Here is the expanded slide deck from my SourceCon 2010 Keynote: Resume Sourcing and Matching – Artificial Intelligence vs. Human Cognition. It contains all of the talking points as text so you are not left guessing as to what I spoke to during the live presentation.

You’ll learn about the intrinsic and often overlooked challenges associated with sourcing resumes (it’s deceptively complex), what artificially intelligent semantic search and match applications claim to do and how they actually work, the limits of artificial intelligence, what people can do that semantic search applications cannot, the 5 levels of semantic search,  the 5 levels of talent mining, and what I think is the ideal candidate sourcing solution.

Are you attending or thinking about attending SourceCon 2010 in San Diego in March?

I am going to be the keynote speaker for the event, and I will be presenting on Artificial Intelligence vs. Human Cognition when it comes to sourcing and matching resumes.

If you’re curious to know what kinds of things I’ll be addressing during the session, here is a sneak peek:

1. The intrinsic and often overlooked challenges associated with sourcing resumes
2. What artificially intelligent semantic search and match applications claim to do and how they actually work
3. The limits of artificial intelligence
4. What people can do that semantic search applications cannot
5. The 5 levels of semantic search
6. The 5 levels of secondary/e-sourcing
7. What I believe would be the ideal candidate sourcing/talent identification solution

If you’ve ever wondered about the fantastic claims that some of the semantic search application vendors on the market make as to how their solution can mimic a senior recruiter when finding candidates, then you will be very interested in hearing what I have to say about the reality of what they can do.

If you’re a sourcer and you’re concerned that your role/position might eventually be replaced by sourcing software, you will be encouraged by my analysis and supporting arguments that explain why the abilities of creative and investigative sourcers will always be in demand – tomorrow and 50 years from now.

I hope you will be able to attend SourceCon 2010 – I know I’m looking forward to it!

If you’re unable to attend, the good news is that the presentations will likely be streamed. Additionally, I plan on posting my expanded slide deck, including all talking points – so you won’t be stuck staring at some pretty pictures wondering what the heck I talked about.

Many individuals and organizations treat the sourcing role and function of recruiting – searching for and identifying potential candidates – as an entry level position, and/or a simple and basic task that does not require much skill or experience. 

I agree.

I believe that it does not take much skill or experience to simply transcribe job titles and required skill keywords into LinkedIn, Monster, or an ATS and click “search.”

However, that oversimplified view of sourcing talent only describes the most basic level of talent identification, of which, I believe there are at least two.

Level 1 Sourcing

Level 1 Sourcing involves little more than taking titles and words from job descriptions and entering them into Internet search engines, social networks, job board resume databases, and applicant tracking systems to search for candidates.   

This is the proverbial “buzzword bingo,” and does not necessarily involve any real understanding (beyond surface level) of the roles, skills, responsibilities, or technologies involved in the hiring profiles or the candidates. These basic search terms produce search results that are then cursorily reviewed for keyword matching. 

Level 1 Sourcing involves a level of matching potential candidates to hiring profiles that is often superficial and generic – very little, if any, interpretive analysis is performed. This level of sourcing can in fact quite easily be performed by “junior” personnel/researchers – almost anyone can match keywords.

Not only can Level 1 Sourcing be performed by junior associates, it can (and often is) outsourced for $5 – $7 an hour.

However, don’t be fooled into thinking you are getting something fantastic for that $5 – $7 an hour – you’re getting exactly what you’ve paid for. Which is Level 1 Sourcing.

In my opinion, there is no need to outsource Level 1 Sourcing, because it does not require any deep understanding of the roles being sourced for, nor does it involve any true analysis or creativity. As such, Level 1 Sourcing is well suited for total automation. Why pay people to match keywords when matching applications can do it for you for considerably less than $5 per hour?

Many people are blissfully unaware of the fact that Level 1 Sourcers from any company will essentially find the same potential candidates as any other Level 1 Sourcer. It’s a simple equation: same keywords = same results. This is one of the reasons why Level 1 Sourcing provides no competitive advantage. 

Additionally, while Level 1 Sourcers can and will find SOME candidates, they will not and can not find ALL potentially qualified candidates available to them in the sources they are searching - that would be impossible, for many reasons that I’ve written about previously that are beyond the scope of this post. 

And finally, Level 1 Sourcers are typically unaware of the people that are in the ATS, job board resume database, or social network that they are searching that their queries did not return. In fact, to them, anyone that they don’t find simply doesn’t exist. 

Level 2 Sourcing

This is the good stuff. Level 2 Sourcing moves well beyond simple keyword matching and most certainly beyond a basic mastery of Boolean logic. 

Boolean logic is easy to learn – after all, there’s only 3 main operators! However, the “magic” of leveraging databases and information systems for talent identification does not lie in the Boolean search operators themselves, but in the following process: 

1. Analyzing, understanding, and interpreting job opening/position requirements - including elements which may or may not be explicitly mentioned in the position description or BQ’s
2. Taking that understanding and intelligently and creatively selecting titles, skills, technologies, companies, responsibilities, terms, etc., to include (or to purposefully exclude!) in a query employing appropriate Boolean operators and query modifiers
3. Analyzing the results of the initial search to assess relevance as well as scanning the results for additional and alternate relevant search terms, phrases, and companies 
4. Based upon the observed relevance of and intel gained from the search results, modifying the search string appropriately and running it again
5. Repeating steps 3 and 4 until an acceptably large volume of highly relevant results is achieved   

The “real” work of creating effective Boolean search strings lies in the interpretive analysis of the need, in determining what terms to include and exclude from searches and in what specific combination, in the analysis of the relevance of the initial search results, and in the adaptive process of learning from the results to creatively refine the Boolean search strings to find well qualified candidates – people who are highly likely to be (or know!) the right match for any particular hiring need. 

