From b8a51f7020bff2cc69657004500818c699778403 Mon Sep 17 00:00:00 2001 From: Luca Foppiano Date: Wed, 14 Feb 2024 07:42:10 +0900 Subject: [PATCH] minor corrections in training data --- .../quantities/corpus/1001.4731.training.tei.xml | 2 +- .../units/corpus/hal-00962359.units.training.tei.xml | 2 +- .../units/corpus/jpa-00232521.units.training.xml | 2 +- .../units/{ => corpus}/materials-14-02583.unit.tei.xml | 0 .../units/corpus/trainingdata0.training.tei.xml | 2 ++ .../units/evaluation/unit-evaluation-corpus.tei.xml | 10 +++++----- 6 files changed, 10 insertions(+), 8 deletions(-) rename resources/dataset/units/{ => corpus}/materials-14-02583.unit.tei.xml (100%) diff --git a/resources/dataset/quantities/corpus/1001.4731.training.tei.xml b/resources/dataset/quantities/corpus/1001.4731.training.tei.xml index f3e6de86..d28d7d9c 100644 --- a/resources/dataset/quantities/corpus/1001.4731.training.tei.xml +++ b/resources/dataset/quantities/corpus/1001.4731.training.tei.xml @@ -45,7 +45,7 @@

With their lower brightness, lack of a dominant ring system, and small solid angle, Uranus and Neptune are used as primary calibrators in infrared through radio wavebands for both ground-and space-based instruments. Spectral coverage in the microwave is somewhat undersampled, however, leaving room for interpretation in atmospheric modeling efforts . The microwave spectra of Uranus and Neptune lack the broad NH 3 absorption centered near 24 GHz which is characteristic of Jupiter and Saturn. At millimeter wavelengths, collision-induced absorption by H 2 is considered the dominant continuum opacity source. CO rotational absorptions have been observed for Neptune in the sub-millimeter (Marten et al. 2005 ). At centimeter wavelengths, ammonia (which is depleted relative to solar nitrogen levels ) and hydrogen sulfide are the main opacity contributors (Deboer & Steffes 1996; Spilker 1995;). Of special note is Uranus' unique 98 ° obliquity, which allows for a slowly changing pole-to-pole panorama of the planet as it moves in its 84year orbit, and may also play a role in determining atmospheric conditions. For selected frequencies, observational databases spanning decades have permitted characterization of the variability of these two planets. Uranus in particular has exhibited long-term whole-disk temperature changes since 1966 which correlate with the viewing aspect of the south pole. After accounting for geometrical solid angle changes, Kramer et al. (2008) reported a gradual temperature drop of order 10% over 20years (1985-2005) at 90 GHz. This is interpreted as a true integrated-disk temperature change, as the 1985 face-on contribution from the bright south pole progressively morphs into a 2007 view dominated by colder equatorial regions. Similar findings, with similar phasing albeit different amplitudes, have been reported in the radio (Klein & Hofstadter 2006) and visible (Hammel & Lockwood 2007). There is weaker evidence that some portion of variability in the light curves is attributable to changes deep in the atmosphere (Klein & Hofstadter 2006); Hofstadter & Butler (2003) used VLA 2 and 6cm " snapshots " of the disk over several years to argue for opacity changes in zonal bands. Neptune's microwave variability is less well documented, although there is ample evidence for variability in the visible and near- IR, as summarized by Hammel & Lockwood (2007). Kramer et al. (2008 find the 90 GHz integrated disk temperature to be constant to within ∼ 8% over 20years. Peak WMAP antenna temperatures for these distant " ice giants " range from roughly 1 mK at W band to ∼0.2mK at K band. The signal-to-noise for individual observations of these objects is low (∼0.3 at W band, ∼0.14 at K band), resulting in large statistical errors in single-season disk temperature determinations. Tables 12 and 13 present the single-season brightness temperatures computed for Uranus and Neptune. Brightness temperatures are listed per frequency rather than per DA as a means of boosting signal-to-noise for those frequencies with multiple DAs. Temperatures for Uranus observing seasons four and five exhibit somewhat larger error bars than other seasons. For these two seasons, there was a reduction in the number of observations available for analysis because data quality checks excluded observations in close proximity to Mars sky coordinates. Sub-WMAP latitudes for Uranus range between −30 ° and 4 ° over our seven-year baseline . Linear correlations of T b against both sub-WMAP latitude and time produced no statistically significant trend other than a flat line. This is not surprising given that the tightest seasonal temperature errors are of order 10%, which encompasses the entire twentyyear range of variation seen in the 90 GHz light curves of Kramer et al. (2008). We performed similar correlations for Neptune, with again the same null variability result. Note, however, that the sub-WMAP latitude for this planet is relatively unchanged over the epochs of observation . With no discernable variability over the observing baseline, we computed seven-year means of the disk brightness for each planet; these are listed as the last line in Tables 12 and 13 . These seven-year means in turn may be compared to observations in the literature and placed in context with the microwave spectra in general. only returned those few acquired in the late 1960s, listed inTable 1 of Gulkis et al. (1978). For a variety of reasons, more recent observations would be preferable, but the older epoch does have the advantage of sharing a similar viewing geometry to that of the WMAP data. These older observations are plotted in gray in the Figure, and agree well with the WMAP measurements, but also do not define a " dip " at high statistical significance. Although suggestive of an interesting atmospheric opacity constraint for future study, the possibility of such a feature has gone unremarked in the literature.

