A 23% compounded average annual rate of return.  Not too bad.

During his career Shelby advised Thomas Dewey on economic matters when he ran for president in 1940 and 1944 and served as Governor Dewey’s First Deputy Superintendent of Insurance from 1944 to 1947 in New York. Later, he served as US ambassador to Switzerland under Presidents Nixon and Ford (1969 – 1975).

A philanthropist, he was head of the Shelby Cullom Davis foundation which disperses funds to higher education and to research on public policy and economics. Princeton, his alma mater, was one of the beneficiaries of his grants. He also served as chairman of the Heritage Foundation.

Shelby got around.

But what the heck does this have to do with Fintech?

In a previous post we found that the Tucson, AZ metro area made our top 15 list of top addressable Fintech Securities & Investments markets on a per employee basis in 2015 (ranked 14^th).

Can a contributing reason for this can be tied directly back to Shelby Cullom Davis?

Securities & investments sector in Tucson, AZ

Tucson, AZ probably does not first come to mind when thinking about hot spots of financial activity. So, what is going on in Tucson?

As we have done in previous posts, let’s start with a high-level view of Tucson’s Securities & Investments sector. As shown below, 62% of the 2015 addressable Fintech market is due to portfolio management. Another 27% is due to securities brokerage. This is generally consistent with the West regional average of 28% for securities brokerage we saw in an earlier post. But it reflects a much higher concentration of potential Fintech spend in the portfolio management sub-sector (62 vs 48%).

Again, our analysis uses ZIP Pointe Market Sizer.  Market Sizer is our Tableau-based market sizing tool based on ZIP Code-level Census data on over 7 million private sector business locations.

In terms of business location size by sub-sector, nearly 80% of the locations are small (1 – 4 payroll employees). This is a recurring theme. The sub-sector is dominated by small locations though these are not necessarily all stand-alone companies (i.e. they could be formal branch locations).

There are only 29 locations with 10 or more payroll employees. A bit over 40% of the locations are involved with securities brokerage, another 26% are portfolio management and another 18% are investment advice (see below).

With respect to “hot spots” of potential Fintech spend per employee, it is clearly the portfolio management sub-sector that stands out, with potential spend ranging from $5,337 to $7,682 per payroll employee. The top hot spot consists of 3 locations with 50 to 99 employees.

Note: The US Census source used by Market Sizer tracks only payroll employees. Independent contractors are not reflected in these numbers. Hence, the actual employment level can be higher, possibly much higher, in sectors that make extensive use of contract labor.

ZIP Code level view

As we have done in prior posts, another way to drill into the addressable Fintech market in Tucson is by ZIP code. The following map shows the ZIP codes in the metro area color coded for addressable market size per employee. At the ZIP Code level, addressable Fintech market size ranges from $1k to $6k per payroll employee.

But there is a distinct hot spot in ZIP Code 85706 (circled in red) at $6k per employee. According to Wealthminder.com, the largest Tucson financial advisory firm (in terms of assets under management) is located in this ZIP Code:  Davis Selected Advisors.

Other top firms in the Tucson metro area (indicated by red arrows) are Tci Wealth Advisors in ZIP Code 85718; Sonora Investment Management in Zip Code 85719; Strategic Equity Management in Zip Code 85715; and Sterling Investment Management in Zip Code 85718.

Which ZIP codes account for the bulk of potential Fintech spend?

We have already discussed two of the top 3: 85706 at 16% and 85718 at 20%. But these are surpassed by ZIP Code 85712 (indicated by the green arrow) which accounts for 27%. These three ZIP Codes together account for 63% of Tucson’s potential Fintech spend in the Securities & Investments sub-sector.

Another way to drill into the addressable Fintech market is on a per business location basis.  Dividing potential spend by the number of locations is another way to normalize the data. This map (shown below) looks quite a bit different and one ZIP Code stands out above all the rest: 85706. Although the ZIP code overall accounts for 16% of total potential Fintech spend, on a per location basis, it ranks at the top with $230k in potential spend per site.

