Search

The development of modern NLP applications often relies on various benchmark datasets containing plenty of manually labeled tests to evaluate performance. While constructing datasets often costs many resources, the performance on the held-out data may not properly reflect their capability in real-world application scenarios and thus cause tremendous misunderstanding and monetary loss. To alleviate this problem, in this paper, we propose an automated test generation method for detecting erroneous behaviors of various NLP applications. Our method is designed based on the sentence parsing process of classic linguistics, and thus it is capable of assembling basic grammatical elements and adjuncts into a grammatically correct test with proper oracle information. We implement this method into NLPLego, which is designed to fully exploit the potential of seed sentences to automate the test generation. NLPLego disassembles the seed sentence into the template and adjuncts and then generates new sentences by assembling context-appropriate adjuncts with the template in a specific order. Unlike the taskspecific methods, the tests generated by NLPLego have derivation relations and different degrees of variation, which makes constructing appropriate metamorphic relations easier. Thus, NLPLego is general, meaning it can meet the testing requirements of various NLP applications. To validate NLPLego, we experiment with three common NLP tasks, identifying failures in four state-of-art models. Given seed tests from SQuAD 2.0, SST, and QQP, NLPLego successfully detects 1,732, 5301, and 261,879 incorrect behaviors with around 95.7% precision in three tasks, respectively.

In this paper, the resistance of garlic digging shovel was used as target, the simulation experiment scheme was designed by quadratic regression orthogonal rotation combination, and the parameters of garlic digging shovel were optimized by EDEM simulation. The regression equation for the resistance of garlic digging shovel was established based on the simulation experiment results. According to the regression equation, the primary and secondary influences of the factors on the resistance were shown as below: the width, length and sharp angle of the shovel had a significant influence on the resistance of the garlic digging shovel; the width of the shovel was greater than the length of the shovel; the length of the shovel was greater than the angle of the shovel; and the angle of the shovel is greater than the speed of the shovel.

To reduce the complex mechanism and difficult control of multi-legged robot, the overall design of the new multi-footed wheel robot was carried out according to the previous research results—multi-footed wheel mechanism. A new multi-wheeled robot model was built by using SolidWorks. External data planning was carried out for dynamic force data curve of the motor. In ADAMS, the kinematics characteristics and data of the multi-footed wheel robot were obtained through simulation analysis of the model by interpolation method. The rationality of mechanism design and driving stability were verified by simulation analysis.

This article studied the effect of different conveyor belt speeds on garlics transportation in the process of garlics harvesting and lateral conveying. The lateral conveying process of garlics is simulated by EDEM, and the optimal transportation speed interval is obtained by simulating different transportation speeds. The simulation results show that the conveying speed range of conveyor belt is 2.46-3.46 m /s when the plant spacing of garlics is 0.1 m, the row spacing is 0.12 m, and the 16 harvesting units of garlics combine harvester harvest at the speed of 2 m/s at the same time, which can effectively avoid the problem of garlics accumulation on the lateral conveyor belt when the conveying speed is too slow, and garlics leakage and garlics secondary damage problems when the delivery speed is too fast. The research can provide some theoretical reference for the design of garlics transversal conveyer and the selection of conveying speed.

The pressure capacity of an ultrahigh pressure cylinder is the bottleneck of the development of ultrahigh pressure vessels. In order to improve the pressure capacity of the cylinder, the finite element software was used to analyse the single-layer thick-walled cylinder and the split cylinder under the same load. Research results show that the split cylinder has higher pressure capacity than that of single-layer thick-walled cylinder. Simulation of the different number subsections of the same size split cylinder show that the number’s increase of the subsections can effectively reduce their Von Mises stress under the same load.

The traditional research and development cycle of garlic operation machinery chassis is long and the production cost is high, which hinders the development of this product. We can find out the condition that sub-functions become independent modules by the judgment matrix. Sub-functions can be divided based on correlation calculation and module division. The modular design of garlic operation machinery chassis is not only conducive to the innovative design and upgrading but also reducing the cost and the products.