by Leif Hanlen
Welcome to the Text Analytics as a Service (t3as) blog.
We want to deliver an open ability in Text Analysis to end users via open source, web-based services. Text Analytics is a significant bottleneck for data analysis: analytics of unstructured text is needed as text data remains largely unused outside research projects.
Text Analytics is an artisanal (or cottage) industry that does not yet lend itself to engineered processes. Text analytics lacks the standardisation required to deploy technology solutions composed of “off the shelf” components.
We want to build a framework that delivers the benefits of open-source text analytics, whilst overcoming the barriers of open data analysis.
How does it work?
Text Analysis as a Service will support multiple sub-projects. The services are hosted on Amazon Web Services, with domain www.t3as.org Projects will deliver:
- Open Source files:
We maintain a private GitHub repository. Projects that reach an appropriate level of maturity, including NICTA open source requirements, will be transitioned to Open Source, under GNU General Public License v3 (GPLv3) and public GitHub
- Application Program Interface hosted on a web service (Amazon Web Services)
These API’s will be useable without commercial license. The hosted API’s are designed to deliver trusted code, but not necessarily high bandwidth. The API’s are versioned, and will support remote interrogation by other services: this allows composition of our web services. If/when load becomes an issue we will develop a free API key access that supports monitoring and potentially limits usage per user.
- Web page as a wrapper
We will host web pages that wrap the API’s to show the services. We expect the web page to give a tutorial in the API calls, as well as allowing developers to duplicate the web page, and experiment with API calls, before developing their own software that calls the API
- Published source code, allowing remote and local deployment
The services work in both remote (as a server-side application) and local (service deployed in-process within the user’s application) modes.