Once names are substituted by roles, the next step is to ascertain negation and represent information captured in the text structure. The discussion here will focus on ProPs; for an example how negation is relevant for finding Factors, see Section 3.3.
Disclosure by Plaintiff p (p is subject)
p disclose
p show
p send_letter show t
F1 not F1 p explain explain t
F1 not F1
Disclosure of Information t (t is subject, passive mode)
t disclosed_
passive
disclosed_
passive_to d F1 not F1
Defendant d finds out about trade secret t (d is subject)
p give give t not F1
d received
received t d visited visited p d access
access_to t not F1
F1
F1 not F1 F1 not F1
not F1
Disclosure in Course of Business Contacts
not F1
Factor F1
Roles:
p plaintiff d defendant
t trade secret information
p and d get_involved_in
negotiations F1 licensing
F1 not F1 not F1 disclose t
disclose_to d
F1 not F1 not F1
send_letter_to d
not F1 F1
give_to d
not F1 F1
not F1
Fig. 8. Manually generated classification tree for Factor F1, Disclosure-in-Negotiations; arrow to the left indicates the feature is present, arrow to the right indicates the feature is absent.
As mentioned above, in the Forcier squib, the sentence “Plaintiff disclosed informa- tion to defendant. . .” contains evidence for the Factor F1, Disclosure-In-Negotiations.
Another sentence from this squib is “Defendant’s patent disclosed most of plaintiff’s information.” It shows that Factor F20, Info-Known-to-Competitors, applies, but is not related to Factor F1. The relevant difference with respect to Factor F1 between these sentences is not the word “patent,” but who disclosed the information and to whom. For Factor F1 to apply, the information has to be disclosed by plaintiff to defendant. If the sentences are represented by ProPs, as (plaintiff disclosed) (disclosed information) (dis- closed to defendant) and (patent disclosed) (disclosed information), respectively, they can be readily distinguished based on who disclosed the information.
We are currently working on how to derive ProPs automatically. To give an intu- ition as to the goal of this research, we manually generated the type of classifier we hope to learn with SMILE. We first collected the sentences that contain evidence for Factor F1 from CATO’s squibs and replaced all names by roles. Then, we abstracted from the actions, which allowed us to derive ProPs. From this representation, we gen- erated the classification tree in Figure 8. Since this was done by hand, we focussed on the positive examples only and did not use a learning algorithm; instead, we relied on a common-sense covering strategy. The sentence “Forcier disclosed the information to Aha!”, represented as (plaintiff disclosed) (disclosed information) (disclosed to defen-
dant), would be classified correctly as an instance of F1 through the leftmost branch in this tree. Without generalizing from the text by manually emulating the techniques introduced in Section 3, it would be impossible to derive this classification tree for F1.
5 Summary
In this paper, we discussed the limitations of IR-based TCBR approaches and how they can be addressed with a state-of-the-art NLP/IE tool, AutoSlog. Specifically, we identified three tasks where NLP/IE can improve TCBR systems. They are (1) extracting names and factual information, (2) preserving information captured in the syntax of the documents, and (3) ascertaining negation. In particular, we discussed how a new type of feature, called Propositional Patterns, can be derived from syntactic relations among words to capture information like “who did what?” These tasks were motivated with an example. We also drew examples from a variety of TCBR systems to demonstrate how these techniques can help overcome the limitations of a BOW representation.
After presenting the general techniques, we focussed on our own TCBR program SMILE for identifying CATO’s Factors in legal cases. The Factors are abstract fact pat- terns, used to compare cases and reason about the differences between partially matched cases. We are working on integrating the three IE-based techniques above and ML meth- ods for better generalizing from the legal case opinions. Our approach in SMILE consists of two steps. First, SMILE generalizes from the text using NLP/IE methods. We replace the names of the parties and product-related information by their role in the lawsuit, derive ProPs to preserve some of the relevant information in the text, and ascertain nega- tion. Second, SMILE uses an ML algorithm to further generalize from examples to more abstract fact patterns in a classifier.
References
Aha, D.; Maney, T.; and Breslow, L. A. 1998. Supporting dialogue inferencing in conversa- tional case-based reasoning. In Proceedings of the Fourth European Workshop on Case-Based Reasoning. Springer. LNAI 1488.
Aleven, V. 1997. Teaching Case-Based Argumentation through a Model and Examples. Ph.D.
Dissertation, University of Pittsburgh.
Ashley, K., and Lenz, M., eds. 1998. Textual Case-Based Reasoning, Papers from the AAAI-98 Workshop. AAAI Press. Technical Report WS-98-12.
Ashley, K., and McLaren, B. 2001. Helping a CBR Program Know What it Knows. In Proceedings of the Fourth International Conference on Case-Based Reasoning. Springer.
Ashley, K. 1990. Modeling Legal Argument, Reasoning with Cases and Hypotheticals. MIT- Press.
Bellazzi, R.; Montani, S.; Portinale, L.; and Riva, A. 1999. Integrating Case-Based and Rule- Based Decision Making in Diabetic Patient Management. In Proceedings of the Third Interna- tional Conference on Case-Based Reasoning. Springer. LNAI 1650.
