+1-312-477-7300
eContext

Why We Love Taxonomy

by

For the taxonomy editors at eContext, categorical organization is all we think about all day, every day. We look for new ways to approach taxonomy and we even find new ways to talk about it. In fact, during interviews, an employee used to pose an unusual question to job candidates: “How do you organize your spice rack?”

 

This simple question worked so well because it was completely unexpected and showed us how good the candidate could be at thinking fast on his or her feet. It’s also the kind of question that people answer honestly, and the answer would say a lot about how well the candidate might do at his or her job here at eContext where we’ve set out to organize, well, everything into a massive taxonomy of social and commercial topics.

 

We got a lot of great answers from candidates who proudly announced their strategy to keep the spice rack organized by size, type, use, and alphabetical order. It’s the kind of organization that’s automatically accessible and intuitive, no matter who’s cooking your kitchen (or shopping in your grocery store, come to think of it.)

 

But we also received well-argued responses that defied our expectations; one candidate grouped his spices according to their frequency-of-use in household favorite recipes. It’s a reminder of how niche-use schemes can have their own value, not to mention how important it is to be cognizant of all the different ways a concept can be classified. Regardless of the approach, a thorough and passionate answer would tell us, “this is a candidate for eContext.”

 

Taxonomy models

 

As we constantly tweak and iterate our taxonomy, we look to other models for inspiration. One taxonomy we love is the Luma Institute’s Taxonomy of Innovation. If you’ve ever wanted a roadmap for bringing an idea to life, all you’ll need to do is follow the steps mapped out in this wonderful interactive taxonomy. Guiding you through the three main stages of innovation —  looking, understanding, and making  —   Luma’s taxonomy helps you explore and understand each step of the process, drilling all the way down to giving you a format to follow for a brainstorming meeting and telling you how many seconds to allow meeting participants to come up with answers.

 

Luma’s taxonomy is a great way for us to help explain to those outside of our taxonomy-loving bubble what we do. It’s sometimes difficult for everyone to see how important it is to be able to organize data into a deep hierarchy on a web taxonomy, but we find almost everybody has had an idea at one point or another for a new product, and they can quickly relate to the Luma taxonomy, with its deep detail, and see the value it brings to a business process by providing a framework for helping ideas come to life.

 

Taxonomy and your business

 

If you have a business that relies on anything digital  —  a website, digital campaigns, or a social media presence, for example — then taxonomy can help you innovate. You can organize any kind of information about your business or your performance into a taxonomy so that you can see the progression of information and how it affects your business.

 

In our industry, the starting point is the basic taxonomy set by the Interactive Advertising Bureau (IAB). This basic taxonomy covers two tiers and most major commerce categories. However, two tiers is a shallow pool when you consider how much detail you need to get to know your customers better through the data you pull.

 

eContext is a deep taxonomy, providing a richer way to dive deep into your data by pulling it into a taxonomy with more than 450,000 nodes. With  up to 21 tiers of specificity by which you can understand your data, you can always zoom in or out to pull the insights you need at any given time. That flexibility will help you make better decisions about how to present information on your site or how to reach out to potential customers in your digital and social campaigns.

 

Weather you’re thinking about how to arrange your spice rack, launch the next great product innovation, or structure your web presence, taxonomy can help. And we’re here to help you build a better, deeper taxonomy.

 

Now, where did we put the cayenne?

by

Last week, one of our favorite things (the library) teamed up with one of our favorite radio shows (The Moth) using one of our favorite technologies (speech-to-text) to grow a taxonomy of storytelling. We’re elated!

If you’re a fan of The Moth, you know it’s “storytelling, live, without notes” — a live event, radio show,  podcast, and more that features real people telling stories on a stage. Stories range from raw and tragic to deeply meaningful and wildly funny.

Since launching in 1997, more than 20,000 stories have been told from Moth stages across the U.S., and now, throughout the world. In an effort to better curate their collection of stories and make them available to people with disabilities such as hearing impairment, The Moth set out to find a partner who could help create transcripts.

This led The Moth into a collaboration with The New York Public Library and a speech-to-text company called PopUp Archive. With a grant from the Knight Foundation, the three organizations collaborated on a very interesting strategy for creating and archiving transcripts of The Moth.

Turning Speech into Natural-Language Text

The problem with speech-to-text programs is, well, sometimes they don’t hear very well. In addition to basic language barriers such as accents, dialects, and fluctuations in speech, which occur all the time, Moth recordings typically take place in a bar where there is a lot of ambient noise, varying levels of audio quality, and other factors that can make a speaker’s words difficult for a machine to accurately recognize. This means that transcripts produced by PopUp Archive would need a good edit, and we’re guessing the Knight Foundation’s grant wasn’t going to cover editing expenses for 20,000 files.

