Extracting ontological relationships (e.g., ISA and HASA) from free-text repositories (e.g., engineering documents and instruction manuals) can improve users' queries, as well as benefit applications built for these domains.

Natural language text corpora are often available as sets of syntactically parsed trees. A wide range of expressive tree queries are possible over such parsed trees that open a new avenue in searching over natural language text. They not only allow for querying roles and relationships within sentences, but also improve search effectiveness compared to flat keyword queries. One major drawback of current systems supporting querying over parsed text is the performance of evaluating queries over large data. In this paper we propose a novel indexing scheme over unique subtrees as index keys. We also propose a novel root-split coding scheme that stores subtree structural information only partially, thus reducing index size and improving querying performance. Our extensive set of experiments show that root-split coding reduces the index size of any interval coding which stores individual node numbers by a factor of 50% to 80%, depending on the sizes of subtrees indexed. Moreover, We show that our index using root-split coding, outperforms previous approaches by at least an order of magnitude in terms of the response time of queries.

Many existing indexes on text work at the document granularity and are not effective in answering the class of queries where the desired answer is only a term or a phrase. In this paper, we study some of the index structures that are capable of answering the class of queries referred to here as wild card queries and perform an analysis of their performance. Our experimental results on a large class of queries from different sources (including query logs and parse trees) and with various datasets reveal some of the performance barriers of these indexes. We then present Word Permuterm Index (WPI) which is an adaptation of the permuterm index for natural language text applications and show that this index supports a wide range of wild card queries, is quick to construct and is highly scalable. Our experimental resultS comparing WPI to alternative methods on a wide range oF wild card queries show a few orders of magnitude performancE improvements for WPI while the memory usage is kept the same for all compared systems.