Lexicon development for speech and language processing (Q1583044)

This volume originated from the fifth European Summer School on Language and Speech Communication, held in Summer of 1997 in Leuven. It responds to the challenge to build natural language and speech processing systems with access to large lexical resources in order to achieve a reasonable degree of coverage and accuracy. The book contains nine contributions and starts with an overview of the field of computational lexicography and lexicology. Two articles focus on the structure and the content of manually created lexica, one concentrating on morphosyntactic and syntactic information, the other focusing on morphophonological and phonological information. Both contributions adopt a declarative constraint-based specification methodology (using HPSG and DATR, respectively) and pay ample attention to the various ways in which lexical generalizations (e.g., by way of inheritance) can be formalized and exploited to enhance the consistency and to reduce the redundancy of lexicons. A complementary perspective is taken in two papers which present techniques for automatically deriving lexical resources from text corpora. The first one provides an overview of inductive data-oriented machine learning methods (as applied to learning pronunciation and gender information), while the other one discusses the use of finite-state automata for recognizing lexical patterns in text, i.e., tokenization, lemmatization, part-of-speech tagging, and shallow, robust parsing of corpora for automated lexicon acquisition. Three articles deal with central issues of providing lexica for spoken language processing applications. One describes the organization and technical prerequisites of speech databases, another focuses on the concrete use of lexica in text-to-speech synthesis systems, and the third concentrates on the particular needs of speech recognition systems dealing with different types of speech (from read to spontaneous) and with different languages. In the final paper, a psycholinguistic perspective is taken on a computational model for morphological processing in the human mental lexicon for visual word recognition.