## Balancing Robustness and Efficiency in Unification-augmented Context-Free Parsers for Large Practical Applications (0)

### Cached

### Download Links

Venue: | Robustness in Language and Speech Technology |

Citations: | 26 - 7 self |

### BibTeX

@INPROCEEDINGS{Rosé_balancingrobustness,

author = {Carolyn Penstein Rosé and Alon Lavie},

title = {Balancing Robustness and Efficiency in Unification-augmented Context-Free Parsers for Large Practical Applications},

booktitle = {Robustness in Language and Speech Technology},

year = {},

publisher = {Kluwer Academic Press}

}

### Years of Citing Articles

### OpenURL

### Abstract

Large practical NLP applications require robust analysis components that can effectively handle input that is disfluent or extra-grammatical. The effectiveness and efficiency of any robust parser are a direct function of three main factors: (1) Flexibility: what types of disfluencies and deviations from the grammar can the parser handle?; (2) Search: How does the parser search the space of possible interpretations, and what techniques are applied to prune the search space?; and (3) Parse Selection and Disambiguation: What methods and resources are used to evaluate and rank potential parses and sub-parses, and how does the parser cope with the extreme levels of ambiguity introduced by its flexibility parameters? In this chapter we describe our investigations on how to balance flexibility and efficiency in the context of two different robust parsers - a GLR parser and a left corner Chart parser - both based on a unification-augmented context-free grammar formalism. We demonstrate how the...

### Citations

505 | Lexical-Functional Grammar: A formal system for grammatical representation
- Kaplan, Bresnan
- 1982
(Show Context)
Citation Context ...y 1000 rules with an accompanying lexicon with approximately 3000 lexical entries. 2 The Grammar Formalism Much attention has been given in the last decade to unification based formalisms such as LFG =-=[16]-=-, GPSG [9] and HPSG [27], as declarative theories for describing the structure of natural language. A fundamental primitive of these formalisms is the unification of feature based representations. Sev... |

432 |
Generalized Phrase Structure Grammar
- Gazdar, Klein, et al.
- 1985
(Show Context)
Citation Context ...s with an accompanying lexicon with approximately 3000 lexical entries. 2 The Grammar Formalism Much attention has been given in the last decade to unification based formalisms such as LFG [16], GPSG =-=[9]-=- and HPSG [27], as declarative theories for describing the structure of natural language. A fundamental primitive of these formalisms is the unification of feature based representations. Several recen... |

293 | Partial parsing via finite-state cascades
- Abney
- 1996
(Show Context)
Citation Context ... only those cases that are consistent with the simplifying assumptions represented by those heuristics. More recent approaches to robust parsing focus instead on shallow or partial parsing techniques =-=[36, 1, 3, 37]-=-. Rather than attempting to construct a parse covering an entire ungrammatical sentence, these approaches attempt to construct analyses for maximal contiguous portions of the input. Thus, they address... |

227 |
The Core Language Engine
- Alshawi, editor
- 1992
(Show Context)
Citation Context ...a general CFG framework, that allows the system to incorporate known CFG parsing algorithms as the core of the system's implementation. Two examples of such systems are the Core Language Engine (CLE) =-=[4]-=- and the Alvey Natural Language Tools (ANLT) [8]. The underlying grammar formalism used by both GLR* and LCFlex is a unification-based grammar formalism that was originally developed for the Generaliz... |

192 | Generalized Probabilistic LR Parsing of Natural Language (Corpora) with Uni�cation-Based Grammars
- Briscoe, Carroll
- 1993
(Show Context)
Citation Context ...tly subsumed subset of the portion of the sentence covered by another packed analysis. For ambiguity resolution, LCFlex uses an adapted version of the statistical disambiguation approach discussed in =-=[17, 7]-=-. In addition to the flexibility that was an integral part of the GLR* parser, LCFlex embodies several other types of flexibility. In addition to skipping, it also has the ability to insert, which can... |

119 |
Efficient Parsing for Natural Languages. A Fast Algorithm for Practical Systems
- Tomita
- 1985
(Show Context)
Citation Context ...e forest, as well as the computed feature structure associated with each of these nodes. 3 The GLR* Parser The GLR* parser is a robust extension of the Generalized LR (GLR) parser developed by Tomita =-=[33]-=-. Tomita's GLR parsing algorithm evolved out of the LR parsing techniques that were originally developed for parsing programming languages in the late 1960s and early 1970s [2]. LR parsers parse the i... |