Unlike Level 1 Sourcing, Level 2 Sourcing involves and in fact requires a deeper understanding of the roles, skills, responsibilities, and technologies involved in the hiring profiles being sourced for. In this regard, Level 2 Sourcing goes well beyond explicit keyword matching and deep into implied experience and capability matching.

In addition to finding all of the candidates that Level 1 Sourcers can find, Level 2 Sourcers can also find the candidates that Level 1 Sourcers have access to, but can not and do not find. Interestingly, all Level 2 Sourcers will not find all of the same candidates, specifically due to their varying experience and creative and analytical ability.

And unlike Level 1 Sourcers, Level 2 Sourcers are acutely aware of the candidates they have not found, because they understand that every Boolean string and search strategy will find some candidates, and exclude others.

Level 2 Sourcing is Not a Junior Role and Cannot Be Automated

It is precisely because of the aforementioned reasons that Level 2 Sourcing cannot be performed by  junior level associates – it is not an entry level role, nor can it be outsourced for $5 – $7 an hour. Okay, it CAN be outsourced at those rates, but you won’t get Level 2 results. Remember, you get what you pay for.

Level 2 sourcing cannot be automated, regardless of what the vendor representatives of “artificial intelligence” resume parse/match applications may claim. This is because Level 2 sourcing requires “a posteriori” knowledge – which comes from experience, which comprises knowledge and skill gained through involvement and exposure. 

Applications do not accumulate experience or gain knowledge or skill, in the true sense of the terms.

AI matching applications essentially perform a form of pattern recognition to classify data through parsing resumes and employing a keyword/phrase taxonomy, which is built based on “a priori” knowledge/information extracted from the patterns and programmed into the matching logic. 

I recently spoke at the PDS Technology Conference and had the honor of seeing Dr. Michio Kaku present on the world of 2020 and beyond. Dr. Kaku believes that “Progress in artificial intelligence may come to a gradual halt around 2020. The two problems facing AI are pattern recognition and common sense.” 

I was happy to hear that Dr. Michio Kaku believes that the employment market of the future will be “dominated by jobs involving common sense (e.g. leadership, judgment, entertainment, art, analysis, creativity) and pattern recognition (e.g. vision and non-repetitive jobs).  Jobs like brokers, tellers, agents, low level accountants and jobs involving inventory and repetition will be eliminated.”

That’s great news for anyone performing Level 2 Sourcing, primarily because it requires creativity, interpretive analysis, judgment, and common sense (a natural understanding based upon experience) - four things that machines and applications are intrinsically incapable of.

Unlike AI matching applications, Level 2 Sourcers intrinsically understand that resumes and social media profiles are imperfect and incomplete representations of the people who created them, and that they often do not explicitly mention all of their skills and experience. As such, Level 2 Sourcers are not only able to find qualified candidates based on the words they use - many can also specifically search for and find people who have experience that they do not mention. In other words, some Level 2 Sourcers can find people based on what they don’t say. This is a skill that can only be developed over time from observation and experience.  

Final Thoughts

Level 1 Sourcing can be performed by entry level associates or be completely automated, as the level of matching produced is superficial and based primarily on explicit keyword and phrase matching. This can be quite sufficient for static and repetitive hiring needs for simple hiring profiles, where title searches will often suffice.

The value and the results provided by Level 1 and Level 2 Sourcing is vastly different - this is why some organizations see the sourcing function as a low level or junior role, simply outsource it for $5 – $7 and hour, or completely automate it. Interestingly, there are sourcers who make $50 to over $100 an hour, and they are worth every penny for the competitive advantage  and value they provide to the organizations they support.

Dr. Michio Kaku would classify Level 1 Sourcing as ”commodity based capital,” in that it is a product that is the same no matter who produces it - man, woman, or machine.

People who perform Level 2 Sourcing are true knowledge workers, whose value is intellectual capital – based in creativity, judgment, analysis, ”common sense” and “a posteriori” knowledge developed over time based upon experience – similar to senior Financial Analysts, Business Analysts, Data Analysts and Business Intelligence Analysts. Level 2 Sourcers produce a product that is quite different based on who produces it, and it cannot be reliably replicated by machines.

To be sure, one could easily break Level 2 Sourcing out to at least 3 different levels, because to lump everything more advanced and sophisticated than Level 1 Sourcing into one broad category is horribly limiting, but for the purposes of this article, it shall suffice.

Human Capital Data data is the sword of the 21st century – those who wield it well are the Sourcing Samurai.

There are many software vendors who offer solutions to HR, recruiting, and sourcing organizations that claim to have automated candidate search and match capability. These applications can take your search, a job description, or an example resume and claim to leverage semantic search, fuzzy logic and/or Artificial Intelligence search technology to return relevant results. 