ZIP Code 85706

Turns out at there are only 3 locations in ZIP Code 85706.  At least one of these (the larger one) is associated with, yep, Davis Selected Advisors. And it may be that the other sites are also associated with Davis as we could not find any other Securities & Investments firms (with at least 10 employees) in this ZIP Code (If anyone knows, please let us know!)

Davis Selected Advisors

So, the Davis name is a major player in the Tucson metro Securities & Investments sector. Any relation to Shelby Cullom Davis perhaps?

Davis Selected Advisers was started in 1969 in New York by Shelby Davis…not Shelby Cullom Davis but his son Shelby Moore Cullom Davis. Following his father’s footsteps, Shelby M.C. Davis graduated from Princeton and rose through the ranks of the New York financial industry – specifically at the Bank of New York.

In the late 1970s the non-portfolio operations of Davis Selected were moved to Sante Fe, N.M. In 1997, a service center was opened in Tucson and personnel were hired locally and moved from Santa Fe. By 2000, Tucson had became Davis Selected’s global headquarters, specifically in ZIP Code 85706 where the company operates out of its own commercial property.

Today, Davis Selected Advisers is led by Shelby’s son, Chris Shelby, grandson of Shelby Cullom Davis.  “Davis Selected controls the nine-fund Davis group, the Selected group of funds, which is a no-load, four-fund group, private accounts of no less than $5 million, and other managed money programs.” (Inside Tucson Business).  Currently, Davis Selected has about $30b in assets under management.

So, there you have it…from Switzerland to the Sonoran Desert! 

Ok, maybe New York to Arizona…

The post Top 15 Metros – From Switzerland to the Sonoran Desert – Fintech Securities & Investments appeared first on KDD Analytics.

Is tone really that important? And if so, can it be measured and scaled?

Why might tone be important?

Analysts and investors are seeking an edge as to where a company is headed next. So, they focus not only on the numbers, but on the words the executives use in their introductory remarks and in the answers to analysts’ questions. Executives can unintentionally or even intentionally “tip their hand” and provide a clue as to the company’s direction.

But aside from making good CNBC commentary, is there a connection between tone and future company performance? Is it possible to detect tone without listening to the call?

Financial textual analysis has matured

Before the advent of natural language programming (NLP), text analytics and enhanced computing power, written transcripts and audio recordings were all analysts had to go on. There is only so much that the human mind can discern. Subtle shifts in tone and word usage can go unnoticed when listening and tone seems lost when reading transcripts of calls you didn’t have time to cover.

But now, new tools have emerged that can help. Going beyond simple sentiment, nuances in tone emerge when analyzed with advanced computational linguistics. A new academic field of inquiry developed, financial textual analysis, out of this ability to decompose earnings transcripts and financial documents into their most granular bits. Academic researchers were (and are) keen to determine if tone can presage future company performance.

Although academic researchers have always been interested in studying financial disclosures, it wasn’t until the early 2000’s that interest in financial textual analysis really took off. The bibliography of a survey of financial textual analysis studies shows the exponential growth in the number of studies starting in the early 2000’s. The authors attribute this growth to the application of “highly evolved” computational linguistics technologies to financial documents and enhanced computing power.

So, does tone matter? Simple answer is YES.

What do these academic studies conclude about tone? Can executives’ tone on earnings calls “tell” analysts something about future company performance?

It is beyond the scope of this post to survey the entire field of financial textual analysis. But we can say the answer is “yes,” tone matters. Here is a taste of what the academic literature has found:

We note that these studies tend to be sophisticated statistical studies that attempt to estimate the relationship between tone and company or market performance after controlling for other factors that may affect such performance (i.e. they tend to be more than simple correlation studies).

