Internal network: Using text analysis + tag matching method to improve recruitment market matching efficiency

Absrtact: White-collar recruitment is a highly fragmented, non-standard market. Standing on one end of the job seekers, the face of a huge amount of recruitment information can not make a choice, only to go without a mind cast, which caused the whole market garbage information flooded. At one end of the recruiting business, you need to face

White-collar recruitment is a highly fragmented, non-standard market. Standing on one end of the job seekers, the face of a huge amount of recruitment information can not make a choice, only to go without a mind cast, which caused the whole market garbage information flooded. At one end of the recruiting business, there is also a huge job resume to sift through thousands of resumes to screen out the most suitable candidates. The situation here is that the more well-known enterprises, the more ineffective resumes will focus on you, resulting in large enterprise HR CV screening costs too high, simple and rough screening process. Not so well-known enterprises, often unable to collect enough resumes, you need to advertise for headhunting, pay additional costs.

The imbalance of attention distribution, the fragmentation of supply and demand information, this is the root of the low efficiency of white-collar recruitment market, and focus on the Internet recruitment network, then try to use text Analysis + tag matching method to make this situation reversed:

For job seekers at one end, the intranet will collect his CV and career appeal (industry, salary, position, etc.), through the text analysis of the resume, to extract the job applicant's label.

For the end of the enterprise, it will collect the company's employment claims, job information, industry attributes, product information and media coverage, the same is the use of text statistics, analysis of the way to extract the label on the enterprise. Because of the complexity of the enterprise data, the internal employ the machine first screen, the artificial tuning method to correct and iterative algorithm. About Two-thirds of the 6,000 tags that are currently extracted are artificially generated.

When highly unstructured job seekers and enterprise information become relatively structured, the hiring engine can establish a more accurate link between the two. The internal network will build a link between the labels, creating a career label map. Job seekers and businesses can find their place on the map, the closer they are, the higher the probability of their potential match.

The typical user experience of an internal network is that job seekers simply submit resumes and job requests, the system will automatically generate a job referral list, which will tell you the characteristics of the business (such as high salary and beauty), and will also tell you the matching score, as a basis for user orientation. While HR simply posts, the system will sort by match, telling you which resumes may need your attention. Because the different types of enterprises to recruit the same position needs different, so the basis of job matching should be "tag combination." For example, a company doing to B software ("to B" + "Software"), and a company doing to C software ("to C" + "Software"), when both release a "product manager" position, the results should be personalized.

The founder of Xiao Heng is a master of computer science at Peking University, after graduating from work, worked in Kyocera and Matsushita Electrical project manager. Xiao Heng, who founded a talent-Dispatch and software outsourcing company for Japanese companies, is said to have performed well. After a number of career adjustments, two of startups were recruited in April 12. We are now looking at the online line in March this year (previously tried to do micro-letter recruitment), the second month to get an agency of millions of Yuan Angel.

During the 11 period, a major change was made to the internal employment Network. Xiao Heng Introduction, the current use of internal network enterprise users have more than 6,000, after the revision of the UV has reached about 10,000, individual users have more than 100,000. Next they will try to reach out to the workplace.

There are some interesting things to do about job matching with big data, such as:

1, can bind the job applicant's social account, through the social data for the job seekers to play a richer label, while assessing his network value. It is clear that well-connected candidates deserve more attention, while in matching positions, BD and market-class jobs should be ranked higher. This is the previous reliance on a paper resume can not be achieved, to do social networking in the workplace, in this direction has done a better example.

2, enterprises can reach the potential candidates, activate the other side of the willingness to change the passive and so on the door for the initiative to recruit people.

3, let the enterprise will be the existing staff background information upload, in the engine run a lap, to build the enterprise existing staff quality model. Combined with the company's historical performance, analyze the effectiveness of the model, and improve the direction, so that more accurate evaluation of the need to attract a certain type of newcomers to enter-this is perhaps a "quantitative hr"?