Non-variable, spatially isolated fixed celestial calibrators for millimeter wavelengths are not common. At high Galactic latitude, bright sources are predominantly identified as some form of AGN/QSO, which are prone to outbursts and variability on a wide range of timescales and frequencies. Brighter sources in the Galactic plane tend to be HII regions (which may not be " point-like " ) or supernova remnants (SNRs) with potential variability. At moderate spatial resolution, confusion with neighboring diffuse and compact sources in the Galactic plane can be an issue for background subtraction. Five sources were chosen for study out of an original list which included some of the brightest sources from Baars et al. (1977) and Ott et al. (1994), in addition to the brightest, least variable objects from the seven-year WMAP source catalog (Gold et al. 2010). The five selected sources are listed inTable 14. Some of the sources which were initially considered but ultimately rejected, primarily because of low background contrast, included 3C286, NGC 7027, 3C84, 3C218, 3C123, and 3C147. There are unfortunately few suitable calibration sources in the Southern hemisphere with a long-term history of observation.

Flux densities of the selected sources are measured from the seven-year sky maps at HEALPix 5 resolution 9 (Nside=512). For each frequency band, the azimuthally symmetrized beam profile (see Section 3.1) is convolved with a sky map pixel to produce a map-based beam template. The sum of the pixel-convolved beam template plus a sloping planar base level is fit to the Stokes I, Q, and U sky map data at each source position, using pixels within 3.5 times the beam width σ (1.5 times the FWHM) in each band. The peak source temperature from each fit is converted from thermodynamic temperature to Rayleigh-Jeans brightness temperature and then translated to a source flux density using a conversion factor Γ that is a weak function of the source spectral index (Jarosik et al. 2010). For point sources, this method of flux measurement is more accurate than the aperture photometry method as used for example by Page et al. (2003b) for Tau A. The WMAP beam profiles have extended wings (Hill et al. 2009), so an aperture radius 3 times the beam FWHM should be used and the results are more susceptible to error due to background confusion. The uncertainty in flux density is calculated as the quadrature sum of (1) map measurement uncertainty, (2) the uncertainty in Γ (0.5% to 0.7% depending on the band), and (3) the 0.2% absolute calibration uncertainty. For Stokes I, map measurement uncertainty is estimated from the rms fit residual in the fit region. The Stokes I residuals generally appear to be dominated by beam asymmetry effects, but also include background confusion and noise. (The source profiles in the maps are asymmetric due to asymmetry in the instantaneous beam profiles and the nonuniform distribution of scan angles over a year for sources away from the ecliptic poles.) For Stokes Q and U, map measurement uncertainty is calculated either from the 1σ source peak uncertainty and base level uncertainty from the beam fitting or as the Q or U flux times the fractional map measurement uncertainty for I, whichever is largest. The latter method is the estimated uncertainty in Q or U due to beam asymmetry effects, assuming the fractional uncertainty is the same in Q or U as it is in I. The flux determination was tested on simulated sky maps containing a population of point sources with no other signals and no noise, generated using WMAP beam window functions as described in Wright et al. (2009). Recovered flux densities were accurate to about 0.1% or better, except in W band where the recovered fluxes tended to be larger than the input fluxes by up to 1%. This is allowed for by including an additional 1% uncertainty term in the quadrature sum for W band. Fractional year-to-year variability for the selected sources has been obtained using sky maps for individual years 1-7. To remove confusion noise from the CMB and Galactic foregrounds, we subtract the seven-year average map from each individual year map for each band. A pixel-convolved beam