The role of frame-based representation in reasoning
Communications of the ACM
An artificial intelligence approach to legal reasoning
An artificial intelligence approach to legal reasoning
Efficient string matching: an aid to bibliographic search
Communications of the ACM
| Hi-index | 0.00 |
We are attempting to extract information automatically from large legal documents. SPADES is an expert system for screening pension plans submitted to the Internal Revenue Service (IRS), a task which has resisted prior automation attempts. Nearly all pension plans are prepared by plan preparation specialists. Most of a pension plan document consists of boilerplate text, which is reused by the preparer in nearly every plan. We describe techniques used for dealing with plans which essentially follow the boilerplate model for a particular preparer, but contain significant deviations.A significant deviation is found to be any extra paragraph in a new plan which was not predicted by the boilerplate model. Other deviations from the boilerplate model can be handled fairly easily. Treatment of a significant deviation is affected by whether the topic of the extra paragraph can be identified. When it can, the logical impact of the extra text may be deduced by the system, or the system may guide an IRS Agent in analyzing the extra paragraph.