The New York Public Library (NYPL) had a solution: Crowdsourcing. Turns out NYPL Labs had developed an interactive transcript editor called Together We Listen, which allows anyone to log on, listen to Moth stories, and assist with editing the transcripts as they go. Volunteers smooth out the transcripts by looking for misspelled words, or misplaced synonyms, putting words and phrases into the context that the speaker intended as they go. Volunteers are also asked to add tags to each file to make it easier to catalogue and categorize the collection.

A Taxonomist’s Dream Come True

We probably don’t have to tell you how much we love this idea. As people who sit in an office all day categorizing words and phrases, we’re thrilled to learn that people are out there volunteering their time to curate stories. The categorization element exemplifies the quiet work that taxonomists do every day in order to curate and maintain information.

The project also illustrates the value of classification on a very practical level. By transcribing Moth stories into text, the organization is making the stories more accessible to all of us, particularly those who may not have been able to enjoy Moth stories before they were available in print. And, because these stories represent such a wide array of topics and audiences, we’ll now be able to reach into the archive and pull out the categories or topics that we want to learn about. Additionally, The Moth will be able to use the catalogue in a number of ways to enrich its offering. For example, they can dig deep and find collections of stories to publish, such as stories from cancer survivors, or humorous stories that will make us laugh. They could pull together various collections to package and sell as educational tools, or deliver favorite celebrity stories online as sponsored content.

We’ve curated more than 450,000 terms, and we’re here to tell you, it can sometimes be a lonely business. Knowing that there’s a growing community out there that loves to categorize, curate, and organize information into a hierarchy makes us smile.

In June, the NYPL is hosting a hackathon and storytelling celebration. We’ll provide more details when we hear about it. In the meantime, be part of the solution by helping to curate these stories!

by

Taxonomy is the heart of eContext. Whether you’re using our technology to expedite machine learning projects, segment audiences by interest, or develop conversational interfaces, a curated subject matter hierarchy is what makes our semantic classification both intuitive and flexible.

Of course, many of the businesses we work with come to us with their own existing topic structures—some homegrown, some open-source. If you need an easy example, just think of any shopping website and the navigation panel you’re sure to find on the left side or top of the screen.

So the obvious question is: if companies already have their own taxonomies, how can they utilize ours? Some prospective customers have expressed concern that, in order to incorporate our text classification capabilities, they’d have to ignore their native structures in favor our ours.

Well, not to worry: there’s no need to throw the baby out with the bathwater. One of the services available to our clients is a taxonomy overlay, which means we map our categories against yours, so subsequent classification efforts are geared towards your unique structure.

How does it work?

Step One: Classify

We pass your existing structure, including any relevant metatags or descriptions, through eContext’s classification engine. The output shows which of your organizational categories correspond directly with our own for the simplest of 1-to-1 conversions. We’ll then review the results and make sure there aren’t any errors.  If we need to build any new categories into the eContext taxonomy, that will happen at this stage as well (though it’s rare that you’ll have a subject in your structure that we don’t have in ours.)

Step Two: Extend

 

We’ve been perfecting our category structure since 2008, working throughout that time to improve in both breadth and depth. Therefore, we typically find that the eContext taxonomy is much more comprehensive than other organizations’, covering a wider variety of subjects and segmenting topics with deeper tiers of subcategories.

Thus, clients receiving an overlay have the option to augment their existing structures with topics or segmentations they may be lacking. Suppose you’re a sporting goods retailer, and your organizational structure goes as deep as, say, Sporting Goods >> Winter Sporting Goods >> Skiing Equipment. eContext could map these categories against our taxonomy as described in step one, but we could also extend this branch to include several subtypes of skiing equipment, such as Ski Poles, Ski Boots, or even Ski Wax.

Step Three: Benefit

With your structure reconciled with ours–and extended where necessary–the vocabulary rules that govern our text classification can automatically be applied without any changes required on your end. Now, any new data you classify via eContext will fit your current setup, which is great if you’re taking in more data than you’re able to sort manually. Besides streamlining your existing operations, taxonomy overlay has the added benefit of future-proofing your data, minimizing complications that arise from incorporating new software or merging with another organization.