plus flat base level is fit to each difference map at each source position, giving a flux difference ∆F i for the ith year for each source in each band. Uncertainty in ∆F i is calculated from the quadrature sum of the source peak uncertainty and the base level uncertainty. The flux difference is divided by the mean flux from the seven-year map to get the fractional flux variation ∆F i /F . For K, Ka, and Q bands, there is a small ( 0.2%) but significant year-to-year variation in the WMAP calibration, which we have measured by correlating each yearly map against the seven-year map (seeFigure 1 of Jarosik et al. 2010). The measured fractional flux variations in K-Q bands are corrected for these calibration variations c i using (∆F i /F ) corrected = (∆F i /F ) measured + (1 − c i ).

-

Source flux densities from the seven-year maps are presented inTable 15. Fractional uncertainties for the Stokes I fluxes are typically 1 to 3%. For some of the sources, the maximum source extent given inTable 14 is not entirely negligible relative to the WMAP beam width in V or W band (FWHM 19.6 ′ in V, FWHM 12.2 ′ in W, Hinshaw et al. 2009). We have estimated the possible error due to source extent for Tau A. The 1.4GHz map of the Tau A region from the NVSS survey (beam size 45 ′′ FWHM, Condon et al. 1998) was smoothed with the symmetrized WMAP V or W-band beam and converted to a resolution 9 (Nside = 512) HEALPix map. The flux determined by our method was found to underestimate the true flux by 1.5% in V band and 3.7% in W band. Spatial variations of the spectral index over Tau A are very small (Morsi & Reich 1987; Bietenholz et al. 1997; Green et al. 2004) so the source extent at W band is probably similar to that at 1.4 GHz. Our V and W-band fluxes for Cas A and 3C58 may also be underestimated by similar amounts. To allow for this, the uncertainty values listed for these three sources inTable 15 nificant secular decrease is seen for Cas A and Tau A. The results are consistent with a frequency independent decrease of about 0.53% per year for Cas A and 0.22% per year for Tau A. Our results for Cas A fall between the ∼ 0.6% per year decrease found by O'Sullivan & Green (1999) Significant yearly flux variation is not seen for Cyg A and 3C58; the rms year-to-year variation is consistent with the uncertainties in each band. The lowest rms variation is in K band, and is 0.27% for Cyg A and 0.33% for 3C58. Carilli & Barthel (1996) give an upper limit of 10% on Cyg A core variability from observations at 5, 15, and 90 GHz over timescales from 10months to 15years. The core contributes ≤ 10% of the total flux at WMAP frequencies (e.g., Robson et al. 1998). For 3C274 there is evidence for year-to-year variability of about 2% in K, Ka, and Q bands. The rms variation is 1.8 to 2.6 times the mean uncertainty in these bands, and the variations are correlated from band to band. Previous observations of 3C274 have shown greater variability. At 90 GHz, Steppe et al. (1988)Wagner et al. 2009), which corresponds to 4% of the total flux. Smaller core flux variations, less than 1% of the total flux, have been observed at lower frequencies (Morabito et al. 1988; Junor & Biretta 1995; Harris et al. 2009). Spectra of the seven-year WMAP fluxes together with previous measurements from the literature are shown in Figures 14 -18. For Cas A, we have scaled the WMAP fluxes, the Archeops fluxes of Desert et al. (2008), the SCUBA fluxes of Dunne et al. (2003), and the BLAST fluxes of Sibthorpe et al. (2010) to epoch 2000 using a secular variation of −0.53% per year. These are plotted with previous measurements that were scaled to epoch 2000 by Hafez et al. (2008) using frequency-dependent scaling. For Tau A, we have scaled previous measurements from Perez et al. (2010) to the epoch of the WMAP data using a secular variation of −0.167% per year at all frequencies. Results are not significantly different if −0.22% per year is used.Table 17 presents parameters from fits to the spectra for WMAP data alone and for the combined data. The two fits are generally consistent within their uncertainties over the WMAP frequency range. Some notes on the individual sources follow.

