TeXTracT: a Web-based Tool for Building NLP-enabled Applications

ALEJANDRO RAGO; FACUNDO R. RAMOS; JUAN I. VELEZ; JORGE ANDRES DIAZ PACE; CLAUDIA MARCOS

Buenos Aires

Congreso; Simposio Argentino de Ingeniería de Software (ASSE 2016) - 45 JAIIO; 2016

SADIO

Over the last few years, the software industry has showed an increasing interest for applications with Natural Language Processing (NLP) capabilities. Several cloud-based solutions have emerged with the purpose of simplifying and streamlining the integration of NLP tech- niques via Web services. These NLP techniques cover tasks such as language detection, entity recognition, sentiment analysis, classification, among others. However, the services provided are not always as exten- sible and configurable as a developer may want, preventing their use in industry-grade developments and limiting their adoption in specialized domains (e.g., for analyzing technical documentation). In this context, we have developed a tool called TeXTracT that is designed to be compos- able, extensible, configurable and accessible. In our tool, NLP techniques can be accessed independently and orchestrated in a pipeline via REST- ful Web services. Moreover, the architecture supports the setup and de- ployment of NLP techniques on demand. The NLP infrastructure is built upon the UIMA framework, which defines communication protocols and uniform service interfaces for text analysis modules. TeXTracT has been evaluated in two case-studies to assess its pros and cons