A study by Price, Doran, Peterson and Bliss finds that “conference call linguistic tone is a significant predictor of abnormal returns and trading volume.” The researchers further find that the Q&A portion of calls has an incrementally greater effect than the introductory portion, which is typically a duplication of the earnings press release. The study concludes that “conference call tone dominates earnings surprises over the sixty trading days following the call.”

Another study by Price with Blau and DeLisle finds a positive relation between companies’ stock returns and earning call tone. But the researchers go on to argue that tone can be subtle and that more sophisticated investors (e.g. short sellers) process the information differently than the average investor. Empirically they find that “short sellers target firms with simultaneous high earnings surprise and abnormally high management tone.”

Druz, Petzev, Wagner and Zeckhauser looked at changes in earnings call tone from quarter to quarter finding that higher negativity of executives’ tone “strongly predicts lower future earnings and greater uncertainty.” Interestingly, the researchers find that decreased negativity only “weakly predicts the opposite.”

Finally, Chen, Demers, and Lev’s research showed that the overall negative tone of earnings call Q&A increased as the morning progressed, dipped slightly after lunch and then rose again until the market close. They also found that the more negative the tone, the more negative the stock returns over the five trading hours after the call. Those negative stock returns associated with negative calls continued downward for up to 15 trading days.

So, it’s clear that “tone matters” and computational linguistics can detect it. Analysts, investors and the market react to the choice of words used by executives in their earnings calls. Investors need to factor this information into their trading strategies, but with care to consider external factors as well.

Boulder Earnings Call Tracker

KDD Analytics is partnering with Boulder Equity Analytics in the development of the Boulder Earnings Call Tracker. This Tableau-based set of dashboards examines earnings call tone among other attributes of earnings calls. It provides our clients with highly relevant signals for sophisticated trading strategies, especially short sellers. On the flip side it provides CEOs and investor relations with unbiased metrics to improve their conference call performance.

The post Negative Earnings Call Tone? Go Short (but not always) appeared first on KDD Analytics.

The disruption is well underway and big money is flowing into the space. CB Insights recently released their quarterly tracking of this segment, Trends in Fintech: Q2 2017.  VC-backed investment in global Fintech companies was $13.5b in 2016.  Up from $2.6b in 2012, a CAGR of 51%.  In the US, VC-backed investment was $5.7b in 2016.  Up an average 33% per year (CAGR) from 2012’s $1.8b.

While these are hefty investment numbers, they beg the question, how large is the addressable market for Fintech?  Public estimates of the “market size” are surprising scarce and difficult to reconcile.  One reason may be that the definition of what constitutes “fintech” has been a moving target.  With new platforms and software services evolving over time, the amount of potential revenue Fintech vendors are chasing has been changing as well.

Fintech market sizing

As a first step, we take a simple macro approach to sizing the market.  Starting with US census data on total revenue earned by two key sectors targeted by Fintech vendors – banking and securities investment – we estimate and decompose the market geographically.  We accomplish this using ZIP Pointe Market Sizer, a Tableau-based market sizing tool based on ZIP Code-level Census data on over 7 million private sector business locations in the US.

US financial services sector

The US Census (NAICS) data defines the financial services sector as consisting of 3 primary sub-sectors: Banking (522000), Securities and Investment (523000) and Insurance (524000).  Since our interest (KDD Analytics is partnering with Boulder Equity Analytics (BEA) in the development of a Fintech offering) is in the sectors primarily involved with financial reporting, analysis and investor relations, we will focus our analysis on the Banking and Securities sub-sectors. The 2015 US Census data for Banking and Securities recorded total revenue of $2.2t.  So how much of that is Fintech, where are those customers located and how fast is the market growing?

Published estimates of Fintech market

As suggested above, public estimates of the US Fintech market size are scarce and difficult to reconcile.  Working from the research we found, the current market is likely between 0.5% and 1.5% of total financial services revenue as reported by the US Census.  Below are several sources we used for our first pass at the market.

