Keywords are not the only thing that makes a page findable

Photo by Sarah Ziegler

Many people believe that keywords are the best way of making content easier to find. While there is some truth to this, it is pretty evident that as semantic search grows, the power of keywords in relation to other influencers diminishes. In short, in the current day, the power of keywords does not always provide the best way of making a piece of content findable. As I continue my path down of translating ideas from David's book to apply to internal enterprise search, I realize more and more that this basic concept is especially true for enterprise search.

Let's dig a little bit. It is a guarantee that when semantic search is involved, the search query always contains words which are not declared keywords on some of the pages returned in the search results. Instead these words come from other locations, from the content itself, from the comments on the content page, from social media that references the page. In addition, if the content or the page supports the ability to rate or like the content, these items can definitely influence the search results.

For employees to truly benefit from semantic in the enterprise, helping them to find the information they are looking for, social capabilities start to really have a huge influence that can't be ignored. While keywords might help, the content of the page, being written well, using the correct nomenclature on the page and allowing people to interact with the content in as many ways as possible becomes a very important factor.

This note was inspired by +David Amerland 's book, Google Semantic Search - Amazon location 2056

I wonder ... Enterprise profile photos in search results

Photo by Eric Ziegler

I know that google has gone away from authorship and providing image previews, but I still wonder if there is some value in providing an image of the authori/authors next to search results. I believe that there is some serious value in the enterprise of showing all authors of a document or a piece of content, so people know who all contributed to the content.

I wonder if there could be some sort of UI design that would provide images of the authors in certain instances (they are authorities on a subject in the enterprise?) and not show the author images in the search results for when they are not the recognized authority on a topic.  Similar to my last post, this technique would most likely drive people to the "higher authority" content.

The only thing that puts some level of doubt into my mind is the changes that Google recently did to their search results. I wonder if they found that the pictures did not add that much to how people found the content they were looking for. I wonder if they determined that having those images did not improve the "trust" that people had for the content.  If that is the case, I wonder if providing profile photos next to the search results would increase or decrease the trust people had related to the content.  

I wonder...

This note was inspired by +David Amerland 's book, Google Semantic Search.

Reputation --> Trust --> Semantic Search in the Enterprise

Photo by Eric Ziegler

Based on what I have been reading I believe that I understand that semantic search is built on the trusting the content that has been written. This trust can be gained via trusting the individual and can the trustworthiness can be increased by people indicating that they trust the content.  

Based on this assumption and assuming that reputation is built based on both digital and analog interactions (especially in a closed environment like the enterprise), how can a complete picture of reputation (and hence trust) be built within the enterprise? And how can a great enterprise search experience be built without the full enterprise reputation picture being created? I would venture to guess that the interactions in digital will approximate the reputation in the analog world, but then again, that is assuming that there is 100% adoption and interaction in the enterprise digital systems. 

Anyone else have any thoughts?

This note was inspired by +David Amerland 's book, Google Semantic Search - Amazon location 1853.

Building Widespread Trust because of Digital Reputation

Photo by Eric Ziegler

In the enterprise, reputation is traditionally built by analog interactions. Even when there are digital interactions,(email, esn, documents, papers, white papers, etc.) a large part of reputation still comes from 1:1, meetings, presentations, , etc. That means that reputation in the enterprise is not just based on a digital interaction. It also means that trust, which is based on reputation, is something that is built on both digital and analog interactions.

What I find interesting is that digital interactions can be so much more dramatically important than the analog interactions in building widespread trust. Why? The power of digital interactions in the enterprise is the reach it provides, allowing employees to build a reputation with employees they never work with and hence gain a level of trust with another employee that would have never been able to occur before that (series of ) digital interaction(s).

This note was inspired by +David Amerland 's book, Google Semantic Search - Amazon location 1842

The Enterprise Knowledge Graph Grows from the Employee Profile

Photo by Eric Ziegler

This blog post has a little more conceptual thought that will probably require time to read David Amerland's book, Google Semantic Search (see the reference below to get an idea of what I was reading to understand the thoughts in this blog post)

The knowledge graph is a semantic search concept that is all about interpreting what is seen in the internet. The same concepts can be applied to the enterprise, and to be honest is probably a lot easier to do. If I understand how the knowledge graph works, the knowledge graph is built by manually and automatically. The automatic part is completed by the level of trust sites are given and the level of trust of information the search engines find. The knowledge graph is grown over time as it determine what is most trustworthy.

