Real Translation Task
		Steps of the Experiment
		Measured Results
		Potential Actual and Future Gains
	Aspects to Improve in the UNL Way
		Common Development Web Platform
		UNL Specification and UW Construction
		Need for Open Source and More Tools
	A New Impetus: The U++ Consortium
		Goals
		Roadmap
		Spreading UNL Usage
	Conclusion
	References
Transfer-Based MT from Spanish into Basque: Reusability, Standardization and Open Source
	Introduction
	General Architecture
		The De-formatter
		The Analyzer
		The Transfer Module
		The Generation
		The Re-formatter
	Formats for Interaction
	Linguistic Data
		Dictionaries
		Grammars
	The Programs
	Evaluation
	Conclusions
	References
Dependency-Based Chinese-English Statistical Machine Translation
	Introduction
	The Dependency Based Translation Model
	The Training of Treelet Mapping
		Data Preparation
		Training of Treelet Translation Probability
		Chinese-English Specific Treatment
	The Decoder
	Experiments and Results
		Training and Test Data
		Experiments
		Discussions
	Conclusions and Future Work
	References
Asymmetric Hybrid Machine Translation for Languages with Scarce Resources
	Introduction
	Natural Language Processing Resources
	Monolingual Natural Language Processing
	Towards Asymmetric Hybrid Translation Machine
	Evaluation Results
	Conclusions
	References
CL-Guided Korean-English MT System for Scientific Papers
	Introduction
	Related Works
	Customizing Korean-English MT System for Scientific Papers
		Korean POS Tagger and Syntactic Analyzer
		Pattern-Based Generation
		Long Sentence Processing
	CL-Guided Machine Translation
		CL Rules for Source Language Rewriting
		Target Language Rewriting Using Example Expressions
	Evaluation
	Conclusion
	References
Comparing and Integrating Alignment Template and Standard Phrase-Based Statistical Machine Translation
	Introduction
	Survey of Phrase-Based SMT
	Comparisons of SP-SMT and AT-SMT
		Sequence Alignment Model
		Feature Functions
		Decoding Process
	Integration AT into SP-SMT
	Experiments
		Corpus
		Feature
		Language
		Integration Method
	Conclusions
Dependency Analysis and CBR to Bridge the Generation Gap in Template-Based NLG
	Introduction
	Case Based Reasoning Techniques and Dependency Analysis
		Case Based Reasoning Techniques and Technologies
		The Role of Taxonomies in Computing Similarity
		Dependency Analysis
	Using Dependency Analysis to Build a Case-Base for Template Selection
		Basic Operation of the Case-Based Template-Selection Module
		The Resources Required: Case Base and Vocabulary
		Constructing the Case Base from the Dependency Trees for the Corpus
	Evaluation and Discussion
	Conclusions and Future Work
Experiments on Generating Questions About Facts
	Introduction
	Why Question Generation?
	AutoTutor
	Related Work
	Our Approach to Question Generation
	The Question Generation Mark-Up Language
		AIML
		Question Generation Mark-Up Language (QG-ML)
		Example of a Category
		Interpreter of the Mark-Up Language
	Evaluation and Experimental Results
		Factual Question Generation
	Future Work
	Conclusions
Expert vs. Non-expert Tutoring: Dialogue Moves, Interaction Patterns and Multi-utterance Turns
	Introduction
	Our Previous Work
	Study of Tutorial Interaction Patterns
		Tutor-Student Interaction Patterns
		Student-Tutor Interaction Patterns
	Study of Multi-utterance Turns
	Conclusions and Future Work
A Competitive Term Selection Method for Information Retrieval
	Introduction
	Term Selection and Weighting
		Entropy
		Transition Point
		Term Enrichment
		Union of Entropy and TP
	Experiments
		Data Description
		Results
	Discussion
Incorporating Passage Feature Within Language Model Framework for Information Retrieval
	Introduction
	Previous Work
		Passage Retrieval
		Language Model in Information Retrieval
	Incorporating Passage Information into the Language Model Framework
	Experiments and Results
		Experiment Design
		Experimental Results
	Conclusion
	References
Enhancing Cross-Language Question Answering by Combining Multiple Question Translations
	Introduction
	Proposed Methods
		Method 1: “Selecting the Best Translation”
		Method 2: “Combining Passages from Several Translations”
		Method 3: “Constructing a Question Reformulation”
	Experimental Results
		Experimental Setup
		Results
	Conclusions and Future Work
	References
The Negative Effect of Machine Translation on Cross–Lingual Question Answering
	Introduction
	State of the Art
	Taxonomy of the MT Errors for CL--QA
		Wrong Word--by--Word Translation
		Wrong Translated Sense
		Wrong Syntactic Structure
		Wrong Interrogative Particle
		Wrong Lexical-Syntactic Category
		Unknown Words
		Wrong Proper Name
	Our Approach to CL--QA
		Solution to Wrong Word--by--Word Translation
		Solution to Wrong Translated Sense
		Solution to Wrong Syntactic Structure
		Solution to Wrong Interrogative Particle
