For years the field of Intellectual Property (IP) has enlisted brilliant minds seeking to understand, explain, and programmatically evaluate the deep secrets of patents. In spite of all this brain power that has been brought to bear against the problem of programmatically evaluating patents, surprisingly few breakthroughs have surfaced. This article discusses the evolution of automated patent evaluation – from reading patents, to counting claims and words, and finally on to today’s state of the art automated patent evaluation techniques.

Introduction 

In the financial world, patents represent an important data point when considering a company’s worth or valuation. As a result, companies both big and small recognize that the presence or absence of patents play a role in their future. Companies without patents must rely on trade secrets and others barriers in the marketplace to protect their livelihood from competitors. Where as companies with patents build equity that can outlast their products, people, technology, and customer base – essentially even if the company closes its doors the value stored in patents lives on. 

Venture capitalists pay particular attention to IP when considering investments in fledgling companies and larger companies consider IP when making strategic purchases of smaller companies. For example, when a company considers acquiring the assets of a company to kick start its product development within an up and coming marketplace, the due diligence of Intellectual Property is vital to ensuring that the investment is sound. Similarly, Market Makers on Wall Street view a company’s IP as a potential indicator of future growth. For example, a recent article in Red Herring pointed out a particular patent held by Google covering advertising in RSS feeds as potentially valuable. 

While the usefulness of any one patent does not always lead to increased revenue or higher company valuations, it usually does result in the build up of a type of currency. It is this currency that becomes more significant over time in defending your company’s technology edge and market share. It is this wide spread interest in IP from the financial, investment, and business communities that has driven individuals and startups to attempt to crack the patent mystery and devise the means to measure its value. 

History of Patent Evaluation

Patent evaluation has historically been synonymous with good old fashioned reading. If one was interested determining whether a patent was any good, they just obtained a paper copy of it from the patent office and then read it. Only patents are not so easy to just read – they can be anywhere from one page long (for design patents) to hundreds of pages long and if that isn’t enough to provide a sufficient obstacle they additionally assume that the one reading it is skilled in the art. In other words, if the patent is about a broadband network device or a software program that manages a billing system for the wireless industry, it will only make perfect sense to person who really knows (or is skilled in the art of) broadband networking or software development within wireless billing respectively. Since the inception of the patent, reading patents (every word of them) was the state of the art in terms of ascertaining its possible value. However, since reading them also requires a skill all its own, that presents an sizable obstacle to outsiders of a particular industry – such as financial or business people. 

As the number of patents active with the patent office increased it eventually became impossible for people to read all the patents within any particular field. Especially as patents got longer and the task of reading all the patents within a particular area became ever more daunting, interested parties began looking for streamlined ways in finding and evaluating patents. Such streamlining methods included skimming patent text for the essence of the patent by reading perhaps just the title, abstract, background and summary of the invention, and all the claims. As some patents can have 100 or more claims, this skimming increasingly reached for even more abbreviated means of assessing the value or quality of a patent by only reading independent claims. The independent claims represent only the key areas of coverage of a patent. While reading a patent is very slow and can be somewhat confusing, it has never gone out of style (especially when it comes to reading the claims section) and remains the only true way to assess the quality or value of a patent. In fact, the claims area of a patent is so critical that the following statement yields one of very few guarantees when it comes to patents, “If it isn’t in the claims, the patent is lame – Peter Sheedy” 

It wasn’t until patents migrated from primarily being printed on paper to the patent office and some patent search services making full electronic texts of patents available online that something could possibly replace reading as a means of evaluating at least some aspect of a patent. As various sections of the patent became available as electronic text, entrepreneurs and scholars began to look into ways of analyzing the patent text in order to form some kind of automated assessment of its value. Table 1.0 describes a sampling of some of the more popular textual aspects of a patent that can be used to approximate its value without reading it.  

Table 1.0 Automated Textual Patent Assessment Field Options 

The availability of full text patents online, has allowed the fields in Table 1.0 to be compared and contrasted programmatically in many different ways depending on the business objective. For example, if the objective is purely financial (should I or should I not invest in this company) the method of analysis is different than if the objective is to acquire technology or competitive advantage. As such, many different tools or approaches have evolved to uniquely address each objective but no such universal approach or tool has surfaced that can address any objective. 

What is a Valuable Patent? 