90 | Parsing algorithms and metrics
- Goodman
- 1996
(Show Context)
Citation Context ...erpretation possible from a given input. Researchers have approached the robust parsing problem from a variety of different directions, including symbolic [28, 17, 3, 25, 23, 20, 18, 12], statistical =-=[5, 24, 26, 30, 10, 19, 31]-=-, and connec1 tionist [11, 6, 14, 15]. While statistical and connectionist approaches are inherently robust, and can often be trained automatically from labeled corpora, symbolic parsers are most capa... |

78 | An information extraction core system for real world german text processing
- Neumann, Backofen, et al.
- 1997
(Show Context)
Citation Context ...uencies and can extract the most complete interpretation possible from a given input. Researchers have approached the robust parsing problem from a variety of different directions, including symbolic =-=[28, 17, 3, 25, 23, 20, 18, 12]-=-, statistical [5, 24, 26, 30, 10, 19, 31], and connec1 tionist [11, 6, 14, 15]. While statistical and connectionist approaches are inherently robust, and can often be trained automatically from labele... |

77 | a natural language using mutual information statistics
- Magerman, Marcus
- 1990
(Show Context)
Citation Context ...erpretation possible from a given input. Researchers have approached the robust parsing problem from a variety of different directions, including symbolic [28, 17, 3, 25, 23, 20, 18, 12], statistical =-=[5, 24, 26, 30, 10, 19, 31]-=-, and connec1 tionist [11, 6, 14, 15]. While statistical and connectionist approaches are inherently robust, and can often be trained automatically from labeled corpora, symbolic parsers are most capa... |

74 |
A fully statistical approach to natural language interfaces
- Miller, Stallard, et al.
- 1996
(Show Context)
Citation Context ...erpretation possible from a given input. Researchers have approached the robust parsing problem from a variety of different directions, including symbolic [28, 17, 3, 25, 23, 20, 18, 12], statistical =-=[5, 24, 26, 30, 10, 19, 31]-=-, and connec1 tionist [11, 6, 14, 15]. While statistical and connectionist approaches are inherently robust, and can often be trained automatically from labeled corpora, symbolic parsers are most capa... |

49 | Deterministic left corner parsing - Rosenkrantz, Lewis - 1970 |

48 | Practical unification-based parsing of natural language
- Carroll
- 1993
(Show Context)
Citation Context ...to incorporate known CFG parsing algorithms as the core of the system's implementation. Two examples of such systems are the Core Language Engine (CLE) [4] and the Alvey Natural Language Tools (ANLT) =-=[8]-=-. The underlying grammar formalism used by both GLR* and LCFlex is a unification-based grammar formalism that was originally developed for the Generalized LR Parser/Compiler [35] [34] at the Center fo... |

45 | Recognizing syntactic errors in the writing of second language learners
- Schneider, McCoy
- 1998
(Show Context)
Citation Context ...plications such as language tutors and grammar checkers that must not only arrive at an analysis for an extra-grammatical input sentence, but also determine what it was that made it extra-grammatical =-=[32]-=-. LCFlex can be set to allow specific combinations of insertions. It also has the ability to perform flexible unification. Where complete flexible unification is inefficient, setting aside a small set... |

45 |
An efficient augmented context-free parsing algorithm
- Tomita
- 1987
(Show Context)
Citation Context ...uage Tools (ANLT) [8]. The underlying grammar formalism used by both GLR* and LCFlex is a unification-based grammar formalism that was originally developed for the Generalized LR Parser/Compiler [35] =-=[34]-=- at the Center for Machine Translation at Carnegie Mellon University. The formalism supports grammatical specification in an LFG framework, that consists of context-free grammar rules augmented with f... |

44 | Incremental finite-state parsing
- Aït-Mokhtar, Chanod
- 1997
(Show Context)
Citation Context ...uencies and can extract the most complete interpretation possible from a given input. Researchers have approached the robust parsing problem from a variety of different directions, including symbolic =-=[28, 17, 3, 25, 23, 20, 18, 12]-=-, statistical [5, 24, 26, 30, 10, 19, 31], and connec1 tionist [11, 6, 14, 15]. While statistical and connectionist approaches are inherently robust, and can often be trained automatically from labele... |

