Intent classification (IC) and Named Entity Recognition (NER) are arguably the two main components needed to build a Natural Language Understanding (NLU) engine, which is a main component of ...conversational agents. The IC and NER components are closely intertwined and the entities are often connected to the underlying intent. Current research has primarily focused to model IC and NER as two separate units, which results in error propagation, and thus, sub-optimal performance. In this paper, we propose a simple yet effective novel framework for NLU where the parameters of the IC and the NER models are jointly trained in a consolidated parameter space. Text semantic representations are obtained from popular pre-trained contextual language models, which are fine-tuned for our task, and these parameters are propagated to other deep neural layers in our framework leading to a faithful unified modelling of the IC and NER parameters. The overall framework results in a faithful parameter sharing when the training is underway, leading to a more coherent learning. Experiments on two public datasets, ATIS and SNIPS, show that our model outperforms other methods by a noticeable margin. On the SNIPS dataset, we obtain a 1.42% improvement in NER in terms of the F1 score, and 1% improvement in intent accuracy score. On ATIS, we achieve 1.54% improvement in intent accuracy score. We also present qualitative results to showcase the effectiveness of our model.
The aim of this thesis is to investigate evidence of heating and radiative feedback in local Gould Belt star-forming regions. I discuss what impact, if any, radiative feedback is having on the star ...formation. I primarily use Submillimeter Common-User Bolometer Array 2 (SCUBA-2) observations from the James Clerk Maxwell Telescope (JCMT) Gould Belt legacy Survey (GBS) of nearby star-forming regions. I analyse this data in conjunction with catalogues of candidate young stellar objects (YSOcs) from mid-infrared surveys with Spitzer IRAC and MIPS surveys. I use the ratio of SCUBA-2 fluxes to calculate dust temperature, given a constant value of dust opacity spectral index, following the method of Reid & Wilson (2005). I employ a two-component beam (2CB) cross convolution to map the temperature of the Serpens MWC 297 region, achieving a resolution of 19.9′′. I employ a convolution kernel to map the temperature of the majority of the JCMT GBS, including the Aquila W40 complex, achieving a resolution of 14.8′′. I use the fellwalker clump finding algorithm to produce a global catalogue of 619 SCUBA-2 850 μm clumps across 26 distinct sub-regions of the JCMT GBS, calculating real temperatures where available. I was the PI of a proposal to observe 12CO 3-2 line emission, with the aim of decontaminating the SCUBA-2 850 μm band. I find 12CO 3-2 line contamination has a significant impact, increasing the dust temperatures calculated per pixel, on average, by 3 K where contamination is less than 10%, and by 16 K where contamination is greater than 10% (in the Aquila W40 complex). I find evidence for 12 outflows in this region, associated with active star formation. I also use archival VLA data to decontaminate both SCUBA-2 bands of free-free emission associated with massive star formation. Where compact free-free sources are sufficiently bright and optically thick, for example the B1.5Ve star MWC 297, their contribution can lead to prominent bright sources at the submillimeter wavelengths detected by SCUBA-2 and lower temperatures around Herbig stars. I present published studies of the Serpens MWC 297 region and the Aquila W40 complex. In both cases I find evidence that the presence of young OB stars is raising the temperatures of nearby clumps. Examining clumps across the JCMT GBS, I find that those clumps isolated from OB stars have a mean temperature of 15±2 K, a value that is consistent with gas temperatures (Friesen et al., 2009) and Bonnor-Ebert sphere models (Kirk et al., 2006). I find no evidence of heating from embedded low-to-medium mass YSOs. Clumps that lie within 3 pc of OB stars have a mean temperature of 21±4 K and O type stars heat clumps over the greatest range. By remodelling the heated clumps with a temperature of 15 K, I calculate that up to 10% of clumps in the JCMT GBS are no longer Jeans unstable, indicating that radiative feedback from OB stars is potentially suppressing fragmentation and allowing for the formation of more massive stars.