I’ve had the opportunity to use and evaluate 4 “big name” semantic and intelligent search/match applications for identifying candidates, and I am currently implementing one of them, including cutomizing the taxonomies and ontologies - so I have quite a bit of hands-on, practical experience with semantic search applications. While I do think that intelligent semantic search and match applications definitely have a place in the sourcing and recruiting process, they should not be looked at as solutions to a problem, challenge, or a deficiency in your skills or your team’s.

Before implementing a semantic search application, it is important to first understand the manual search process, how automated semantic search works, and the intrinsic limitations of of semantic search applications. 

People Do the Work, Computers Move Information

Eiji Toyoda, the fifth president of Toyota Motor Corporation, who collaborated Taiichi Ohno to fine tune the concept of Kaizen as well as to develop the core concepts of the ‘Toyota Way’, explains brilliantly: ”Society has reached the point where one can push a button and be immediately deluged with…information. This is all very convenient, of course, but if one is not careful there is a danger of losing the ability to think. We must remember that in the end it is the individual human being who must solve the problems.” 

As I have written before, thinking is the most critical step in the candidate sourcing process, and regardless of “Artificial Intelligence” and semantic search marketing hype, applications do not have any true cognitive power, nor do they have the ability to be creative or learn as people do. Thus I could not agree more with Jeffrey Liker’s (author of The Toyota Way) assessment - ”People do the work, computers move information.”

Do Not Automate What You Cannot Already Perform Manually

If you are not a highly proficient Talent Miner, capable of manually querying databases and systems with Boolean logic and a high degree of precision to find not only the obvious and easy to find candidates, but also the those residing in Hidden Talent Pools – how can you even HOPE to begin to evaluate what an automated semantic search solution is claiming to do?

You can’t.

If you’re going to buy and use product or service of ANY kind, and you don’t really understand what it does or even exactly how it does it (beyond the marketing hype), and you can’t tell if it REALLY does everything it claims to – how and at what level can you determine if the product or service truly meets your needs and will provide true value?  

If you or your organization struggles with the challenge of finding the right candidates at the right time on a consistent basis, implementing an automated search and match application will not magically solve this “problem” for you. The real underlying problem is likely that you or your organization currently does not #1 possess the right skills or the right people who are highly proficient at candidate sourcing and/or #2 have documented highly effective candidate sourcing processes and best practices that are consistently trained to and followed. No software application can fix either or both of those issues.

Jane Beseda, Group VP at Toyota, believes that it is best to “First work out the manual process, and then automate it. Try to build into the system as much flexibility as you possibly can so you can continue to kaizen the process as your business changes” because “…you can kaizen (continually improve) people processes very easily, but it is hard to kaizen a machine (or application).” I agree wholeheartedly.

Master the Sourcing Process Manually, THEN Introduce Automation

If you or the people on your team don’t fully understand how to effectively leveraging technology for talent identification and you can’t perform it manually, I do not recommend implementing an automated search/match solution. It is critical to #1 First develop your skills and ability (or your team’s) to manually source candidates, #2 Document your sourcing best practices and processes, #3 Make sure that they are consistently trained and applied, and #4 Strive to continually improve them. THEN you can go about evaluating automated search and match solutions because you will actually have the ability to truly assess and understand what the products do, you will be able to determine whether or not they meet your needs and can provide real value to your organization, and you will be able to assess how you can possibly best leverage them into your sourcing efforts.

For example, if you don’t really comprehend the concept of semantic search and how it can be applied in candidate sourcing (e.g., you can manually leverage semantic search in your Boolean strings at an expert level or you have someone on your team that can), you won’t be able to tell whether or not a application claiming to leverage semantic search is really effectively doing so, if it will provide any real value to your sourcing efforts, or if it is finding all of the best possible matches in the sources it is searching.

Speaking of Semantic Search

Many automated candidate search and match solutions claim to leverage semantic search. I’ve written many articles on the topic of how semantics can be leveraged for sourcing candidates – so I won’t go into too much detail here – but semantics refers to the study of meaning, as inherent at the levels of words, phrases, and sentences.

Sourcers and recruiters can leverage semantics in their sourcing efforts to more quickly find more relevant results. When the search results match the intended MEANING of the search (the INTENT), there is a semantic similarity between the INTENT of search and its results. In other words – you get what you’re looking for. When search results simply match the search terms but not the INTENT of the search, there only a lexical similarity between the search and its results. In other words – the words match, but you don’t get what you’re looking for.

For example, if I were manually searching for Linux systems administrators – that is the INTENT of my search – to find people who have been primarily responsible for administering Linux systems, regardless of exactly how they express that experience in their resume. However, with basic keyword and title search, I will find a mix of relevant results (people who have been responsible for Linux administration) and irrelevant lexical-only match results (people who happen to mention the search terms of Linux and systems and administraion somwehere in their resumes or profiles, but who have not been primarily responsible for Linux systems administration).

However, as a thinking human being, I can truly learn from the search results and continually improve and adapt my search strategies and keywords to leverage the language used specifically by people who have been primarily responsible for administering Linux systems (using semantics – the actual meaning of specific words and phrases), ensuring more of the results returned actually match the INTENT of my search.

So How Do Applications Achieve Semantic Search?