+

Source flux densities from the seven-year maps are presented inTable 15. Fractional uncertainties for the Stokes I fluxes are typically 1 to 3%. For some of the sources, the maximum source extent given inTable 14 is not entirely negligible relative to the WMAP beam width in V or W band (FWHM 19.6 ′ in V, FWHM 12.2 ′ in W, Hinshaw et al. 2009). We have estimated the possible error due to source extent for Tau A. The 1.4GHz map of the Tau A region from the NVSS survey (beam size 45 ′′ FWHM, Condon et al. 1998) was smoothed with the symmetrized WMAP V or W-band beam and converted to a resolution 9 (Nside = 512) HEALPix map. The flux determined by our method was found to underestimate the true flux by 1.5% in V band and 3.7% in W band. Spatial variations of the spectral index over Tau A are very small (Morsi & Reich 1987; Bietenholz et al. 1997; Green et al. 2004) so the source extent at W band is probably similar to that at 1.4 GHz. Our V and W-band fluxes for Cas A and 3C58 may also be underestimated by similar amounts. To allow for this, the uncertainty values listed for these three sources inTable 15 nificant secular decrease is seen for Cas A and Tau A. The results are consistent with a frequency independent decrease of about 0.53% per year for Cas A and 0.22% per year for Tau A. Our results for Cas A fall between the ∼ 0.6% per year decrease found by O'Sullivan & Green (1999) Significant yearly flux variation is not seen for Cyg A and 3C58; the rms year-to-year variation is consistent with the uncertainties in each band. The lowest rms variation is in K band, and is 0.27% for Cyg A and 0.33% for 3C58. Carilli & Barthel (1996) give an upper limit of 10% on Cyg A core variability from observations at 5, 15, and 90 GHz over timescales from 10months to 15years. The core contributes ≤ 10% of the total flux at WMAP frequencies (e.g., Robson et al. 1998). For 3C274 there is evidence for year-to-year variability of about 2% in K, Ka, and Q bands. The rms variation is 1.8 to 2.6 times the mean uncertainty in these bands, and the variations are correlated from band to band. Previous observations of 3C274 have shown greater variability. At 90 GHz, Steppe et al. (1988)Wagner et al. 2009), which corresponds to 4% of the total flux. Smaller core flux variations, less than 1% of the total flux, have been observed at lower frequencies (Morabito et al. 1988; Junor & Biretta 1995; Harris et al. 2009). Spectra of the seven-year WMAP fluxes together with previous measurements from the literature are shown in Figures 14 -18. For Cas A, we have scaled the WMAP fluxes, the Archeops fluxes of Desert et al. (2008), the SCUBA fluxes of Dunne et al. (2003), and the BLAST fluxes of Sibthorpe et al. (2010) to epoch 2000 using a secular variation of −0.53% per year. These are plotted with previous measurements that were scaled to epoch 2000 by Hafez et al. (2008) using frequency-dependent scaling. For Tau A, we have scaled previous measurements from Perez et al. (2010) to the epoch of the WMAP data using a secular variation of −0.167% per year at all frequencies. Results are not significantly different if −0.22% per year is used.Table 17 presents parameters from fits to the spectra for WMAP data alone and for the combined data. The two fits are generally consistent within their uncertainties over the WMAP frequency range. Some notes on the individual sources follow.

