Googler Norvig Provides Some Clues about Smart Software
March 8, 2015
If you are into the Google, you will want to view the 30 minute talk by Dr. Peter Norvig. He is a heavy hitter in smart software.
Not art. A visualization. Source: Peter Norvig, Google, 2015.
To get the program navigate to this link. Don’t bother to register for the slides, you have to provide a “real” name and then the link generates a blank page. Nifty.
The key point for me was, and I am paraphrasing:
Yep, Google can learn from examples. This is good for Google and a validation of Autonomy’s training methods.
For more insights about program optimization, Google’s approach to self driving autos, and advertising, view the video. Norvig’s operative word is “suggestive.” Wolfram Alpha is, as I understand the video, handcrafted and not Googley.
Stephen E Arnold, March 8, 2015
Estimating Enterprise Search Returns
March 8, 2015
I read an interesting statement about financial forecasts in “Investigator Was Told Women Made More Money for Kleiner.” Keep in mind I am not sure about the accuracy of information disseminated in the course of legal proceedings. One has to be there and be part of a legal team involved in the matter to get a sense of the “truth.”
Attorney Stephen Hirschfeld, hired by Kleiner, Perkins, Caufield, and Byers to probe allegations of discrimination against the firm, testified in a San Francisco courtroom today that then-partner Trae Vassallo had run the numbers on both male and female partners. According to her calculations, the women came out ahead of the men in multiple measures of profitability. Even so, the men forecasted higher revenues for their investments than the women, Hirschfeld said Vassallo told him.
I have highlighted the assertion/factoid that caught my attention. It would seem that the wild and crazy estimates about the revenue from search and content processing investments cut across gender. For example, there are the folks at Attivio, BA Insight, and Coveo who have to generate sustainable revenue and healthy profits for their stakeholders. But these outfits’ estimates pale in comparison with the billions IBM Watson will spin in the time it takes to get a stack of pancakes at IHOP. Isn’t IBM’s chief estimator a female?
Stephen E Arnold, March 8, 2015
Aberdeen Consulting Labors to Pump up the Watson Balloon
March 7, 2015
I read “IBM Watson and Answering the Questions of the World with Cognitive Computing.” Darned amazing mid tier consulting firm dream spinning it is. Here’s the paragraph I noted, which is a quote from an IBM Watson guru named Rob High, the chief technical officer for Watson:
…We’re going to see this cognitive computing capability be brought down deeper into things that we do on a daily basis. I think we’re going to find this to be the dominant form of computing in the future, especially given that to personalize all of those things is not something you can conceivably do if you had to program all the logic around that for each individual person. These systems are only going to be able to achieve that kind of personalized value if they’re able to learn, learn about you, learn about your way of interpreting the world and the way that you envision the world and the priorities that are important to you, that perhaps you find useful in how you conduct your life. That’s why I think that’s the role that cognitive computing is going to have for us, is to provide that degree of personalization.”
I can hear personalized trumpet fanfares…almost, maybe. I think I hear the heavy breathing of the number one trumpeter.
Now where’s the real sound, the sound of cash registers ringing as companies spend for Watson’s wizardry?
Stephen E Arnold, March 7, 2015
Funnelback: Getting Quieter and Quieter
March 7, 2015
Prior to some management changes, Funnelback was on my radar. The company seemed active in the UK and Australia. For 2015, the company has commenced operations in the US. The firm’s office is in Santa Monica. There is a Web site refreshing at www.funnelback.com. The last big news items I noticed was that Funnelback became a Crown Commercial Service Supplier on G-Cloud 6. Was Funnelback a bit of a one man band. When the drummer departed, the search focus softened. Worth watching.
Stephen E Arnold, March 7, 2015
SharePoint and Search: Questions Arise
March 6, 2015
SharePoint search has delivered the best of times for consultants who get paid to make the system work. For users, SharePoint has been a contributor to bad findability times.
I read “Excuse Me SharePoint: A Crossroads or an On-Ramp?” Let me cut to the main point: No one knows. I know that I don’t want to be road kill in the busy intersection of high expectations and massive cost overruns.
I have an opinion. But first, let me call your attention to this statement from the write up:
They [a cadre of SharePoint “experts] acknowledged enterprise users’ frustrations, which Holme called more of a communication problem than an IT problem. In the past, Microsoft was way behind the industry in implementing new features and has gone to implementing them so rapidly that an item a company demoed yesterday might be gone today. The focus tends to be on the end user, which isn’t always the most useful for an enterprise. And in 2015, a lot of organizations are still trying to figure out SharePoint 2013.
