This invention relates to the field of human relations and more particularly to a system for screening applicants for hire.
Currently, many hiring organizations pour through countless résumés looking for qualified candidates to fill open positions. It has been calculated that the cost to find and hire a qualified candidate is often over 4,000.00 US Dollars.
In the past, the primary tool for making preliminary matches between candidates and open positions is a résumé along with a cover letter. The résumé often cites facts about the candidate such as education/training, prior positions and activities, knowledge/abilities, etc. Generally, the same résumé is sent to all companies with open positions, but the cover letter is tailored to each individual company being sought. The cover letter tells the hiring manager why the candidate would like to work for the company. Often relating education and experience to the needs of the open position.
For many companies, a human resource manager reads many written cover letters and résumés, looking for those that match the open position, creating a shorter list of candidates. The hiring manager then reviews this shorter list of résumés and written cover letters to produce an even shorter list, then, the jobseekers on this even shorter list are contacted, possibly for an initial discussion or to schedule an interview. Next, each candidate is interviewed by the hiring manager, human resource manager, and/or other person in the company. Hopefully, the even shorter list has candidates that are good matches for the open position, as the cost for each interview is quite high, including the time for someone to schedule the interview and, sometimes, transportation and lodging costs for the candidate, especially when the candidate lives a distance from the company location.
Recently, automated tools have been deployed to recognize text in the printed résumé and cover letter (e.g., optical character recognition) and the recognized text is scanned and parsed for certain elements that are desired by the hiring manager. For example, if the position is for a sales person, then résumés with keywords/phrases of “outgoing” or “enjoy working with clients” will obtain higher scores in the automated analysis. If, instead, the position is writing software, then résumés with keywords/phrases of “works independently” will obtain a higher score. Such tools help the human resource manager by narrowing down a huge number of résumés to a more manageable number.
Such résumé scanning tools work help with online submissions and paper submissions, but are incapable of parsing a video/audio or audio cover letter that is prepared to tell the human resource manager and/or hiring manager why the jobseeker thinks they are a good match for the open position. Existing scanning programs are now only capable of scanning a paper résumé, but not the video and/or audio. The existing scanners are easy to fool by jobseekers who know what keywords are needed to improve their chances of being called for a phone interview or an in-person interview.
What is needed is a system that will process a résumé and a video and/or audio cover letter to extract skills and attributes of the jobseeker and analyze the skills and attributes with respect to open position for a job.
In one embodiment, a system using artificial intelligence for selecting candidates for an open position is disclosed including. The system includes a computer with software running on the computer. The software receives the open position and a plurality of inputs from jobseekers. Each input includes a résumé and a video cover letter, the video cover letter has audio and video. The software parses each résumé into résumé data for an associated jobseeker; parses the open position into open position requirements; parses the video from each of the video cover letters for the associated jobseeker and extracts video attributes from the video; and parses the audio from each of the video cover letters for the associated jobseeker and extracts audio attributes from the audio. For each of the jobseekers, the software feeds the résumé data, the open position requirements, the video attributes and the audio attributes into an artificial intelligence engine that uses a knowledge base to evaluate the résumé data, the video attributes and the audio attributes of each associated jobseeker with respect to the knowledge base and outputs a score for each of the associated jobseekers. The scores of the jobseekers are sorted and filtered into a ranked list and the ranked list of the associated jobseekers is provided as output.
In another embodiment, a method for using artificial intelligence to select candidates for an open position is disclosed. The method includes receiving the open position and a set of applications from jobseekers. Each application includes a résumé and a video cover letter and each video cover letter includes audio and video. Next, parsing each résumé into résumé data for an associated jobseeker; parsing the open position into open position requirements; parsing the video from each video cover letter and extracting video attributes from the video; and parsing the audio from each video cover letter and extracting audio attributes from the audio. Now, for each application, the résumé data, the open position requirements, the video attributes, and the audio attributes are fed into an artificial intelligence engine. The artificial intelligence engine evaluates the application using a knowledge base and the artificial intelligence engine outputs a score for each application. The scores are sorted and filtered, thereby ranking the applications and generating a ranked list of jobseekers.
