The Okapi project’s main purpose is to architect a set of building blocks for the creation of larger open source localization and translation tools. But many Okapi components are generic enough to be of interest to the text mining, natural language processing, and text retrieval communities. Okapi’s many text filters (HTML, Properties, XML (ITS XPath-based rules), OpenXML, ODF, Regex etc.) provide a straightforward way to access the text of multiple document formats. Its document events and pipeline can be made to integrate with other frameworks such as UIMA, LingPipe, OpenPipeline, OpenNLP, GATE, and Lucene. The advantage of Okapi’s text filters is that not only is text extracted, but all non-textual formatting is preserved. It is possible to decompose a document into events, process them via the pipeline, and then rebuild the input document without loss. Structural information can be added to Okapi document events so that tables, lists, links, titles etc. are grouped together and treated as a unit. This is useful when context based on a “universal” document structure is needed. The Okapi event model supports user configurable annotations, similar to UIMA, but simpler and more restricted in scope. User can annotate spans of text or add new resources such as translation memory matches, terminology, token types, or part of speech information.
ACOPOST is a set of freely available POS taggers modeled after well-known techniques. The programs are written in C (aiming for extreme portability and code correctness/safety) and run under various Unix flavors (and probably even under Windows). ACOPOST currently consists of four taggers that are based on different frameworks: Maximum Entropy Tagger (MET), Trigram Tagger (T3, based on Hidden Markov Models), Error-driven Transformation-based Tagger (TBT or Brill Tagger), and Example-based tagger (ET).
foma is a compiler, programming language, and C library for constructing finite-state automata and transducers for various uses. It has specific support for many natural language processing applications such as producing morphological analyzers. Although NLP applications are probably the main use of foma, it is sufficiently generic to use for a large number of purposes. It comes with an xfst-compatible interface and regular expression language. The library contains efficient implementations of all classical automata/transducer algorithms: determinization, minimization, epsilon-removal, composition, and boolean operations. More advanced construction methods are also available: context restriction, quotients, first-order regular logic, transducers from replacement rules, etc.
TWSI is software that produces lexical substitutions in context for over 1000 frequent nouns. It processes English text. This functionality is realized by a supervised word sense disambiguation system, which is trained by sense-labeled occurrences of target words. A classification model is trained for each word, and used to decide which sense an unseen occurrence most likely belongs to. Associated with senses are lists of substitutions, which are injected into the text using inline annotation.