Incorporating a User Model into an Information Theoretic Framework for Argument Interpretation

Abstract

Abstract.We describe an argument-interpretation mechanism based on the Minimum Message Length Principle [1], and investigate the incorporation of a model of the user's beliefs into this mechanism. Our system receives as input an argument entered through a web interface, and produces an interpretation in terms of its underlying knowledge representation -a Bayesian network. This interpretation may differ from the user's argument in its structure and in its beliefs in the argument propositions. The results of our evaluation are encouraging, with the system generally producing plausible interpretations of users' arguments.

Building user argumentative models

“…Zukerman et al [18] incorporate a user model in the process of argument evaluation, that is arguments that a user receives from other ones; but this work excludes the argument generation. The user model is represented by a Bayesian network.…”

“…Next, we successively perform the same action, but with each ancestor (steps [17][18][19][20], and create a new node in the taxonomy that represents the item, whose parent are the ancestors generated previously (step 21). Following the example, the new ancestor of anc 1 is has_goal(user(U ), goal(discuss_topic(T ))) (the same for anc 2 ); and finally, we replace goal(discuss_topic(T )) with goal(G) and obtain the most general expression of the initial condition.…”

A Probabilistic Approach for Argument Interpretation

“…15 Our previous experience shows that when users interact freely with the system, their arguments may not test behaviours of interest, and their assessment of the interpretations may be influenced by their experience with the web interface (Zukerman et al, 2003a). …”

“…Each BN in the system can support a variety of scenarios, depending on the instantiation of the evidence nodes. The murder mystery used for this paper is represented by means of a 32-node BN, which is a less detailed version of the 85-node BN used in (Zukerman, 2001;Zukerman et al, 2003a). 4 Figure 6 shows our 32-node BN: the observable evidence nodes are boxed, and the goal node [GreenMurderedBody] is circled.…”

“…We use the following formula to express the salience of a node in terms of its access type and its frequency and recency (Zukerman et al, 2003a). This formula yields the access score of a node, assigning a high score to nodes that were recently accepted or mentioned by a user.…”