We want to build something like the computer in StarTrek which can do complex queries based on natural language speech.
- See what the existing state of the art tools can do to parse a question / sentence. We will test out tools from Existing Resources and results will be documented in the wiki.
Make a system which can run a command which is stated as a query
- "What is the length of students?"
Make a system which can run simple queries that don't need a bunch of joins
This is a list of existing attempts to do this which we would do well to use as a jumping off point.
- Nguyen, D. Q., Nguyen, D. Q., & Pham, S. B. (2014). Ripple Down Rules for Question Answering, 0, 21. http://doi.org/10.3233/SW-150204
- Androutsopoulos, I., Ritchie, G. D., & Thanisch, P. (1995). Natural Language Interfaces to Databases - An Introduction. Journal of Natural Language Engineering, (709), 50. http://doi.org/10.1017/S0269888900005476
- Li, F., & Jagadish, H. V. (2014). Constructing an Interactive Natural Language Interface for Relational Databases. Proceedings of the VLDB Endowment, 8(1), 73–84. Retrieved from http://www.eecs.umich.edu/eecs/about/articles/2015/VLDB_Best_Paper.pdf
- Natural language to SPARQL Queries
- Natural language to SQL Queries
- TextRazor Easy to try with a demo
- Bluemix Needs a sign-up but has a free plan of 1000 free NLU items per day
- Google Natural Language API The article on Morphology and Dependecy Trees(https://cloud.google.com/natural-language/docs/morphology) gives an idea of what POS's they're capable of tagging
Our goal was to start with JSON files, so I have started by populated the example-datasets folder with JSON files from JsonStudio.
- companies.json
- enron.json
- stocks.json
- world_bank.json
- zips.json