Training dense passage representations via contrastive learning (CL) has been shown effective for Open-Domain Passage Retrieval (ODPR). Recent studies mainly focus on optimizing this CL framework by improving the sampling strategy or extra pretraining. Different from previous studies, this work devotes itself to investigating the influence of conflicts in the widely used CL strategy in ODPR, motivated by our observation that a passage can be organized by multiple semantically different sentences, thus modeling such a passage as a unified dense vector is not optimal. We call such conflicts Contrastive Conflicts. In this work, we propose to solve it with a representation decoupling method, by decoupling the passage representations into contextual sentence-level ones, and design specific CL strategies to mediate these conflicts. Experiments on widely used datasets including Natural Questions, Trivia QA, and SQuAD verify the effectiveness of our method, especially on the dataset where the conflicting problem is severe. Our method also presents good transferability across the datasets, which further supports our idea of mediating Contrastive Conflicts.

Though offering amazing contextualized token-level representations, current pretrained language models take less attention on accurately acquiring sentence-level representation during their self-supervised pre-training. However, contrastive objectives which dominate the current sentence representation learning bring little linguistic interpretability and no performance guarantee on downstream semantic tasks. We instead propose a novel generative self-supervised learning objective based on phrase reconstruction. To overcome the drawbacks of previous generative methods, we carefully model intra-sentence structure by breaking down one sentence into pieces of important phrases. Empirical studies show that our generative learning achieves powerful enough performance improvement and outperforms the current state-of-the-art contrastive methods not only on the STS benchmarks, but also on downstream semantic retrieval and reranking tasks. Our code is available at https://github.com/chengzhipanpan/PaSeR.

The study area is a marginal oilfield with tilted oil-water-contact. The target reservoir is located more than 5000m deep with an oil column of merely 1m. The primary challenge is to accurately land the well in this thin interval without penetrating the water zone. Faced with complex lithology in a low resistivity pay zone, formation evaluation poses obvious complications. Steering the production lateral within the thin oil column, while performing real-time reservoir evaluation, is the key to succeed economic objectives. A combined while-drilling technology, including multi-layer bed-mapping and multi-function Logging-While-Drilling (LWD) tools were employed to address these challenges and by the end of 2014, the pilot project had successfully placed the lateral within the target zone, securing an in zone Key-Performance-Indicator (KPI) ratio of over 90% despite encountering lateral heterogeneity. By achieving this KPI, the drilling cycle was shortend by almost one-third. The authors believe that the fit-for-purpose solution can be applied in similar blocks to achieve repeated success.

Though offering amazing contextualized tokenlevel representations, current pre-trained language models actually take less attention on acquiring sentence-level representation during its selfsupervised pre-training. If self-supervised learning can be distinguished into two subcategories, generative and contrastive, then most existing studies show that sentence representation learning may more benefit from the contrastive methods but not the generative methods. However, contrastive learning cannot be well compatible with the common token-level generative self-supervised learning, and does not guarantee good performance on downstream semantic retrieval tasks. Thus, to alleviate such obvious inconveniences, we instead propose a novel generative self-supervised learning objective based on phrase reconstruction. Empirical studies show that our generative learning may yield powerful enough sentence representation and achieve performance in Sentence Textual Similarity (STS) tasks on par with contrastive learning. Further, in terms of unsupervised setting, our generative method outperforms previous state-of-the-art SimCSE on the benchmark of downstream semantic retrieval tasks.

Horizontal wells with open-hole completions are very common in N field Middle East, over 80% of the well completions are open-hole completions. However, in recent years’ production, some of the open-hole wells face several challenges from the perspective of reservoir management, such as early-stage water/gas breakthrough and inefficient drainage around the wellbore. These challenges are caused by the existence of high permeability streaks and the heel-to-toe effect. Therefore, aiming at these challenges, the present work proposed the lower completion strategy and established the lower completion workflow to implement inflow control devices (ICD) or autonomous inflow control devices (AICD) with packers, and selected some representative wells in N oilfield to find out the lower completion optimization results of the whole well life. The principles of screening and selecting the pilot candidate wells in the N field middle east are operable, duplicable, and valuable. The pilot wells need to be the representatives in both geology and engineering, scheduled to be drilled recently and able to be evaluated from numerical simulation. For the comprehensive production analysis, the static model was used for sensitivity study and optimization and the dynamic model was used for the full-field assessment. The sensitivity and uncertainty of design parameters include compartment length, nozzle size, cross-section area and flow coefficient. Then, the static model was coupled with the dynamic model to obtain the optimal lower completion settings of the whole well life. The dynamic sector model that included injection wells and production wells was derived from a full field model. By using the optimized ICD/AICD parameters and deploying the tools correctly based on reservoir heterogeneity, the reservoir performance such as water cut, gas-oil ratio and cumulative oil production can be obtained to validate the optimization results. Several cases were studied to demonstrate the production improvement and economic values with the optimized lower completion. It is shown that the cumulative oil production will increase, the production plateau will extend and the overall NPV will be higher. This work can give guidance to the selection of the most fit-for-purpose lower completion technology for different well objectives when facing different kinds of challenges.