In another embodiment, a method for using artificial intelligence to select candidates for an open position is disclosed. The method includes receiving the open position and a set of applications from jobseekers. Each application includes a résumé and a video cover letter and each video cover letter includes audio and video. The method continues with parsing the open position into open position requirements and keywords; parsing each résumé into résumé data for an associated jobseeker; parsing the video from each video cover letter and extracting video attributes from the video; and parsing the audio from each video cover letter and extracting audio attributes from the audio with respect to the keywords from the open position requirements. Then, for each application, the résumé data, the open position requirements, the video attributes and the audio attributes are fed into an artificial intelligence engine. The artificial intelligence engine evaluates the application using a knowledge base and outputs a score for the application. The scores of the applications are sorted and filtered, thereby ranking the applications and generating a ranked list of jobseekers.
The invention can be best understood by those having ordinary skill in the art by reference to the following detailed description when considered in conjunction with the accompanying drawings in which:
Reference will now be made in detail to the presently preferred embodiments of the invention, examples of which are illustrated in the accompanying drawings. Throughout the following detailed description, the same reference numerals refer to the same elements in all figures.
In general, the system for applicant screening provides tools for a human resource manager and/or hiring manager to help reduce a large number of job applicants down to a shorter list of candidates based upon attributes extracted from a video, audio, or audio/video segment produced by each candidate in view of attributes sought by a job description of the open position(s). Although any attribute is anticipated and included herein, examples of attributes include friendly, cooperative, easy-going, curt, rambles, concise, verbose, team player, introverted, extroverted, self-starter, quick-learner, good communicator, professionalism, visual focus and attention, clarity of expression, subject matter expertise, inclusion of cultural attributes, experience attributes, skill-based attributes, etc.
Throughout this description, the term, “reviewer” refers to any person (e.g., human resource manager, hiring manager) who will review a set of jobseekers with respect to one or more open positions. “Jobseeker” refers to a person having interest in one or more of the open positions.
The term “résumé” refers to a document (paper or electronic) that contains applicant qualifications. The “résumé” is one input from the jobseeker informing the reviewer of data about the jobseeker, for example, employment history, education, skills, talents, and data such as birthdate, residence address, abilities and knowledge, etc. Although the résumé is often delivered in paper or electronic form, there is no limitation as to how such data is delivered to the reviewer. For example, at times, data similar to the résumé are entered into an online submission form.
“Open position” refers to description of a position in the company that needs to be filled. The open position typically describes the position and activities of the position as well as requirements (e.g., open position requirements) and/or desires of someone who might fill the open position. For example, an open position for a janitor might say that the person will need to sweep and dust, move light-weight equipment, looking for a candidate that works well alone and is conscientious. An open position for a software coder might say that the person will program cell phone applications that communicate with a server over Wi-Fi, looking for a candidate that has experience in C++, experience with Android and IOS operating systems, works well with a team, etc.
The term “cover letter” refers to a letter often prepared by the jobseeker to a prospective employer indicating why the jobseeker is a good match to the open position. The term “video cover letter” refers to a short (usually one minute) audio/video segment which includes audio and/or video of the jobseeker as they describe themselves, sometimes with respect to the open position.
Referring to
The server computer 500 has access to data storage 512. Although one path between the user computer 10/10A and the server computer 500 is shown going through the network 506 as shown, any known data path is anticipated. For example, the Wi-Fi transceiver 96 (see
The server computer 500 transacts with software runs on the user computers 10/10A through the network(s) 506. The system for applicant screening runs on the server computer 500, receives multiple résumés 19 and/or video cover letters 17, culls multiple résumé's 19 and/or video cover letters 17, and reports a reduced list of candidates to the user computers 10/10A.
In
Referring to
The example user computer 10/10A represents a typical device used to access the system for applicant screening. This exemplary user computer 10/10A is shown in its simplest form. Different architectures are known that accomplish similar results in a similar fashion, and the present invention is not limited in any way to any particular computer system architecture or implementation. In this exemplary user computer 10/10A, a processor 70 executes or runs programs in a random-access memory 75. The programs are generally stored within a persistent memory, storage 12, and loaded into the random-access memory 75 when needed. The processor 70 is any processor, typically a processor designed for phones. The random-access memory 75 is interfaced to the processor by, for example, a memory bus 72. The random-access memory 75 is any memory suitable for connection and operation with the selected processor 70, such as SRAM, DRAM, SDRAM, RDRAM, DDR, DDR-2, etc. The storage 12 is any type, configuration, capacity of memory suitable for persistently storing programs and data, for example, flash memory, read only memory, battery-backed memory, hard disks, etc. In some exemplary user computers 10, the storage 12 is removable, in the form of a memory card of appropriate format such as SD (secure digital) cards, micro-SD cards, compact flash, etc.
