Text Analysis Software Tool Based on the Enhanced Annotated Suffix Tree Method

<p>Report 35 pages, 4 chapters, 7 illustrations, 2 tables, 35 sources, 5 appendices.</p><p><strong><em>Key words:</em></strong><em> text analysis, algorithms on strings, annotated suffix trees, suffix arrays, synonym extraction, concept graphs.</em></p><p>The subject of work is a statistical text analysis software tool based on the Enhanced Annotated Suffix Tree (AST) method. While the original AST methodology is a very promising tool for a great variety of text analysis applications, it has proven to be not efficient enough in terms of running time and space consumption to be able to process large-scale text collections. Besides, numerous experiments have shown that it often produces irrelevant results due to its insensibility to language-specific features such as synonyms.</p><p>Our goal is to develop a modification of the original AST method that would allow to overcome its drawbacks mentioned above as well as to make a reliable software implementation of the corresponding algorithms. Several experiments have shown that our new AST methodology implementation is superior to the previous ones in terms of both memory consumption (the new implementation uses 10 times less memory) and runtime of the core algorithms. This has been primarily achieved due to the usage of a new data structure, namely the suffix arrays, instead of the traditional but heavyweight suffix trees.</p><p>The software tool is distributed as an open-source application under the MIT license. It has been registered as a Python package EAST in the Python Package Index system and can be installed on every machine supporting Python. It not only can be used as a Python library in other projects, but also provides the end user with a convenient command line interface, which allows one to perform the main operations of the AST text analysis methodology in an easy way.</p><p>The software product has been successfully integrated with the software for newsfeeds collection and analysis currently developed at a student group &ldquo;Text analysis and visualization methods&rdquo; at NRU HSE.</p><p>Further work includes refining the synonym extraction algorithm configuration and introducing more AST methodology applications to our software.</p>