A connectionist approach to conceptual information retrieval
By Richard K. Belew. 1987
This report proposes that recent advances using low-level connectionist representations offer new possibilities to those interested in free text information retrieval (IR). The AIR system demonstrates that this representation suits the IR domain well, particularly the special problems attending the more sophisticated forms of conceptual retrieval required in legal applications. Also, the natural way in which connectionist representations allow learning means that AIR can avoid the high costs associated with manual indexing while providing comparable results. The paper begins by motivating the importance of legal information retrieval, from the perspectives of both the Law and artificial intelligence (AI). Our approach is then compared to traditional methods for IR, and to more recent work using higher-level symbolic representations from AL After a brief introduction to connectionist representations in general, the AIR system is presented. The paper closes with evidence that this system does, in fact, begin to support the use of those “open textured” concepts that make the Law both a very difficult and a very illuminating domain for AI research.
Subfields: Information Retrieval & Search
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Information Retrieval & Search
- Jimmy J. Lin
- Rodrigo Nogueira
- Andrew Yates
- Zeynep Akkalyoncu Yilmaz
- Sebastian Hofstätter
- Allan Hanbury
- Zhuyun Dai
- Wei Yang
Citations in the corpus
- Pretrained Transformers for Text Ranking: BERT and Beyond (2020, methodology): “Like applications of information retrieval techniques to the medical and biomedical domain, the first search systems for the legal texts date to at least the 1960s [Robins, 1967], with retrieval techniques based on neural networks appearing as early as the late 1980s [Belew, 1987].”