Due to the polysemy and complexity of Chinese language, machine reading comprehension has always been a challenging task. To improve semantic understanding robustness models, we propose model that utilizes adversarial training algorithms Permuted Language Model (PERT). Firstly, employ PERT pre-training embed paragraphs questions into vector space obtain corresponding sequential representations. Secondly, use multi-head self-attention mechanism extract key textual information from sequence...

Aiming at the problem of long soil–water characteristic curve (SWCC) testing times and difficulty prediction accuracy in complex environments, this paper establishes a SWCC model based on neural network machine learning algorithm which can take into account influence multiple factors such as temperature, deformation, salinity. The input layer reflect physical properties soil external environment, while suction is taken an variable, turn directly obtain water content under corresponding...

Apple orchards, as an important center of economic activity in forestry special crops, can achieve yield prediction and automated harvesting by detecting locating apples. Small apples, occlusion, dim lighting at night, blurriness, cluttered backgrounds, other complex scenes significantly affect the automatic estimation To address these issues, this study proposes apple detection algorithm, “YOLOv5-ACS (Apple Complex Scenes)”, based on YOLOv5s. Firstly, space-to-depth-conv module is...

The core of the classic collaborative filtering algorithms about similar calculation are designed on basis "user-item matrix" model. This paper proposes an improved algorithm cube" model, which takes care factor data produced when user purchased item. attaches corresponding weight to date factor, and then is used similarity. method increases accuracy recommendation system significantly.

Constructing click models and extracting implicit relevance feedback information from the interaction between users search engines are very important to improve ranking of results. Using neural network model users' behaviors has become one effective methods construct models. In this paper, We propose a new Filter-Enhanced Transformer Click Model (FE-TCM) for web search. The uses as backbone feature extraction add filter layer innovatively. Firstly, in order reduce influence noise on user...

At present, most of association rule mining are static algorithms. However, in real Web applications, new transactions usually incrementally inserted into database, the previous algorithms couldn’t satisfy rapid growth like this. In this paper, we proposed incremental called Pre-FP based on PUFP-tree and Per-large-itemsets. This algorithm don’t need rescan original database when surpass fixed threshold, especially each time small fraction inserted, can achieve a good execution efficiency for...

To better understand users’ behavior patterns in web search, numerous click models are proposed to extract the implicit interaction feedback. Most existing heavily based on information model user behaviors, ignoring impact of explicit between queries and documents search sessions. In this paper, we fully consider topic relevance sessions propose a novel relevance-aware (TRA-CM) for search. TRA-CM consists estimator an examination predictor. The predictor context encoder. predictor, utilize...

The value of data is becoming increasingly important in the Big Data era, and users are eager to retrieve high-quality data. However, current popular data-retrieval systems still adopt technical framework traditional document retrieval, they lack user-oriented quality-control judgment mechanisms. Accordingly, this study first summarizes research on data-quality (DQ) characteristics, user DQ behavioral characteristics. Second, based decision theory cognitive science, a model interaction...

Based on probability, the paper recommends a new inference method for validating trusted forensics. Using graph theory, original data is formalized to abstract data, and shown as some nodes in graph. The operation actions forensics methods are abstracted edge sets. With these, using probability of each node calculated its credence. these methods, we can finally deduce evidence

For improving the validity and efficiency of 3D reconstruction engineering drawings, this paper presents a new method feature recognition based on multi-granularity information acquisition. In method, by analyzing line-frames combined primitive views hidden semantics in drawing, relationship graphics primitives can be identified, then imitating partition figuration analysis, view loop separation individual is realized inertial search with some priority. This algorithm actualized AutoCAD 2008...

Based on management experimental method, the paper studies logistics customer complaint management. There are two purposes in our research. One is to analyze factors influencing complaining behaviors and service staffs how deal with complaints. The other choose suitable staffs. study helpful improve abilities of service, capable afford theoretical guide for selecting high level enterprises.

Current e-education resource and service is simple without integrative design, which cannot satisfy the needs of e-education. This paper introduces in information age development, illustrates advantages Web-based university libraries, analyzes necessity platform (WEP) construction, explains components WEP function characteristics, reports on analysis problems construction proposes relevant solutions.

The evaluation index weight is the key factor to determine reasonable result. In order effectively overcome ambiguity and uncertainty of analytic hierarchy process in importance scoring comparison, this paper combines intuitionistic fuzzy set propose a new subjective determination method. This method uses characteristics that can express three states support, opposition hesitation at same time, fully reflects fuzziness human judgment comparison process, more reasonably accurately....

The extension of intelligent library knowledge service is the further development and upgrading digital service. It supports comprehensive ability in teaching, scientific research, social management, cultural construction development. should change from resource-driven to service-oriented. study found that 52 percent thought libraries provide innovative services,45 it would help readers understand high technology,23 helped them find jobs, 14 improved their job skills.