As software improves our ability to search documents for ideas and concepts, instead of only for text strings, in-house lawyers will not only cope with discovery requests more effectively but will also more effectively be able to find useful material for their practice. KMWorld, Feb. 2011 at S9, has an interesting foreshadowing by an executive at EMC SourceOne of three directions being pursued under the rubric of conceptual search.
- “Word sense disambiguation” capabilities help figure out the actual meanings of words and their fundamental concepts. For example, I think the software uses knowledge of grammar to help.
- “Latent semantic analysis” uses vectorial semantics (“documents and queries are represented as vectors within a linear algebra matrix”). If you look at several related documents and link key words in them to each other as a network, the software makes more sense out of them.
- “Local co-occurrence statistics” is a “technique that counts the number of times pairs of terms appear together (co-occur) within a given text range or document set. Words frequently linked together enrich the software’s “understanding” of their conceptual place.