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.

Double replacement gene targeting for the production of a series of mouse strains with different prion protein gene alterations

We have developed a double replacement gene targeting strategy which enables the production of a series of mouse strains bearing different subtle alterations to endogenous genes. This is a two-step process in which a region of the gene of interest is first replaced with a selectable marker to produce an inactivated allele, which is then re-targeted with a second vector to reconstruct the inactivated allele, concomitantly introducing an engineered mutation. Five independent embryonic stem cell lines have been produced bearing different targeted alterations to the prion protein gene, including one which raises the level of expression. We have constructed mice bearing the codon 101 proline to leucine substitution linked to the human familial prion disease, Gerstmann-Straussler-Scheinker syndrome. We anticipate that this procedure will have applications to the study of human inherited diseases and the development of therapies.

Primary structure of the speC gene encoding biosynthetic ornithine decarboxylase in Escherichia coli

A 2.91-kb fragment of the Escherichia coli chromosome containing the speC gene, encoding biosynthetic ornithine decarboxylase (ODC) was sequenced. The speC gene is encoded by a 2133-bp ORF; the deduced amino-acid sequence contains 711 residues whose predicted molecular mass is 79,505 Da.

Progesterone and EGF inhibit mouse mammary gland prolactin receptor and beta-casein gene expression

Regulation of mouse mammary gland long-form prolactin receptor (PRL-RL) mRNA levels by progesterone and epidermal growth factor (EGF) and the relationship between PRL-RL and beta-casein gene expression were examined in vivo and in vitro. PRL-RL and beta-casein mRNA levels increased approximately 6- and 15-fold from the pregnant to the lactating period, respectively, when normalized to the level of beta-actin mRNA. Ovariectomy of pregnant mice rapidly reduced the serum concentration of progesterone and increased the level of PRL-RL and beta-casein mRNAs approximately three- and fourfold compared with sham-operated animals 24 h after the operation. Injection of progesterone, but not estrogen, inhibited the increase in both mRNA levels. PRL-RL and beta-casein mRNA levels in cultured mammary epithelium increased in response to insulin, hydrocortisone, and prolactin, whereas progesterone or EGF caused inhibition. The combination of EGF and progesterone produced a greater inhibition than either hormone alone. These results indicate that both progesterone and EGF serve as negative regulators of lactogenesis.

Cloning and sequencing of the cDNA encoding the murine mammary gland long-form prolactin receptor

The nucleotide sequence of a 1992-bp cDNA encoding the long form of the murine mammary prolactin receptor (PRL-R) has been determined. The deduced 68-kDa protein has high sequence identity with long forms of prolactin receptors from rat ovary and rabbit mammary gland.

Chromosomal aberration types in cells at the second division after irradiation in G1 or G2

Cultures of JU-56 cells were irradiated in either G1 or G2 and examined either in their first post-irradiation metaphase (in diploids) or in their second post-irradiation metaphase (in colcemid-induced tetraploids). The timing of fixation, together with tritiated thymidine pulse labelling, allowed selection for scoring only metaphases of cells that were in the G1 or G2 phase of the cell cycle during irradiation. With G1 irradiation it was found that many of the aberrations observed at the first division were not present (as derivatives) at the second division, and also that new aberrations were found at the second division, which were not derived from aberrations at the first division. Clonogenic survival was also measured in populations of cells irradiated in G1. It was found that cells containing chromosomal aberrations at the first division were not numerous enough to explain lack of survival. When frequencies of aberrations following G2 irradiation scored at the second division were compared with those scored at the first, there was a significant increase in dicentrics as compared with their progenitor asymmetrical chromatid interchanges, and of mirror-image dicentrics as compared with their progenitor sister unions. A substantial number of sister unions were also observed at the second division. We conclude that some aberrations are lost during the interphase between the first and the second post-irradiation metaphase and that new chromosomal aberrations arise during the second post-irradiation interphase.

Time sequence of events leading to chromosomal aberration formation

Investigations have been carried out which have measured the influence of the repair polymerases on the yield of different types of chromosomal aberrations. The studies were mainly concerned with the effect of inhibiting the polymerases on the yield of aberrations. The polymerases fill in single strand regions, and the fact that their inhibition affects the yield of aberrations suggests that single strand lesions are influential in aberration formation. The results indicate that-- 1. There are two actions of polymerases in clastogenesis. One is in their involvement in a G2 repair system, in which the pair of chromatids is concerned, and which does not yield aberrations unless the inhibition is still operating when the cells enter mitosis. The second also operates in G1 and S, and is such that when repair is inhibited, further damage accrues. 2. The second action is affected by inhibiting polymerase but operates even when the repair enzymes are active. 3. The production of chromosomal exchanges involves a series of reactions, some of which are reversible. 4. The time span over which the reactions occur is much longer than has been envisaged previously (e.g., most of a cell cycle).