One of the many thoughts I had on the concept of the knowledge graph in the enterprise is that it can be grown systematically based on employee profiles. The links to project sites, links to primary contacts, links to topics that are related are often found in employee profiles. So Building out an enterprise profile is key to creating a social network of connections. The profile is a key piece of information that can help the knowledge graph grow - growing through the connections of topics to employees and employees to employees and then back from employees to topics.

This note was inspired by +David Amerland 's book, Google Semantic Search - Amazon location 1727

The Personal Touch of an Individual

Photo by Eric Ziegler

As I read through different books, each books gets me thinking and the note I create might not make perfect sense.  And sometimes when I review my notes, a note causes multiple thoughts to occur that are really not related. Yesterday's and today's post are both the outcome of the same note from the same location in David's book.

Enterprises are often very quick to use organizational or departmental names when posting an article or piece of content, instead of posting the content with an individual's name. While this is viable and there can be argued some great reasons to do this, there are definitely reasons to minimize the approach of using the non-descript, non-personal organizational name.

Using an individual's name and associating it with the content provides that personal touch that an organization name or identity does not. And if you do it correctly, the arguments for using an organizational name becomes less of an issue (not going into the details on that subject here). As a company creates content and applies a person's name to the content, the person becomes a voice of the department or organization, and the individual becomes the voice and authority on the subject or topic.  In addition, the employees will embrace and interact with the content more often because they feel a person is talking to them, versus some non-descript organization.

This comment was inspired by +David Amerland 's book, Google Semantic Search - Amazon location 1457

Multiple Authors, Authority and SEO

Photo by Sarah Ziegler

As I read through different books, each books gets me thinking and the note I create might not make perfect sense.  And sometimes when I review my notes, a note causes multiple thoughts to occur that are really not related.Today's and tomorrow's posts are both the outcome of the same note from the same location in David's book.

One of the issues with authorship occurs when more than one person is responsible for the blog post, document, wiki page. This is especially exposed when the last person that modified the document appears as the author of the document or content. Thankfully in the enterprise, there is a solution already in place to help resolve this issue (at least in most instances). Most internal collaboration and intranet systems include a mechanism identify each of the authors via history and versioning. Based on this history, the content can be attributed to each of the authors.

Enterprise search systems can use this extra meta data to increase authorship rank, trust and authority of that person on the subject while also influencing the page rank of other content from the same author on the same subject.

This comment was inspired by +David Amerland 's book, Google Semantic Search - Amazon location 1457

Page Rank, Authority and Enterprise Search

As explained in David's book, authority is used to help determine the rank of a piece of content.  And page rank is most likely influenced by using the items that David highlights in his book. Specifically:

• Who created the content
• What else that person has created in the past 
• The content creator’s social media connections
• The content creator’s online activity with further content
• The content creator’s interaction with other people
• How the content this person created was received in a social media setting
• The content’s quality, authority, and originality.
• The content’s stylistics (language level, reading difficulty, paragraph length, use of headings and subheadings, overall length, embedded links, supportive links in footnotes, citations, images, and any multimedia embedded in it.

I am not willing to completely read between the lines on this, but I sense that there could be a hint of not only knowing what content was created in the past by the person, but actually what content has the person created on the same subject in the past. If I do or do not read between the lines, I am thinking that authority can be taken to an extra layer of granularity within the enterprise.  What I mean is, authority can actually be assigned to employees for a specific subject area.  

Even in the enterprise, a page rank on a subject can still be applied using the bullets above with a couple of small adjustments.  Page rank would be influenced based on the person's previous content created on the subject, including both writing and social interactions on the subject.  

So, by building on the original thoughts in David's book, the ideas on determining the rank of a piece of content depends on not just the general authority of the person that created the content, but can be strengthened based on the authority the employee has on the subject the content is about.  (btw, I could have completely gone down the path that page rank should be based on the subject of the page, so it becomes more granular and is a subject page rank - this concept is much more difficult to do).

This comment was inspired by +David Amerland 's book, Google Semantic Search - Amazon Location 1351

Enhancing the Enterprise Community via Search and Social

Photo by Eric Ziegler

This comment was inspired by +David Amerland 's book, Google Semantic Search.

David mentions that communities back in the day numbered in the 100's and that in those communities it was easy to reach out and contact another person since most people knew each other.  In many companies (not all) the same thing happens, employees are numbered in the 100's. While companies can and are in many cases similar to the the small communities mentioned in David Amerland's book, it doesn't mean that employees can just go over and talk to another person to get the answer. Employees do not know everyone in the enterprise as they would in a community.  The dynamics in the enterprise is different than a community. So, some of the best ways of getting answers to question employees don't know is to use search and to use social.

