<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-4020">
  <Title>The Second Release of the RASP System</Title>
  <Section position="4" start_page="0" end_page="78" type="intro">
    <SectionTitle>
2 Components of the System
</SectionTitle>
    <Paragraph position="0"> RASP is implemented as a series of modules written in C and Common Lisp, which are pipelined, working as a series of Unix-style filters. RASP runs on Unix and is compatible with most C compilers and Common Lisp implementations. The public release includes Lisp and C executables for common 32- and 64-bit architectures, shell scripts for running and parameterising the system, documentation, and so forth. An overview of the system is given in Figure 1.</Paragraph>
    <Section position="1" start_page="0" end_page="77" type="sub_section">
      <SectionTitle>
2.1 Sentence Boundary Detection and
Tokenisation
</SectionTitle>
      <Paragraph position="0"> The system is designed to take unannotated text or transcribed (and punctuated) speech as input, and not simply to run on pre-tokenised input such as that typically found in corpora produced for NLP purposes. Sentence boundary detection and tokenisation modules, implemented as a set of deterministic finite-state rules in Flex (an open source re-implementation of the original Unix Lex utility)  and compiled into C, convert raw ASCII (or Unicode in UTF-8) data into a sequence of sentences in which, for example punctuation tokens are separated from words by spaces, and so forth.</Paragraph>
      <Paragraph position="1"> Since the first release this part of the system has been incrementally improved to deal with a greater variety of text types, and handle quotation appropriately. Users are able to modify the rules used and recompile the modules. All RASP modules now accept XML mark up (with certain hard-coded assumptions) so that data can be pre-annotated--for example to identify named entities--before being passed to the tokeniser, allowing for more domain-dependent, potentially multiword tokenisation and classification prior to parsing if desired (e.g. Vlachos et al., 2006), as well as, for example, handling of text with sentence boundaries already determined.</Paragraph>
    </Section>
    <Section position="2" start_page="77" end_page="77" type="sub_section">
      <SectionTitle>
2.2 PoS and Punctuation Tagging
</SectionTitle>
      <Paragraph position="0"> The tokenised text is tagged with one of 150 part-of-speech (PoS) and punctuation labels (derived from the CLAWS tagset). This is done using a first-order ('bigram') hidden markov model (HMM) tagger implemented in C (Elworthy, 1994) and trained on the manually-corrected tagged versions of the Susanne, LOB and (subset of) BNC corpora. The tagger has been augmented with an unknown word model which performs well under most circumstances. However, known but rare words often caused problems as tags for all realisations were rarely present. A series of manually developed rules has been semi-automatically applied to the lexicon to ameliorate this problem by adding further tags with low counts to rare words. The new tagger has an accuracy of just over 97% on the DepBank part of section 23 of the Wall Street Journal, suggesting that this modification has resulted in competitive performance on out-of-domain newspaper text. The tagger implements the Forward-Backward algorithm as well as the Viterbi algorithm, so users can opt for tag thresholding rather than forced-choice tagging (giving &gt;99% tag recall on DepBank, at some cost to overall system speed). Recent experiments suggest that this can lead to a small gain in parse accuracy as well as coverage (Watson, 2006).</Paragraph>
    </Section>
    <Section position="3" start_page="77" end_page="77" type="sub_section">
      <SectionTitle>
2.3 Morphological Analysis
</SectionTitle>
      <Paragraph position="0"> The morphological analyser is also implemented in Flex, with about 1400 finite-state rules incorporating a great deal of lexically exceptional data. These rules are compiled into an efficient C program encoding a deterministic finite state transducer. The analyser takes a word form and CLAWS tag and returns a lemma plus any inflectional affixes. The type and token error rate of the current system is less than 0.07% (Minnen, Carroll and Pearce, 2001). The primary system-internal value of morphological analysis is to enable later modules to use lexical information associated with lemmas, and to facilitate further acquisition of such information from lemmas in parses.</Paragraph>
    </Section>
    <Section position="4" start_page="77" end_page="78" type="sub_section">
      <SectionTitle>
2.4 PoS and Punctuation Sequence Parsing
</SectionTitle>
      <Paragraph position="0"> The manually-developed wide-coverage tag sequence grammar utilised in this version of the parser consists of 689 unification-based phrase structure rules (up from 400 in the first release).</Paragraph>