The influence of growth medium on the yield of X-ray-induced chromatid exchanges in the presence and absence of aphidicolin

The frequency of exchanges in JU56 cells irradiated in the G2 phase in the presence and absence of the polymerase inhibitor aphidicolin (APC), and in the presence of a range of concentrations of cysteine was measured. It was found that in the absence of cysteine, incubation for 2 h with APC had no effect on the yield. Addition of cysteine at concentrations of 50-250 mg/l reduced the frequency of exchanges, and at these concentrations the frequency was increased by incubation with APC. At higher concentrations, the yield was reduced and incubation with APC did not elevate it. In following experiments, it was found that incubation with cysteine for a period of longer than 10 minutes was necessary before APC affected the yield of exchanges.

DNA polymerase delta mediates increase in exchange production by X-radiation in human lymphocytes moving from G0 to G1

Earlier work of several laboratories established that the yields of radiation-induced ring and dicentric chromosomes are greater when human peripheral blood lymphocytes are irradiated in GH1 some hours after phytohemagglutinin stimulation than if they are irradiated in G0 before stimulation. Post-treatment of lymphocytes irradiated in G0 with the DNA polymerase inhibitor aphidicolin, which is effective against both pol alpha and pol delta, produces a similar increase in ring and dicentric yield. We found that aphidicolin post-treatment was much less effective in increasing ring and dicentric yield increases in cells irradiated in G1 four to five hours after stimulation. Because we had earlier found specific inhibitors of DNA pol alpha ineffective in producing increased yields in either G0 or G1 lymphocytes, we conclude that much of the G0 to G1 increase in yields is mediated by pol delta.

Cyclic AMP inhibits and putrescine represses expression of the speA gene encoding biosynthetic arginine decarboxylase in Escherichia coli

The speA gene of Escherichia coli encodes biosynthetic arginine decarboxylase (ADC), the first of two enzymes in a putrescine biosynthetic pathway. The activity of ADC is negatively regulated by mechanisms requiring cyclic AMP (cAMP) and cAMP receptor protein (CRP) or putrescine. A 2.1-kb BamHI fragment containing the speA-metK intergenic region, speA promoter, and 1,389 bp of the 5' end of the speA coding sequence was used to construct transcriptional and translational speA-lacZ fusion plasmids. A single copy of either type of speA-lacZ fusion was transferred into the chromosomes of Escherichia coli KC14-1, CB806, and MC4100, using bacteriophage lambda. The speA gene in lysogenized strains remained intact and served as a control. Addition of 5 mM cAMP to lysogenic strains resulted in 10 to 37% inhibition of ADC activity, depending on the strain used. In contrast, the addition of 5 or 10 mM cAMP to these strains did not inhibit the activity of beta-galactosidase (i.e., ADC::beta-galactosidase). Addition of 10 mM putrescine to lysogenized strains resulted in 24 to 31% repression of ADC activity and 41 to 47% repression of beta-galactosidase activity. E. coli strains grown in 5 mM cAMP and 10 mM putrescine produced 46 to 61% less ADC activity and 41 to 52% less beta-galactosidase activity. cAMP (0.1 to 10 mM) did not inhibit ADC activity assayed in vitro. The effects of cAMP and putrescine on ADC activity were additive, indicating the use of independent regulatory mechanisms. These results show that cAMP acts indirectly to inhibit ADC activity and that putrescine causes repression of speA transcription.

Plateletpheresis with the COBE spectra single needle access option

A group of modifications, including a reservoir bag in the return circuit, has been devised to allow single needle plateletpheresis with the COBE Spectra. We compared the number and quality of platelets collected from 10 subjects in paired donations with single and dual needle protocols. There was no evidence of hemolysis with either protocol. Mean (+/- SD) platelet yields were 3.61 +/- 1.47 x 10(11) with two needles and 3.31 +/- 1.31 x 10(11) with the single needle procedure (P = .13). Mean leukocyte levels (standard manual counting chamber) were 1.0 +/- 1.7 x 10(7) and 1.2 +/- 1.0 x 10(7), respectively (p = .78). pH values during storage were acceptable in both groups of concentrates, and there were no significant differences between the single needle and dual needle concentrates in morphology scores or beta-thromboglobulin levels at 0, 1, 3, or 5 days of storage. Thus the single needle modification produced platelets that were comparable in quantity and quality to those from the standard dual needle plateletpheresis protocol.