Also connected to the processor 70 is a system bus 82 for connecting to peripheral subsystems such as a cellular network interface 80, a graphics adapter 84 and input/output devices 92 such as mice, keyboards, etc. The graphics adapter 84 receives commands from the processor 70 and controls what is depicted on the display 86.
In general, some portion of the storage 12 is used to store programs, executable code, and data, etc. In some embodiments, other data is stored in the storage 12 such as audio files, video files, text messages, etc.
The peripherals shown are examples, and other devices are known in the industry such as Global Positioning Subsystems, speakers, microphones, USB interfaces, cameras, microphones, Bluetooth transceivers, Wi-Fi transceivers 96, image sensors, temperature sensors, etc., the details of which are not shown for brevity and clarity reasons.
In some embodiments, a network interface 80 connects the user computer 10/10A to the network 506 through any known or future protocol such as Ethernet, Wi-Fi, GSM, TDMA, LTE, etc., through a wired or wireless medium 78. There is no limitation on the type of connection used. In such, the network interface 80 provides data and messaging connections through the network 506, connecting the user computer 10/10A to other computer systems such as the Internet and to the server computer 500. In some embodiments, remote storage is accessible through the network 506, for example, cloud storage.
Referring to
This exemplary server computer 500 is shown in its simplest form. Different architectures are known that accomplish similar results in a similar fashion and the present invention is not limited in any way to any particular computer system architecture or implementation. In this exemplary computer system, a processor 570 executes or runs programs in a random-access memory 575. The programs are generally stored within a persistent memory 574 and loaded into the random-access memory 575 when needed. The processor 570 is any processor, typically a processor designed for computer systems with any number of core processing elements, etc. The random-access memory 575 is connected to the processor by, for example, a memory bus 572. The random-access memory 575 is any memory suitable for connection and operation with the selected processor 570, such as SRAM, DRAM, SDRAM, RDRAM, DDR, DDR-2, etc. The persistent memory 574 is any type, configuration, capacity of memory suitable for persistently storing data, for example, magnetic storage, flash memory, read only memory, battery-backed memory, magnetic memory, etc. The persistent memory 574 is typically interfaced to the processor 570 through a system bus 582, or any other interface as known in the industry.
Also shown connected to the processor 570 through the system bus 582 is a network interface 580 (e.g., for connecting to a data network 506), a graphics adapter 584 and a keyboard interface 592 (e.g., Universal Serial Bus—USB). The graphics adapter 584 receives commands from the processor 570 and controls what is depicted on a display 586. The keyboard interface 592 provides navigation, data entry, and selection features.
In general, some portion of the persistent memory 574 is used to store programs, executable code, and data, etc.
The peripherals are examples and other devices are known in the industry such as pointing devices, touch-screen interfaces, speakers, microphones, USB interfaces, Bluetooth transceivers, Wi-Fi transceivers, image sensors, temperature sensors, etc., the details of which are not shown for brevity and clarity reasons.
Referring to
In some embodiments, the system for applicant screening parses the résumé 100 into a basic format called JavaScript Object Notation (JSON). JavaScript Object Notation is a data-interchange format that is easy for humans to read and write and is easy for computer programs to parse and generate. JavaScript Object Notation is based on a subset of the JavaScript Programming Language Standard ECMA-262 3rd Edition—December 1999. JavaScript Object Notation is a text format that is completely language independent but uses conventions that are familiar to programmers of the C-family of languages, including C, C++, C#, Java, JavaScript, Perl, Python, and many others. These properties make JavaScript Object Notation an ideal data-interchange language. JavaScript Object Notation is built on two structures: A first is collection of name/value pairs. In various languages, this is realized as an object, record, struct, dictionary, hash table, keyed list, or associative array; and a second is an ordered list of values. In most languages, this is realized as an array, vector, list, or sequence.
Referring to
Referring to
Referring to
Now, each response 209A/209B/209N is reviewed 202 by the reviewer (e.g., HR person). As one can imagine, if n is 100, the reviewer must review 100 responses 209A/209B/209N. As each written cover letter is typically around one page, this requires reading 100 résumés and 100 written cover letters to review all 100 responses 209A/209B/209N. Once the review 202 is complete, the reviewer selects 203 the best response 209A/209B/209N and makes an offer 206 (e.g., mails or emails an offer letter). If the jobseeker accepts 205 the offer, the process is complete, but if the jobseeker rejects 206 the offer, the above selection 204 and offer 206 is repeated with the next best response 209A/209B/209N, etc. This process is long, tedious, and prone to error (e.g., selecting a less-than qualified candidate) due to fatigue of the reviewer as reading countless résumés 19 and written cover letter becomes boring after a time. Some prior art systems use résumé and cover letter parsers, but many jobseekers are aware of these machines and know how to beat the system. This outdated process results in applicants knowing how to work the system raising to the top of the pile, even though they may not actually have the stated job requirements.