To provide relevant search results, various automated search and match applications claim to:

• Have “pseudo-AI” – these perform semantic “concept matching” based on a taxonomy/ontology (lists of terms, equivalent terms and their relationships)
• Perform “natural language” search – using pre-programmed phrases and/or proximity matching
• Execute “context-aware” semantic matching of jobs and resumes
• Perform fuzzy matching – returning results matching words that were not specifically searched for by the user, but that the application “thinks” is likely to be relevant
• Have “full AI” – the software is designed with algorithms to create relationships between words, abbreviations and phrases dynamically and without human intervention

Applications Are Not Mind Readers

No software application is capable of determining the INTENT of any search – it is critical to recognize and understand this, because it’s at the very core of semantic matching, although it’s not often written about when it comes to sourcing candidates. No application can actually *know* what you’re looking for – only YOU know what you’re trying to find, which is a person who has had some sort of specific experience performing certain responsibilities, often in specific environments.

Essentially, automated search/match applications take either your search terms, a job description, and/or an example resume and make their “best guess” as you your intent. Don’t get me wrong – some applications do a very good job of taking what you give them and providing some relevant matches. However, you must always be aware that these are just that – guesses – and on top of it all, they’re guesses made by an application, not a person.

Output is Limited by Input

Search/match aplications are only as good as their input - if a user is not especially adept at crafting search strings the quality of the results the application can produce will be limited. Additionally, we are all aware the resumes and job descriptions are FAR from the best representations of skills and experience, so the results produced by an application interpreting a resume or a job order is instrinsically limited.

There is no replacement for the cognitive and interpretive power of the human mind – people are much more capable of “reading between the lines” of resumes and job descriptions and getting to the true essence of the required skills and experience to determining how to best approach sourcing efforts to find the right people. Applications can claim to do this, but it’s an apples to oranges comparison.

Let’s also remember that a sourcer or recruiter can discuss the position’s requirements with the hiring manager (in most cases – if not, they should be able to!) – this information is typically beyond what any example resume or written job description can (or does) convey. Armed with this critical information and detail, a sourcer or recruiter can translate what they’ve gathered verbally from the hiring manager/team into a sourcing strategy, down to the Boolean search string level. The last time I checked, an application cannot do this. There is no replacement for good analytical and interpretive ability – it’s precisely why you can’t fully automate a business analyst’s, a data analyst’s or a financial analyst’s job.

Part of the instrinsic challenge faced by automated search and matching applications is that they are expected to make the leap from “give me what I said” (match the words) to “give me what I want” – matching your intent. Considering that an application is incapable of “understanding” your intent, how can it hope to deliver results that match your intent when your intent cannot be interpreted from an example resume, a job description, or a poorly defined search? 

Intrinsic Limitations of Automated Semantic Search/AI

Be aware that any automated semantic search/match application is only as good as its programming, taxonomy, and ontology:

• Pre-programmed lists of “relevant” or related keywords may in fact not be relevant or related to your specific search, and can get outdated quickly
• Fuzzy matching is by definition “approximate” or “inexact” matching, the very opposite of “precise”
• Applications are essentially “guessing” the intent of your search based on the keywords, resume, or job description you feed it – but finding the best talent is not a guessing game
• Some systems will return results with related words – which may in fact NOT be relevant to your specific need – will you be able to determine which are and which aren’t?
• Applications that claim to “learn” may not actually improve the relevance of results over time

The Tough Questions

Are you capable of evaluating an application “under the hood?”  When a vendor tells you that their application is “trained to identify the equivalent meanings of terms found in resumes,” do you know how to get to the bottom of exactly how their application accomplishes this, and whether or not their application can actually do it for your specific hiring profiles? Could you manually run searches to objectively evaluate the vendor’s claims? If you don’t already possess the ability to manually source candidates from information systems, what will you do when your automated search/match application fails to produce the right results, or enough of them? Do you have the ability to search your own database to find the candidates that the search/match application can not find/misses? 

My Experience and Opinion

I’ve had the opportunity to use a number of automated semantic search/match applications. I’m also currently implementing one and I’m involved in customizing the taxonomies and ontologies. I’m very excited about the features and capability of the semantic search application I’m using. However - I’ve found it does have it’s limitations, and that there is no perfect solution that magically produces the best candidates available with the push of a button, nor can an automated intelligent/semantic search application fully replace a skilled sourcer or recruiter. Augment and empower - yes. Replace – no.

As someone who is highly proficient in manual Boolean and semantic search, from everything I have gathered, most automated search and match applications simply automate basic sourcing best practices, but none that I have used do so completely or flawlessly.

In fact, some search/match apps appear to do little more than heavily leverage common title and skill terminology matching – which any sourcer or recruiter of average skill can accomplish. Other applications go overboard and get sloppy in the process, suggesting and incorporating alternate search terms it *thinks* are related and relevant to your search – and in some cases the terms are actually NOT relevant to the intent of the user’s search, causing more harm than good. But if you don’t know what to look for (past the marketing and past the huge cloud of “related” search terms the app suggests) and you don’t know how to get to the core of the search logic and the true relevance based on the intent of your search, you might just actually believe you have the solution to all of your candidate sourcing problems.

While some industries can benefit heavily from automated search and match applications – for example,  those with simple and highly consistent titles, lexicology, and phraseology(such as finance and accounting), others vary widely and change rapidly (such as Information Technology) and pose serious challenges for vendors of automated search and match solutions.

