Journals
  Publication Years
  Keywords
Search within results Open Search
Please wait a minute...
For Selected: Toggle Thumbnails
Difference between Multi-modal vs. Text Pre-trained Models in Embedding Text
SUN Yuchong, CHENG Xiwei, SONG Ruihua, CHE Wanxiang, LU Zhiwu, WEN Jirong
Acta Scientiarum Naturalium Universitatis Pekinensis    2023, 59 (1): 48-56.   DOI: 10.13209/j.0479-8023.2022.074
Abstract326)   HTML    PDF(pc) (2107KB)(187)       Save
This paper provides quantitative comparison between the text embedding of a text pre-trained model (i.e., RoBERTa) and a multi-modal pre-trained model (i.e., WenLan). Two quantitative comparison methods are proposed, in an embedding space: representing the semantics of a word using the set of ?-nearest words to it and then analyze the semantic changes of the word in the two spaces using the Jaccard similarity of the two sets; forming pairs between each word and its nearest ? words to analyze the relationship. The results show that the multi-modal pre-training brings more semantic changes for more abstract words (e.g., success, love), and the multi-modal pre-trained model can better differentiate antonyms and discover more hypernyms or hyponyms, while text pre-training works better in finding synonyms. Moreover, multi-modal pre-trained model can construct a more extensive associative relationship between words.
Related Articles | Metrics | Comments0
Hydrology Effect Evaluation of Green Infrastructure Based on HSPF: A Case Study on Dalang River Basin
XING Yinghua, CHENG Xiang, QIN Huapeng, ZHAO Zhijie
Acta Scientiarum Naturalium Universitatis Pekinensis    2018, 54 (5): 1053-1059.   DOI: 10.13209/j.0479-8023.2018.019
Abstract859)   HTML    PDF(pc) (17362KB)(156)       Save

Taking Dalang River Basin, Shenzhen city of Guangdong Province as an example, HSPF model was used to simulate hydrological effects of rainfall runoff under different scenarios. The results showed that runoff rate of the efficacy maximization, economical and moderate scenario were decreased by 34.9%, 14.2% and 28.5% than that in background scenario. The peak value of these three scenarios were lower 40.5%, 19.8% and 33.0% than that in background scenario. Base flow of these three scenarios were higher 88.9%, 11.1% and 44.4% than that in background scenario. The economical scenario didn’t reach good effect. The effect of moderate scenario was better than economical scenario and inferior to efficacy maximization scenario.

Related Articles | Metrics | Comments0
Temporal and Spatial Distribution Characteristic Research of Water Environmental Capacity in Moyang River Basin
CHENG Xiang, ZHAO Zhijie, QIN Huapeng, SONG Baomu, YU Xiangying, HE Kangmao
Acta Scientiarum Naturalium Universitatis Pekinensis    2016, 52 (3): 505-514.   DOI: 10.13209/j.0479-8023.2016.029
Abstract1325)   HTML    PDF(pc) (2249KB)(1255)       Save

This study takes Moyang River basin, which is lack of hydrologic data, as the research object to simulate the temporal and spatial distribution of water flow through HSPF (hydrological simulation program-Fortran) model, and calculates the temporal and spatial distribution of chemical oxygen demand (COD) and ammonia nitrogen water environmental capacity using one-dimensional steady-state water quality mathematical model. Results show that 1) HSPF model’s yearly and monthly hydrological errors are below 15%, Nash-Sutcliffe coefficient is over 0.9; the relative error of the water quality model is around 10%, Nash-Sutcliffe coefficient over 0.8. 2) Under 90%, 50% and 10% assurance rate, COD capacity of Moyang River basin is 164500 t, 218400 t and 249700 t respectively, and ammonia nitrogen is 5100 t, 8800 t and 11400 t respectively. Affected by seasonal runoff fluctuations, difference of water environment capacity between dry season and wet season is obvious. The variation of water environment capacity in January is minimum and June the maximum. 3) Bearing capacity of main stream of Moyang River is larger than that of the primary and secondary tributaries. Therefore, in some basins where hydrological data is lack, HSPF model can be applied to simulate the hydrology, analysis time and space distribution of water environmental capacity, and provide guidance for the establishment of total amount control scheme of water environmental capacity.

Related Articles | Metrics | Comments0
Concept Template of Combining Attributes with Attributive Values
CHENG Xianyi,SHI Quan,SHEN Xuehua,TIAN Yuhe
Acta Scientiarum Naturalium Universitatis Pekinensis   
Abstract631)      PDF(pc) (445KB)(277)       Save
According to the task of Chinese discourse analysis and practice, combined with traditional study, the authors propose clause as basic discourse unit and give its definition from the structure, function, form etc. The authors analyse the The authors extract triad 〈concept, attribute, property value〉 based on ontology. Concept is represented as a vector which is combined with attributes and attributive values based on vocabulary clustering. Comparasion between conceptual template based on attributes and based on attributive values is performed. The study shows that the concept template of combination attributes with attributive values is superior to any single template.
Related Articles | Metrics | Comments0
Automatic Extraction of Coastline from TM Image Integrating Texture and Spatial Relationship
ZHOU Yanan,ZHU Zhiwen,SHEN Zhanfeng,CHENG Xi
Acta Scientiarum Naturalium Universitatis Pekinensis   
Verifiable Multi-secret Sharing Scheme Based on Homogeneous Linear Recursion
CHEN Yangkui,YU Jia,CHENG Xiangguo,HAO Rong,LIU Hongyan,LI Xin
Acta Scientiarum Naturalium Universitatis Pekinensis   
Abstract616)            Save
A new multi-secret sharing scheme based on homogeneous linear recursion is proposed, and then it is converted into a verifiable scheme. In the distribution phase, very few of public values are needed to publish. In the recovery phase, each participant only needs to submit a pseudo shadow instead of his secret shadow, and his secret shadow cannot be disclosed. When secrets are changed, secret shadows don?t need to be redistributed, which makes secret shadow able to be used multiple times. The proposed scheme has many advantages, for example, the secret shares can be used multiple times and the scheme publishes very few parameters as well as the reconstructed polynomial has a low degree. This makes the proposed scheme more efficient. Therefore, it better satisfies demands of various applications.
Related Articles | Metrics | Comments0