36 | An ef�cient implementation of the head-corner parser - Noord - 1997 |

30 | Understanding Spontaneous Speech - Ward |

24 | Robust processing of real-world naturallanguage texts
- Hobbs, Appelt, et al.
- 1992
(Show Context)
Citation Context ...y drive the analysis process. The earliest approaches to robust parsing involved hand-coded grammar specific heuristics for selecting a subset of analyses to extend when each input word was processed =-=[13, 23, 22, 21]-=-. With these approaches, the work of making the parser robust and efficient must be redone by hand for every new grammar developed. The use of hand-coded heuristics also constrains these approaches to... |

22 |
A Grammar Based Robust Parser For Spontaneous Speech
- Lavie
- 1995
(Show Context)
Citation Context ...uencies and can extract the most complete interpretation possible from a given input. Researchers have approached the robust parsing problem from a variety of different directions, including symbolic =-=[28, 17, 3, 25, 23, 20, 18, 12]-=-, statistical [5, 24, 26, 30, 10, 19, 31], and connec1 tionist [11, 6, 14, 15]. While statistical and connectionist approaches are inherently robust, and can often be trained automatically from labele... |

22 |
An Efficient Chart-based Algorithm for Partial-Parsing of Unrestricted Texts
- McDonald
- 1992
(Show Context)
Citation Context ...y drive the analysis process. The earliest approaches to robust parsing involved hand-coded grammar specific heuristics for selecting a subset of analyses to extend when each input word was processed =-=[13, 23, 22, 21]-=-. With these approaches, the work of making the parser robust and efficient must be redone by hand for every new grammar developed. The use of hand-coded heuristics also constrains these approaches to... |

21 | A connectionist architecture for learning to parse
- Henderson, Lane
- 1998
(Show Context)
Citation Context ...ers have approached the robust parsing problem from a variety of different directions, including symbolic [28, 17, 3, 25, 23, 20, 18, 12], statistical [5, 24, 26, 30, 10, 19, 31], and connec1 tionist =-=[11, 6, 14, 15]-=-. While statistical and connectionist approaches are inherently robust, and can often be trained automatically from labeled corpora, symbolic parsers are most capable of performing deep and detailed a... |

21 |
PARSEC: A Connectionist Learning Architecture for Parsing Spoken Language
- Jain
- 1991
(Show Context)
Citation Context ...ers have approached the robust parsing problem from a variety of different directions, including symbolic [28, 17, 3, 25, 23, 20, 18, 12], statistical [5, 24, 26, 30, 10, 19, 31], and connec1 tionist =-=[11, 6, 14, 15]-=-. While statistical and connectionist approaches are inherently robust, and can often be trained automatically from labeled corpora, symbolic parsers are most capable of performing deep and detailed a... |

21 | Adaptive Parsing: Self-Extending Natural Language Interfaces - Lehman - 1992 |

21 | Recent Advances in JANUS: a Speech Translation System
- Woszczyna, Coccaro, et al.
- 1993
(Show Context)
Citation Context ...s and planned future research. All of the evaluations reported in this paper were conducted the English Spontaneous Scheduling Domain (ESST) Grammar developed for the JANUS speech translation project =-=[39, 38]-=-. The grammar is primarily semantic and consists of approximately 1000 rules with an accompanying lexicon with approximately 3000 lexical entries. 2 The Grammar Formalism Much attention has been given... |

20 |
Robust Partial-Parsing Through Incremental, Multi-Algorithm Processing
- McDonald
- 1991
(Show Context)
Citation Context |

20 | Janus 93: towards spontaneous speech translation - Woszczyna - 1994 |

18 | Robust Interactive Dialogue Interpretation
- Rosé
- 1998
(Show Context)
Citation Context |

18 |
The Generalized LR Parser/Compiler - Version 8.4
- Tomita
- 1990
(Show Context)
Citation Context ... Language Tools (ANLT) [8]. The underlying grammar formalism used by both GLR* and LCFlex is a unification-based grammar formalism that was originally developed for the Generalized LR Parser/Compiler =-=[35]-=- [34] at the Center for Machine Translation at Carnegie Mellon University. The formalism supports grammatical specification in an LFG framework, that consists of context-free grammar rules augmented w... |