Check Your Reasoning

If you’re looking to implement an automated candidate search/match application, simply ask yourself why – and dig down to the very root of it. Is it because Boolean searching is “too hard?” Is it because you think the application can reduce your costs by reducing headcount? Do you think a search/match app can speed up the talent identification cycle?

I’ll let you in on a little secret – Boolean search isn’t that hard with the proper training. It’s more science than art, really. Which is good news, because manual talent mining via Boolean search strings can be broken down to repeatable steps, including the interpretive and analytical processes, which can be continually and dynamically improved – whereas an application cannot (easily, or at all). 

Also – semantic search/match applications are not a replacement for people, because the applications don’t actually perform real work or deliver value to the end customer. When I refer to “work” and “value,” I’m referring to how Lean and the Toyota Production System define those terms. Recall Jeffrey Liker’s assessment, “People do the work, computers move the information.” The author of The Toyota Way has also explained that ”…the only thing that adds value in any type of process…is the physical or information transformation of that product, service, or activity into something the customer wants.”

By customer, Jeffrey means the END customer – the internal hiring manager or the external client. Automated search/match apps don’t transform or produce a product that the end client wants – which is a talented person who is a great match for their need. Automated search/match applications produce resumes and profiles (i.e. “computers move the information”) and the sourcers and/or recruiters analyze and transform those resumes/profiles into fully screened, closed, and qualified candidates for the end customer (i.e., people do the work). 

My Suggestions

Automated search/match applications definitely have their place in world of sourcing and recruiting. I think they are best used to facilitate and augment the talent identification efforts of sourcers and recruiters. Some can be especially useful in quickly and simultaneously searching multiple online sources of resumes and social media profiles and parsing large volumes of results into your internal ATS/CRM – this quick and permanent human capital data capture can be a major benefit of using some search/match solutions.

Ultimately, I feel that sourcers and recruiters using AI/semantic search applications should utilize the results they produce and return as ”suggested reading” - but I would never rely solely on an application to exhaustively identify top talent, just as no one would trust a plane full of people to take off and land on autopilot. 

Whether or not you are considering purchasing an automated search and match application or service, or you’ve already implemented one, I strongly urge you to #1 First develop your skills and ability (or your team’s) to manually source candidates, #2 Document your sourcing best practices and processes, #3 Make sure that they are consistently trained and applied, and #4 Strive to continually improve them. THEN asses the best way to leverage an automated matching solution to augment your talent identification efforts.

Do not take a poorly functioning sourcing process or team and expect to fix it using an automated search/match application. Fine tune your sourcing process and best practices, develop your sourcers and recruiters with exceptional training, and then surgically insert matching technology to enhace them. Technology is a tool that exists to support your people and your processes – it is not a solution to a problem nor a replacement for a process or a person. 

Conclusion

I am in no way against inserting semantic search applications into the sourcing function (remember, I’m using one now!) – but I feel it must be done for the right reasons AND with a full understanding and capability of performing the sourcing processes manually, else you cannot continually improve your sourcing processes, nor will you be capable of picking up sourcing where the application fails to deliver.  

If finding some candidates is your end goal, then you can feel comfortable using automated search and match solutions to do all of the “heavy lifting” for you or your associates when it comes to talent identification. However, if you goal is to find all of the best candidates, then you should use an automated matching application as a tool to support your sourcers and recruiters who are well trained and effective in manual sourcing best practices.

Caveat Emptor

If you are looking to purchase an automated candidate search/match solution, or you’ve already implemented one, and you do not have the expertise or experience associated with assessing and implementing such solutions – I STRONGLY urge you to seek a neutral, third party HR/Recruiting technology consultant and involve them in the process. It is all too easy to be sold by a vendor’s marketing and messaging – but you are at a distinct disadvantage if the vendor knows more than you about the sourcing and matching function – be it manual OR automated.

Be wary – do not seek to automate a process that which you do not fully understand how to perform manually.

Partial List of Vendors of Matching Technology:

• Pure Discovery: http://www.purediscovery.com/
• Actonomy: http://www.actonomy.com/
• Semetric (Engenium): http://www.krollontrack.com/semetric/
• TalentSpring: http://www.talentspring.com/
• Sovren: http://www.sovren.com/
• BurningGlass: http://www.burning-glass.com/
• ResumeMirror: http://www.talenttech.com/

Are you a vendor and want to be added to this list? Would you like me to evaluate your product? Let me know.  Thanks!

I define E-sourcing as leveraging information systems for active talent identification – searching the Internet, social media, job board resume databases, and applicant tracking systems to find candidates. The proper use of technology in the sourcing and recruiting process should increase your efficiency, productivity, and effectiveness.  I’ve created the SlideShare presentation below to cover a number of different ways for you to maximize your ability to find more of the right people more quickly, to accelerate and enable your recruiting efforts.

Click on the presentation below to review:

• Boolean operators and common query modifiers
• Searching LinkedIn, Twitter, and Facebook
• X-Ray searching Social Media
• Search automation and aggregation
• Semantic search: manual and artificial intelligence matching solutions
• Search ROI – a comparison of the searchability and data depth of the Internet, Social Media, Resume Databases, and ATSs
• Talent Warehouse concepts

Semantic search has been a hot topic in Internet search for a number of years now and it continues to generate quite the buzz. For example, Google just recently rolled out semantic search capability. However, when I talk about semantic search, I’m not referring to the semantic web or “web 3.0.”