For me, SharePoint is an opportunity to make money. Customers drink the Microsoft Seattle latte and believe three things:
- SharePoint is the operating system for the organization. Hey, everyone uses Word. SharePoint is just like that.,
- SharePoint does many things really, really well: Ease of use, document management, search, collaboration, etc.
- SharePoint search is the state of the art in finding concepts, people, facts, you name it.
The reality is that SharePoint does many things, but none of them is exactly what the customer believes. Most of the functions can be made to work with sufficient money, expertise, time, and management patience.
The problem is that consultants want to sell their SharePoint expertise. Those engineers with hard won SharePoint expertise, like and Oracle database administrator, have little incentive to explain certain aspects of the SharePoint decision. Users are clueless and senior managers pre-occupied with sales, litigation, their compensation package, and personnel issues.
Getting the truth about SharePoint costs, complexities, weaknesses is difficult. When it comes to search, the number of third party alternatives makes one thing clear—SharePoint search is not as good as third party solutions.
So what? Well, you get to spend more money for a utility that should work. That’s good for the third party vendors. For others? Well, like the future of SharePoint, no one knows or no one is saying.
Stephen E Arnold, March 6, 2015
Explaining Markov Chains
March 6, 2015
Do you know what a Markov chain is? If not read about “Markov Chains” on the Circuits of Imagination blog:
“A Markov chain is a set of transitions from one state to the next; Such that the transition from the current state to the next depends only on the current state, the previous and future states do not effect the probability of the transition. A transitions independence from future and past sates is called the Markov property.”
This boils down to Markov chains are a way to explain patterns that happen over time and were once used to document human behavior. The chains are not the best way to model human behavior, because they only exist in the present. They do not take into account past or future experiences, otherwise called “memoryless.” The chains can only rely on the action that previously occurred
Markov chains are useful to identify abnormal behavior in systems that don’t exhibit the Markov Property. How? If the system keeps making the wrong decisions based of its program, then it can be diagnosed and repaired. The post explains how the Markov chains are used in coding and provides an example to illustrate how developers can recognize them.
Whitney Grace, March 06, 2015
Sponsored by ArnoldIT.com, developer of Augmentext
A Message from the COO Connotate
March 6, 2015
Tom Williams is the chief operating officer of Connotate, the industry leading platform for harvesting and monetizing content. Like most organizations, Connotate is looking to attract attention and connect with people on a personal level by using social media. Videos have rapidly become the favored medium, because they are easily digestible.
On Connotate’s Youtube channel, there are two videos why content is important and how Connotate can help people use it to their advantage. The first video “Tom Williams, COO of Connotate: Why They Must Evolve” explains that old fashioned information methods no longer work in today’s world:
“Information service providers find themselves ill-equipped to deal with the challenges of incorporating web data. As you look at the scale from tens to hundreds to thousands, the old stare and compare, handwritten scripts, and off the shelf tools no longer apply. These techniques have a low barrier entry, but they don’t scale.
The next video “Tom Williams, COO of Connotate: Why Consider Connotate” explains how Connotate can help organizations harness their content and make money from it. These are powerful brief messages that get straight to the point of how Connotate can help.
Whitney Grace, March 06, 2015
Sponsored by ArnoldIT.com, developer of Augmentext
Enterprise Search: Roasting Chestnuts in the Cloud
March 6, 2015
I read “Seeking Relevancy for Enterprise Search.” I enjoy articles about “relevance.” The word is ambiguous and must be disambiguated. Yep, that’s one of those functions that search vendors love to talk about and rarely deliver.
The point of the write up is that enterprise content should reside in the cloud. The search system can then process the information, build an index, and deliver a service that allows a single search to output a mix of hits.
Sounds good.
My concern is that I am not sure that’s what users want. The reason for my skepticism is that the shift to the cloud does not fix the broken parts of information retrieval. The user, probably an employee or consultant authorized to access the search system, has to guess which keywords unlock the information in the index.
Search vendors continue to roast the chestnuts of results lists, keyword search, and work arounds for performance bottlenecks. The time is right to move from selling chestnuts to those eager to recapture a childhood flavor and move to a more efficient information delivery system. Image source: http://www.mybalkan.org/weather.html
That’s sort of a problem for many searchers today. In many organizations, users express frustration with search because multiple queries are needed to find information that seems relevant. Then the mind numbing, time consuming drudgery begins. The employee opens a hit, scans the document, copies the relevant bit if it is noted in the first place, and pastes the item into a Word file or a OneNote type app and repeats the process. Most users look at the first page of results, pick the most likely suspect, and use that information.