In some situations of the prior art, each response 209A/209B/209N is parsed by this resume parser machine before a subset of the responses 209A/209B/209N are provided to the reviewer.
Referring to
If the jobseeker accepts 225 the offer, the process is complete, but if the jobseeker rejects 226 the offer, the above selection 223 and offer 224 is repeated with the next best response 210A/210B/210N, etc. In some embodiments, feedback is provided to the system for applicant screening (e.g., to processes 220) from the selection 223, offer 224, acceptance 225, and rejection 226. The feedback provides for learning by the system for applicant screening. For example, if the open position is seeking an outgoing candidate and during selection 223, the reviewer rejects a certain applicant (response 210A/210B/210N), indicating that this candidate does not appear to be outgoing, the system for applicant screening feeds this back into a learning process to modify how the attribute of “outgoing” is determined and weighed.
Also, in some embodiments, irrespective of whether the jobseeker is selected for the open position or note, the jobseeker will receive feedback. For example, the jobseeker receives feedback if their application doesn't make it to the top of the applicants that are reviewed, or if if the jobseeker's application is reviewed and the application doesn't make it through the human evaluation (e.g., by a reviewer), and the jobseeker receives feedback if the jobseeker is not offered the open position.
Referring to
In some embodiments, the evaluation 242 includes the score 272 (see
In some embodiments, feedback is provided to the system for applicant screening from the evaluation 242. This feedback is used by the system for applicant screening for training the intelligence of the system for applicant screening. Such feedback includes whether the jobseeker approves 244, has redone their résumé 19 and video cover letter 17 and/or specific feedback such as indicating a certain attribute indicated by the system for applicant screening is not true, etc.
Referring to
Referring to
In some embodiments, there is a single knowledge base 280, learning from all reviewers regarding how to score candidates based upon résumés 19 and video cover letters 17 (applications 240) from jobseekers. In other embodiments, there are multiple knowledge bases 280, one knowledge base 280 per organization such as a company, a class of companies, a school, a company division, a company department, a reviewer, a hiring manager, etc. In this embodiment, a knowledge base 280 associated with the target organization is loaded into the artificial intelligence engine 270. In some such embodiments, the target organization is selected based upon the origin of the open position 201 and/or requirements parsed from the open position 201.
Although the system for applicant screening is described with respect to filling an open position 201, it is fully anticipated that the disclosed candidate scoring process 260 of the system for applicant screening be used in other similar applications such as scoring and filtering candidates for admission into schools such as colleges, scoring and filtering candidates for certain degree programs such as masters' degrees or PhD programs, scoring and filtering candidates for promotions within an organization, etc. Further, it is fully anticipated that the disclosed candidate scoring process 260 of the system for applicant screening be used to evaluate members of a team to determine cultural fit, to recommend training, and to improve and assess a team's performance with respect to the team having the right blend of personalities.
In some embodiments, the open position 201 includes screening questions 199 that are answered by the jobseeker during the creation of the video cover letter 17. In such embodiments, the video and audio of the video cover letter 17 are parsed (see below) with respect to the screening questions 199.
In some embodiments, the jobseeker receives jobseeker feedback 273 from the artificial intelligence engine 270, providing feedback on how the jobseeker performed in the creation of the video cover letter 17 and answering the screening questions 199, if any.
The application from a jobseeker includes a résumé 19 and video cover letter 17 (application 240). The résumé 19 and video cover letter 17 for each jobseeker parsed separately. The résumé 19 is parsed by a résumé parser 262. For some résumés 19 that are on paper or in PDF format, the résumé 19 is first run through character recognition (e.g., OCR), then parsed into a normalized format such as JavaScript Object Notation, providing data elements regarding the jobseeker.