This is a pretty high-level rundown; there may be unique considerations (and advantages) in performing a taxonomy overlay for your business. We’d love to discuss with you, so feel free to contact us with any questions.

by

In our last post, we discussed classification in general, arguing that structured labeling helps businesses leverage more information with fewer resources. For the next couple of entries, we’re going to focus more specifically on semantic classification, which is eContext’s core specialty. eContext labels massive volumes of content by subject matter, so our clients can better manage and learn from their data without having to manually take stock of each individual item. These Twitter posts mention NFL players; this video is about Silicon Valley; this customer survey is focused on retail jewelry, etc, etc. As previously mentioned, our clients typically access our classification by way of the eContext API. Before digging deeper into concrete use cases, we thought we would answer a few of practical questions that come up when new clients use the API for the first time.

What does the eContext API actually do?

The eContext API allows subscribers to map their content to eContext’s Categories. A category could represent a thing, person, place, product, service, or an abstract concept. Categories are organized in a huge taxonomy that houses specific topics beneath broader ones; each category exists in a single location within this structure. A user submits a batch of content (such as a list of social posts or URLs) and receives, for each individual item, a list of corresponding categories that includes the following information:

  • Category name
  • Category ID
  • Category path
    • for example, the category path for “Breaking Bad” is Arts & Entertainment::Movies & Television::Movie & TV Products::TV::Drama TV Shows::Breaking Bad
  • A few statistics on the category’s prevalence within recent social conversations
    • the importance of the category within recent social conversations
    • the percentage of conversations in the past 28 days that have been about the category

 

What are the different functions available through the eContext API?

Clients can make different classifications calls depending on the kind of data they need to label. This is because eContext uses different language processing for different types of content. The core classification functions of the API are:

  • Classify/Text – The most basic functionality, designed for free-form text.
  • Classify/Social – Used for social media posts and other user generated content. The language processing used for Classify/Social is optimized for short-to-medium length content, and also includes functionality to consider usernames and hashtags.
  • Classify/Keywords – This function is optimized for very short text strings. While other calls can map content to multiple results, Classify/Keywords will give each keyword a single, best-possible category. This allows keywords to be bucketed into discrete groups without duplicates.
  • Classify/URL – Users submit a list of URLs and eContext labels the topics that appear on each of those web pages. Our processing method ignores advertising and other irrelevant elements, focusing only on the core content of the page.
  • Classify/HTML – An alternative to Classify/URL, this call allows users to hand-select which page elements are to be classified.

In addition, the eContext API offers a few ancillary functions, allowing users to see a list of eContext’s top-tier categories, obtain keywords that have been pre-mapped to individual categories, and check their own usage information. For a more tech-oriented guide to the eContext API, feel free to check out our documentation.

What can I do with eContext Classifications?

It depends on your goals, but broadly speaking, applications for semantic classification can be divided into two groups: utility and insight. UTILITY APPLICATIONS: eContext classifications are used to facilitate some other (usually automated) process.  Examples include:

  • tagging browseable content for intuitive navigation, related content suggestions, etc.
  • deriving user profiles that can be used to automate personalization
  • organizing data in CRM so an organization can quickly source information by topic
  • automatic content filtering for improved relevance and/or brand safety

INSIGHT APPLICATIONS: eContext classifications are used to analyze large-scale digital activity, typically for marketing and research purposes. Examples include:

  • discovering the distinctive interests of a target audience
  • identifying optimal channels for efficient media buying
  • content ideation to publish media that resonates with consumers

Of course, many of our clients use eContext in ways that combine these “utility” and “insight” benefits. For instance, an online retailer might classify product descriptions so they’re easier to search, but can also get the added value of topic-by-topic conversion stats. In customer service, classification can help dispatch tickets to the best representative, but it can also help the company analyze the kinds of problems their users are experiencing. Check back soon for an in-depth look at how clients are putting eContext to use.[/vc_column_text][/vc_column][/vc_row]

by

Recently, we talked a bit about the concept of topic intelligence: the ability to identify which specific subjects are mentioned within your text data. Aside from improved tagging and retrieval, topic intelligence can teach you about the behavior of consumers who interact with your content.

Usually in our examples, the user behaviors we discuss are positive ones: engagements, conversations, those actions that publishers and retailers seek to encourage. If you’re a purveyor of consumer content, keen awareness of that content’s many subject matters can help you figure out which topics resonate with your audience.

But what about those behaviors that organizations (and many users) find undesirable? Can you arm yourself against trolling, abuse — any of those actions which you may hope to discourage — by understanding, on a topic-by-topic basis, where those behaviors tend to occur?

This week, a handful of UK journalists grappled with just that question. Unusually for a news outlet of its kind, the Guardian turned the flashlight on itself — specifically, on the user comment threads that appear beneath most of its articles — as part of a series entitled “The Web We Want”. Comment sections in general are notorious for negative user interaction; the goal of the Guardian’s article was to spot significant trends that could identify when and where users might post abusive comments.