No, you say.
I suggest you conduct the type of research my team and I have been doing for many years. Expert searchers are a rare species. Today’s employees perceive themselves as really busy, able to make decisions with “on hand” information, and believe themselves to be super smart. Armed with this orientation, whatever these folks do is, by definition, pretty darned good.
It is not. Just don’t try telling a 28 year old that she is not a good searcher and is making decisions without checking facts and assessing the data indexed by a system.
What’s the alternative?
My most recent research points to a relatively new branch or tendril of information access. I use the term “cyberosint” to embrace systems that automatically collect, analyze, and output information to users. Originally these systems focused on public content like Facebook, Twitter posts, and Web content. Now the systems are moving inside the firewall.
The result is that the employee interacts with reports generated with information presented in the form of answers, a map with dynamic data showing where certain events are now taking place, and in streams of data that go into other systems such as a programmatic trading application on Wall Street.
Yes, keyword search is available to these systems which can be deployed on premises, in the cloud, or in a hybrid deployment. The main point is that the chokehold of keyword search is broken using smart software, automatic personalization, and reports.
Keyword search is not an enterprise application. Vendors describe the utility function as the ringmaster of the content circus. Traditional enterprise search is like a flimsy card table upon which has been stacked a rich banquet of features and functions.
The card table cannot support the load. The next generation information access systems, while not perfect, represent a significant shift in information access. To learn more, check out my new study, CyberOSINT.
Roasting chestnuts in the cloud delivers the same traditional chestnut. That’s the problem. Users want more. Maybe a free range, organic gourmet burger?
Stephen E Arnold, March 6, 2015
AlchemyAPI: Beefing Up Watson
March 5, 2015
Why does Watson need beefing up? I have been inundated with information about Watson the game show winner, Watson the recipe maker, Watson the cancer fighter, and Watson the developer’s Eagle Scout.
IBM acquired a company involved in smart software and predictive analytics. That’s great for the owners of AlchemyAPI. I just am curious why the analytics tools IBM has developed itself, the SPSS toolset, the analytic components like WebFountain on the shelf in an IBM office somewhere are not enough.
At any rate, the news is presented in “IBM Buys AlchemyAPI to Boost Watson Computing Unit.” The write up reports in the best spirit of recycling IBM PR:
The purchase is designed to boost IBM’s push into more human-like computing services, based around its Watson technology, which can sift huge amounts of data, learn from the results and respond to spoken questions.
I quite like the phrase “IBM is trying to build a big business around Watson.”
No kidding. What does desperation smell like? The odor of cold muffins and warm laptops in a Manhattan office?
When it comes to delivering an integrated suite of service based on predictive analytics and other next generation goodies, I am not sure just buying stuff is the optimal approach. In my opinion, IBM seems to be struggling with the whole Watson offering. Executives unable to land deals with the velocity of Recorded Future, RedOwl, and some other hot outfits may believe that an acquisition is just what Dr. Watson needs.
We’ll see if a startup can power the new smart software economy which Google also covets and Hewlett Packard is chasing with the Autonomy black box of goodies.
Stephen E Arnold, March 5, 2015
MovieGraph from Senzari Offers a Better Way to Find Movies
March 5, 2015
The article on Dataversity titled Creating Detailed Semantic Graphs Around Video Content with MovieGraph suggests a possible breakthrough in video sense making. MovieGraph is the platform of entertainment data company Senzari. Chief Operating Officer Demian M. Bellumio spoke to the methods utilized by MovieGraph, which include machine learning and an API for recommendations. The article continues to refer to Bellumio’s statements,
“Senzari focused on metadata while building MovieGraph. He also said that Senzari trained machine learning algorithms to break down the narratives of movies, extracting the data with precision across each element. The company designed their own matrix for cataloging movies; MovieGraph uses machine learning techniques to semantically tag and organize every movie and TV show across hundreds of dimensions. Senzari also added proprietary narrative features to MovieGraph such as setting, conflict, symbols or tones present in a film.”
The possibilities for recommendations seem much more targeted than the Netflix model, which often makes suggestion based on categories that are too wide and abstracted to be accurate. The article mentions that since Netflix only recently closed its public API, MovieGraph may be in a position to fill that gap. MusicGraph is also built to work with MusicGraph, another Senzari platform. Content creators in particular might find the crossover to be useful in terms of finding appropriate content for their projects.
Chelsea Kerwin, March 05, 2015
Sponsored by ArnoldIT.com, developer of Augmentext
-
- Subscribe to Beyond Search
Feature archive
News archive
-
