Extraction and linking

[bolandka]

• much more efficient context extraction using lucene's highlighter package
• more efficient pattern searching using LucenePatternSearcher, added tokenization algorithm and whitespaceAnalyzer
• textualReferences consist of whole sentences instead of fixed context size (Context size #5)
• number of words used in patterns is adjustable and independent of context size
• generalized (further) punctuation and years in patterns for higher recall
• optimized lucene queries for better performance
• prohibit only generation of less general patterns in FrequencyBasedBootstrapping, not of equally general patterns
• fixed ReliabilityBasedBootstrapping (fix computation of reliability scores (InfolisPattern.getReliable) #15, fix reliability-based bootstrapping (ReliabilityBasedBootstrapping) #16)
• added startpage parameter for text extraction to enable ignoring title pages
• updated libraries (java, lucene, ..)
• adapted code to changes in the dara repository
• filtering resourceType: dataset when querying dara (ReferenceResolving: better filter for dara results #75)
• creation of temporary files in tokenizer and bibliographyExtractor if no output directories are specified
• posting of intermediate results to temporary datastores (objects have to be posted to the datastore but cannot be deleted #87)
• added class to extract dois from texts (add class to extract DOIs #80)
• added a very basic importer for springer a++ documents
• added package for importing and exporting annotated data (some features still missing)
• some more optimizations, fixes and cleaning (e.g. extractBib property in Bootstrapping ignored when input files are txt instead of pdf #79, NPE in ApplyPatternAndResolve if called with queryServices variant #77, NullPointerException when trying to retrieve reliability of QueryService #53)