We decided to treat the 70m comments that have been left on the Guardian — and in particular the comments that have been blocked by our moderators — as a huge data set to explored rather than a problem to be brushed under the carpet.

The Research

The Guardian looked at comments that were labeled abusive and took note of the articles those comments appeared beneath. Researchers classified each article according to two different features:

  • The gender (and other demographic characteristics) of the article’s author
  • The article’s subject matter

The gender flag, it should be noted, was particularly striking. According to the investigation, “articles written by women got more blocked (ie abusive or disruptive) comments across almost all sections.” Unsurprisingly, articles written by straight, white males received the lowest percentage of blocked comments.

The examination of subject matter yielded significant results as well — and it is in this line of inquiry that we see the benefits of topic intelligence:

We also found that some subjects attracted more abusive or disruptive comments than others. Conversations about crosswords, cricket, horse racing and jazz were respectful; discussions about the Israel/Palestine conflict were not. Articles about feminism attracted very high levels of blocked comments. And so did rape.

The message is clear, if already intuitive: articles about divisive or turbulent subjects are more likely to draw abusive user comments.

It’s worth noting that the Guardian tags its articles with a fairly high degree of specificity — instead of simply relying on broad, traditional categories like sports and fashion (though those distinctions exist as well), the articles are tagged more granularly, with niche topics such as crosswords and horse racing. Even a sensitive subject like rape is tagged, because developing an awareness of subject matters means recognizing any topic, regardless of whether or not it’s pleasant to talk about. Because the Guardian organizes its content both comprehensively and with specificity, its researchers were able to develop a clearer understanding of the topics that are typically met with abusive comments.

The Response

So what do you do with this kind of intel? For one thing, publishers who can predict abuse on a topic-by-topic level are better equipped to moderate user behavior more efficiently, or even learn when to turn off commenting altogether:

The Guardian has already taken the decision to cut down the number of places where comments are open on stories relating to a few particularly contentious subjects, such as migration and race. This allows moderators to keep a closer watch on conversations that we know are more likely to attract abuse.

Beyond moderators, the Guardian’s journalists can also benefit from an awareness of troll-magnet topics. Style, tone, rhetoric — authors make innumerable decisions regarding how best to communicate with their audiences, and even if it doesn’t affect their overall approach to telling  a story, knowing which subjects tend to draw abusive comments can inform that decision making with an eye towards clear and uncompromised communication. Even marketing and PR teams, working to put this content in front of new readers, can be more effective at their jobs with these kind of findings. After all, the last thing you want is for potential users to associate your content that’s “crude, bigoted, or just vile.”

Going Deeper with Topic Intelligence

Crucially, the report included links to a sibling article that provided some details on the researchers’ methodology. The comment section of this article (fittingly) contained constructive advice from readers on how this research could be strengthened.

For what it’s worth, we at eContext have the strongly-held belief that increased granularity always leads to increased clarity. Perhaps in subsequent investigations, the Guardian can go a little deeper into their topics to try and uncover even more helpful correlations between subject matter and negative user interactions. In addition, organizing tags to a hierarchical structure could foster awareness at different tiers of specificity: Which topics within “Technology” are most likely to draw abusive comments? In articles dealing with the Israel/Palestine conflict, are there certain co-occurring subjects that reliably inhibit unwanted behaviors? These kinds of questions and others require an even more exhaustive tagging system, one in which specific tags are nested under broader categories for increased flexibility and coherence.

Finally, it would certainly be useful to classify the abusive interactions themselves. Which topics tend to appear in blocked or deleted comments? Do those topics align with the referring articles’ content or do they tend to differ substantially compared to the comments deemed appropriate? As long as you have the means to structure and learn from this kind of information, we firmly believe that the more behavioral data you can attain, the more equipped you’ll be to make constructive decisions.

On the whole, we found the Guardian’s efforts to be a terrific example of an organization using topic intelligence to foster a more positive experience for their audience. We hope other outlets will follow this example and classify their inventory to gain a more comprehensive understanding of content.  

by

We continue to hear more about personalization for social media channels. This past week, the Instagram blog had a very simple announcement: “To improve your experience, your feed will soon be ordered to show the moments we believe you will care about the most.”

There’s a big promise in that simple announcement. Instagram is letting us know that it plans to invest more in getting to know what users like and dislike, and place a higher priority on understanding how to deliver a more relevant experience.