The aim of this thesis is to investigate evidence of heating and radiative feedback in local Gould Belt star-forming regions. I discuss what impact, if any, radiative feedback is having on the star ...formation.I primarily use Submillimeter Common-User Bolometer Array 2 (SCUBA-2) observations from the James Clerk Maxwell Telescope (JCMT) Gould Belt legacy Survey (GBS) of nearby star-forming regions. I analyse this data in conjunction with catalogues of candidate young stellar objects (YSOcs) from mid-infrared surveys with Spitzer IRAC and MIPS surveys.I use the ratio of SCUBA-2 fluxes to calculate dust temperature, given a constant value of dust opacity spectral index, following the method of Reid & Wilson (2005). I employ a twocomponent beam (2CB) cross convolution to map the temperature of the Serpens MWC 297 region, achieving a resolution of 19.900. I employ a convolution kernel to map the temperature of the majority of the JCMT GBS, including the Aquila W40 complex, achieving a resolution of 14.800 . I use the fellwalker clump finding algorithm to produce a global catalogue of 619 SCUBA-2 850 µm clumps across 26 distinct sub-regions of the JCMT GBS, calculating real temperatures where available.I was the PI of a proposal to observe 12CO 3-2 line emission, with the aim of decontaminating the SCUBA-2 850 µm band. I find 12CO 3-2 line contamination has a significant impact, increasing the dust temperatures calculated per pixel, on average, by 3 K where contamination is less than 10%, and by 16 K where contamination is greater than 10% (in the Aquila W40 complex). I find evidence for 12 outflows in this region, associated with active star formation.I also use archival VLA data to decontaminate both SCUBA-2 bands of free-free emission associated with massive star formation. Where compact free-free sources are sufficiently bright and optically thick, for example the B1.5Ve star MWC 297, their contribution can lead to prominent bright sources at the submillimeter wavelengths detected by SCUBA-2 and lower temperatures around Herbig stars.I present published studies of the Serpens MWC 297 region and the Aquila W40 complex. In both cases I find evidence that the presence of young OB stars is raising the temperatures of nearby clumps. Examining clumps across the JCMT GBS, I find that those clumps isolated from OB stars have a mean temperature of 15±2 K, a value that is consistent with gas temperatures (Friesen et al., 2009) and Bonnor-Ebert sphere models (Kirk et al., 2006). I find no evidence of heating from embedded low-to-medium mass YSOs.Clumps that lie within 3 pc of OB stars have a mean temperature of 21±4 K and O type stars heat clumps over the greatest range. By remodelling the heated clumps with a temperature of 15 K, I calculate that up to 10% of clumps in the JCMT GBS are no longer Jeans unstable, indicating that radiative feedback from OB stars is potentially suppressing fragmentation and allowing for the formation of more massive stars.
Radiative feedback can influence subsequent star formation. We quantify the heating from OB stars in the local star-forming regions in the JCMT Gould Belt survey. Dust temperatures are calculated ...from 450/850 micron flux ratios from SCUBA-2 observations at the JCMT assuming a fixed dust opacity spectral index \(\beta=1.8\). Mean dust temperatures are calculated for each submillimetre clump along with projected distances from the main OB star in the region. Temperature vs. distance is fit with a simple model of dust heating by the OB star radiation plus the interstellar radiation field and dust cooling through optically thin radiation. Classifying the heating sources by spectral type, O-type stars produce the greatest clump average temperature rises and largest heating extent, with temperatures over 40 K and significant heating out to at least 2.4 pc. Early-type B stars (B4 and above) produce temperatures of over 20 K and significant heating over 0.4 pc. Late-type B stars show a marginal heating effect within 0.2 pc. For a given projected distance, there is a significant scatter in clump temperatures that is due to local heating by other luminous stars in the region, projection effects, or shadowing effects. Even in these local, `low-mass' star-forming regions, radiative feedback is having an effect on parsec scales, with 24% of the clumps heated to at least 3 K above the 15 K base temperature expected from heating by only the interstellar radiation field, and a mean dust temperature for heated clumps of 24 K.
We present observations of the Cepheus Flare obtained as part of the James Clerk Maxwell Telescope (JCMT) Gould Belt Legacy Survey (GBLS) with the SCUBA-2 instrument. We produce a catalogue of ...sources found by SCUBA-2, and separate these into starless cores and protostars. We determine masses and densities for each of our sources, using source temperatures determined by the Herschel Gould Belt Survey. We compare the properties of starless cores in four different molecular clouds: L1147/58, L1172/74, L1251 and L1228. We find that the core mass functions for each region typically show shallower-than-Salpeter behaviour. We find that L1147/58 and L1228 have a high ratio of starless cores to Class II protostars, while L1251 and L1174 have a low ratio, consistent with the latter regions being more active sites of current star formation, while the former are forming stars less actively. We determine that, if modelled as thermally-supported Bonnor-Ebert spheres, most of our cores have stable configurations accessible to them. We estimate the external pressures on our cores using archival \(^{13}\)CO velocity dispersion measurements and find that our cores are typically pressure-confined, rather than gravitationally bound. We perform a virial analysis on our cores, and find that they typically cannot be supported against collapse by internal thermal energy alone, due primarily to the measured external pressures. This suggests that the dominant mode of internal support in starless cores in the Cepheus Flare is either non-thermal motions or internal magnetic fields.