Time Synchronous Chart Parsing of Speech Integrating Uniication Grammars with Statistics

We present an active chart parser which parses left connected wordgraphs in a strictly time synchronous way. The parser performs a beam search on the possible paths through the word graph and on the possible derivations of the uni-cation grammar simultaneously. A metric is given to assign scores to edges, taking into account the whole left context thereby combining acoustic probabilities, n-gram probabilities and uniication grammar probabilities. A specialized model for the derivation of typed uniica-tion grammars is introduced. Diierent ways of coupling the parser with an LR beam decoder in an online time synchronous fashion are de-ned and several experimental results are presented. Two top down and one bottom up method are investigated. In bottom up mode, the decoder sends word hypotheses as they are found from left to right, while the parser keeps step. In verify mode, the decoder is always a frame ahead, while the parser veriies received hypotheses, providing language information to the decoder. In predict mode, the parser is a frame ahead, sending possible successor information to the decoder. The latter uses this information to restrict its search space. Finally a method to maximally reduce multiple paths in a left connected word graph produced by a beam decoder ist presented, which can be used in all of the three strategies.