The Instagram blog also notes, “you may be surprised to learn that people miss on average 70 percent of their feeds.” Instagram knows that the social universe is getting ever more crowded, and they have to do a better job of grabbing and holding our attention. eConsultancy reports that Instagram users spend 21 minutes a day on the app. That’s a lot of time, but if we assume that a typical Instagrammer breaks up her 21 minutes into smaller intervals throughout the day, and that she spends those intervals paging quickly through her friends’ posts, checking her own for likes and comments, and taking in the occasional ad, it’s clear how crowded those 21 minutes really are. Bore her and she’ll just page through faster, missing even more of what she logged on to see.

 

Better social media personalization

 

To improve the experience, Instagram told users, they will now re-order Instagram feeds so users can “see the moments we believe you will care about the most.” We’ll infer that Instagram is also tweaking their feed to deliver the ads with which users are most likely to engage.

And that’s OK. Personalization for social media should be laser-focused on delivering to us what we want. If our social media channels know us so well that they can know who our favorite musician is and show us a video from that concert we wanted to see, as Instagram promises, that’s just as good as your best friend writing to tell you, “did you see this on YouTube last night?” And if Instagram can get just as good with delivering ads, it’s the same as your friend writing to tell you, “tickets go on sale tomorrow for the show, you in?”

 

Getting to that personal level

 

The trick is how to get there. After all, in real life and in social media, it takes a while to build a friendship, and there’s a lot of trust and sharing that has to go on. But it’s through that dual process of sharing and extrapolation that you can get to the heart of the relationship.

There are a lot of ways to parse data in order to get to know a user better. Smart players in social are using a wide variety of cues to make better decisions about what to serve up to users. From user profiles, listening tools, and basic clickstream data to more sophisticated approaches based on natural language processing and machine learning, social sites are able to draw a more accurate picture of who wants what.

eContext enhances that process in a few ways. First, we can help you pull all of your data into a single data set in real time. That way, you can see the big picture from all your data with very little effort on your part.  We also map all of that data to a deep taxonomy so that you can drill through tiers to get better insights on user interest and make smarter predictions about user intent. Our taxonomy has been created and curated by real people who understand the difference between a reference to jaguars at the zoo and Jaguars on the car lot. We deliver 21 tiers of categories, with 450,000 nodes to help you deliver the kind of personalization experience that keeps visitors engaged. Whether you’re a social site or a publisher looking to deliver better content and more relevant ads, or you’re researcher trying to deliver more relevance, we can help you pull and use data.

Why do we do it? We want our social feed to be more relevant, too.

by

If you sit on huge amounts of data and text, a text classification for publishers can provide new ways to make more accurate suggestions for readers and better pairings of content with promotions and ads. Text classification can also help publishers repurpose and repackage content so that they’re always getting as much as possible out of their most valuable asset.

Being able to better classify content may also be the key to winning the content game in the long term.  

A tight marketplace for publishers

The Pew Research Center’s State of the News Media 2015 report shows that only five companies generated 61 percent of total domestic digital ad revenue in 2014. Google, Facebook, Microsoft, Yahoo and AOL took the lion’s share of revenue. The report also showed that revenues and spending are up for publishers, and indicated that the market is getting more crowded and competitive each year.

Publishers who want a slice of the remaining pie need to be smart and aggressive. They must find ways to deliver quality content in ever more engaging ways while pairing content with the right kind of advertising for a better all-around experience.

Publishers have the text and data they need to analyze and optimize their efforts, but it can be a challenge to pull all that data together to make sense of it and really use it to its highest advantage.

A vertical classification system for publishers

There are a number of products and services available to publishers to help them organize their content or pull insights from their social channels. As good as these insights can be, the publisher is still left with the same problem: Data and text in a wide range of formats interspersed throughout editorial, marketing, and business systems with no way to connect them.

eContext makes the connection. We help publishers pull that data through a single filter, analyze the data, and categorize all of it under a unified classification system that goes 1000 times deeper than the industry-standard IAB system to 21 tiers, and in 25 top categories. This can be especially helpful to publishers who cover a wide range of topics across a number of verticals.

With a unified classification system, editors and digital marketers are all viewing the property through a single lens. Editors can make amazingly better on-page recommendations to keep readers on the site, and the digital team can benefit from deeper classification to increase engagement and clicks.

There will always be giants at the table taking big pieces of the pie for themselves, but they had better watch out — publishers with sharp elbows are bellying up to the table to take their fare share. Armed with smarter tools like eContext, they now have a good opportunity.

 

Watch a video to find out more or schedule a demo with us.

by

This post is the fourth in a series on the eContext API and how our clients get value out of semantic classification. To learn more about our technology and its uses, feel free to get in touch.

