Technical

The following ATLAS components provide the above-described ATLAS functionalities:

1. search - this component, typically started first, provides and marshals the access to the underlying Lucene indexes. The component provides the search and indexing functionalities to the rest of the ATLAS components. The "search" component exposes its services via remote OSGi. Therefore, it is important to start it first, as the Zeroconf autodiscovery did not work well in the ATLAS environment.
2. i-publisher - this component provides the Graphical User Interface and website rendering tools to all ATLAS functionalities. All services provided by the "i-publisher" component require the remote "search" services. The location of the remote search services is defined in the "services.xml" file, located in the "com.tetracom.atlas.remote.edef" plugin.
3. pre-proc - this component is responsible for the extracting of text from various document formats, detecting its language and sending it to the appropriate language processing queue. The component communicates with the rest of ATLAS via asycnronous messages.
4. processing - this component wraps a UIMA languaguage processing chain. It accepts messages from a language queue, processes the text and sends the annotated CAS objects in an "output" queue. The component communicates with the rest of ATLAS via asycnronous messages. At least one "processing" component should be started for each of the languages supported in ATLAS. Usually, 1 GB heap should be enough for a single thread of a language processing chain.
5. tmstore - this component stores the already annotated documents in a fusion of a relational database and Lucene indexes. Services, provided by the "tmstore" component, require the remote "search" services. The location of the remote search services is defined in the "services.xml" file, located in the "com.tetracom.atlas.remote.edef" plugin. The component communicates with the rest of ATLAS via asycnronous messages.
6. autocat - the component provides automatic categorization services to ATLAS. The two tasks performed by this component, model building and categorization prediction, are executed asyncronously, thus the component communicates with the rest of ATLAS via asycnronous messages.
7. summarizer - the component provides text summarization services to ATLAS. The component communicates with the rest of ATLAS via asycnronous messages.

The diagram below depicts the actual installation of the ATLAS components and the communication channels between them.