An Ernst and Young report commissioned by HM Treasury (UK) estimated the California and New York markets for Fintech were $7.1b and $8.4b in 2015.  Given that the US Census reported total revenue for NAICS 52 (all US Finance and Insurance) to be $432.5b and $568.1b in 2015, the EY estimates for Fintech represent a share of total revenue of 1.6% and 1.5%.

A 2016 Citibank report came in a bit lower with an estimate that 1% of North American banking revenue has migrated to a digital model.  Citibank projects that by 2020, this share will rise to 10%, then 17% by 2023 (see figure).

At the low end of the range, in their Annual Report, JPMorgan Chase disclosed it spent $600m on Fintech in 2016 out of a $9.5b technology spend, or 6.3%.  Extrapolating to the US banking sector, this implies a Fintech market size of $5.3b or 0.33% of 2015 banking sector revenue.

Additionally, JPMorgan’s Jamie Dimon commented on relationships with several independent Fintech companies highlighting the difficulty in separating its share of traditional banking revenue.  We believe this will become even more challenging in the future as large banks absorb “Fintech” business models and technology into their operations.  As Fintech gets more press, there will be even more marketing pressure to talk up the initiatives while it is unlikely that they will report them as separate line items on financial disclosures.

Fintech market size

Based on our initial survey of the research, we propose to use a revenue weighted average of the two sectors of .86%, 1% of revenue for Banking and .5% of revenue for Securities. We applied this percentage of .86% to the 2015 US Census Total revenue of $2.2t.

Conclusion: our baseline assumption for the Fintech share of the 2015 Banking and Securities market is $18.8b.

We invite the Fintech community to chime in on whether our assumptions make sense.  We will include any updates in future articles as we delve deeper into the Fintech market.

Top Fintech metro areas

Since BEA is a Fintech vendor, we want to know where the $18.8b in customers are located.  It is not surprising that, based solely on absolute numbers, the usual metro areas are at the top of the list:

Rank – Top Ten Metro Areas by Fintech Market Size

1.    New York-Newark-Jersey City, NY-NJ-PA
2.    Dallas-Fort Worth-Arlington, TX
3.    Los Angeles-Long Beach-Anaheim, CA

However, there is a fair amount of geographic variation as shown below:

Moreover, this variation becomes even more interesting when we drill into the data and examine Fintech markets by size of company, by revenue per employee, by growth rate, by region, by financial sector, etc.  Over the course of several articles we will be sharing our findings.

So stay tuned!

This post was written with Tom Marsh, CTO at Boulder Equity Analytics (BEA).

The post Sizing the Fintech addressable market appeared first on KDD Analytics.

Analysts under time and deadline pressure need to know that the information distilled by an AI solution is relevant to their context and is not simply the result of key word searches.  Understanding context is foundational to the collaborative AI solution offered by our partner BEA.

In another guest article, Tom Marsh, CTO at Boulder Equity Analytics (BEA), talks about context and why, without it, computer algorithms will never be able to truly know “you”.

Since Robert Scoble and Shel Israel just released their new book “The Fourth Transformation”, I decided to revisit “Age of Context“, a global survey of the contributions to the forces influencing technology.  The five forces were mobile, social media, data, sensors and location.  Scoble called these the “five forces of context”, the future of computing.  The five forces are still there but hardly tamed and in the rear view mirror.

Revisiting this topic three years later (see my ai-one post), there is still a lot of work to do in mainstream applications of cognitive computing.  For our BEA clients, it remains a critical challenge to building analytics for analysts under time and deadline pressure, faced with exploding amounts of information.

Why is context so important

First, context is fundamental to our ability to understand the text we’re reading and the world we live in.  When reading a sentence, you draw on the semantics of the words, the sentence, the paragraph, the context of the page, chapter, book and prior works or conversations.  Added to your ability to understand the sentence is your education and experience, and your reasons and objectives for reading it.  This diagram from Chris Campion’s blog is instructive in this regard.

