Integration of speech with natural language understanding.

The integration of speech recognition with natural language understanding raises issues of how to adapt natural language processing to the characteristics of spoken language; how to cope with errorful recognition output, including the use of natural language information to reduce recognition errors; and how to use information from the speech signal, beyond just the sequence of words, as an aid to understanding. This paper reviews current research addressing these questions in the Spoken Language Program sponsored by the Advanced Research Projects Agency (ARPA). I begin by reviewing some of the ways that spontaneous spoken language differs from standard written language and discuss methods of coping with the difficulties of spontaneous speech. I then look at how systems cope with errors in speech recognition and at attempts to use natural language information to reduce recognition errors. Finally, I discuss how prosodic information in the speech signal might be used to improve understanding.

A perspective on early commercial applications of voice-processing technology for telecommunications and aids for the handicapped.

The Colloquium on Human-Machine Communication by Voice highlighted the global technical community's focus on the problems and promise of voice-processing technology, particularly, speech recognition and speech synthesis. Clearly, there are many areas in both the research and development of these technologies that can be advanced significantly. However, it is also true that there are many applications of these technologies that are capable of commercialization now. Early successful commercialization of new technology is vital to ensure continuing interest in its development. This paper addresses efforts to commercialize speech technologies in two markets: telecommunications and aids for the handicapped.

User interfaces for voice applications.

This paper discusses some of the aspects of task requirements, user expectations, and technological capabilities that influence the design of a voice interface and then identifies several components of user interfaces that are particularly critical in successful voice applications. Examples from several applications are provided to demonstrate how these components are used to produce effective voice interfaces.

Mycoplasmas and oncogenesis: persistent infection and multistage malignant transformation.

Oncogenic potential of human mycoplasmas was studied using cultured mouse embryo cells, C3H/10T1/2 (C3H). Mycoplasma fermentans and Mycoplasma penetrans, mycoplasmas found in unusually high frequencies among patients with AIDS, were examined. Instead of acute transformation, a multistage process in promotion and progression of malignant cell transformation with long latency was noted; after 6 passages (1 wk per passage) of persistent infection with M. fermentans, C3H cells exhibited phenotypic changes with malignant characteristics that became progressively more prominent with further prolonged infection. Up to at least the 11th passage, all malignant changes were reversible if mycoplasmas were eradicated by antibiotic treatment. Further persistent infection with the mycoplasmas until 18 passages resulted in an irreversible form of transformation that included the ability to form tumors in animals and high soft agar cloning efficiency. Whereas chromosomal loss and translocational changes in C3H cells infected by either mycoplasma during the reversible stage were not prominent, the onset of the irreversible phase of transformation coincided with such karyotypic alteration. Genetic instability--i.e., prominent chromosomal alteration of permanently transformed cells--was most likely caused by mutation of a gene(s) responsible for fidelity of DNA replication or repair. Once induced, chromosomal alterations continued to accumulate both in cultured cells and in animals without the continued presence of the transforming microbes. Mycoplasma-mediated multistage oncogenesis exhibited here shares many characteristics found in the development of human cancer.

Mouse Col18a1 is expressed in a tissue-specific manner as three alternative variants and is localized in basement membrane zones.

We have isolated overlapping cDNAs encoding the N-terminal non-triple-helical region of mouse alpha 1(XVIII) collagen and shown that three different variants of alpha 1(XVIII) collagen exist. Each of the three variants shows characteristic tissue-specific expression patterns. Immunohistochemical studies show positive staining for alpha 1(XVIII) collagen along the basement membrane zones of vessels in the intestinal villi, the choroid plexus, skin, liver, and kidney. Thus, we conclude that alpha 1(XVIII) collagen may interact (directly or indirectly) with components in basement membrane zones or on the basal surface of endothelial/epithelial cells.

The yeast and mammalian isoforms of phosphatidylinositol transfer protein can all restore phospholipase C-mediated inositol lipid signaling in cytosol-depleted RBL-2H3 and HL-60 cells.

The mammalian phosphatidylinositol transfer proteins (PITP) and the yeast Saccharomyces cerevisiae PITP (SEC14p) that show no sequence homology both catalyze exchange of phosphatidylinositol (PI) between membranes compartments in vitro. In HL-60 cells where the cytosolic proteins are depleted by permeabilization, exogenously added PITPalpha is required to restore G protein-mediated phospholipase Cbeta (PLCbeta) signaling. Recently, a second mammalian PITPbeta form has been described that shows 77% identity to rat PITPalpha. We have examined the ability of the two mammalian PITPs and SEC14p to restore PLC-mediated signaling in cytosol-depleted HL-60 and RBL-2H3 cells. Both PITPalpha and PITPbeta isoforms as well as SEC14p restore G protein-mediated PLCbeta signaling with a similar potency. In RBL-2H3 cells, crosslinking of the IgE receptor by antigen stimulates inositol lipid hydrolysis by tyrosine phosphorylation of PLCgamma1. Permeabilization of RBL cells leads to loss of PLCgamma1 as well as PITP into the extracellular medium and this coincides with loss of antigen-stimulated lipid hydrolysis. Both PLCgamma1 and PITP were required to restore inositol lipid signaling. We conclude that (i) because the PI binding/transfer activities of PITP/SEC14p is the common feature shared by all three transfer proteins, it must be the relevant activity that determines their abilities to restore inositol lipid-mediated signaling and (ii) PITP is a general requirement for inositol lipid hydrolysis regardless of how and which isoform of PLC is activated by the appropriate agonist.