The social force model is one of the most commonly used models for microscopic modeling of pedestrian movement and crowd evacuation. A large number of researchers have conducted continuous research on it, which makes the social force model suitable for pedestrian movement and crowd evacuation modeling in a variety of complex environments. The prerequisite for the effective use of the social force model in actual scenarios is the need to check and verify the social force model according to the application scenario. Therefore, it is necessary to confirm the correct realization of the social force model structure from the perspective of theory and practical application. The expected velocity parameter has an important impact on the movement of the crowd, including evacuation time and flow rate, so the expected velocity also needs to be considered during model validation. In addition, the performance of the social force model is also sensitive to the expected rate, measured by entropy. However, research and analysis on the desired rate setting is still lacking. In this paper, the optimal selection of the expected rate sub-model in the social force model is carried out in a data-driven manner, so that the social force model can more realistically reflect the movement of crowds, so that it can be better used in crowd control based on modeling and simulation.

With the increasing scale of the Internet medical market, a large number of platforms have started to provide free medical diagnosis services to users. The number of questions on medical Q&A platforms is growing rapidly, and there are a lot of repetitive and similar questions, which can no longer meet the demand of users to get fast and accurate answers only by the manual answers of medical experts. To solve the above problems, this paper proposes a Chinese medical short text matching model based on fine-tuning BERT-Attention-BiLSTM (FBAB), use fine-tuning BERT model for representation of text, the attention mechanism is used to obtain the interaction information between sentences, use the Bi-directional Long Short-Term Memory (BiLSTM) to fuse the contextual information of the 2 sentences., and the feature information is further extracted by maximum pooling and mean pooling to obtain the sentence-level semantic representation, which is input to the prediction layer after stitching, and the softmax classification function is used to calculate the probability that the two sentences are similar. Experiments show that compared with the classical deep short text matching model, the FBAB model proposed in this paper shows better results on the medical short text corpus, which proves the effectiveness and feasibility of the model.

This paper utilizes the BERTbased-BiLSTM-CRF models to complete Chinese named entity recognition tasks, including finetuned and unfinetuned BERT models. Use the pre-training model BERT(Bidirectional Encoder Representations from Transformers), a BiLSTM(Bi-directional Long Short-Term Memory) network and CRF(Conditional Random Field) to perform NER(Named Entity Recognition) on Chinese. Tested on the people-daily-ner-pretreatment corpus, compared with other models, the BERTbased models can effectively identify entity information, and the evaluation metrics on the dataset have been improved.

Named entity recognition (NER) is an important research direction in natural language processing (NLP). Traditional machine learning algorithms in NER have problems such as low accuracy, highly dependent feature design, poor domain adaptability, and inability to handle the different contexts of multiple meanings of the term in recognizing Chinese entities. Based on these problems, this paper adopts a method based on the BERT-CRF model in Chinese NER. The BERT preprocessing language model generates word vectors that represent contextual semantic information, automatically extract numerous word-level features and semantic features in text, and decodes through the CRF layer generates entity tag sequences. In this paper, the BERT model has been fine-tuned to make the model perform better on NER tasks, and the experimental verification is carried out on the People’s Daily dataset, and the F1 value reaches 94.5%.

MOTIVATION: The next-generation high-throughput sequencing technologies, especially from Illumina, have been widely used in re-sequencing and de novo assembly studies. However, there is no existing software that can simulate Illumina reads with real error and quality distributions and coverage bias yet, which is very useful in relevant software development and study designing of sequencing projects. RESULTS: We provide a software package, pIRS (profile-based Illumina pair-end reads simulator), which simulates Illumina reads with empirical Base-Calling and GC%-depth profiles trained from real re-sequencing data. The error and quality distributions as well as coverage bias patterns of simulated reads using pIRS fit the properties of real sequencing data better than existing simulators. In addition, pIRS also comes with a tool to simulate the heterozygous diploid genomes. AVAILABILITY: pIRS is written in C++ and Perl, and is freely available at ftp://ftp.genomics.org.cn/pub/pIRS/.