A pragmatic approach to translation: Vocabulary alignment through Multiagent Interaction and Observation

Aus International Center for Computational Logic
Wechseln zu:Navigation, Suche

A pragmatic approach to translation: Vocabulary alignment through Multiagent Interaction and Observation

Vortrag von Paula Chocrón
Talk by: Paula Chocrón

Institute: Artificial Intelligence Research Institute (IIIA-CSIC), Barcelona, Spain

Abstract: "Every speaker that has been abroad knows that understanding a foreign language is easier when performing simple, well-defined interactions. For example, it is easier to ask for a coffee than to discuss politics in a language we do not master. In this talk I will discuss how this idea can be applied to help achieve meaningful communication in artificial multi-agent systems. In open, heterogeneous environments, it is likely that interlocutors with different backgrounds use different languages. Can the contextual information about the tasks being performed be used to learn a translation that allows agents to interact?"

I will start by presenting a notion of context that is based on the formal specifications of the tasks performed by agents. I will then show how this context can be used by agents to align their vocabularies dynamically, by learning mappings from the experience of previous interactions. In doing so, we will also rethink the traditional approach to semantic matching and its evaluation, tackling the following questions: What does it mean for agents to "understand each other"? When is an alignment good for a particular application? Finally, I will present an application to agents that interact using procedural protocols obtained from the WikiHow website, showing how they can infer a translation between English and Spanish without using external resources.