      <Paragraph position="1"> The preterminals to this grammar are the PoS and punctuation tags2. The terminals are featural descriptions of the preterminals, and the non-terminals project information up the tree using an X-bar scheme with 41 attributes with a maximum of 33 atomic values. Many of the original  rules have been replaced with multiple more specific variants to increase precision. In addition, coverage has been extended in various ways, notably to cover quotation and word order permutations associated with direct and indirect quotation, as is common in newspaper text. All rules now have a rule-to-rule declarative specification of the grammatical relations they license (seex2.6). Finally, around 20% of the rules have been manually identified as 'marked' in some way; this can be exploited in customisation and in parse ranking.</Paragraph>
      <Paragraph position="2"> Users can specify that certain rules should not be used and so to some extent tune the parser to different genres without the need for retraining.</Paragraph>
      <Paragraph position="3"> The current version of the grammar finds at least one parse rooted in S for about 85% of the Susanne corpus (used for grammar development), and most of the remainder consists of phrasal fragments marked as independent text sentences in passages of dialogue. The coverage of our WSJ test data is 84%. In cases where there is no parse rooted in S, the parser returns a connected sequence of partial parses covering the input. The criteria are partial parse probability and a preference for longer but non-lexical combinations (Kiefer et al., 1999).</Paragraph>
    </Section>
    <Section position="5" start_page="78" end_page="78" type="sub_section">
      <SectionTitle>
2.5 Generalised LR Parser
</SectionTitle>
      <Paragraph position="0"> A non-deterministic LALR(1) table is constructed automatically from a CF 'backbone' compiled from the feature-based grammar. The parser builds a packed parse forest using this table to guide the actions it performs. Probabilities are associated with subanalyses in the forest via those associated with specific actions in cells of the LR table (Inui et al., 1997). The n-best (i.e. most probable) parses can be efficiently extracted by unpacking subanalyses, following pointers to contained subanalyses and choosing alternatives in order of probabilistic ranking. This process backtracks occasionally since unifications are required during the unpacking process and they occasionally fail (see Oepen and Carroll, 2000).</Paragraph>
      <Paragraph position="1"> The probabilities of actions in the LR table are computed using bootstrapping methods which utilise an unlabelled bracketing of the Susanne Treebank (Watson et al., 2006). This makes the system more easily retrainable after changes in the grammar and opens up the possibility of quicker tuning to in-domain data. In addition, the structural ranking induced by the parser can be re-ranked using (in-domain) lexical data which provides conditional probability distributions for the SUBCATegorisation attributes of the major lexical categories. Some generic data is supplied for common verbs, but this can be augmented by user supplied, possibly domain specific files.</Paragraph>
    </Section>
    <Section position="6" start_page="78" end_page="78" type="sub_section">
      <SectionTitle>
2.6 Grammatical Relations Output
</SectionTitle>
      <Paragraph position="0"> The resulting set of ranked parses can be displayed, or passed on for further processing, in a variety of formats which retain varying degrees of information from the full derivations. We originally proposed transforming derivation trees into a set of named grammatical relations (GRs), illustrated as a subsumption hierarchy in Figure 2, as a way of facilitating cross-system evaluation.</Paragraph>
      <Paragraph position="1"> The revised GR scheme captures those aspects of predicate-argument structure that the system is able to recover and is the most stable and grammar independent representation available. Revisions include a treatment of coordination in which the coordinator is the head in subsuming relations to enable appropriate semantic inferences, and addition of a text adjunct (punctuation) relation to the scheme.</Paragraph>
      <Paragraph position="2"> Factoring rooted, directed graphs of GRs into a set of bilexical dependencies makes it possible to compute the transderivational support for a particular relation and thus compute a weighting which takes account both of the probability of derivations yielding a specific relation and of the proportion of such derivations in the forest produced by the parser. A weighted set of GRs from the parse forest is now computed efficiently using a variant of the inside-outside algorithm (Watson et al., 2005).</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>