Second, if you broaden the challenge it overlaps with personal intelligent agents (Siri, Alexa, Cortana, Google Now), the bigger problem of complexity.  The inability to provide context has always made it difficult for computers and people to understand each other.  Three years ago Scoble felt Google Glass could be that enabling breakthrough and now maybe VR will be the answer.

People and the language used to describe the world is a complex system.  No matter how much data is crunched or how sophisticated the technology (e.g., new generation convolutional neural nets), you can’t be reduced to an algorithm.  Claiming personalization, these tools create approximations of you based on people like you, not a true you.

There’s a difference, especially when applied to automating the decisions of experts in investing, reinsurance contracts or road-maps for new technologies. The intelligent agents that make those decisions need to have a complete model of a complex world.

The five forces of context as the foundation

Pete Mortensen also addressed the problem of context at the same time in his article “The Future of Technology Isn’t Mobile, It’s Contextual.”  Mortensen argues that the five forces are finally giving computers the foundational information needed to understand “your context” and that context is expressed in four graphs.  These data graphs are
 
•    Social (friends, family and colleagues)
•    Interest (likes & purchases)
•    Behavior (what you do & where)
•    Personal (beliefs & values)
 
At BEA, we build the foundations for context by extracting Mortensen’s graphs from the complex data generated by your digital activity, your domain and a complete view of the material you’re researching.  We’ve been using our technology to process and store unstructured and structured text in diverse domains to deliver a representation of that knowledge through powerful visualizations of the models that before now existed only on the page and in your head. 

The five forces of context – and beyond

So first, we get the context right.  Then we add metadata and attributes using NLP, entity extraction, sentiment and other proprietary linguistic algorithms.  This is processed, indexed and stored via python middle-ware that allows us to extract all possible attributes for each paragraph. 

Running in the cloud and integrated with big data sources and ecosystems of existing APIs and applications, we can quickly create and test your investment models or add intelligence to old ones.  Our interactive Tableau visualizations respect the analyst’s expertise, neutralize human bias while acknowledging the limitations of the technology, never delivering “black box” results that aren’t transparent and verifiable.

You work with concepts and use context constantly in your offline life.  We work to bring the same intuitive and human experience to your work as an analyst, without compromising your privacy or creating an approximation of you from the data of others.
 

Tom

@tom_semantic

KDD Analytics and Boulder Equity Analytics are partnering to deliver collaborative artificial intelligence to the financial and competitive analysis industries.

The post Context, Then Concepts, Words Last appeared first on KDD Analytics.

Language and the human brain

How many geese in this picture?  You didn’t need to count. Your answer probably contained the words “approximately”, “average”, “almost”, “sort of”, “guesstimate” or “about”.  One of the most powerful features of your brain is that it does not treat language as math, a series of binary yes or no formal constructs.  Humans are masters of writing the same idea in many ways, understanding what you meant even if you didn’t say it perfectly.  You also know when someone is being so careful with their words, they’re lying (we’re all tested on this one daily). This critical skill is used by analysts all the time.

Many technical approaches to textual analysis try to convert sentences into math or use the presence of specific words to return a yes or no result.  Our approach at BEA overcomes this issue with software agents that compare arrays (models) of patterns and return “similarity scores”.  These normalized scores allow us to determine whether a paragraph is about the same as another and by how much. 

In my last post, I discussed the nature and definition of concepts and how our solution is built to find them. Whether it’s a concept you’ve created or an example you’ve found in a document, if you can’t describe it you can’t find it.

Describing concepts

You can spot a concept when you read one, but learning to describe one isn’t as easy as it sounds and its rarely exact. Our AI platform enables us to create and teach “intelligence agents” to “read” documents to score similarity to concepts.

The first step in the process is to provide examples in a simple text box. This is surprisingly difficult the first time because our brain is never reading text without injecting our own education, bias and assumptions into the process.  Unfortunately the software only sees exactly what you provide it, no more, no less. When one user asked us how to teach the concept of “fraud” to our agents, we had to take a step back.

