Figure 2 - uploaded by Richard Crouch
Content may be subject to copyright.
Source publication
Type checking defines and constrains system output and intermediate representations. We report on the advantages of introducing multi- ple levels of type checking in deep parsing sys- tems, even with untyped formalisms.
Citations
... To anchor the discussion, we focus on regression testsuites designed for a grammar-based questionanswering system (Bobrow et al., 2007). The Bridge system uses the XLE (Crouch et al., 2008) parser to produce syntactic structures and then the XLE ordered rewrite system to produce linguistic semantics (Crouch and King, 2006) and abstract knowledge representations. Abstract knowledge representations for passages and queries are processed by an entailment and contradiction detection system which determines whether the query is entailed, contradicted, or neither by the passage. ...
In complex grammar-based systems, even small changes may have an unforeseeable impact on overall system performance. Re-gression testing of the system and its com-ponents becomes crucial for the grammar engineers developing the system. As part of this regression testing, the testsuites themselves must be designed to accurately assess coverage and progress and to help rapidly identify problems. We describe a system of passage-query pairs divided into three types of phenomenon-based test-suites (sanity, query, basic correct). These allow for rapid development and for spe-cific coverage assessment. In addition, real-world testsuites allow for overall per-formance and coverage assessment. These testsuites are used in conjunction with the more traditional representation-based re-gression testsuites used by grammar engi-neers.
In this paper we present a method for greatly reducing parse times in LFG parsing, while at the same time maintaining parse accuracy. We evaluate the methodology on data from English, German and Norwegian and show that the same patterns hold across languages. We achieve a speedup of 67% on the English data and 49% on the German data. On a small amount of data for Norwegian, we achieve a speedup of 40%, although with more training data we expect this figure to increase.
In this paper we illustrate and underline the importance of making detailed linguistic information a central part of the process of automatic acquisition of large-scale lexicons as a means for enhancing robustness and at the same time ensuring maintainability and re-usability of deep lexicalised grammars. Using the error mining techniques proposed in (van Noord, 2004) we show very convincingly that the main hindrance to portability of deep lexicalised grammars to domains other than the ones originally developed in, as well as to robustness of systems using such grammars is low lexical coverage. To this effect, we develop linguistically-driven methods that use detailed morphosyntactic information to automatically enhance the performance of deep lexicalised grammars maintaining at the same time their usually already achieved high linguistic quality.