And if that is the case, what are you doing to ensure that search is built in a way that finds useful information and how are you preparing your organization to embrace social?

Reducing the Gap through Semantic and Social

Photo by Eric Ziegler

This comment was inspired by +David Amerland 's book, Google Semantic Search

As I think about semantic search and the value it can provide to an enterprise, I realize very quickly that enterprise semantic search is all about reducing the gaps and the distance between employees, groups, teams, departments, divisions. And combining the power of enterprise social with semantic search is even more powerful that either on their own.  These thoughts lead me to believe that it is all about getting people to realize and recognize that interacting, finding, discovering and interacting again is all about reducing the gap.

Search in the Enterprise and HR Systems

Image: Eric Ziegler

Three weeks ago I started a series of blog posts about search engines in the enterprise.  My premise was that the search engines in the enterprise are not as good as the search engines in the internet.  I do have to say though, that this is most likely not completely the fault of the vendors that provide such tools, but more about the difference on how people create content for consumption on the internet vs. the enterprise.  This was the premise of my first post.  In the next set of posts, I started to propose that there are ways around the behaviors of employees for people on the internet and the search engines used in the enterprise could adapt to improve the search experience.  In my second post, I discussed how using the context of the employee can provide enterprise search engines a boost in providing improved search results. In my last blog post, I started to provide more details on what I mean by employee context by discussing connections of employees (e.g. following each other). I provided several ways connections can improve search results.  

In this blog post, I plan on discussing another part of how employee context can improve search results.  The Human Capital Management (HCM) profile is my focus for this blog post.  Companies have a wealth of profile information on each employee.  This profile information comes from the Human Resource or Human Capital Management (HCM) systems.  HCM systems contain data that captures who each employee works for and who each employee works with.  These systems also know what each employee's job title, where they are located (building, country, etc.) along with having the employees entire job history.   

HCM Connections : As discussed in my previous blog post, connections provide information that can improve search results. HCM systems provide many different types of connections.  The first connection is between employee and boss.  The second connection is the connection between peers on a team or within a department.   While the employee might not be following their boss or the people they work with, they still have connections with these people.  Bosses, employees and their peers all work together on projects, documents or presentations.   Using similar reasoning as the Directly Following example in my previous blog post, search results can be improved by these HR releated connections. The content created by a manager or by a peer should get bump in relevance because of the relationship between that employee and the person doing the search.

Location : Another piece of information that often comes from the HCM systems is the location of the person.  When I talk about location I mean, the country and city the person works.  I also mean where the office the person sits in, assuming they don't work from home.  If they work from home, this information is typically captured also.  Each of these locations can be used to improve the search results.  For example, if the employee is located in Belgium, and searches benefits information, search results should be returned in context, and not return a link to the Japanese benefits content.    Or if the person searches for what is being served for lunch today, the lunch menu for the company cafeteria that is closest to his building (if it is not actually in his building) should be the top result returned.  Again, search results in context.

While I highlight only two types of data from the HCM system, there is the potential for a log of other information that could be used to improve the search results for the employee.  Of course there are concerns that need to be addressed.  If there is personal information about the employee, there are privacy or security concerns.  But if careful planning occurs and the correct legal and security teams are consulted, the data from the HCM system can dramatically improve search results for each employee.

What other types of HCM data could be used to improve search engines in the enterprise?  What other types of connections can make search engines better?

Enterprise Search Failure - Connections

Image: Eric Ziegler

How often do you hear someone say, "why doesn't our enterprise search work as well as Google search? Bing?" or "Why can't I find the content I want  to find." or "I can't believe our search engine sucks." or "Our enterprise has a very small fraction of the content that Google searches and I still can't find the content I am looking for." or "FIX IT!".

In my previous blog posts, I started a series of  posts about internal enterprise search and how it is not as good as internet search.  In my first post, I provided an overview of how internal social interactions can improve internal search engine results.  In my second post, I discussed how using the context of the employee can provide enterprise search engines a boost in providing improved search results.  As a definition for what I meant by employee context I proposed that employee context is made up of a wide variety of types of information, including HR system information, social profile information, social connections and social interactions.   In this blog post, I plan to discuss how the power of social connections can be used to improve search results within the enterprise.  