Way back when we were in R&D here at eContext, you could find, at any given time, up to 200 language and subject-matter experts, working on a project basis, curating the vocabulary rules that became the core of eContext’s classification. When the mass development process was over, most of these guys moved on–some to new employment, others to post-grad academics, etc.

I can’t tell you how many times we’ve heard back from one of these former employees, saying something like: “Man… this place where I am right now? They could definitely use what we were building.”

Our last few posts have really been devoted to consumer-oriented applications of eContext: market research, personalization, improved navigation for publisher and ecommerce properties, etc. And yet, while increased convenience and relevance for consumers should absolutely be emphasized, we can’t overlook the benefits of semantic classification for enterprise-level knowledge management.

Imagine you work for a company with a sub-par system of organizing and retrieving documents (which, incidentally, is probably true for most of you).  You’re tasked with updating all of your organization’s training and development materials, but since those documents aren’t all in one place, you have to do a little digging. To make matters worse, prior documents have used different terminology to refer to the same concepts, meaning you have to perform separate searches for things like “human resource development” and “corporate education”, among many others.

This kind of gophering isn’t just tedious. It’s expensive. According to research conducted by IDC, inefficiencies in working with documents can cost an organization $19,732 per information worker per year and result in a 21.3% decrease in organization’s overall productivity.

No one, neither a consumer nor an employee, should ever have to conduct multiple searches to look up one common topic. It’s outdated; we have the technology. So below are a few basic steps to use eContext for enterprise document management. If you’d like to know more about the nuts and bolts of integrating classification in your organization’s existing architecture, that’s a conversation we’d love to have with you.

Much like in consumer-oriented content delivery, organizing enterprise materials with semantic classification really involves two separate phases: labeling the documents themselves, then aligning those labels with user navigation methods.

Step One: Classify Content with the eContext API

Broadly speaking, there are a few API calls you’ll want to use here, depending on the type of content to be labeled:

  • Classify/html – for HTML content
  • Classify/social – for most user-generated content, including long-form documents, archived enterprise messaging, emails, etc.
  • Classify/url – for any digital document that exists on a discoverable web page (this blog post, for example)
  • Classify/keywords – for extremely short text strings, when you want to limit classification to one best-match topic
  • Classify/text – for any general text data not satisfied by the above

Each API call (except for classify/keywords) will generate a list of labels describing the topics of the classified document; these labels can then be appended as part of the document’s metadata.

Step Two: Classify Searches for Apples-to-Apples Retrieval

eContext can be used to classify user searches in real-time, using the classify/keywords API call. For example, the search strings “leadership development” and “leadership instruction” would both be classified to the eContext topic, “Leadership Training”. Remember: eContext’s topics are organized hierarchically, so depending on how you opt to integrate topic classification into your existing infrastructure, you could retrieve content that matches your topic exactly, or include content that matches any of your topic’s subtypes.

Another feature of classify/keywords is the ability to constrain classification to one or more specified verticals. This is particularly helpful in select knowledge management applications where a given subject matter can be assumed. For example, in an organization that deals entirely in the buying and selling of automotive parts and accessories, searches can be automatically be classified to the most relevant automotive category.

While plenty of digital innovation is geared towards consumers, providing more convenience in exchange for increased sales and satisfaction, the technology of the office can often get overlooked, especially in small and medium sized businesses. Organizing your documents by topic–a portable, intuitive, and evergreen system–means employees can save their sanity and companies can save their cash.

by

This post is the third in a series on the eContext API and how our clients get value out of semantic classification. To learn more about our technology and its uses, feel free to get in touch.

In the previous installment of “The Adventures of the eContext API User” (working title), I touched on different ways businesses can better understand their audiences’ interests—self-disclosed through billions of digital interactions—and discussed a few ways to profit from that insight.

I have to admit, that post was a little bit guilty of a flaw that’s typical in conversations about data: the straightforward argument that “more data = more relevance = happier users.” While this is generally true, especially when catering to entire audiences, things get a little more complicated when you talk about individuals. Today, we’re going to talk about single user profiles—not just how to divine a consumer’s interests, but how to leverage that information with care.

For the sake of clarity, let’s discuss this issue using a typical eContext use case:

  • The client is media group with several digital magazines, mainly offering articles and video content on a wide variety of different topics
  • The client’s primary source of revenue comes from selling ad impressions
  • Client’s goal is to facilitate better individual personalization—tailoring recommended content to a user’s demonstrated interests—in order to cross-promote sibling properties and keep the user on client-owned websites for longer browsing sessions

Step One: Classify Existing Content

If you’ve seen any of our previous posts, you’ll probably be familiar with this part: our client will pass their content through the eContext API and receive precise, structured data on the topics covered by that content.