Nucleotide sequence and analysis of the speA gene encoding biosynthetic arginine decarboxylase in Escherichia coli

The DNA sequence of a 3.23-kilobase fragment of the Escherichia coli chromosome encoding biosynthetic arginine decarboxylase (ADC) was determined. This sequence contained the speA open reading frame (ORF) as well as partial speB and metK ORFs. The ADC ORF is 1,974 nucleotides long; the deduced polypeptide contains 658 amino acids with a molecular size of 73,980 daltons. The molecular weight and predicted ADC amino acid composition are nearly identical to the amino acid analysis of purified ADC performed by Wu and Morris (J. Biol. Chem. 248:1687-1695, 1973). A translational speA-lacZ fusion, pRM65, including 1,389 base pairs (463 amino acids) of the 5' end of speA was constructed. Western blots (immunoblots) with beta-galactosidase antisera revealed two ADC::beta-galactosidase fusion proteins in E. coli bearing pRM65: 160,000 and 156,000 daltons representing precursor and mature hybrid proteins, respectively. The predicted amino acid sequence of ADC contains a region of six amino acid residues found in two bacterial diaminopimelic acid decarboxylases and three eucaryotic ornithine decarboxylases. This conserved sequence is located approximately eight amino acids from the putative pyridoxal phosphate-binding site of ADC and is predicted to be involved in substrate binding.

DNA polymerase alpha does not mediate G0-G1 increase in yield of X-ray-induced exchange aberrations in human peripheral blood lymphocytes

We report experiments to test the hypothesis that the increased yield of dicentric chromosomes observed in human peripheral blood lymphocytes treated with X-rays during the G1 phase of their first cell cycle, as compared with the yield when the cells are treated in their G0 phase prior to phytohemagglutinin stimulation, is a manifestation of the recently-reported conversion of an inactive form of DNA polymerase alpha to its active form as the PHA-stimulated cells pass from G0 into G1 (Sylvia et al., 1988). The specific polymerase alpha inhibitor butylphenyl deoxyguanosine was used as an X-ray post-treatment. The results show that polymerase alpha is not involved.

Thyrotropin alters the utilization of thyroglobulin's hormonogenic sites

We injected rabbits and guinea pigs with bovine thyrotropin (TSH) daily for 3 days, while controls received saline. All animals received sodium [125I]iodide on the second day, and thyroglobulin was purified from the thyroids of each group by gel filtration. Hormonogenic tryptic peptides from each S-cyanoethylated thyroglobulin preparation were isolated by high performance liquid chromatography, and their amino acid sequences were determined, permitting their localization within the thyroglobulin polypeptide chain by comparison with cDNA-derived sequences from bovine and human thyroglobulins. Thyroglobulins from the saline-injected rabbits and guinea pigs contained the same four major hormonogenic sites, designated A-D, previously described (Dunn, J. T., Anderson, P. C., Fox, J. W., Fassler, C. A., Dunn, A. D., Hite, L. A., and Moore, R. C. (1987) J. Biol. Chem. 262, 16948-16952). In both species, sites A and C were the major loci for thyroxine and triiodothyronine, respectively. However, site D in the guinea pig had a greater ratio of [125I]thyroxine to [127I]thyroxine than did site A, whereas the reverse was true in the rabbit. TSH administration produced the following changes in thyroglobulins of both species, relative to controls: 1) an increase in the ratio of [125I]triiodothyronine to [125I] thyroxine (rabbit, 0.29 versus 0.17; guinea pig, 0.19 versus 0.08), with the increase in triiodothyronine principally at site C; 2) a marked increase in 125I/127I and in thyroxine formation at site D (14.1% of thyroglobulin's thyroxine versus 9.8% in rabbits, 24 versus 13% in guinea pigs); 3) a corresponding decrease in thyroxine formation at site A (33 versus 43% in rabbits, 30 versus 46% in guinea pigs); and 4) a sharp increase in conversion of thyroglobulin's N-terminal 125I-labeled approximately 20 kDa hormone-rich iodopeptide, which contains site A, to a 125I-labeled approximately 15-kDa (rabbit) or 125I-labeled approximately 13-kDa (guinea pig) form, reflecting probable peptide bond cleavage. Our results show that TSH alters both the structure of the thyroglobulin molecule and the priority of utilization of its hormonogenic sites. We conclude that these changes are important to TSH's enhancement of thyroid hormone synthesis.

An investigation using inhibition of G2 repair of the molecular basis of lesions which result in chromosomal aberrations

Cultures of JU56 cells were irradiated with 2.5 Gy X-rays and 16 h later the cultures were exposed to a moderately inhibitory dose of 1-beta-D-arabinofuranosylcytosine (ara-C) or aphidicolin (APC) and to colcemid, for 2 h. The c-metaphases collected for examination had therefore been exposed to X-rays in G1 or early S, and to the repair inhibitors APC and ara-C during the latter half of G2. It was found that treatment of cells irradiated early in cell cycle, that is, in G1 and early S, with APC or ara-C in G2, (1) reduced the frequency of chromatid and chromosome exchanges below that of cells treated with X-rays alone, (2) produced no more chromatid breaks and gaps than were seen in unirradiated cells, (3) increased the number of chromosome fragments and gaps in a more than additive fashion, and (4) produced only an additive effect, by comparison with the effect of X-rays and drug given separately, on the total number of chromosomal aberrations.