Natural language generation in expert systems
This thesis will be conducted in close work with startup www.inspirient.com.
Inspirient is a Berlin-based startup run by former graduates of FU Berlin’s Computer Science department that is specializing in automated, cloud-based data analytics with an intuitive, math-free user experience. We want to push analytic capabilities to non-expert users, who, coincidentally, tend to be the ones who know which insights are most valuable. To achieve this objective we are developing an Artificial Intelligence (AI) that automates business analytics and displays the resulting insights as easy-to-understand visuals.
Our visuals contain two key pieces of information, firstly, and most importantly, the title describing the insight, and secondly the supporting chart. Therefore, automatically and intelligently generating titles is key to delivering value and achieving an exceptional user experience. To this end, the objective of this thesis is to investigate and implement a suitable software architecture for Natural Language Generation for Inspirient’s AI.
A master thesis will thus comprise the following items:
- Review of natural language generator design principles, architectures, and current approaches
- Design, adaptation, and implementation of algorithms and data structures to suitably generate natural language text for the domain of business analytics, incl. internationalization
- Quantitative evaluation of fidelity of output text given real-world input datasets
If you are interested in this thesis and would like to know more, please reach out to Dr. Guillaume Aimetti at email@example.com, or visit us at www.inspirient.com. From the FU side, please contact Dr. Matthias Wählisch.