When I mention connections, I am referring to the idea that one employee "follows" another employee.  This following is similar to those external social networking sites such as facebook, twitter and google+.   But how can these connections between employees be used to improve search results?  Below are several illustrations of how connections can improve search results for each employee.  For each of the examples below, I am using the following base example to illustrate my point:  

Joe, an employee, follows five other employees and has 15 employees following him (Joe).  In addition, the five employees Joe is following, follow a combined 20 more people (some with multiple people following the same person).   Each of these connections (both direct and indirect) can play a key role in improving Joe's search results.  

Directly following:  Joe is directly following five other employees.  Content created, modified or interacted with (comments, likes, tagging, bookmarking, etc.) by these people has a higher importance to Joe then other employees.  Think about it, Joe is following these people for a reason.  So why is the content these people create, modify and interact with not given higher relevance when providing search results to Joe?

Directly being followed: Similar reasoning can be used for the content created by the 15 people following Joe, but it goes in reverse.   Joe does not realize that the 15 people are creating good content.   Joe is not following any of these people, but they are following Joe because he creates content that is related to ideas each of them are interested in.  Because of this, the content these people create has a higher chance of being valuable to Joe, he just doesn't know it.  But there is a caveat to this, Joe is not following these people, either because he has knowingly not followed these people or he has not discovered these people.   Because of this, the content created by these 15 employees should not receive as much of a relevance boost as content created by the direct followers.  

Indirectly Following: Again, similar reasoning can be used for the indirect followers.  Joe is following 5 people who are following a total of 20 people that Joe is not following.  Since Joe trusts and follows those 5 people, there is some merit and a higher chance that the content created by the 20 indirect employees will be of higher relevance and importance to Joe.  Because of this, the content created by these 20 employees should get a relevance boost in the search results. But just as the 15 employees that follow Joe, there is a caveat.   Joe is not following these people, either because he has knowingly not followed these people or he has not discovered these people.   Because of this the content created by these 20 employees should not receive as much of a relevance boost as content created by the direct followers.

With these improved relevance boosts for people direclty following, directing being followed and indirectly following, Joe searches on the term Java and receives results with an improved relevance boost for the 5 employees he is following, the 15 employees that are following him and the 20 indirect employees. In addition, the content from the 5 direct employees has the largest relevance boost, with the 15 employees following Joe and the 20 indirect employees he is following improving the relevance but not with as much of a boost in relevance.  

With these search engine algorithm improvements, the search results have just gotten tremendously improved, making searching for content a better experience for Joe and every other employee.  In future blog posts, I plan on continuing to discuss topics of how to improve internal enterprise search engine results. Future topics include reviewing how HR systems, employee profiles and social interactions can be used to improve search results.  So check back periodically to hear my thoughts on how enterprise search can be much better than it is today.

Enterprise Search and Context

Image: Breen Jones (link)

In my previous blog posts I talked about enterprise search and the reasons why search was not as good in the enterprise as it is in the enterprise.  The post was about why enterprise searches fail and that using social internal to an enterprise, enterprise search can be greatly improved.  +Joachim Stroh (@Joachimstroh) posted a comment on my blog that adds to my ideas.  He states:

"The traditional view is that content is coming from carefully curated repositories that are tied to corporate taxonomies that are tied to search facets. Way too complicated...."  

My response is absolutely correct! My thoughts are that if an employee has to do something beyond what is necessary for the job it probably won't happen.

So if employees are not helping other employees to find the valuable content they created, how do you improve search in the enterprise?  In my previous blog post, I mentioned that social business / E2.0 are both valuable ways to improve search results.  But there must be other ways to improve search.  In addition, social is way deeper than I could ever talk to in one blog post.  In this blog post, I will attempt to outline how enterprise search can be improved by providing search results based on the context of the employee performing the search.

As I did in my last blog post, I want to first analyze what are some of the tricks that internet search engines use to improve search results for each end user.  Internet search engines provide search results based on the context of who the person is that is searching.  They use the location of the person and provide results that are located closer to their current location.  Internet search engines keep track of each person's recent interactions and uses those interactions (which search results did they click) to improve search results. Internet search engines have now started the trend of incorporating end users social interactions to improve the results.  They are using a persons connections (relationships, who they are following, who they interact with most often) along with a host of other social information to improve search results. Each of these improvements to internet search is about providing search results in the context of person doing the search.  For example, when I search on a term, the potential that the search results are different for me vs. you are getting higher and higher.  These changes to how internet search engines provide search results to people is very powerful and a great way of providing the best results possible when there is a humongous amount of content in the internet.

