Building Data-Driven Dialogue Systems (Paper a Day) — Tracking the state of a conversation is a whole sub-genre of its own, which goes by the name of dialogue state tracking or DSTC. It is framed as a classification problem: given current input to the dialogue state tracker plus any relevant external knowledge from other sources (e.g., the timetable information from our previous example), the goal is to output a probability distribution over a set of predefined hypotheses, plus a special “REST” hypothesis that captures the probability that none of the others are correct.
Crypto Storage Cheat Sheet (OWASP) — This article provides a simple model to follow when implementing solutions to protect data at rest.
Learn faster. Dig deeper. See farther.
Join the O'Reilly online learning platform. Get a free trial today and find answers on the fly, or master something new and useful.
Goods: Organizing Google’s Data Sets (PDF) — Goods extracts metadata ranging from salient information about each data set (owners, timestamps, schema) to relationships among data sets, such as similarity and provenance. It then exposes this metadata through services that allow engineers to find data sets within the company, to monitor data sets, to annotate them in order to enable others to use their data sets, and to analyze relationships between them.
Edward Snowden’s Life as a Robot (NY Mag) — After a while, you stop noticing that he is a robot, just as you have learned to forget that the disembodied voice at your ear is a phone. Snowden sees this all the time, whether he is talking to audiences in auditoriums or holding meetings via videoconference. “There’s always that initial friction, that moment where everybody’s like, ‘Wow, this is crazy,’ but then it melts away,” Snowden told me.