Risk and opportunity in legal issues

In the financial filings we focus on, concepts are often expressed indirectly, through a description of the act or its consequences.  The objective of the analyst is to find indications of wrongdoing or their cover up in company filings, earnings calls, news and social media. The critical evidence is never a clear statement like “I just gave my friend inside information on our earnings announcement so they could trade ahead of our disclosure.”

People don’t recite the definition of the crime when they talk about it. The language used is always more subtle and disguised. It also varies dramatically from one context to another (email vs. interview transcript for example). Tweets and text messages are full of acronyms, slang, phrases and partial sentences.

The guilty party is usually aware of the act and tries to avoid being discovered. He is more likely to say, “Hi Dave, here are some stats you might find interesting.” Is he talking about the company earnings report or his fantasy football team? If he says, “we’ll have a big surprise for you tomorrow”, is it a surprise birthday party or a merger announcement?

Kant’s tree concept (again)

To underscore this point, let’s revisit the example of the tree concept from my last post. The concept of a tree abstracted from descriptions of many trees is clear enough. The challenge in language analytics and many (most) language processing problems is that we are looking for the indirect effect or consequences resulting from the existence or actions of the tree.

This is easier to understand with examples. “The fall colors in New England are beautiful this time of year”.  Or “We need to get some shade for the yard at the nursery.” The concept of the tree is there but if I was searching for trunks, roots, branches and leaves, the “criminal” tree would escape detection!

Context is key

The context of the concept is critical.  Defining the context for the software can be complex.  But in most cases the patterns are there once you clear away the noise.

Capturing context can be as simple as setting filters.  For example, documents between specific individuals during specific time periods during which they had access to the information, resources and counterparts needed to commit wrongdoing.  Metadata from documents and entity attributes like organization, location and titles are commonly used for this purpose.

Where possible, we extract context from the source documents (and the source) to ensure that the more complex contextual factors are incorporated automatically into the intelligence agent. For example, the same person in the same period does not use the same language on twitter as they do in email.  Our powerful “almost engine” is what makes our system resilient and with our user in charge of how tight to set the “almost” meter, it adapts to a range of problems.

Is “big data” the answer?

We are all guilty of injecting our own bias and filters into understanding language. Good technical solutions capture the richness and subtlety in the context of the communications and ensure consistency of review.  This is not a problem solved by more and bigger data. 

We believe analysts have more than enough information.  They just don’t have enough time to find what matters.  BEA can help with that.

Want to know how we would use “almost” to solve your problems?  Drop us a line.

Tom @tom_semantic

KDD Analytics and Boulder Equity Analytics are partnering to deliver collaborative artificial intelligence to the financial and competitive analysis industries.

The post Good Concept Detection Requires an “Almost Engine” appeared first on KDD Analytics.

Our partner at BEA takes textual analysis a step further and teaches artificial intelligence (AI) software to find concepts and themes, not just words.  It’s one thing to find all occurrences of the word “decline” in an earnings call transcript.  It is another thing altogether to understand the concept of “decline” within the context of a paragraph.  Is it a decline in sales?  or a decline in bad accounts?

In another guest article, Tom Marsh, CTO at Boulder Equity Analytics (BEA), talks about keyword vs. theme detection and why “concepts are key, not words”.

A critical skill for the analyst during earnings season is detecting changes in the key indicators or themes for a company and its peers. Keyword detection is often passed off as theme detection but it’s not and the difference is critical.  Here at BEA, teaching software (AI) to find themes buried in SEC filings, earnings calls and press coverage from investor relations is a critical technology advantage.

First, understand the terms. Analysts tell us that with all the buzzwords and claims by vendors, it’s hard to understand the difference between real and apparent performance.  For us, the goal is to replicate an expert analyst’s ability to read and understand a document, whether its a filing, earnings call or interview.