Here are the different API calls the client would use for different kinds of content:

  • The URL of every client-owned webpage will be submitted via Classify/url, giving the client a list of the most prevalent topics on each page.
  • The client’s properties feature comment sections at the end every article; this user-generated content is submitted via Classify/social.
  • For video content, the client has a few different options:
    • Relying on Classify/url to classify the elements surrounding the video: the title, description, etc.
    • If the client has a reliable transcript of a video, that transcript can be passed through either Classify/social or Classify/text.
    • If the client can’t get a reliable transcript AND classifying the video-adjacent elements is insufficient, eContext can still help. Because we operate our own transcription models, we can process and classify video content in-house. Note that this is not an API call, but an agreed-upon project to be carried out by eContext.

The client will then metatag every page, article, video, and message board post with eContext’s topic classifications. Remember: our topics are nested within a giant subject matter hierarchy. That means an article comparing the best brands of hair spray won’t just be labeled with Hair Spray, but also Beauty, Hair Care, Hair Care Products, and Styling Products. This hierarchical structure is going to come in handy, as I’ll discuss later in this post.

Step Two: Classify New Content, Search Queries

Obviously, the client can’t just call it a day after labeling all of its existing content. New material is being published daily, so the client sets up real-time access to Classify/url and Classify/social in order to stay up-to-date.

In addition, the client will use Classify/keywords to label the topic of each search conducted across its properties. Remember, the client is using these classifications to both learn about users and deliver content that matches their interests, so keeping track of what a user is searching for will be a key source of information (especially if the user is searching for a topic that client doesn’t have content about!)

Step Three: Building Profiles

Most sophisticated digital properties have a method of keeping track of users, whether through cookies, an IP address, a site-specific user account, or social media authentication. The important thing is that the client can keep track of all the content consumed by a user over time, any searches he or she makes, and the messages he or she posts to the website’s comment sections.

Because all of those artifacts have now been classified to topics thanks to eContext, the user’s activity yields an active, real-time profile that communicates the user’s interests. Here’s a basic visual representation of what such a profile might look like:

Now, there are a lot of different ways to rank what a user is “most” interested in. How to weigh the topics of a user’s searches against the topic he or she ends up reading about? Should rankings decay over time to reflect a user’s developing interests? If a user spends an above-average time on a particular article, should that article’s topics be prioritized? These are questions best answered on a case-by-case basis per the goals of an individual client. However, feel free to contact us if you ever want to go over a few recommended practices.

Before Step Four, Let’s Talk about Relevance

I’m a pretty big nerd, no doubt about it. If I’m on Buzzfeed and I see a link on the homepage to “The Best Easter Eggs in ‘Stranger Things’ ”, there’s a pretty good chance I’ll click.

On the other hand, if I see “The Best Easter Eggs in ‘An Age of Kings’ ”, I’ll probably be a little creeped out. That’s because An Age of Kings isn’t some hot new Netflix show; it’s a BBC Shakespeare adaptation from 1960 that’s only available on DVD, and the only reason I would see a link like that is if the website somehow knew that I’d recently purchased it.

All of the different little personalizations in our digital lives—suggested purchases, your home marked on Google Maps, that video Facebook makes for your friendiversaries—are a reflection of who you are in the real world. As that reflection achieves more and more granularity, it approaches a kind of online uncanny valley; it’s so accurate that it starts to get disconcerting.

Here’s my point. eContext gives clients the opportunity to understand users with unparalleled precision, but the real challenge is finding the right balance between broad and granular personalization.

High-Fidelity Personalization

Pro: Guarantees relevance

I love it when Amazon knows exactly what I’m looking for!

Con: Potentially unsettling

Why am I seeing ads for maternity wear?! We haven’t even told our friends!

Low-Fidelity Personalization

Pro: Yields suggestions that might not be 100 percent relevant, but are close in a way that can broaden existing interests

Spotify’s Discover Weekly playlists are awesome…  I never knew I liked EDM!

Con: Increased risk of unhelpful recommendations

I bought some measuring cups last week, so now the internet thinks I’m a bakery.

With this delicate give and take and mind, we’ll now consider how our client can leverage the user profiles they’ve obtained thanks to eContext.

Step Four: Delivery, Feedback, and Calibration

Across its digital magazines, the client can now take stock of a visitor’s interests and offer up content to match. More than that though, thanks to the hierarchical nature of eContext topics, the client is free to calibrate exactly how precisely to offer up those recommendations.