A slightly curved spectrum gives a better fit to the combined data than a power law (chi-squared per degree of freedom χ 2 ν = 1.20 compared to χ 2 ν = 2.13 for a power law). The flattening of the spectrum with increasing frequency was previously noted by Hafez et al. (2008). This may be consistent with observations of spatial variations of the spectrum in Cas A (e.g., Wright et al. 1999, Anderson & Rudnick 1996). Wright et al. (1999) presented spectra of 26 brightness peaks from maps with 7 ′′ resolution at 1.5, 5, 28, and 83 GHz. The data were mostly consistent with power-law spectra, with spectral indices ranging from -0.75 to -0.95. (For comparison, the overall spectral index from a power-law fit to our integrated spectrum from 1.4 to 93 GHz is -0.73). Such a variation will lead to curvature in the integrated spectrum. Wright et al. (1999) also found curvature for some of the brightness peaks with the spectra progressively flattening at the higher frequencies, and noted that such curvature is expected from models of particle acceleration in cosmic ray modified shocks (Reynolds & Ellison 1992). Within the WMAP frequency range, all of the epoch 2000 scaled fluxes are consistent within the uncertainties with the fit to the combined data. This includes the WMAP fluxes, the absolutely calibrated fluxes from Janssen et al. (1974) at 22.29 GHz and Mason et al. (1999) at 32 GHz, the 33 GHz flux from Hafez et al. (2008), which is calibrated using the five-year WMAP Jupiter temperature, and the 86 GHz flux from Liszt & Lucas (1999), which is calibrated using the Ulich (1981) Jupiter temperature. Above 300 GHz, there is excess emission above that expected for synchrotron emission, which has most recently been interpreted as emission from cool dust by Sibthorpe et al. (2010). The 353 and 545 GHz fluxes from Archeops (Desert et al. 2008) are much higher than the 600 GHz flux from BLAST (Sibthorpe et al. 2010) and the 353 and 666 GHz fluxes from SCUBA (Dunne et al. 2003). The Archeops measurements were made with a larger beam (∼ 12 ′ compared to ∼ 20 ′′ or better for SCUBA and BLAST) and appear to be affected by dust emission that is not associated with Cas A. The WMAP 23 GHz (K band) polarization map for Cas A exhibits unexpected structure.

Figure 19 shows seven-year mean intensity and polarization images centered on each source at each of the five WMAP frequency bands. These images are 4.15 ° on a side, have not been background subtracted, and are scaled such that brighter pixels are black. The polarization (P) image for Cas A shows an irregular ∼ 0.15mK ring at a radial distance roughly 40 ′ − 50 ′ from the source position. The angular extent of Cas A is ∼ 5 ′ (Table 14), leaving the reality of the ring feature in question. We have attempted to simulate this feature under the hypothesis that it is an artifact introduced by a combination of beam and source spectrum characteristics. Cas A is a steep-spectrum source. As a result of effective frequency differences between the two K-band radiometers (Jarosik et al. 2003b), the K11 radiometer (fed by the axial OMT port) has an FWHM roughly 3% wider than that of K12 (which is fed by the lateral OMT port), and the peak observed signal in K11 is a few percent higher than that of K12. Simulated beam maps for the two K-band radiometers were generated separately using Jupiter data as a template, and then individually scaled to peak values representative of Cas A. The difference between the two beams as a function of azimuthal angle and radial distance from beam center can be used to compute a rough estimate of induced Cas A polarization signal, under the assumption of complete scan-angle coverage.Figure 20 shows the results of such a simulation, which produces a feature with an approximately correct peak position (near 50 ′ ), but slightly wider and 30%-40% brighter than that shown in the data image. A more complete simulation would include scan-angle coverage effects, which in this scenario are presumed responsible for the gaps in the ring. This is the only known instance of an apparent artifact in WMAP polarization data.

