David Tcheng

Released in 1999, was developed by the Automated Learning Group (ALG) at NCSA. has been used to solve many problems for both industry (e.g., Sears, Caterpillar, etc.) and government agencies (e.g., NSF, NASA, NIH, etc.). Academic uses include bioinformatics, seismology, hydrology, and astronomy. uses a data flow paradigm where a program is a network (directed graph) of processing modules. Modules can be primitive, defined as a single piece of source code that implements a single well defined task, or can be nested meaning it is defined as a network of previously defined modules. Decomposition of programs into modules that implement a well defined input-output relationship promotes the creation of reusable code. Nesting modules into higher-level modules helps to manage complexity. parallelizes across any number different computers by simply running a copy of "D2K Server" on each available machine. The software distribution comes as a basic package, with core modules capable of doing general purpose data-mining, as well as such task-specific add-on packages as text analysis (T2K), image analysis (I2K), and now music analysis (M2K).

Bei Yu

Similar to many data-mining tools, has implemented a number of automatic classification and clustering algorithms. Compared to the commercial text mining tools, for example SAS Text Miner, has richer NLP preprocessing tools, especially after its integration with GATE. Tools include: stemmer, tokenizer, PoS-tagger, data cleaning and named-entity extraction tools. The clustering visualization is tailored for thematic analysis. On one hand, provides a text mining platform for the HC community. On the other hand, is also a platform to automate the HC research results and thus facilitate their applications to the text mining community in general. For example, most of the text mining tasks are still topicality oriented, but the affect analysis has emerged in the last couple of years. The affect of a document includes the subjectivity/objectivity, the positive/neutral/negative attitude, and the strength of emotions, etc. Some researchers have adapted stylistic analysis techniques from HC to analyze customer reviews. The found non-thematic features can also be used as predictors for document genre, readability, clarity and many other document properties.

Geoffrey Rockwell

has released an alpha of the portal and will have the beta ready by June 2005. The portal is designed to allow researchers to run tools (which can be local or remote web services) on texts (which can be local or remote.) The portal has been designed to work with other systems like in three ways:

Stephen J. Ramsay

Tamarind began with the observation that the most basic text analysis procedure of all — search — does not typically operate on the text archive itself. It operates, rather, on a specially designed data structure (typically an inverted file or pat trie index) that contains string locations and byte offsets. Tamarind's primary goal is to facilitate access to analytical data gleaned from large-scale full text archives. Our working prototype of Tamarind, for example, can quickly generate a relational database of graph properties in a text which can in turn be mined for structural information about the texts in question. Tamarind creates a generalized database schema for holding text properties and allows you to specify this structure as one that should be isolated and loaded into the database. Work is proceeding on a module that will allow the user to load a Tamarind database with millions of word frequency data points drawn from several gigabytes of encoded data. Unlike existing tools, this newest module includes information about where those counts occur within the tag structure of the document (something that is impossible to do without the raw XML). For the purposes of this project, we intend to use and as the primary clients for Tamarind data stores.

J. Stephen Downie

M2K is being developed to provide the Music Information Retrieval (MIR) community with a mechanism to access a secure store of copyright-sensitive music materials in symbolic, audio and graphic formats. is a set of open-source, music-specific, modules jointly developed by members of the project and the wider MIR community. modules include such classic signal processing functions as Fast Fourier Transforms, Spectral Flux, etc. In combination with 's built-in classification functions (e.g., Bayesian Networks, Decision Trees, etc.), the modules allow MIR researchers to quickly construct and evaluate prototype MIR systems that perform such sophisticated tasks as genre recoginition, artist identification, audio transcription, score analysis, and similarity clustering.