So how does enterprise search learn from internet search?  Enterprise searches can do every one of these techniques to improve search in the enterprise.  And the amazing thing contextual search results is that I believe this is the one spot where enterprise search can do a much better job than internet search.  Enterprise's have a wealth of basic information that they know about employees.  While the internet search engines can learn from personal profiles, enterprise's can use information from the human resource systems along with personal profile information (internal and potentially even external (privacy could be an issue here)).  Enterprise search engines can learn from the interactions people have every day.  For example, the systems employees use and interact with each day can provide a wealth of information on how an employee does their job and what systems they use most. But interactions don't just happen with systems of record, but they also can occur on internal social platforms.  These social platforms can provide a large amount of information about who people are connected to, what interactions they have with other people, and the types of topics they are most interested in.   Lastly, by combining many of these pieces of information together, you can determine an employees expertise, which can also influence and improve search results for that employee.

In future blog posts, I plan on talking about the internal search engine topics.  This topic is so large and has so much potential that I will be creating several blog posts on how "employee context" can improve enterprise search.  Topics will include: the employee profile, the influence of interactions, the connections of employees and expertise of employees.  Each of these topics will review how they have the ability to improve search results for the employee.  So check back periodically to hear my thoughts on how enterprise search can be much better than it is today.

Enterprise Search Failure

Image by: Phillie Casablanca - link)

How often do you hear someone say, "why doesn't our enterprise search work as well as Google search? Bing?" or "Why can't I find the content I want  to find." or "I can't believe our search engine sucks." or "Our enterprise has a very small fraction of the content that Google searches and I still can't find the content I am looking for." or "FIX IT!".

Fix it sounds so easy, but it isn't that easy.  Internet search is dramatically different than enterprise search.  While there are many technical  differences, the biggest difference is not technical by behavioural.  People creating content in the internet are creating the content with the hope that others will find it.  These people, creating content on the internet, have personal incentives (marketing, personal brand, selling a product) to have other people find the content they are creating.  They use many different techniques to improve the finability of their content.  So, what techniques do they use to improve their content?

They tag their content, they use Search Engine Optimization (SEO) sites to improve their content for search engines.  They use social media to  market their material, which also improves the findability of their content.  The more references to their content, the more likely the content will  show up in a search result.  By using social media, people are attempting to garner a loyal following that promotes their content (think free advertising).  And if they can use social media to have their content cross linked and promoted by other people, the chances that more people (directly connections and indirect connections) are going to see their content, which in the end is exactly what they are attempting to do.

Let's contrast this with what happens in the enterprise.  In the enterprise, employees create content typically for a specific group of people, not  intending for the content to be found.  In many cases, employees are not creating content to be shared. (There are exceptions but those are for  specific groups setup to share corporate information, such as a corporate communications teams).  In contrast, the vast majority of the enterprise is creating documents and information and not thinking about how the information could be found or discovered. They are not sharing and there are no  incentives (personal or corporate) to share the information.  In many organizations the tools the employees use to create content and store the content is not even setup to allow for the content to be easily found.  Enterprise search engines often are not even searching the repository where the content is stored.  While the technical connections are not available, the employee is not helping the rest of the company to find their content either.  For example, when the employee wants to let someone else know about their content, what tool do they first turn to?  eMail.  If you are lucky they send a link to the content.  If not, they send the attachment in the email.  The only people that benefit from the later is the sender and the recipient.  In addition, in both scenarios, search engines don't learn and improve relevancy when eMail is used.

What are some ideas on how to improve the search results in the enterprise?   Incorporating social media techniques into the enterprise is one method of improving search results in the enterprise.   By providing a platform for people to collaborate, share, and to tout what they are working on is just one step in improving search results.   If an employee can use these tools and the search engine can learn from the enterprise social media interactions, the search results will improve.  By providing these hooks, such as tagging of content, cross linking of the content, and getting more people to interact with each each other and with communities will improve the search results, much the same way social media improves the search results in the internet.

But that is not the biggest thing that needs to occur.  While a technology can be put in place to allow people to interact, interactions don't occur because of the technology.  Companies have to change their culture;  companies have to get their employees to think from a sharing perspective. They have to create information and documents with the realization that others can find it and learn from it.  Companies need to find the incentive for the employees to partipate and be active and to share.