I’m not so excited about whether or not Google can correctly disambiguate my search on the word “bark” and figure out if I mean “the sound of a dog,” “the skin of a tree,” or “a three-masted sailing ship.” I’m interested in an passionate about semantic search techniques and applications specifically geared towards enabling more effective and efficient talent identification and acquisition.

So exactly what is semantic search?

That’s an excellent question!  If you run a search on Google or even a semantic search engine such as Hakia or Powerset for the phrase “semantic search,” you’ll find many confusing results and very little in the way of an easy-to-understand explanation of the concept. I’ll attempt to do my best to explain semantic search in a way that is pertinent to sourcing and recruiting.

Quite simply, semantic search can be defined as search techniques and/or applications that attempt to return results that more closely match the “meaning”  or intent of the search rather than simply returning results that match the words of the search. “Semantics” refers to the study of meaning, as inherent at the levels of words, phrases, and sentences.

What You Want vs. What You Say

Most sourcers and recruiters search information systems with queries that simply return a collection of words – words that do not have any associative meaning and that are not guaranteed to return relevant results. Relevance can be defined as the extent to which a search result matches the information need of the person executing the search. In other words, relevant results are what you want – highly relevant results are search results that match exactly what the searcher is looking for.

It’s safe to assume that sourcers and recruiters who are searching information systems to identify candidates are actually trying to find people that have specific skills, qualifications, abilities, and experience. However, just because certain words appear in a person’s resume or profile – it does not mean that the person has been primarily responsible for working with those words (typically skills, technologies, etc.). It’s important to be able to create Boolean search strings that return results that you actually want – not just the words you used in the search.

Lexical vs. Semantic Similarity

There is often a critical difference between the semantic similarity between a search and its results vs. the lexical similarity between a search and its results. When search results simply match the search terms but not the intended meaning of the search, there only a lexical similarity (the words match) between the search and its results. When the search results match the intended MEANING of the search, there is a semantic similarity between the search and its results.

Sourcers and recruiters leveraging semantic search are seeking to increase the probability that they will return search results of people who can actually do the job they are looking to hire for – a strong semantic match – not just people who happen to mention the words of the search somewhere in their resume or profile (lexical matches only, aka – false positives).

Automated Semantic Search

When I refer to “automated semantic search” for sourcing and recruiting, I’m referring to applications and systems that claim to perform semantic search for you. Semantic search applications and engines take your search terms and attempt to “understand” the intent of your search (i.e., what you’re really looking for) and return relevant results.

For example – various semantic search applications claim to:

• Have “pseudo-Artificial Intelligence” – performing concept matching based on a back end list of keywords (e.g., CPA is associated with accounting, J2EE is associated with Java, etc.)
• Execute “context-aware” semantic matching of job descriptions and resumes
• Perform fuzzy search/matching – returning results that are likely to be relevant
• Have “full Artificial Intelligence” – the software is designed with algorithms to create relationships between words, abbreviations and phrases dynamically and without human intervention

Partial List of Semantic Vendors (with links):

• Pure Discovery
• Actonomy
• Semetric (Engenium)
• TalentSpring
• Sovren
• BurningGlass
• ResumeMirror
• Autonomy

Limitations of Automated Semantic Search

While vendors of semantic search solutions are happy to sell you on the idea that their applications can do all of the work for you in quickly identifying highly qualified candidates, you need to be aware that there are several limitations of automated semantic search.

Semantic search applications that come with pre-programmed lists of “relevant” keywords may in fact return results that are not relevant to your search.  Also, a back-end list of keywords can get outdated quickly in some industries.

Fuzzy matching by it’s very definition is “approximate” or “inexact” matching. Does that sound like a good thing? In other words, fuzzy search can return results that are *likely* to be relevant to a search argument even when the search does not exactly ask for it.  But who determines the likelihood? Not you.

While fuzzy search/match can help with misspellings and researching unfamiliar terms, I have found it to have limited usefulness for those who know *exactly* what they are searching for.

Many semantic search solutions employing “artificial intelligence” often return results that mention words RELATED to the terms you were searching for – which are words OTHER than the words you actually searched for. While that can sound like a good thing, I have found that in many cases these words that the application “thinks” are relevant are in fact NOT relevant to your specific search. This can actually produce more false positive results that do not match the intent of your queries, wasting your time instead of increasing your productivity.

(Wo)Man vs. Machine – Human Cognition vs. Artificial Intelligence

You must take a moment to reflect and realize that an application that claims to perform semantic search is essentially trying to *guess* the intent of your query – it’s taking your search terms, example resume, or job description, and guessing what might be relevant to you.  Never lose sight of the fact that the operative word in “artificial intelligence” is “artificial” – these applications have no true interpretive/cognitive power.

When you are conducting a search for candidates, only YOU can determine what results are relevant to you – or in other words – what kinds of candidates actually are capable of performing the role you are hiring for. As we all know, resumes and social media profiles are often poor and inaccurate representations of candidate’s abilities and experience. Talented and experienced sourcers and recruiters go WAY beyond buzzword matching and quickly and intuitively apply significant interpretive analysis when reviewing search results, effectively “reading between the lines” to identify talent.

