Dustin Arthur Smith

An interface for targeted collection of common sense knowledge using a mixture model

We present a game-based interface for acquiring common sense knowledge. In addition to being interactive and entertaining, our interface guides the knowledge acquisition process to learn about the most salient characteristics of a particular concept. We use statistical classification methods to discover the most informative characteristics in the Open Mind Common Sense knowledge base, and use these characteristics to play a game of 20 Questions with the user. Our interface also allows users to enter knowledge more quickly than a more traditional knowledge-acquisition interface. An evaluation showed that users enjoyed the game and that it increased the speed of knowledge acquisition.

Managing implicit assumptions in natural language interfaces

A persons choice of what to communicate and how to communicate it depends on the information he or she believes is shared with the audience. This presents a challenge for natural language interfaces, because it is hard for people to predict what information they share with the interface and how it will use this information to interpret their text. This is especially difficult for pragmatic-level assumptions supplied by the interpreter that go beyond the information in the surface text, because these assumptions are negotiated in dialogue and frequently revised or redacted. We have built a calendaring interface that allows users to communicate English event descriptions. This constrained task gives us a clear criteria for communication success and failure. Failures are opportunities to acquire and revise assumptions: to collect lexical and semantic knowledge from a variety of users. By lowering the interaction barrier so end users can contribute to the linguistic interpretation process, we can collect culture-specific lexical and semantic knowledge directly from the members of the cultural group who possess it. This knowledge is essential for the pragmatic task of deriving what a speaker meant from what they said. The goal of this research is to make the assumptions involved with interpreting natural language explicit to the user. Using a model of language generation and interpretation based on planning and plan recognition, we capture, through user contributions, word definitions and commonsense assumptions - and we represent both as belief-changing actions. Using visualizations and a direct manipulation interface, users can access the interpretation status, inspect which assumptions were made, and suggest or modify existing assumptions. With the aim of providing the functionally equivalent of the negotiation stage in interpersonal dialogue, we evaluate the interface by how it allows users to revise and extend assumptions toward successful interpretation.

Action planning with commonsense knowledge

Understanding other peoples goals is an essential part of interpersonal interactions. This capability enables a person to naturally predict another persons future actions in a situation and produce appropriate joint or shared actions. In like manner, a human-like planning agent (or sociable robot) should be able to understand the users action goal and come up with subgoal-based plans to achieve the goal. In this paper we focus on how the agent can automatically construct the subgoal-based action hierarchy corresponding to the users high-level goal. As a first step, we implement an action-planning engine based on ConceptNet, and indicate the drawbacks of using ConceptNet for this purpose. Also, we present the structure of a new goal-oriented commonsense-reasoning knowledgebase for the agents action-goal representation and action planning.