Multi-stage Training with Improved Negative Contrast for Neural Passage Retrieval

In the context of neural passage retrieval, we study three promising techniques: synthetic data generation, negative sampling, and fusion. We systematically investigate how these techniques contribute to the performance of the retrieval system and how they complement each other. We propose a multi-stage framework comprising of pre-training with synthetic data, fine-tuning with labeled data, and negative sampling at both stages. We study six negative sampling strategies and apply them to the fine-tuning stage and, as a noteworthy novelty, to the synthetic data that we use for pre-training. Also, we explore fusion methods that combine negatives from different strategies. We evaluate our system using two passage retrieval tasks for open-domain QA and using MS MARCO. Our experiments show that augmenting the negative contrast in both stages is effective to improve passage retrieval accuracy and, importantly, they also show that synthetic data generation and negative sampling have additive benefits. Moreover, using the fusion of different kinds allows us to reach performance that establishes a new state-of-the-art level in two of the tasks we evaluated.

Aligning Cross-lingual Sentence Representations with Dual Momentum Contrast

In this paper, we propose to align sentence representations from different languages into a unified embedding space, where semantic similarities (both cross-lingual and monolingual) can be computed with a simple dot product. Pre-trained language models are fine-tuned with the translation ranking task. Existing work (Feng et al., 2020) uses sentences within the same batch as negatives, which can suffer from the issue of easy negatives. We adapt MoCo (He et al., 2020) to further improve the quality of alignment. As the experimental results show, the sentence representations produced by our model achieve the new state-of-the-art on several tasks, including Tatoeba en-zh similarity search (Artetxe andSchwenk, 2019b), BUCC en-zh bitext mining, and semantic textual similarity on 7 datasets.

Improving Faithfulness in Abstractive Summarization with Contrast Candidate Generation and Selection

Despite significant progress in neural abstractive summarization, recent studies have shown that the current models are prone to generating summaries that are unfaithful to the original context. To address the issue, we study contrast candidate generation and selection as a model-agnostic post-processing technique to correct the extrinsic hallucinations (i.e. information not present in the source text) in unfaithful summaries. We learn a discriminative correction model by generating alternative candidate summaries where named entities and quantities in the generated summary are replaced with ones with compatible semantic types from the source document. This model is then used to select the best candidate as the final output summary. Our experiments and analysis across a number of neural summarization systems show that our proposed method is effective in identifying and correcting extrinsic hallucinations. We analyze the typical hallucination phenomenon by different types of neural summarization systems, in hope to provide insights for future work on the direction.

studying The random variables according to Neutrosophic logic

We present in this paper the neutrosophic randomized variables, which are a generalization of the classical random variables obtained from the application of the neutrosophic logic (a new nonclassical logic which was founded by the American philosopher and mathematical Florentin Smarandache, which he introduced as a generalization of fuzzy logic especially the intuitionistic fuzzy logic ) on classical random variables.

Testing weak form efficiency in Damascus securities market: Acomparisation study with Amman, Muscat securities markets

This paper aims to test weak form efficiency in Damascus , Amman , Muscat securities market .It examines daily stock return index during ( 1 - 4- 2010 ) , ( 31 - 12 - 2016 ) using normal distribution test , runs test , autocorrelation test , unit root test , variance ratio test , auto regressive integrated moving average test

Using of t student’s test to compare the variations of magic Latin square design in microbiological experiments

This research was conducted to study the effectiveness of the variations of magic latin square design, to reduce the value of the experimental error, in experiences of microbiological (Lactobacillus acidophilus), and to improve the activity of random rectangles as one source of the variations of magic latin square design. Where they were conducting a random distribution of (6) treatments supposed to (36) experimental unit test has repeated the process of distribution (150) times in order to obtain magic latin squares realized the conditions which show per treatments once in the row and once in the column, and once in each rectangle within the one design.

Evaluation performance of Some Lentil (Lens esculenta L.) Genotypes under irrigated and rain-fed agriculture conditions

The research was conducted at Abo-Jarash farm, faculty of agriculture, Damascus university during the two growing season 2011-2012 and 2012-2013 in order to evaluate genetic variability of some lentils genotypes based on some physiological and productivity traits associated with drought tolerance. The experiment was laid out according to factorial randomized complete block design with three replications.

استخدام تحليل التباين في دراسة تأثير موعد الزراعة و الكثافة النباتية على إنتاج محصول القمح

This work deals with the analysis of variance basis .It starts from monofactor analysis of variance to bifactor analysis of variance . Then , I applied the already mentioned theoretical study to actual issue in the agriculture field . I have chosen wheat and studied the best date to plan it and the best Plant production through a detailed study based o SPSS program . then , I showed the results of each variable in a detailed . Finally , this work can be a basis of another work using other analysis methods and other statistical in the analysis of variance .

The Relation Between Thrombus , Spontaneous Echo Contrast , And Left Atrial Appendage Velocity In Patients With Ischemic Stroke

Cardiogenic embolism can be identified in at least 15% of ischemic stroke patients, left atrial appendage is believed to be the place where thrombi is performed, then goes to systemic circulation causing embolic accident . In this search we study the function of left atrial appendage by pulse Doppler within the proximal third of the appendage in patients with ischemic stroke. Thrombus was defined as a discrete echocardiographically dens mass within the body of atrium or the appendage, with a different echocardiographically density than the adjacent endothelium . Spontaneous echo contrast was defined as high density flow due to low flow condition , which remains stable with changes in gain setting . In conclusion , the velocities which recorded in patients with ischemic stroke were lower in comparison with other patients , and it was significant lower in patients with spontaneous echo contrast or atrial fibrillation .

Variance and genetic and phenotypic correlations among productivity and its components in Cotton

Seven cotton genotypes (Rkka5, Aleppo1-33, Aleppo90 – Aleppo 118, Aleppo 40, Deir22, and Line124), were used to estimate variance and genetic and phenotypic correlations between seed cotton productivity and its components [earliness in flowering and maturity(day), plant height (cm), number of vegetative and reproductive branches, number of bolls per plant, 100 seed weight (g), average boll weight (g), lint percentage (%), and plant yield (g)], to be used as selection indices for higher yield performance in breeding programs. The experiment was carried out in Al-Ghab region (Hama) in 2012- 2013 growing season using randomized complete block design (RCBD) with three replications.