Branting, L. 1999. Reasoning with Rules and Precedents - A Computational Model of Legal Analysis. Kluwer Academic Publishers.
Br¨uninghaus, S., and Ashley, K. 1997. Using Machine Learning for Assigning Indices to Textual Cases. In Proceedings of the Second International Conference on Case-Based Reasoning.
Springer. LNAI 1266.
Br¨uninghaus, S., and Ashley, K. 1999. Bootstrapping Case Base Development with Anno- tated Case Summmaries. In Proceedings of the Third International Conference on Case-Based Reasoning. Springer. LNAI 1650.
Burke, R.; Hammond, K.; Kulyukin, V.; Lytinen, S.; Tomuro, N.; and Schonberg, S. 1997.
Question-Answering from Frequently-Asked Question Files: Experiences with the FAQ-Finder System. AI Magazine 18(1):57–66.
Burke, R. 2000. A Case-Based Approach to Collaborative Filtering. In Proceedings of the Fifth European Workshop on Case-Based Reasoning. Springer. LNAI 1898.
Burton, W. 1992. Legal Thesaurus. Simon & Schuster Macmillan.
Daniels, J., and Rissland, E. 1997. Finding Legally Relevant Passages in Case Opinions. In Proceedings of the Sixth International Conference on AI and Law. ACM Press.
Daniels, J. 1997. Retrieval of Passages for Information Reduction. Ph.D. Dissertation, University of Massachusetts, Amherst, MA.
Fellbaum, C., ed. 1998. WordNet: An Electronic Lexical Database. MIT Press.
Jurafsky, D., and Martin, J. 2000. Speech and Language Processing. Prentice Hall.
Leake, D., and Wilson, D. 1999. Combining CBR with Interactive Knowledge Acquisition, Manipulation and Reuse. In Proceedings of the Third International Conference on Case-Based Reasoning. Springer. LNAI 1650.
Leake, D., ed. 1998. Case-Based Reasoning: Experiences, Lessons & Future Directions. MIT Press.
Lenz, M.; H¨ubner, A.; and Kunze, M. 1998. Textual CBR. In Lenz, M.; Bartsch, B.; Burkhard, H.-D.; and Wess, S., eds., Case-Based Resaoning Technology. Springer. LNAI 1400.
Racine, K., and Yang, Q. 1997. Maintaining Unstructured Case Bases. In Proceedings of the Second International Conference on Case-Based Reasoning. Springer. LNAI 1266.
Riloff, E. 1995. Little Words Can Make a Big Difference for Text Classification. In Proceedings of the 18th Annual International ACM SIGIR Conference on Research and Development in Information Retrieval. ACM Press.
Riloff, E. 1996. Automatically Generating Extraction Patterns from Untagged Text. In Proceed- ings of the Thirteenth National Conference on Artificial Intelligence. AAAI Press.
Sonderland, S.; Aronow, D.; Fisher, D.; Aseltine, J.; and Lehnert, W. 1995. Machine Learn- ing of Text Analysis Rules for Clinical Records. Technical Report CIIR TC-39, University of Massachusetts, Amherst, MA.
Strunk, W., and White, E. 1979. The Elements of Style. Macmillan Publishing Co.
Weber, R.; Aha, D.; Sandhu, N.; and Munoz-Avila, H. 2001. A textual case-based reason- ing framework for knowledge management applications. In Proceedings of the Ninth German Workshop on Case-Based Reasoning. Shaker Verlag.
Weber, R. 1998. Intelligent Jurisprudence Research. Ph.D. Dissertation, Federal University of Santa Catarina, Brazil.
D.W. Aha and I. Watson (Eds.): ICCBR 2001, LNAI 2080, pp. 90-104, 2001.
© Springer-Verlag Berlin Heidelberg 2001
Attribute Graph Approach
Edmund K. Burke1, Bart MacCarthy2, Sanja Petrovic1, and Rong Qu1
1School of Computer Science and Information Technology, The University of Nottingham, Nottingham, NG8 1BB, U.K
{ekb, sxp, rxq}@cs.nott.ac.uk http://www.cs.nott.ac.uk/
2School of Mechanical, Materials, Manufacturing Engineering and Management, The University of Nottingham, Nottingham, NG7 2RD, U.K
http://www.nottingham.ac.uk/school4m {Bart.MacCarthy@nottingham.ac.uk}
Abstract. An earlier Case-based Reasoning (CBR) approach developed by the authors for educational course timetabling problems employed structured cases to represent the complex relationships between courses. The retrieval searches for structurally similar cases in the case base. In this paper, the approach is further developed to solve a wider range of problems. We also attempt to retrieve those cases that have common similar structures with some differences. Costs that are assigned to these differences have an input upon the similarity measure. A large number of experiments are performed consisting of different randomly produced timetabling problems and the results presented here strongly indicate that a CBR approach could provide a significant step forward in the development of automated systems to solve difficult timetabling problems. They show that using relatively little effort, we can retrieve these structurally similar cases to provide high quality timetables for new timetabling problems.
1. Introduction