		Solution to Wrong Lexical-Syntactic Category
		Solution to Unknown Words
		Solution to Wrong Proper Name
	Evaluation
	Conclusion and Future Work
Using Clustering Approaches to Open-Domain Question Answering
	Introduction
	Sentence Clustering for Cluster-Based Language Model
		Main Idea
		One-Sentence-Multi-Topic
		Experiments with Sentence Retrieval
	Pattern-Similarity-Based Clustering to Learn Answer Patterns
		Architecture for Answer Pattern Learning
		Sentence Clustering and Pattern Extraction
		Vertical Clustering
		Horizontal Clustering
		Experiments with Unsupervised Answer Pattern Learning
	Conclusion and Future Work
	References
A Little Known Fact Is . . . Answering Other Questions Using Interest-Markers
	Introduction
	Answering ``Other'' Questions
		Finding the Wikipedia Article
		Extracting Target-Specific Interest Markers
		Finding Interesting Sentences
		Ranking Interesting Sentences
		Universal Interest Markers
	Results and Analysis
	Related Work
	Summary and Future Work
Adapting the JIRS Passage Retrieval System to the Arabic Language
	Introduction
	Retrieving Passages in Arabic
	The Arabic-JIRS Passage Retrieval System
	Experiments and Results
		Conclusions and Further Work
		Test-Bed for Arabic Question Answeringhttp://www.dsic.upv.es/ybenajiba
		Preliminary Results
Using Question-Answer Pairs in Extractive Summarization of Email Conversations
	Introduction
	Previous and Related Work
	The Data
	Extractive Summarization
	Question-Answer Pair Detection
	Integrating Question-Answer Pairs with Extractive Sentences
	Postprocessing Extracted Sentences
	Email Summarization Interface
	Conclusion and Future Work
NEO-CORTEX: A Performant User-Oriented Multi-Document Summarization System
	Introduction
	Background and Related Works
	System Overview
		Similarity
		Overlap
		Final Sentence Ranking
		Evaluating Summary Quality
		Tuning the Parameters for the DUC Task
	Adaptations for DUC 2006 Task
		Managing the Topics
		Finding the Best Metrics for DUC 2006
		Managing the Sentence Length
	Results
	Conclusion and Future Work
Event-Based Summarization Using Time Features
	Introduction
	Related Work
	Event Representation on Time Line
	Event Weighting and Sentence Selection
	Experiments and Discussion
		Preliminary Evaluation on Two Clusters
		Evaluation on Ten Clusters
		Discussion
	Conclusions and Future Work
	References
NLP-Based Curation of Bacterial Regulatory Networks
	Introduction
	A Regulatory Network Extraction System
	A Markup Language for Mining Bacterial Regulatory Networks
	Analyzed Corpora, RegulonDB and the Manual Curation Process
	Evaluating the Network Extraction Task
	Evaluating Manual/Automatic Curation Strategies
	References
	Discussion
Exploiting Category Information and Document Information to Improve Term Weighting for Text Categorization
	Introduction
		Feature Selection and Term Weighting
		Category Information and Document Information
	Intuitionistic Term Weighting Factor Constraints
	A Partial Probability Model for Term Weighting
		Model Formulation
		Global Term Weighting Factor
		Relation with idf
		Statistical Characteristics
	Experiments
		Results and Discussions
		Comprehensible Samples
	Conclusions
On the Impact of Lexical and Linguistic Features in Genre- and Domain-Based Categorization
	Introduction
	Genres and Domains
	Methodology
		Development of a Pilot Corpus
		Feature Selection for Scientific Texts
		Classifiers Used
		Evaluation Framework
	Experimentations
		Domain Classification
		Genre Classification
		Further Analysis: Micro vs. Macro-precision
	Analysis of the Discriminatory Descriptors
		Domain Descriptors
		Genre Descriptors
	Conclusion
Clustering Narrow-Domain Short Texts by Using the Kullback-Leibler Distance
	Introduction
	The Kullback-Leibler Distance
	Description of the Corpora
		The $CICLing-2002$ Corpus
		The $hep-ex$ Corpus of CERN
		The KnCr Corpus of MEDLINE
		Preprocessing
		Description of the FSTs Used
	Experimental Results
		Performance Measurement
	Results
	Conclusions
A Mixed Trigrams Approach for Context Sensitive Spell Checking
	Introduction
	A Mixed Trigrams Approach
		Mixed Trigrams
		Confusion Sets
		Levensthein Distance
		Method
		Conditional Probability Estimation for the Central Word
	Experimental Settings
		Algorithm Training
		Test Data
		Performance Measures
	Experimental Results
		Hit and False Positive Rates
		Coverage
	Conclusions and Future Work
Combining Methods for Detecting and Correcting Semantic Hidden Errors in Arabic Texts
	Introduction
	Semantic Hidden Errors
	Detecting Semantic Irregularities
		Co-occurrence-Collocation Method
		Context-Vector Method
		Vocabulary-Vector Method
		Latent Semantic Analysis Method
		Voting Method
	Correcting Semantic Errors
	Context of Work
	Testing and Results
		Evaluation of the Detection Component
		Evaluation of the Correction Component
	Conclusion
	References
Author Index