Text Structuring and Categorization When Summarizing Legal Cases

Conclusions

An initial text categorization and structuring is useful for many purposes of text analysis including automatic abstracting. The recognition of the text category, and of relevant and insignificant text components is an important first step when intellectually abstracting. Automating this process was especially useful for controlling the overload of present and future court decisions.

Knowledge of the discourse structures of criminal cases proved to be very helpful in automatically extracting relevant information from the cases and in automatically abstracting them. The patterns of the discourse involved in an automatic categorization and structuring of the legal cases are complex, but the number of patterns is limited, hence our choice of a manually constructed knowledge base for analyzing the cases. A powerful formalism for representing the knowledge is needed. It has been shown that a representation as a text grammar is very promising. Our research is a step towards generic representations of knowledge about discourse patterns.

However, in spite of the potential of the knowledge of discourse patterns in text analysis, intertextual analysis of the constitution of texts in terms of types and genres is underdeveloped in the legal field.