While there’s more, in this post I want to make sure we understand each other when we use the term “theme”, “topic” or “concept”.

What is a “concept”?

Since we claim to teach software agents to find “concepts”, let’s check the definition of the term “concept” to make sure we are using it correctly. While I didn’t expect this to lead me all the way back to Philosophy class with references to Kant, Locke, Mill etc., our approach and use of this term are fundamentally consistent with the excerpts below from Wikipedia.

Concept– definition from Wikipedia

A concept is a general idea, or something conceived in the mind.

Notable definitions:

John Locke‘s description of a general idea corresponds to a description of a concept. According to Locke, a general idea is created by abstracting, drawing away, or removing the uncommon characteristic or characteristics from several particular ideas. The remaining common characteristic is that which is similar to all of the different individuals.

John Stuart Mill argued that general conceptions are formed through abstraction. A general conception is the common element among the many images of members of a class. “…When we form a set of phenomena into a class, that is, when we compare them with one another to ascertain in what they agree, some general conception is implied in this mental operation” (A System of Logic, Book IV, Ch. II).

Philosopher Arthur Schopenhauer argued that concepts are “mere abstractions from what is known through intuitive perception, and they have arisen from our arbitrarily thinking away or dropping of some qualities and our retention of others.” (Parerga and Paralipomena, Vol. I, “Sketch of a History of the Ideal and the Real”).

By contrast to the above philosophers, Immanuel Kant held that the account of the concept as an abstraction of experience is only partly correct. He called those concepts that result from abstraction “a posteriori concepts”.

A concept is a common feature or characteristic. Kant investigated the way that empirical a posteriori concepts are created.

“The logical acts of the understanding by which concepts are generated as to their form are:

1. comparison, i.e., the likening of mental images to one another in relation to the unity of consciousness;

2. reflection, i.e., the going back over different mental images, how they can be comprehended in one consciousness; and finally

3. abstraction or the segregation of everything else by which the mental images differ …

In order to make our mental images into concepts, one must thus be able to compare, reflect, and abstract, for these three logical operations of the understanding are essential and general conditions of generating any concept whatever. For example, I see a fir, a willow, and a linden. In firstly comparing these objects, I notice that they are different from one another in respect of trunk, branches, leaves, and the like; further, however, I reflect only on what they have in common, the trunk, the branches, the leaves themselves, and abstract from their size, shape, and so forth; thus I gain a concept of a tree.”

— Logic, §6

Optimization of AI software

We worked with our ai-one partner to optimize their AI for this task. 

At the core, our application processes each line of text much the way our brains do it, learning the patterns of language, the “key” words, their importance and the words most closely associated with them. The AI provides commands to extract as an array those key words and associations, their direction and values (strengths).

While the ability to score the similarity of concepts is important, my observation from years of applying it to problems from NASA (Topic Mapping pg.198) to SwissRe is that it’s even more proficient at filtering out the noise, giving lowest values to the unimportant words and associations.

Filtering is fundamental to our brain’s ability to find the topic that’s important.

Locke describes a concept as an idea “created by abstracting, drawing away, or removing the uncommon characteristic or characteristics”. This is very close to the way our solution builds a model of a concept after learning from the examples provided to teach it. The fingerprint we extract from that text is an array that represents the concept in the same way. The similarity score for that comparison is a powerful attribute we use in a number of ways to deliver a great user experience.

Building powerful qualitative analytics for financial analysts and investors starts with the right core technology. Finding concepts buried inside documents is the first part and foundation of extracting actionable insight.

Now that I think about it, maybe Philosophy 101 wasn’t a liberal arts waste of money after all.

Tom

@tom_semantic

KDD Analytics and Boulder Equity Analytics are partnering to deliver collaborative artificial intelligence to the financial and competitive analysis industries.

The post Concepts are Key, Not Words appeared first on KDD Analytics.