Just because the user has shown an interest in Hair Spray doesn’t mean that the client must restrict recommendations to that topic. That could get old, redundant, or even a little too close for comfort. If Hair Spray suggestions aren’t getting traction, the client may set its recommendation algorithms to consider the broader categories like Hair Care or even Beauty. By using eContext’s broad-to-granular topic structure and consistently calibrating towards the right level of specificity, the client can continually flex its personalization strategy and optimize visitors’ time-on-site.

by

Apple has made their commitment to privacy very clear, and the company maintains a full microsite devoted to explaining its privacy policies.  Now its data vs privacy as Apple is under pressure from the FBI to help unlock the iPhone of the San Bernardino terrorist, with the FBI director making a personal plea to the company for help.

 

The Apple question is surely fueling discussions this week at Mobile World Congress. Taking place in Barcelona, the show is the hot spot for mobile phone manufacturers and the technologies that support the industry. One of the show sessions, Putting Privacy at the Core of Digital, features leading experts from AVG Technologies, Facebook,  Meeco, and Telfonica weighing in on the push-and-pull issues that are unique to the mobile industry. The use of customer data generated from devices, the balance of power between privacy controls offered by the provider and the true privacy of users, and the ever-present issue of third-party data. To keep services affordable for the consumer, manufacturers and developers are under great pressure to access more and more private data to sell to advertisers.

 

Data anonymization

 

Mobile phone manufacturers aren’t blind to the need for better privacy standards, so they will often implement data anonymization, a process that strips data of identifiable information so that, for instance, if the data is sold to a third-party for use in marketing, individual people could not be identified in the data set.

 

However, as reported by Wired in a story titled, Anonymized Phone Location Data Not So Anonymous, Researchers Find, researchers at MIT and the Universite Catholique de Louvain in Belgium who looked at 15 months’ worth of anonymized mobile phone data found that “it took very few pieces of data to uniquely identify 95 percent of the users” and trace activity to a specific individual.

 

As consumers gain more and more awareness of just how much of their personal data is available through their phone, pressure will mount for companies to show, as Apple does, that they have a strong commitment to protecting privacy.

 

Data and text classification: insights that respect privacy

 

There’s an easier way for many companies to get the answers they want about a certain population, and these answers can be attained without violating privacy.

 

If you have access to user generated content–for example, social posts, blogs, messages, or search queries–a comprehensive topic classifier can help you uncover and aggregate valuable information about what’s being said, while ignoring the details that betray the identity of the content’s author. What’s more valuable: learning that Danny in Spokane is thinking of getting a hoverboard, or that hoverboard enthusiasts show a distinctive interest in Samsung products and the website Mashable? As you look for the patterns of what a population or an audience is talking about, you can better understand what their passions are, and make informed decisions about how to market to them in the future – without having to scrape their profiles, track them with cookies, or follow their every move online.

 

eContext provides a complex and very useful framework for this kind of audience analysis. While some social listening tools can deliver similar insights, they are typically narrowly focused, rely heavily on one-by-one keyword contain searches, and barely scratch the surface with two or three tiers of classification offered for data. Consider Hillary Clinton’s media buying team. Right now, they’re trying to put together this summer’s ad budget, and mobile ads will be huge for them. They might want to find out what kind of sports Hillary Clinton’s followers are into so they can target ads at popular events during the summer. If the team uses a social listening tool to query a stream of social data, they will get some broad answers like “soccer” and “baseball”.

 

With eContext the team can pull that same data stream and combine it with any other available data they may have so that there is a larger corpus to query. We’re guessing Hillary’s team has a lot of data. eContext pulls all of their data, from any source, through its filter in real-time, and the result is deep classification of that data into a taxonomy that delves 21 tiers deep and delivers up to 450,000 categories. Now, Hillary’s team has some specific answers — They’ll get information on which football and baseball teams that Hillary Clinton supporters like best, so they can plan mobile campaigns more accurately, and share their findings with ground teams so that Hillary can show up at a few baseball games this summer where her supporters are. Knowing where your audience will be helps you know where the ads — and in this case, the candidate — should be.

 

And Hillary can do it without irking her audience by compromising their privacy. After all, nobody knows better than Hillary the sting you feel when perfect strangers go through your email for personal gain.

Ready to learn more?

Schedule a demo

Chicago

116 West Hubbard Street
Suite 201
Chicago, Illinois 60654
United States
+1-312-477-7300

London

167-169 Great Portland Street
Floor 5
London
W1W 5PF
United Kingdom
+44 (0)20 7834 5000

INQUIRIES

[email protected]

Ready to learn more?

schedule a demo