Search Is Simple: Factoid Question Answering Made Easy
April 10, 2015
I know that everyone is an expert when it comes to search. There are the Peter Principle managers at Fortune 100 companies who know so much about information retrieval. There are the former English majors who pontificate about next generation systems. There are marketers who have moved from Best Buy to the sylvan manses of faceted search and clustering.
But when one gets into the nitty gritty of figuring out how to identify information that answers a user’s question, the sunny days darkens, just like the shadows in a furrowed brow.
Navigate to “A Neural Network for Factoid Question Answering Over Paragraphs.” Download the technical paper at this link. What we have an interesting discussion of a method for identifying facts that appear in separate paragraphs. The approach makes use of a number of procedures, including a helpful vector space visualization to make evident the outcome of the procedures.
Now does the method work?
There is scant information about throughput, speed of processing, or what has to be done to handle multi-lingual content, blog type information, and short strings like content in WhatsApp.
One thing is known: Answering user questions is not yet akin to nuking a burrito in a microwave.
Stephen E Arnold, April 10, 2015
Comments
One Response to “Search Is Simple: Factoid Question Answering Made Easy”
-
- Subscribe to Beyond Search
Feature archive
News archive
-
Web Site
Search Is Simple: Factoid Question Answering Made Easy : Stephen E. Arnold @ Beyond Search