15 |
Incremental parsing by modular recurrent connectionist networks
- Jain, Waibel
- 1990
(Show Context)
Citation Context ...ers have approached the robust parsing problem from a variety of different directions, including symbolic [28, 17, 3, 25, 23, 20, 18, 12], statistical [5, 24, 26, 30, 10, 19, 31], and connec1 tionist =-=[11, 6, 14, 15]-=-. While statistical and connectionist approaches are inherently robust, and can often be trained automatically from labeled corpora, symbolic parsers are most capable of performing deep and detailed a... |

15 |
Information Based Syntax and Semantics
- Pollard, Sag
- 1987
(Show Context)
Citation Context ...ompanying lexicon with approximately 3000 lexical entries. 2 The Grammar Formalism Much attention has been given in the last decade to unification based formalisms such as LFG [16], GPSG [9] and HPSG =-=[27]-=-, as declarative theories for describing the structure of natural language. A fundamental primitive of these formalisms is the unification of feature based representations. Several recent practical sy... |

14 | Design and Development of Spoken Natural-Language Dialog Parsing Systems
- Hipp
- 1992
(Show Context)
Citation Context |

13 |
Adaptive language acquisition in a multi-sensory device
- Sankar, Gorin
- 1993
(Show Context)
Citation Context |

13 | Parsing real input in JANUS: a concept-based approach - Mayfield, Gavalda, et al. - 1995 |

13 | Concept-based Speech Translation - Mayfield, Ward, et al. |

11 | Fertility models for statistical natural language understanding
- Pietra, Epstein, et al.
- 1997
(Show Context)
Citation Context |

10 | Recovering from parser failures: A hybrid statistical and symbolic approach
- Rosé, Waibel
- 1996
(Show Context)
Citation Context |

10 | A model for robust processing of spontaneous speech by integrating viable fragments
- Worm
- 1998
(Show Context)
Citation Context ... only those cases that are consistent with the simplifying assumptions represented by those heuristics. More recent approaches to robust parsing focus instead on shallow or partial parsing techniques =-=[36, 1, 3, 37]-=-. Rather than attempting to construct a parse covering an entire ungrammatical sentence, these approaches attempt to construct analyses for maximal contiguous portions of the input. Thus, they address... |

9 | Spoken Dialogue Interpretation with the DOP Model
- Bod
- 1998
(Show Context)
Citation Context |

8 |
The Interplay of Syntactic and Semantic Node Labels in Partial Parsing
- McDonald
- 1993
(Show Context)
Citation Context |

8 | An efficient distribution of labor in a two stage robust interpretation process - Rosé, Lavie - 1997 |

6 | FeasPar--A Feature Structure Parser Learning to Parse Spontaneous Speech
- Buo
- 1996
(Show Context)
Citation Context |

5 | Robust parsing with the head-corner parser
- Noord
- 1996
(Show Context)
Citation Context ... only those cases that are consistent with the simplifying assumptions represented by those heuristics. More recent approaches to robust parsing focus instead on shallow or partial parsing techniques =-=[36, 1, 3, 37]-=-. Rather than attempting to construct a parse covering an entire ungrammatical sentence, these approaches attempt to construct analyses for maximal contiguous portions of the input. Thus, they address... |

4 |
Efficiently parsing large corpora
- McDonald
- 1993
(Show Context)
Citation Context ...y drive the analysis process. The earliest approaches to robust parsing involved hand-coded grammar specific heuristics for selecting a subset of analyses to extend when each input word was processed =-=[13, 23, 22, 21]-=-. With these approaches, the work of making the parser robust and efficient must be redone by hand for every new grammar developed. The use of hand-coded heuristics also constrains these approaches to... |

4 |
93: Towards spontaneous speech translation
- JANUS
- 1994
(Show Context)
Citation Context ...s and planned future research. All of the evaluations reported in this paper were conducted the English Spontaneous Scheduling Domain (ESST) Grammar developed for the JANUS speech translation project =-=[39, 38]-=-. The grammar is primarily semantic and consists of approximately 1000 rules with an accompanying lexicon with approximately 3000 lexical entries. 2 The Grammar Formalism Much attention has been given... |

1 | DeterministicTechniques for Efficient Non-deterministic Parsers - Lang - 1974 |