As such, I recommend that the results produced by automated semantic search and “artificial intelligence” matching applications be used for “suggested reading” – kind of like when you are searching for something to buy on Amazon.com and Amazon.com suggests products “you might also like.” But please don’t count on semantic search applications to replace talented sourcers and recruiters. I have found that these solutions tend to work best  with title matching and searching for “cookie cutter” roles. However, it must be said that even the most junior sourcer or recruiter can run a search for titles and find candidates.

A Final Word of Caution Regarding Automated Semantic Search

You must always keep technology in perspective. I’d like to quote a passage from The Toyota Way Fieldbook by Jeffrey K. Liker and David Meier:

“It is not enough to show in the abstract that IT can automate a process or provide more or better information. It must be clear how it will add value and support a well-thought-out and time-tested process. Typically, the process is done well manually before it is automated. The technology supports human decision making-it does not replace it. And the technology should not be used as an excuse to stop thinking and lose focus on kaizen.”

I could not have said it better myself! Semantic search engines and applications should not be seen as solutions to replace or eliminate human thought, analysis and decision making – they should be used to support human decision making.

What I see as EXTREMELY DANGEROUS is sourcing and recruiting organizations who are excited to implement semantic search/artificial intelligence matching solutions expecting them to “solve the problem” of the intrinsic challenges associated with leveraging information systems and human capital data for talent identification.

If you don’t have anyone on your team or in your organization who is highly proficient in MANUAL semantic search, let alone “standard” Boolean search/Talent Mining, how will you be able to assess if the automated solution is actually doing what the vendor is claiming it can do? How will you be able to test and determine if the automated solution is actually finding all of the available candidates, let alone the best available?

YOU CAN’T!

It is critical that you and your team first master the processes and best practices of manually searching and analyzing human capital data before you fast-forward and attempt to automate a process you don’t understand in the first place. If my word of caution isn’t enough for you, take Toyota’s advice – they’ve proven that they know a thing or two about technology and process automation.

Manual Semantic Search

While the semantic web and semantic search engines employing artificial intelligence/concept matching get a lot of buzz in the HR, recruiting and staffing industry, the big secret is that you don’t have to rely on a search engine, system or an application to perform semantic search for you.

Yes – sourcers and recruiters can perform semantic search manually.  I’ve referred to this concept as user-generated/user-defined semantic search in previous posts. Instead of letting an application take an artificially educated guess as to the intent of your search and give you results it *thinks* are relevant to you, you can create Boolean search strings that go beyond simply trying to match the words themselves and attempt to delve into the meaning implied by the words – targeting candidates based on what they DO, not just what they SAY.

Examples of User-Defined Semantic Search

#1 BASIC: adding functional/responsibility-related terms to your searches

• admin*, config*, deploy*, manag*, implement*, audit*, reconcil*

#2 MID-LEVEL: NEAR operator (Monster, Exalead…)

• (implement* or deploy*) NEAR SAP
• reconcil* NEAR (accounts or statements)

#3 ADVANCED: configurable proximity and variable term weighting (Exalead, Lucene, dtSearch)

• Exalead: (implement* or deploy*) NEAR/5 SAP

If the NEAR operator and configurable proximity searching are new concepts to you, or you’d like to learn more about how to achieve manual semantic search, I recommend that you read these 5 posts (click on links):

• Semantic Search for Sourcers and Recruiters
• Semantic Search for Sourcers and Recruiters Round 2
• Semantic Search Using the NEAR Boolean Operator
• Achieving Semantic Search Without Proximity Operators
• Extended Boolean: Proximity and Weighting

Benefits of Semantic Search

Quite simply, the reason why semantic search can be such a big deal to sourcers and recruiters is that when executed properly, it can:

#1 Significantly reduce sourcing time by reducing/eliminating “false positive” results of candidates who are not likely to be qualified

#2 Increase a sourcer’s/recruiter’s ability to quickly find more appropriately qualified candidates

#3 Enable sourcers and recruiters to move beyond “buzzword bingo” and identify talent based on what they are capable of DOING, not just the words they SAY in their resume/profile

Conclusion

Semantic search IS a big deal for sourcers and recruiters. Every day, more information about more people becomes available in the form of human capital data that can be found in resumes, social media profiles, Tweets, blog posts, press releases, etc. The end goal of leveraging information systems for talent identification (fancy speak for searching for candidates) is to quickly find the RIGHT people – people who are capable of performing the role you are sourcing/recruiting for.

One effective way of doing this is through a combination of manual and automated semantic search – leveraging search techniques and applications that increase your ability to find people who are more likely to meet or exceed your hiring requirements by tapping into the power of the meaning inherent at the levels of words, phrases, and sentences.

Be sure to fully understand and master the concepts and best practices of manual semantic search before jumping to automate a process you don’t fully understand and cannot evaluate properly or effectively. While it can be tempting to try and skip the challenges associated with Talent Mining via Boolean search, trying to replace or eliminate human decision making in the process of talent identification and acquisition would be a HUGE mistake!

Over the years, I have had the opportunity to evaluate several of the “big name” resume and job matching applications that claim to use artificial intelligence and I can say that the claim that they can find the same resumes that an “experienced recruiter” would choose is both accurate and inaccurate.

