Radiative and predissociative lifetimes of the V′=0 level of the A2Σ+ state of SH and SD from chemical and spectroscopic studies Radiative and predissociative lifetimes of the V′=0, 1, and 2 levels of the A 2Σ+ state of OH and OD From a laser induced fluorescence (LIF) experiment on a molecular beam of CH, we have obtained the band c hyperfine constants, the rand r D spin-rotation constants as well as accurate values for the rotational constants B, D, and H for the C 2l; + , V = 0 state. From measurements of the linewidths, that are partially caused by predissociation, and by comparing relative line intensities, we determined different lifetimes for upper (F I ) and lower (F2) p-doublet states of the C 2l; + s~te. For the FI states we find a constant lifetime of 3.7 ± 1.0 ns, that is independent of N, while for the F2 states we observed an increase in lifetime for higher N up to 8.0 ± 1.5 ns for N= 11.

No abstract

The spectrum of the c'II, u = 0-a'A, u = 0 band of the NH molecule at X = 324 nm has been investigated under high resolution by laser-induced fluorescence in a molecular beam. From an analysis of the spectra we obtained: the magnetic dipole interaction constants aN,u and the electric quadrupole constants eQq,,2 for both electronic states, the improved values for the Adoubling constants 4, qf', and qf for the c'Il state, and rotational constants for both electronic states up to a third-order centrifugal distortion. Also, the Adoubling in the a 'A state could be determined. 0 1986 Academic PPS, Inc. ' We regret omitting Ref. (14) on the A'II-X3Z system in our previous work.

No abstract

Speech recognition systems are expensive to train, mostly due to the high cost of annotating training data. We previously proposed an iterative training algorithm [1], which sought to improve speech recognition by automatically selecting a subset of the available humanly transcribed training data, thereby improving error rates without incurring additional transcription cost. We suggest one improvement to that "selective sampling" algorithm and show that we are able to reduce the error rate on a particular alphadigit recognition problem from 10.3% to 9.5%. We then extend the iterative training algorithm to work with untranscribed speech, guiding selection of speech that is then transcribed. We show, on a particular alphadigit recognition problem, that it is possible to match the baseline error rate while only incurring 25% of the transcription cost.