Search

This paper introduces our proposed system for the MIA Shared Task on Cross-lingual Open-retrieval Question Answering (COQA). In this challenging scenario, given an input question the system has to gather evidence documents from a multilingual pool and generate from them an answer in the language of the question. We devised several approaches combining different model variants for three main components: Data Augmentation, Passage Retrieval, and Answer Generation. For passage retrieval, we evaluated the monolingual BM25 ranker against the ensemble of re-rankers based on multilingual pretrained language models (PLMs) and also variants of the shared task baseline, re-training it from scratch using a recently introduced contrastive loss that maintains a strong gradient signal throughout training by means of mixed negative samples. For answer generation, we focused on language- and domain-specialization by means of continued language model (LM) pretraining of existing multilingual encoders. Additionally, for both passage retrieval and answer generation, we augmented the training data provided by the task organizers with automatically generated question-answer pairs created from Wikipedia passages to mitigate the issue of data scarcity, particularly for the low-resource languages for which no training data were provided. Our results show that language- and domain-specialization as well as data augmentation help, especially for low-resource languages.

Study objectives To report the efficacy and findings of a large-scale preventive screening program for severe acute respiratory syndrome-associated coronavirus (SARS-CoV) using amplification of the virus from a nasopharyngeal swab (NPS) obtained from the health-care workers (HCWs). Design A prospective observational study. Setting A medical center in Taiwan. Participants Two hundred thirty HCWs. Intervention NPS examination for the presence of SARS-CoV by two nested reverse transcription-polymerase chain reaction (RT-PCR) assays. Measurements and results During the outbreak of severe acute respiratory syndrome (SARS), NPS polymerase chain reaction screening of HCWs for SARS-CoV was performed. SARS-CoV was examined by two nested RT-PCRs and a quantitative RT-PCR. Serum-specific antibodies were assessed by enzyme immunoassay and indirect immunofluorescence. We monitored 230 HCWs, including 217 first-line HCWs and 13 non–first-line HCWs. One hundred ninety first-line HCWs and 13 non–first-line HCWs had negative results in both nested RT-PCR assays. Two first-line HCWs who were positive on both nested RT-PCR assays had SARS. They had 16,900 ± 7,920 copies (mean ± SD) of RNA per milliliter in the NPS and had detectable anti-SARS antibodies. The remaining 25 first-line HCWs were negative for the first nested RT-PCR but positive for the second nested RT-PCR. Their corresponding titers were 338 ± 227 copies of RNA per milliliter; antibodies developed in none of these 25 HCWs. The expression and function of angiotensin-converting enzyme-2 were not different among these HCWs. This study shows that colonization of SARS-CoV occurred in 25 of 217 well-protected first-line HCWs on a SARS-associated service, but they remained seronegative. Conclusion With the second RT-PCR assay more sensitive than the first RT-PCR assay, we are able to show that approximately 11.5% of well-protected HCWs exposed to SARS patients or specimens may have colonization without seroconversion. Only those with significant clinical symptoms or disease would have active immunity. Thus, regular NPS screening for nested RT-PCR assays in conjunction with a daily recording of body temperature in all first-line HCWs may provide an effective way of early detection.

The autoSCAN-W/A (W/A; Dade Behring Microscan Inc., West Sacramento, Calif.) and Vitek AutoMicrobic System (Vitek AMS; bioMérieux Vitek Systems, Inc., Hazelwood, Mo.) are both fully automated microbiology systems. We evaluated the accuracy of these two systems in identifying nonglucose-fermenting gram-negative bacilli. We used the W/A with conventional-panel Neg Combo type 12 and Vitek GNI+ identification systems. A total of 301 isolates from 25 different species were tested. Of these, 299 isolates were identified in the databases of both systems. The conventional biochemical methods were used for reference. The W/A correctly identified 215 isolates (71.4%) to the species level at initial testing with a high probability of ≥85%. The Vitek GNI+ correctly identified 216 isolates (71.8%) to the species level at initial testing with a high probability of ≥90%. After additional testing that was recommended by the manufacturer's protocol, the correct identifications of the W/A and Vitek GNI+ improved to 96.0 and 92.3%, respectively. The major misidentified species were Sphingomonas paucimobilis and Agrobacterium radiobacter in the W/A system and Acinetobacter lwoffii , Chryseobacterium indologenes , and Comamonas acidovorans in the Vitek GNI+ system. The error rates were 4.0 and 7.6%, respectively. The overall accuracy for both systems was above 90% if the supplemental tests were applied. There was no significant difference in accuracy ( P > 0.05) between the two systems.

Adenoviruses are a common etiology of respiratory tract infections in children, with several serotypes responsible for most epidemic respiratory infections. This study examined the molecular epidemiology and clinical features of pediatric adenoviral infections in a 1-year period.From May 1999 to April 2000, virus specimens collected from children with respiratory tract infections were identified. The presence of adenovirus was confirmed by direct fluorescent staining, and viral types were determined by polymerase chain reaction sequencing.Adenoviruses were identified from 272 children (mean +/- standard deviation age, 48.3 +/- 30.5 months), 227 (83.5%) of whom were aged 6 years or younger. Inpatients were younger than outpatients (44.1 +/- 30.7 months vs 53.0 +/- 29.4 months; p = 0.006). The commonest serotype identified was serotype 3 (164 patients; 60.3%), 73.1% of which were identified between September 1999 and January 2000. Serotype 3 was more common in inpatients (p = 0.015), while serotypes 1, 2, 5, and 6 were more common in outpatients (p = 0.021). Patients with pneumonia were younger than those with other infections (31.8 +/- 20.2 months vs 50.3 +/- 31.0 months; p = 0.001). Most of the children (90.1%) had fever for a mean of 3.80 +/- 2.65 days before seeing a doctor. The clinical manifestations were similar regardless of the serotype.Adenovirus serotype 3 caused the most adenovirus infections in autumn and winter of 1999 to 2000. The children were mostly preschool age and required hospital admission.

PCR-restriction fragment length polymorphism (RFLP) analysis is a nonprobe method for the rapid identification of Mycobacterium species. We demonstrate the separation of DNA or restriction fragments digested from the mycobacterial gene encoding the 65-kDa heat shock protein ( hsp65 ) by capillary electrophoresis (CE). By using a pair of unlabeled primers, Tb11 and Tb12, and only one restriction enzyme, HaeIII, we investigated a total of 52 reference and clinical strains encompassing 12 Mycobacterium species. The electrophoretic separation of high-resolution CE required Mycobacterium species. Beyond the agarose and polyacrylamide gel electrophoresis, high-resolution CE provides an alternative for rapid identification of Mycobacterium species that is feasible for automation and routine use without the need for costly probes.

Early detection and identification of SARS-CoV-infected patients and actions to prevent transmission are absolutely critical to prevent another SARS outbreak. Antibodies that specifically recognize the SARS-CoV spike and nucleocapsid proteins may provide a rapid screening method to allow accurate identification and isolation of patients with the virus early in their infection. For this reason, we raised peptide-induced polyclonal antibodies against SARS-CoV spike protein and polyclonal antibodies against SARS-CoV nucleocapsid protein using 6x His nucleocapsid recombinant protein. Western blot analysis and immunofluorescent staining showed that these antibodies specifically recognized SARS-CoV.