Jacques Farré

Computer Aided Correction and Extension of a Syntactic Wide-Coverage Lexicon

The effectiveness of parsers based on manually created resources, namely a grammar and a lexicon, rely mostly on the quality of these resources. Thus, increasing the parser coverage and precision usually implies improving these two resources. Their manual improvement is a time consuming and complex task: identifying which resource is the true culprit for a given mistake is not always obvious, as well as finding the mistake and correcting it. Some techniques, like van Noord (2004) or Sagot and Villemonte de La Clergerie (2006), bring a convenient way to automatically identify forms having potentially erroneous entries in a lexicon. We have integrated and extended such techniques in a wider process which, thanks to the grammar ability to tell how these forms could be used as part of correct parses, is able to propose lexical corrections for the identified entries. We present in this paper an implementation of this process and discuss the main results we have obtained on a syntactic wide-coverage French lexicon.

A Bounded Graph-Connect Construction for LR-regular Parsers

Parser generation tools currently used for computer language analysis rely on user wisdom in order to resolve grammar conflicts. Here practical LR(0)-based parser generation is introduced, with automatic conflict resolution by potentially-unbounded lookahead exploration. The underlying LR(0)-automaton item dependence graph is used for lookahead DFA construction. A bounded graph-connect technique overcomes the difficulties of previous approaches with empty rules, and compact coding allows to precisely resume right-hand contexts. Resulting parsers are deterministic and linear, and accept a large class of LR-regular grammars including LALR(k). Their construction is formally introduced, shown to be decidable, and illustrated by a detailed example.

Shift-Resolve parsing: simple, unbounded lookahead, linear time

This paper introduces a mechanism for combining unbounded lookahead exploration with linear time complexity in a deterministic parser. The idea is to use a resolve parsing action in place of the classical reduce. The construction of shift-resolve parsers is presented as a two-step algorithm, from the grammar to a finite nondeterministic automaton, and from this automaton to the deterministic parser. Grammar classes comparisons are provided.

Bounded-connect noncanonical discriminating-reverse parsers

The precision of right-hand context covering for conflict resolution is improved over previous NDR parsers, resulting in acceptation of wider subsets of LR-regular and LR-nonregular grammars, including all LALR(k) grammars for a given k. Parser generation combines a new form of DR items and subgraph connections of bounded length, without the need to implement subgraph copies. The NDR parser, whose algorithm remains essentially unchanged, is presented as an extended two-stack pushdown automaton. The technique is illustrated with a detailed example.

A Basis for Looping Extensions to Discriminating-Reverse Parsing

We present a noncanonical extension to the Discriminating Reverse parsing method, which accepts non-LR grammars. In cases of parsing conflict, actions are deferred and a mark is virtually pushed onto the parsing stack. Then, locally-canonical DR parsing resumes until sufficient right context is read to resolve the initial conflict. Marks code coverings of the right contexts that are compatible with the actions in conflict. A suboptimal solution for such a coding is proposed, which is computed from the DR automaton itself. The stack vocabulary is enlarged with the mark set, but no new state is added to the basic DR automaton. Moreover, conflict resolution basically uses the DR parser. The method determines at construction time whether all the conflicts can be resolved, and only produces deterministic parsers.

Reducing acyclic cover transducers

Finite languages and finite subsequential functions can be represented by possibly cyclic finite machines, respectively called cover automata and cover transducers. In general, reduced cover machines have much fewer states than the corresponding minimal machines, yielding a compact representation for lexicons or dictionaries. We present here a new algorithm for reducing the number of states of an acyclic transducer.

Reduction Techniques for Acyclic Cover Transducers

Finite languages and finite subsequential functions can be represented by possibly cyclic finite machines, respectively called cover automata and cover transducers. Reduced cover machines can have fewer states than the corresponding minimal machines, yielding a compact representation for lexicons or dictionaries. We present here a new algorithm for reducing the number of states of an acyclic transducer.

Implementation and Application of Automata

Invited Lectures.- Languages Recognizable by Quantum Finite Automata.- The Language, the Expression, and the (Small) Automaton.- Technical Contributions.- Minimization of Non-deterministic Automata with Large Alphabets.- Simulating Two-Dimensional Recognizability by Pushdown and Queue Automata.- Component Composition Preserving Behavioural Contracts Based on Communication Traces.- Strong Retiming Equivalence of Synchronous Schemes.- Prime Normal Form and Equivalence of Simple Grammars.- An Incremental Algorithm for Constructing Minimal Deterministic Finite Cover Automata.- Finite Automata and Unions of Regular Patterns with Bounded Constant Segments.- Inside Vaucanson.- Deterministic Recognition of Trees Accepted by a Linear Pushdown Tree Automaton.- Shorter Regular Expressions from Finite-State Automata.- Wind in the Willows - Generating Music by Means of Tree Transducers.- On Deterministic Catalytic Systems.- Restricting the Use of Auxiliary Symbols for Restarting Automata.- A Class of Rational n-WFSM Auto-intersections.- Experiments with Deterministic ?-Automata for Formulas of Linear Temporal Logic.- Computing Affine Hulls over and from Sets Represented by Number Decision Diagrams.- Tree Automata and XPath on Compressed Trees.- Deeper Connections Between LTL and Alternating Automata.- The Structure of Subword Graphs and Suffix Trees of Fibonacci Words.- Observations on Determinization of Buchi Automata.- The Interval Rank of Monotonic Automata.- Compressing XML Documents Using Recursive Finite State Automata.- Non-backtracking Top-Down Algorithm for Checking Tree Automata Containment.- Size Reduction of Multitape Automata.- Robust Spelling Correction.- On Two-Dimensional Pattern Matching by Finite Automata.- Poster Abstracts.- Incremental and Semi-incremental Construction of Pseudo-Minimal Automata.- Is Learning RFSAs Better Than Learning DFAs?.- Learning Stochastic Finite Automata for Musical Style Recognition.- Simulation of Soliton Circuits.- Acyclic Automata with Easy-to-Find Short Regular Expressions.- On the Equivalence Problem for Programs with Mode Switching.- Automata and AB-Categorial Grammars.- On a Class of Bijective Binary Transducers with Finitary Description Despite Infinite State Set.

Practical Nondeterministic DR(k) Parsing on Graph-Structured Stack

A new approach to parse context-free grammars is presented. It relies on discriminating-reverse, DR(k), parsers, with a Tomita-like nondeterminism-controlling graph-structured stack (GSS) algorithm. The advantage of this generalized discriminating-reverse (GDR) approach over GLR lies on the possibility of using DR(k) parsers, which combine full LR(k) parsing power with a small number of states even for k > 1. This can greatly reduce nondeterminism originating from limited parsing power, as it is typical of the restricted direct LR parsers (SLR, LALR) commonly used in Tomitas algorithm. Furthermore, relying on a DR parser allows a GSS that associates nodes to symbols instead of direct-LR states, and makes easier computation of the shared forest. Moreover, DR(k) parsers have been shown to be linear for LR(k) grammars, and the DR(k) parser efficiency has been practically found to be very similar to direct LR(k) parsers. The paper first presents the nondeterministic DR(k) generation algorithm (for non-LR(k) grammars). Then, it discusses the corresponding adaptation of the GSS algorithm and shows how the shared forest computation is naturally handled.