Analyzing the Effectiveness of Pruning and Grouping Methods Used in Literature-Based Discovery Tools

LBD tools enable the establishment of relationships between concepts appearing in scientific articles in the biomedical field and the generation of new hypotheses via the examination of these existing relationships. In this paper, we study the effectiveness of generally accepted grouping and eliminating logics used in LBD tools. This work is performed in the context of Lit2Info, a system that we have developed as an LBD tool. Significant performance variations over that of popular pruning and grouping methods are reported. One of our findings is that there is no positive or negative impact of pruning of closely related terms in links, which is one of the pruning methods commonly applied to decrease the number of irrelevant hypotheses. More importantly, we find that the pruning methods used in this connection actually lead to a decrease in the effectiveness of the hypotheses generation process.