Word embedding method based on pre-training still has some defects in the stability and the quality of low-frequency words. The authors propose a new word embedding method based on Hownet. First, based on the sememe independence assumption, all sememes of Hownet are specified in an Euclidean Space’s standard orthogonal basis to initialize all sememe vectors. Secondly, utilizing the relationship between word and sememe defined in the Hownet, each word vector representation can be regarded as a subspace projection by related sememes. Finally, a deep neural network model is put forward to learn word representations. The experimental results indicate that proposed word embedding method based on Hownet obtained comparable results in the two standard evaluation tasks including the word similarity computation and the word sense disambiguation.
A simple approach to estimating Gini coefficient based on Lorenz curve is proposed to solve the problem that the concentration index replacing the Gini coefficient results in deviation from imbalance measurement of spatial distributions and size distributions. In view of the scale-free distribution phenomena in the complex social and economic system, a logarithmic function of convex Lorenz curve is derived from the pure Zipf’s distribution by means of Euler’s formula for the sum of harmonic sequence. Then an approximate formula to estimate Gini coefficient can be constructed by using the parameters of the logarithmic Lorenz curve model. The formula is applied to the cities in Beijing, Tianjin and Hebei (Jing-Jin-Ji) region, and the Gini coefficients of 22 years are evaluated by the night lighting data. The results show that there is a significant difference between the Gini coefficient and the centralization index. A conclusion can be reached that the centralization index is applicable to the distributions with characteristic scales, while the proposed formula are suitable for the scale-free distributions. This work will help researchers to understand the scopes of application of the imbalance measurements and provide a reference for further developing the direct estimation methods of Gini coefficient.
To study the effects of land use pattern change on habitat quality in the background of urbanization, InVEST model was applied to calculate the contributions of habitat threats (urban area, rural construction area, main roads and agriculture land) to habitat degradation, and to assess habitat quality changes between 1990 and 2010 in Beijing. The results showed that the most severe habitat degradation problems occurred in Haidian, Chaoyang, Shijingshan and Fengtai districts, Yongding River, Chaobai River, and the mountain-plain border area (at an elevation about 75-100 m). Agriculture lands had the largest effect on habitat degradation, followed by urban areas, rural construction areas and main roads. During these two decades, the contribution of agriculture lands decreased from 77.79% to 61.15%, but the contribution of urban areas increased from 18.10% to 31.54%. Between 1990 and 2000, habitat quality declined dramatically in Haidian, Chaoyang, Shijingshan and Fengtai districts, which were located at the rim of Dongcheng and Xicheng districts. The total habitat quality value decreased by 2.3%. During 2000 and 2010, the deterioration of habitat quality expanded to most of the plain areas. In the mountain area of Fangshan and the mountain-plain border area (at an elevation about 75-100 m) in Miyun, Huairou, Changping and Fangshan, the value decreased significantly. However, habitat quality had been restored notably in northern and western mountain areas. The total habitat quality value declined by 1.2%, which indicated that habitat deterioration had been mitigated to some extent compared with the last decade. For habitat conservation and urban planning in the future, more attention should be paid to mountain-plain border areas. The current study played an active role in land use policy-making and biodiversity conservation.