The fluxes from Janssen et al. (1974) at 22.29 GHz, Wright & Birkinshaw (1984 at 89 GHz, and Wright & Sault (1993) at 94 GHz are consistent with the WMAP results. The flux from Hafez et al. (2008) at 33 GHz is lower than WMAP Ka-band flux by 2.7σ, taking both flux uncertainties into account. We found that excluding the Hafez et al. (2008) flux from the power-law fit to the combined data improved χ 2 ν from 2.89 to 0.79, so it was excluded for the fit plotted inFigure 15 and the fit parameters given inTable 17. Most of the measurements above 100 GHz inFigure 15 are fluxes summed over the core and two hot spots in the radio lobes. These are probably valid measurements of the total flux, since the contribution of extended emission from the steep-spectrum lobes appears to be small or negligible at these frequencies. At 230 GHz, Salter et al. (1989b) found that the integrated flux over the entire source was only about 10% greater than the summed flux of the hot spots and core, which did not amount to a significant detection of emission from the lobes.

diff --git a/resources/dataset/units/corpus/hal-00962359.units.training.tei.xml b/resources/dataset/units/corpus/hal-00962359.units.training.tei.xml index d46daa38..3c07c907 100644 --- a/resources/dataset/units/corpus/hal-00962359.units.training.tei.xml +++ b/resources/dataset/units/corpus/hal-00962359.units.training.tei.xml @@ -24,7 +24,7 @@ mm mm mm - N/mm 2/min + N/mm 2/min mm percent days diff --git a/resources/dataset/units/corpus/jpa-00232521.units.training.xml b/resources/dataset/units/corpus/jpa-00232521.units.training.xml index b7236929..aa0914d9 100644 --- a/resources/dataset/units/corpus/jpa-00232521.units.training.xml +++ b/resources/dataset/units/corpus/jpa-00232521.units.training.xml @@ -5,7 +5,7 @@ cm cm cm-1 - cm-1 + cm−1 cm K K diff --git a/resources/dataset/units/materials-14-02583.unit.tei.xml b/resources/dataset/units/corpus/materials-14-02583.unit.tei.xml similarity index 100% rename from resources/dataset/units/materials-14-02583.unit.tei.xml rename to resources/dataset/units/corpus/materials-14-02583.unit.tei.xml diff --git a/resources/dataset/units/corpus/trainingdata0.training.tei.xml b/resources/dataset/units/corpus/trainingdata0.training.tei.xml index 829f32b8..4ea925bd 100644 --- a/resources/dataset/units/corpus/trainingdata0.training.tei.xml +++ b/resources/dataset/units/corpus/trainingdata0.training.tei.xml @@ -28,7 +28,9 @@ V/cm 223455,6V/cm μg + ms−2 μl + ls−1 pH pH pH diff --git a/resources/dataset/units/evaluation/unit-evaluation-corpus.tei.xml b/resources/dataset/units/evaluation/unit-evaluation-corpus.tei.xml index c73b50b1..33429dab 100644 --- a/resources/dataset/units/evaluation/unit-evaluation-corpus.tei.xml +++ b/resources/dataset/units/evaluation/unit-evaluation-corpus.tei.xml @@ -616,7 +616,7 @@ meV∕T Ω∕mm mV - wt.%--> + wt.% pJ/m GByte∕platter μM @@ -935,7 +935,7 @@ K^2∕meV^2 Å−3 TeV - V^2/K^2 + V^2/K^2 hole m−3 monolayers at. cm−3 @@ -1138,7 +1138,7 @@ W∕mm^2 A−1 m μN∕m - kg m{−1} s−2--> + kg m{−1} s−2 N m−1 s−1 nm V−1 cm^2V s @@ -1402,7 +1402,7 @@ counts∕mm∕s^2 kV∕dc nm g∕cm^3 - ions∕cm^2/s + ions∕cm^2/s meV∕cell J∕g∕°C gm∕h @@ -1574,7 +1574,7 @@ m^3 mol−1 poise μm−3 - Ω □−1--> + Ω □−1 mV∕A m rad dB mm−1