Outline of the Ephyra Source Code
Ephyra main classes
- Ephyra2 (most recent version of Ephyra): Ephyra/src/info/ephyra/Ephyra2.java
- Ephyra2 with answer projection on a text corpus: Ephyra/src/info/ephyra/trec/Ephyra2Corpus.java
QA pipeline
- Modules:
- Question analysis: Ephyra/src/info/ephyra/questionanalysis/
- Query generation: Ephyra/src/info/ephyra/querygeneration/
- Search Ephyra/src/info/ephyra/search/
- Answer extraction and selection: Ephyra/src/info/ephyra/answerselection/
- Standardized components:
- Query generators extend Ephyra/src/info/ephyra/querygeneration/generators/QueryGenerator.java
- Knowledge miners (for unstructured sources) extend Ephyra/src/info/ephyra/search/searchers/KnowledgeMiner.java
- Knowledge annotators (for (semi)structured sources) extend Ephyra/src/info/ephyra/search/searchers/KnowledgeAnnotator.java
- Filters (for answer extraction and selection) extend Ephyra/src/info/ephyra/answerselection/filters/Filter.java
- Data structures:
Other useful resources and tools
- NLP:
- Sentence segmentation, tokenization, POS tagging, phrase chunking, full syntactic parsing, NE tagging, stemming, verb form conversion: Ephyra/src/info/ephyra/nlp/
- Function words, irregular verbs, word frequencies: Ephyra/src/info/ephyra/nlp/indices/
- Semantic parsing: Ephyra/src/info/ephyra/nlp/semantics/
- Ontologies for term extraction and expansion: Ephyra/src/info/ephyra/nlp/semantics/ontologies/
- Tools for TREC: Ephyra/src/info/ephyra/trec/
- Ephyra with answer projection on a text corpus
- Evaluation tools
- Parsers for TREC question and pattern files
- Target preprocessing
- Resolution of coreferences in TREC questions
- Preprocessing tools for several text corpora: Ephyra/src/info/ephyra/indexing/
- AQUAINT newswire corpus
- AQUAINT2 newswire corpus
- Blog06 crawl of blogs
- Collection of useful tools: Ephyra/src/info/ephyra/util/
- Conversion of regular expressions
- HTML conversion
- File access
- Caching
- String manipulation and comparison
- Message printing and logging: Ephyra/src/info/ephyra/io/
- A sandbox to test individual components: Ephyra/src/info/ephyra/test/
![(please configure the [header_logo] section in trac.ini)](http://www.cs.cmu.edu/~nico/ephyra/doc/images/ephyra.jpg)