Photo: Scott Ingram Photography

From my experience, most AI matching applications can return some well matched resumes based on an example resume or job description. However, some of the results that are returned are definitely NOT good matches, although I can see why they were returned in the results. This is expecially prevalent when searching for job descriptions/resumes/hiring profiles in which many different types of candidates can mention the same words in their resumes.

What I’ve found from my own extensive hands-on experience testing these applications is that in addition to some well matched resumes, every AI matching application I have tested also ranks some results relatively highly who match the keywords, but not the “essence” of the job or resume. By “essence” I mean what the person will be responsible for/what the candidate has been primarily responsible for. If the search results don’t match the “essence” of the example resume or the job descritpion – they are of little-to-no value. Although in all fairness, junior sourcers or recruiters who are not very proficient at running Boolean searches may similarly struggle in creating Boolean search strings that return a high percentage of results of candidates who match the “essence” of an example resume or job description. Hopefully your staffing organization does not consist of only junior and/or sourcers or recruiters who are not proficient at creating effective and precise Boolean queries.

I believe the root of the limitation of AI matching apps is that essentially, all an AI matching application is capable of doing is finding resumes that it “thinks” are matched, based on it’s algorithms and pre-programmed search logic. These applications are simply taking words from an example resume or a job description and matching them to words that someone (the person or team who programmed the application) decided was related and relevant. In other words – regardless of marketing hype – the applications are simply buzzword matching, based on someone else’s programming. I’d argue that this is also essentially true for applications that make the claim that they “learn” as they are used. Let’s not forget the operative word in “Artificial Intelligence” is “Artificial.”

What should be especially disturbing to most people is that there is no objective way to assess whether the matching technology finds the BEST, or ALL of the potentially qualified candidates in the database or system. In my opinion, finding SOME matches is not a significant accomplishment, and certainly does not qualify an AI matching application to claim to be able to replace or be as effective as a human in the role of a sourcer or recruiter.

I think a good way to think about what all of these solutions do is to take a look at Amazon’s “you might also like” feature. While searching Amazon for a particular product, Amazon’s AI/matching engine tries to suggest other products you might be interested in based on the data collected from other online shoppers who have looked for or purchased similar products. Sometimes the suggestions are accurate – meaning you might actually be interested in the suggested product, but sometimes the suggestions are WAY off. What can we expect? After all – it’s just a computer program trying to “think” like a human (let alone YOU specifically).

Pretty much everyone in the staffing industry is aware of how imperfect resumes are in terms of their ability to represent a person’s experience and capabilities. The same is true of job descriptions. So you can pick your poison – allow a human sourcer or recruiter to interpret the the job description or the resume of the “ideal” candidate and create a search in an attempt to find qualified candidates, or allow an AI matching application to perform some exact and “fuzzy” buzzword matching.

Let’s get real here – a software application does not have the analytical and interpretive power of human cognition. With no example resume and a bare-bones job description, a good sourcer or recruiter can still find well qualified candidates. A software application cannot do the same thing – it can only “work” when given text to work with. Poor job descriptions will throw AI matching apps for a major loop and effectively render them useless. And while some job descriptions come with plenty of words – pages worth – in many cases all of the words actually don’t in fact accurately describe the position, it’s requirements, or define the ideal candidate.

Matching resumes is equally fraught with issues. Two candidates of similar skills, experience, and capability often have very different resumes. After all – the people we are looking to hire are not professional resume writers, nor should we expect them to be. Experienced sourcers and recruiters can “read between the lines” using cognitive processes and interpet resumes for skills, experience, and capabilities that are not explicitly mentioned in the resumes. This is something that an AI matching application may claim to do, but is hard pressed to prove. Let’s remember – AI matching apps can only work with text they are given – both the example resume or job description and the resumes it searches for. Although the people who program these applications can draw relation between multiple words/terms that can represent the same thing, they cannot possibly account for the unpredictable variety of ways in which a person can express – explicitly or implicitly – skills, experience, and capability.

AI matching applications do have value – in my opinion they provide “suggested reading.” They can certainly be used to assist junior sourcers/recruiters, or sourcers and recruiters who are not particularly adept at creating effective Boolean search strings. However, I would never implicitly trust that the matching engine is finding all of and the best matches in your database. The funny thing is – you’ll never know if this is the case or not…unless you’re really good at running searches, because then you can run your own searches and see if you can find more and better candidates that the matching engine did not find. Which is what I do.

This should go without saying, but one thing you should always keep in mind is that that all of these vendors with AI matching engines are SELLING you a product. They know that creating effective Boolean search strings can be hard to learn so they want to sell their solution as something that relieves sourcers and recruiters of having to run searches to make matches.

In my opinion, an ideal solution for most staffing organizations would include a resume/candidate database/ATS with BOTH a fully configurable search interface that supports full Boolean logic and ideally extended Boolean (with at the very least configurable proximity and variable term weighting), AND an AI matching engine – not one at the expense of the other. AI resume and job matching is great, but it is not a magical solution that replaces the human cognition of talented sourcers and recruiters. While AI matching apps may help junior recruiters, or those people who simply aren’t “wired” to run Boolean searches, it is not something anyone or any staffing organization should trust with making all of their matches. That would be scary!