Natural Language Processing with Java 1st Edition by Richard Reese ISBN 1784391794 9781784391799

1. Introduction to Natural Language Processing (NLP)

• What is NLP?
• Why use NLP?
• Challenges in NLP
  • Why is NLP difficult?
• Survey of NLP Tools
  • Apache OpenNLP
  • Stanford NLP
  • LingPipe
  • GATE
  • UIMA
• Overview of Text Processing Tasks
  • Finding parts of text
  • Sentence segmentation
  • Named entity recognition (NER)
  • Part of Speech (POS) tagging
  • Text classification
  • Relationship extraction
  • Combining approaches for better results
• Understanding NLP Models
  • Identifying the task
  • Selecting a model
  • Building and training the model
  • Verifying the model
  • Using the model
• Preparing Data
• Summary

2. Finding Parts of Text

• Understanding Parts of Text
• Tokenization
  • What is tokenization?
  • Uses of tokenizers
• Tokenization Techniques in Java
  • Using Scanner class
  • Specifying delimiters
  • Using split method
  • Using BreakIterator
  • Using StreamTokenizer
  • Using StringTokenizer
• Performance Considerations in Java Tokenization
• NLP Tokenizer APIs
  • OpenNLP Tokenizer
  • SimpleTokenizer
  • WhitespaceTokenizer
  • TokenizerME
  • Stanford Tokenizer
  • PTBTokenizer
  • DocumentPreprocessor
• Tokenization Pipelines
• Using LingPipe Tokenizers
• Training a Tokenizer
• Normalization Techniques
  • Converting to lowercase
  • Removing stopwords
  • Using LingPipe for stopwords removal
  • Stemming (Porter Stemmer and LingPipe)
  • Lemmatization (Stanford and OpenNLP)
• Summary

3. Finding Sentences

• Sentence Boundary Detection (SBD)
  • The SBD process
  • Challenges in SBD
  • Rules of LingPipe’s HeuristicSentenceModel
• Java-Based Sentence Detection
  • Using regular expressions
  • Using BreakIterator
• NLP APIs for Sentence Detection
  • OpenNLP: SentenceDetectorME
  • Stanford API
  • LingPipe: IndoEuropeanSentenceModel
• Training a Sentence Detection Model
  • Evaluating using SentenceDetectorEvaluator
• Summary

4. Finding People and Things (Named Entity Recognition)

• Challenges in Named Entity Recognition (NER)
• Techniques for NER
  • Lists and regular expressions
  • Statistical classifiers
  • Using regular expressions with Java and LingPipe
• Using NLP APIs for NER
  • OpenNLP for NER
  • Stanford NER
  • LingPipe NER models
  • Training an NER model
• Entity Accuracy and Evaluation
• Processing Multiple Entity Types
• Summary

5. Detecting Part of Speech (POS)

• POS Tagging Process
  • Importance of POS tagging
  • Challenges in POS tagging
• Using NLP APIs for POS Tagging
  • OpenNLP: POSTaggerME, Chunking, and Dictionary class
  • Stanford POS: MaxentTagger and pipeline
  • LingPipe POS tagging: HmmDecoder with Best-First and NBest tags
• Training a POS Model
• Summary

6. Classifying Texts and Documents

• Overview of Text Classification
  • Sentiment analysis
  • Text classification techniques
• Using APIs for Classification
  • OpenNLP: Training and using DocumentCategorizerME
  • Stanford: ColumnDataClassifier for sentiment analysis
  • LingPipe: Text classification, sentiment analysis, and language identification
• Summary

7. Using Parsers to Extract Relationships

• Understanding Parse Trees
• Types of Relationships
• Extracting Relationships Using NLP APIs
  • OpenNLP and Stanford API
  • LexicalizedParser and TreePrint classes
  • GrammaticalStructure for word dependencies
  • Coreference resolution
• Applications in Question-Answer Systems
• Summary

8. Combined Approaches

• Data Preparation
  • Using Boilerpipe for HTML extraction
  • POI for Word document extraction
  • PDFBox for PDF text extraction
• Building Pipelines
  • Stanford NLP pipeline
  • Using multiple cores in a pipeline
  • Creating pipelines for text search