The video of the video cover letter 17 is parsed by a video parser 263 to gather video attributes (e.g., video-related attributes) such as smiling, looking away, looking at the camera, enthusiasm. . . . The audio of the video cover letter 17 is parsed by an audio parser 264 to gather audio attributes such as vocal tones, enthusiasm, excitement. . . . In some embodiments, entity analysis is used. In such, natural language processing of the audio is performed with respect to known entities to extract specific keywords. For example, if the video cover letter 17 was recorded with respect to an open position 14, then the audio portion of the video cover letter 17 is analyzed, looking for certain keywords that were stated in the open position 14 such as specific skills like programming in C++ or skills using spreadsheets. Such entity analysis also finds titles of prior experiences such as job titles or collegiate activities. In some embodiments, the audio is analyzed for sentiment. Again, natural language processing analyzes text, computational linguistics and biometrics to identify, extract, and quantify affective states and subjective information from the audio of the video cover letter 17. For example, sentiment analysis will report intonation and clarity attributes of the jobseeker.
An open position 14 (e.g., job post), is parsed 265 to determine what data elements are sought by the hiring organization. For example, the open position 14 is parsed 265 to determine requirements such as minimum education, degree required, certain amounts of experience, certain skills, etc. As discussed above, certain keywords from the open position 14 are used in the analysis of the audio portion of the video cover letter 17. For example, if the open position 14 includes a skill requirement of “programming language: C++” then, during the analysis of the audio, it is noted whether the audio includes mention of “C++.”
The parsed data from the résumé 19 (e.g., resume requirements), the attributes parsed from the video cover letter 17 (e.g., video attributes and/or audio attributes), and the requirements parsed from the open position 14 (e.g., open position requirements) are fed to the artificial intelligence engine 270. The artificial intelligence engine 270 analyzes the data from the résumé 19 and the attributes parsed from the video cover letter 17 in view of the requirements from the open position 14 and generates a score 272 that represents how well the associated jobseeker matches the open position 14. Although the value system used for the score 272 is arbitrary, in one embodiment, the score 272 is a value between zero and 100, where zero is a score 272 indicating that the jobseeker is not a match for the open position 14 and a score 272 of 100 indicates that the jobseeker is a strong match for the open position 14.
Referring to
Referring to
In some embodiments, feedback 282 is provided to the artificial intelligence engine 270 and the artificial intelligence engine 270 updates the knowledge base 280 based upon this feedback 282. For example, if a particular jobseeker is rated with the highest possible score 272, but the reviewer does not agree that that jobseeker is the best match, feedback 282 is provided to the artificial intelligence engine 270 indicating such and, in some embodiments, the feedback 282 is provided with details such as this jobseeker does not match a certain requirement of the open position 14 (e.g., this jobseeker lacks a required education or seems not to match a desired attribute such as “outgoing.” Another example of feedback occurs when the jobseeker is offered the open position 14 and accepts the open position 14, reinforcing the assumptions and decisions made by the artificial intelligence engine 270. Likewise, if the jobseeker is offered the open position 14 and turns down the offer, there may be knowledge to learn as to why the offer was denied, for example, the jobseeker was over-qualified, etc. When available, the feedback process 284 receives and processes the feedback 282 to make appropriate changes to the knowledge base 280.
Referring to
In
In
In some embodiments, feedback 282 is provided to the artificial intelligence engine 270 from various stages of the process and the artificial intelligence engine 270 updates the knowledge base 280 based upon this feedback 282. For example, if a particular jobseeker is rated with the highest possible score 272, but the reviewer does not agree that that jobseeker is the best match, feedback 282 is provided to the artificial intelligence engine 270 indicating such and, in some embodiments, the feedback 282 is provided with details such as this jobseeker does not match a certain requirement of the open position 14 (e.g., this jobseeker lacks a required education or seems not to match a desired attribute such as “outgoing.” Another example of feedback occurs when the jobseeker is offered the open position 14 and accepts the open position 14, reinforcing the assumptions and decisions made by the artificial intelligence engine 270. Likewise, if the jobseeker is offered the open position 14 and turns down the offer, there may be knowledge to learn as to why the offer was denied, for example, the jobseeker was over-qualified, etc.
Equivalent elements can be substituted for the ones set forth above such that they perform in substantially the same manner, in substantially the same way, for achieving substantially the same result.
It is believed that the system and method as described and many of its attendant advantages will be understood by the foregoing description. It is also believed that it will be apparent that various changes may be made in the form, construction, and arrangement of the components thereof without departing from the scope and spirit of the invention or without sacrificing all of its material advantages. The form herein before described being merely exemplary and explanatory embodiment thereof. It is the intention of the following claims to encompass and include such changes.