Nate Blaylock, \Retroactive Recognition of Interleaved Plans for Natural Language Dialogue," Technical Report 761, University of Rochester, Department of Computer Science, December 2001. 46  Home/Search   Document Details and Download   Summary   Related Articles   Check

....uniquely identified, resulting in reduced effectiveness of the potential advantages. Collagen [Rich00] is a typical examlv using this approach. Alternatively, it is possible to try to fit the user s utterance in all available parsing framng and then pick that one that best catches the utterance [Blaylock01] More advanced approaches use semvMC[ networks andmdv9[ passing to understand the relations between the user s utterances [Norvig87] ormv9 a logical derivation of how the user s actionsmtio fit in the user s plans and goals [Wilensky02] We discussed these inmv[ detail in our previous ....

Blaylock, N. (2001). Retroactive recognition of interleaved plans for natural language dialogue. Technical Report 761, University of Rochester, Departmr, ofCom9H[v Science, Deceme,v Available Internet: http://www.cs.rochester.edu/u/blaylock/Pubs.

....that more complex devices actually form, as will be discussed below. Plan recognition has also been used by dialogue systems especially in the TRAINS (and subsequently TRIPS) system at the University of Rochester. The most recent work on plan recognition from 14 that project is described in [Bla01] The plan recognition system is focused mainly on inferring user plans either bottom up (actions rst) or top down (goals rst) and it has been recently enhanced to model synergies that can arise from plans being interleaved. Interleaved plans are those in which one action is part of several ....

Nate Blaylock. Retroactive recognition of interleaved plans for natural language dialogue. Technical report, Department of Computer Science, University of Rochester, December 2001.

....can use the system to complete the task. While these plan based approaches to intention recognition have proved successful in the past, the commitment to continuous understanding requires a migration to incremental processing. Most plan recognition systems expect a complete speech act as input ([Bla01]) which makes them unsuitable for our purposes. The sole exceptions found in [Bla01] are [NMiH 99] already discussed in section 2.5, and a line of Bayesian modelling developed in [CG91] and 49 [CG93] While this latter approach does not explicitly attempt incremental intention recognition, ....

....recognition have proved successful in the past, the commitment to continuous understanding requires a migration to incremental processing. Most plan recognition systems expect a complete speech act as input ( Bla01] which makes them unsuitable for our purposes. The sole exceptions found in [Bla01] are [NMiH 99] already discussed in section 2.5, and a line of Bayesian modelling developed in [CG91] and 49 [CG93] While this latter approach does not explicitly attempt incremental intention recognition, its belief nets are updated incrementally, and there is some promise that it could be ....

Nathan Blaylock. Retroactive recognition of interleaved plans in natural language dialogue, 2001. Area Paper for University of Rochester Department of Computer Science. Unpublished.

No context found.

Nate Blaylock, \Retroactive Recognition of Interleaved Plans for Natural Language Dialogue," Technical Report 761, University of Rochester, Department of Computer Science, December 2001. 46

....recipes, and creating and executing solutions. The details of our collaborative problem solving model can be found in [Allen et al. 2002; Blaylock, 2002] Implementation of the plan and intention recognizer is still a work in progress. Our preliminary plan recognition system is described in [Blaylock, 2001]. We are currently expanding this recognizer to perform intention recognition using our collaborative problem solving model. 6.2 Domain Plans Our collaborative problem solving model is domain independent. It utilizes a domain dependent plan library, which provides the system with knowledge of ....

Nate Blaylock, "Retroactive Recognition of Interleaved Plans for Natural Language Dialogue," Technical Report 761, Department of Computer Science, University of Rochester, December 2001.

....to try to mimic the human plan recognition behavior found in the experiments described above. BELIEVER was given a priori knowledge of the world and about specific plans. It would then be given an observation, from which it would build a plan hypothesis for the actor. This section is based on [Blaylock, 2001]. Other experiments also asked subjects to recall the events or to predict what the next action would be. This particular set of experiments, however, is the most relevant to our discussion. BELIEVER matched incoming observations into a pre built expected plan structure. The expected plan ....

....an intention recognition algorithm for our model (see Section 5 below) and we must be able to adequately deal with these issues. This section describes our implementation of a plan recognition system which partially overcomes the shortcomings of speed, ambiguity and mistaken system inference [Blaylock, 2001]. We will first describe the plan recognition system and then how it addresses each of these issues. In the Section 5.2 we discuss plans to extend this system to handle intention recognition for our collaborative problem solving model. Plan Recognizer Overview In this section we give a basic ....

[Article contains additional citation context not shown here]

Nate Blaylock, "Retroactive Recognition of Interleaved Plans for Natural Language Dialogue," Technical Report 761, University of Rochester, Department of Computer Science, December 2001. 46

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC