JOURNAL OF SHANDONG UNIVERSITY(NATURAL SCIENCE) ›› 2021, Vol. 56 ›› Issue (1): 83-90.doi: 10.6040/j.issn.1671-9352.4.2020.131

•   • Previous Articles     Next Articles

Research on matching resumes and positions based on Arc-LSTM

Fei-fei XU1(),Yun-jie XU2   

  1. 1. College of Computer Science and Technology, Shanghai University of Electric Power, Shanghai 200090, China
    2. Process and IT Platform Software Development BU, COMAC Shanghai Aviation Industrial(Group), Shanghai 200232, China
  • Received:2020-06-15 Online:2021-01-01 Published:2021-01-05

Abstract:

An automatic scheme is proposed, which combines the deep neural network Arc-LSTM with Word2Vec model. In this paper, the CBOW model is used to extract resume text features. A new deep neural network, Arc-LSTM is put forward, which is optimized by ArcReLU. The experimental results show that Arc-LSTM can effectively improve the classification accuracy and convergence speed, and achieve the exact match between Chinese resumes and positions.

Key words: deep learning, activation function, job matching, ArcReLU, Arc-LSTM

CLC Number: 

  • TP18

Fig.1

ArcReLU activation function"

Fig.2

Arc-LSTM framework map"

Table 1

Resume category"

职位类别 职位名称
管理及产品岗 数据总监, 分析挖掘, 算法、平台研发
算法工程 推荐系统, 搜索排序, 语音语义, 计算机视觉, 自动驾驶, 深度学习, 卫星导航, 机器人, 通信算法, 其他
数据分析 商业、经营分析, 用户、产品分析, 统计分析, 数据挖掘, 广告, 风控, 量化
数据开发 数仓架构, 大数据ETL, 传统ETL, 大数据开发
平台架构 大数据架构, 平台工具开发, 大数据运维

Table 2

Data set"

数据集 条目数 属性 决策属性
Car Evaluation 1 728 buying, maint, doors, persons, lug-boot, safety class values
Adult 9 502 age, workclass, fnlwgt, education, marital-status, relationship, race sex
Avila 12 647 intercolumnar distance, upper margin, lower margin, exploitation, row number, modular ratio, interlinear spacing, weight, peak number class

Table 3

Dictionary mapping"

实际标题 映射类别
名字 姓名
姓名 姓名
性别 性别
工作地点 城市
工作经验 经验
工作经历 经验
出生日期 年龄
教育背景 教育背景
教育情况 教育背景

Fig.3

Sample of Chinese resume"

Fig.4

CBOW model"

Fig.5

Results of word embedding"

Table 4

Calculated consumption of each model on Car Evaluation"

No LSTM Re-LSTM Arc-LSTM
1 0:04:14.76 0:03:58.32 0:04:18.14
2 0:05:37.39 0:05:15.26 0:05:31.86
3 0:05:02.01 0:05:16.63 0:05:25.29
4 0:05:10.52 0:05:01.50 0:05:18.12
5 0:05:01.46 0:04:57.82 0:05:02.50
6 0:05:16.82 0:05:05.30 0:05:20.50
7 0:05:13.66 0:05:03.84 0:05:16.57
8 0:05:13.38 0:05:00.40 0:05:17.63
9 0:05:19.55 0:05:13.35 0:05:21.16
10 0:05:17.39 0:04:59.50 0:05:16.65
AVG 0:05:08.69 0:04:59.19 0:05:12.84

Fig.6

ROC comparison chart"

Table 5

AUC of each model on Car Evaluation"

评判指标 LSTM Re-LSTM Arc-LSTM
AUC 0.789 1 0.799 8 0.834 4

Table 6

Calculated consumption of each model on Car Evaluation %"

评判指标 LSTM Re-LSTM Arc-LSTM
训练精度均值 77.39 77.98 79.42
测试精度均值 76.50 77.54 79.22

Table 7

Calculated consumption of each model on Adult"

No LSTM Re-LSTM Arc-LSTM
1 0:13:30.83 0:12:04.69 0:13:36.12
2 0:13:57.20 0:13:06.58 0:12:37.52
3 0:13:00.34 0:12:14.62 0:13:15.56
4 0:15:09.22 0:14:07.00 0:16:32.00
5 0:13:49.67 0:13:15.03 0:14:16.82
6 0:14:51.88 0:13:50.40 0:12:45.10
7 0:14:37.62 0:14:01.51 0:13:17.88
8 0:14:48.64 0:14:01.78 0:14:40.52
9 0:16:17.18 0:16:25.57 0:16:55.70
10 0:15:20.20 0:15:31.25 0:16:14.50
AVG 0:15:22.79 0:13:51.84 0:14:24.20

Fig.7

ROC comparison chart"

Table 8

AUC of each model on Adult"

评判指标 LSTM Re-LSTM Arc-LSTM
AUC 0.793 1 0.804 8 0.815 1

Table 9

Calculated consumption of each model on Adult %"

评判指标 LSTM Re-LSTM Arc-LSTM
训练精度均值 77.19 78.41 80.18
测试精度均值 77.12 78.23 79.74

Table 10

Calculated consumption of each model on Avila"

No LSTM Re-LSTM Arc-LSTM
1 0:59:15.31 0:57:52.23 0:58:19.54
2 0:56:59.87 0:54:82.63 0:55:09.32
3 1:12:32.65 1:11:43.98 1:12:13.55
4 0:06:21.82 1:02:51.02 1:04:02.02
5 1:06:14.37 1:02:00.46 1:05:08.46
6 0:54:07.49 0:52:07.87 0:52:29.16
7 0:53:44.59 0:51:32.24 0:56:51.12
8 0:52:35.66 0:49:16.91 0:52:32.26
9 0:56:21.39 0:54:33.67 0:55:39.54
10 1:13:36.99 1:10:53.54 1:11:43.56
AVG 1:01:17.54 0:58:49.08 0:59:69.85

Fig.8

ROC comparison chart"

Table 11

AUC of each model on Avila"

评判指标 LSTM Re-LSTM Arc-LSTM
AUC 0.742 3 0.755 2 0.773 7

Table 12

Calculated consumption of each model on Avila %"

评判指标 LSTM Re-LSTM Arc-LSTM
训练精度均值 75.66 76.69 78.94
测试精度均值 75.05 76.59 78.37

Table 13

Experimental Results of each model on Resume %"

算法 准确率 召回率 F1值
LSTM 76.68 67.26 71.21
Re-LSTM 81.41 71.43 73.51
Arc-LSTM 87.83 76.15 77.18
1 KATZ B, LIN J. Selectively using relations to improve precision in question answering[C]//Proceedings of the EACL-2003 Workshop on Natural Language Processing for Question Answering. Budapest: EACL, 2003: 43-50.
2 朱倩, 程显毅, 韩飞. 汉语句子语义三维表示模型[J]. 智能系统学报, 2009, 4 (2): 122- 130.
ZHU Qian , CHENG Xianyi , HAN Fei . A three-dimensional representative model of Chinese sentence semantics[J]. CAAI Transactions on Intelligent Systems, 2009, 4 (2): 122- 130.
3 张宜浩, 朱小飞, 徐传运, 等. 基于用户评论的深度情感分析和多视图协同融合的混合推荐方法[J]. 计算机学报, 2019, 42 (6): 1316- 1333.
ZHANG Yihao , ZHU Xiaofei , XU Chuanyun , et al. Hybrid recommendation approach based on deep sentiment analysis of user reviews and multi-view collaborative fusion[J]. Chinese Journal of Computers, 2019, 42 (6): 1316- 1333.
4 GULCEHRE C, MOCZULSKI M, DENIL M, et al. Noisy activation functions[C]//Proceedings of the 33rd International Conference on Machine Learning. New York: ICML, 2016.
5 GOMAR S, MIRHASSANI M, AHMADI M. Precise digital implementations of hyperbolic tanh and sigmoid function[C]//Conference on Signals, Systems & Computers. Alberta: IEEE, 2016: 1586-1589.
6 廖祥文, 陈泽泽, 桂林, 等. 基于多任务迭代学习的论辩挖掘方法[J]. 计算机学报, 2019, 42 (7): 1524- 1538.
LIAO Xiangwen , CHEN Zeze , GUI Lin , et al. An argumentation mining method based on multi-task iterative learning[J]. Chinese Journal of Computers, 2019, 42 (7): 1524- 1538.
7 许赟杰, 徐菲菲. 基于ArcReLU函数的神经网络激活函数优化研究[J]. 数据采集与处理, 2019, 34 (3): 517- 529.
XU Yunjie , XU Feifei . Optimization of activation function in neural network based on ArcReLU function[J]. Journal of Data Acquisition & Processing, 2019, 34 (3): 517- 529.
8 WU H . Global stability analysis of a general class of discontinuous neural networks with linear growth activation functions[J]. Information Sciences, 2009, 179 (19): 3432- 3441.
doi: 10.1016/j.ins.2009.06.006
9 李亚超, 熊德意, 张民. 神经机器翻译综述[J]. 计算机学报, 2018, 41 (12): 2734- 2755.
doi: 10.11897/SP.J.1016.2018.02734
LI Yachao , XIONG Deyi , ZHANG Min . A survey of neural machine translation[J]. Chinese Journal of Computers, 2018, 41 (12): 2734- 2755.
doi: 10.11897/SP.J.1016.2018.02734
10 陈建廷, 向阳. 深度神经网络训练中梯度不稳定现象研究综述[J]. 软件学报, 2018, 29 (7): 2071- 2091.
CHEN Jianting , XIANG Yang . Survey of unstable gradients in deep neural network training[J]. Journal of Software, 2018, 29 (7): 2071- 2091.
[1] Chang-ying HAO,Yan-yan LAN,Hai-nan ZHANG,Jia-feng GUO,Jun XU,Liang PANG,Xue-qi CHENG. Dialogue generation model based on extended keywords information [J]. JOURNAL OF SHANDONG UNIVERSITY(NATURAL SCIENCE), 2019, 54(7): 68-76.
[2] LIU Biao, LU Zhe, HUANG Yu-wei, JIAO Meng, LI Quan-qi, XUE Rui. Comparative study on neural network structures in power analysis [J]. JOURNAL OF SHANDONG UNIVERSITY(NATURAL SCIENCE), 2019, 54(1): 60-66.
[3] LIU Ming-ming, ZHANG Min-qing, LIU Jia, GAO Pei-xian. Steganalysis method based on shallow convolution neural network [J]. JOURNAL OF SHANDONG UNIVERSITY(NATURAL SCIENCE), 2018, 53(3): 63-70.
[4] PANG Bo, LIU Yuan-chao. Fusion of pointwise and deep learning methods for passage ranking [J]. JOURNAL OF SHANDONG UNIVERSITY(NATURAL SCIENCE), 2018, 53(3): 30-35.
[5] LIU Ming, ZAN Hong-ying, YUAN Hui-bin. Key sentiment sentence prediction using SVM and RNN [J]. JOURNAL OF SHANDONG UNIVERSITY(NATURAL SCIENCE), 2014, 49(11): 68-73.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
[1] QIN Zhao-yu,LIU Shi-lian*,YANG Yin-rong,LIU Fu-jun,LI Jian-yuan,SONG Chun-hua . Technology exploration for proteomics analysis in hepatopancreas of shrimp (Fenneropenaeus chinensis) with white spot syndrome[J]. J4, 2007, 42(7): 5 -08 .
[2] ZHANG Fang-guo. Elliptic curves in cryptography: past, present and future…[J]. J4, 2013, 48(05): 1 -13 .
[3] XUE Yan-Bei, YANG Bei, CHEN Zhen-Xiang. Structural health monitoring of civil engineering based on wavelet analysis[J]. J4, 2009, 44(9): 28 -31 .
[4] DING Mei,FENG Jun-e and WANG Zhi-hong . The minimum time problem of crossing river with restricted conditions[J]. J4, 2007, 42(3): 23 -28 .
[5] HE Xiao-li,FU Wen-qing and LI Sheng-gang . Local connectedness of L-pretopological spaces[J]. J4, 2007, 42(8): 58 -61 .
[6] XING Jian-min . LR-normal orthogroup congruence on regular semigroups[J]. J4, 2006, 41(1): 41 -44 .
[7] YU Wen-guang1, HUANG Yu-juan2. Ruin probability in an interfered multi-type risk model of a variable ruin limit[J]. J4, 2011, 46(3): 58 -62 .
[8] DING Wei-ping1,2,3, WANG Jian-dong2, DUAN Wei-hua2, SHI Quan1. Research of cooperative PSO for attribute reduction optimization[J]. J4, 2011, 46(5): 97 -102 .
[9] . Asymptotic behavior of  nonoscillatory solutions of second  order nonlinear differential equation with perturbation[J]. J4, 2009, 44(2): 19 -23 .
[10] ZHU Li-ping, LI Hong-qi, YANG Zhong-guo, LIU Qiang. An information extraction method for scientific literature introduction[J]. JOURNAL OF SHANDONG UNIVERSITY(NATURAL SCIENCE), 2015, 50(07): 23 -30 .