Unfortunately, patents cannot be evaluated accurately on an individual basis. Certainly one can examine various well accepted aspects of a patent like citations, longevity, company size, etc. as described previously to derive a value of an individual patent. Evaluating patents in this way can yield a very basic measure of a patent’s value in isolation. For example, consider a patent covering digital rights from Microsoft or IBM or Apple. If one calculates some kind of value for this patent that looks attractive the actual value perceived may not have anything to do with the patent’s innovation quality or technical merit but rather the fact that it is held by a large well known company. The same digital rights patent in the hands of an individual or a small unknown startup will also have value only it just wouldn’t command the same price or interest as if the owner were larger and well known. It is important to understand that the more patents any one firm has the greater the overall power of such resulting Intellectual Property currency. As a result, a number of low to mediocre patents could be just as good (perhaps even better) than if you only had one great (or highly relevant) patent. There is just something to be said about having a dozen patents with 20 method claims. 

Aside from the obvious things that can impact a patent’s value aforementioned, the most important aspect of a patent is not what it contains by itself, but rather what it contains relative to other patents in existence covering similar areas. This group of similar patents is called a cohort and successful identification of a cohort (all related patents) is perhaps the most important task in being able to programmatically derive the value of any patent. Cohorts allow its member patents to be compared, contrasted, and ranked. In the case of the previous example of a digital rights patent, the patent could be ranked within the top 5% of the cohort or among the bottom 5% based on its content - regardless of its owner. Such ranking, when done properly, can help investors and business people better understand the patent owner’s standing within a particular marketplace. However how does one identify a cohort? 

Cohorts can be identified by keyword searching, semantic analysis or to a lesser extent by either assembling composite citations or similar class code comparisons. Programmatically, assembling a cohort using composite citations or class codes is the easiest way to gather a list of similar patents. However, the patent office is at best inconsistent in assigning proper class codes to patents so this field cannot be used effectively. Using composite citations is also a fairly easy way of assembling a list of similar patents. Composite citations use the forward citations identified in each patent as a means of assembling a list of all patents pertaining to the subject of a patent according to the patent filer as well as the patent office. What makes this method of cohort generation unreliable is the fact that new patents don’t often list all available prior art in their patent citations as a matter of practice. Again, the more patent citations they list, the higher the likelihood the patent will not be able to withstand litigation. 

Determining a list of keywords that best describe the area, searching patents for those keywords, and manually eliminating those patents that do not pertain to the subject is the most common method of cohort identification. Another means of accomplishing this is by finding a sample of one to three highly relevant patent(s) and then performing semantic analysis against this sample. Semantic analysis is a means of analyzing words in a patent in terms location and frequency and then using linear algebra along with a method called Singular Value Decomposition (SVD) to generate a relevancy measure for each patent. Through this relevancy measurement and its sub-products, one can use a small sample of relevant patents to generate a listing of patents that are similar to this patent – the result is a highly relevant cohort. The use of keywords is the most accepted method of cohort identification and semantic analysis has been increasingly used to offset or compliment traditional keyword search techniques as a means of further identifying the cohort.  

Some patent search services (such as Patent Café) use semantic analysis as their primary means of cohort identification with some respectable degree of success. However, for reasons similar to that of why patents cannot be evaluated based on forward citations, companies filing patents don’t always follow the most logical path. Deception in filing is still widely practiced. For example, words can be changed so that a rarely used word is inserted rather than what most logically should be residing in its place. When this happens, it presents challenges to semantic analysis to uncover the stem of such words as well as all the alternate definitions of such words. When such deception is used, semantic analysis can be tricked to not include such items within the cohort. However, since keywords generally start out much more broadly than semantic analysis and then resort to manual exclusion, such steps can generally lead to the inclusion of such deceptive patents. 

Quantity versus Quality 

Some companies amass large numbers of mediocre patents as a means of exercising something called the “thud factor.” The thud factor can project a type of scare tactic allowing its owners to wield a big stick in an effort to get potential licensees to comply with paying fees as opposed to fighting it out in an expensive litigation. When companies amass the thud factor, it becomes increasingly difficult to litigate because of the increased possibility that some aspect of their portfolio hits pay dirt. 

Patent trolls on the other hand seek out very specific great (or highly relevant) patents and than focus on certain industries they know well in order to extract licensing revenue from companies operating within these industries. All patent trolls need is one patent within any one area followed by a history of successfully licensing this patent to companies operating in the area. While this strategy works well for this particular type of patent holder, the masses generally rally around a strategy of accumulating a number of patents covering a wide range of technologies and uses within their area of business focus for the purpose of creating a defensive portfolio. 

Note that a patent is not a permission to use a particular innovation; rather it is a negative right – or the right to prevent someone else from using the innovation. As a result, filing for patents and/or acquiring them is a logical step in securing your company’s business by permitting you to seek license revenue from other companies that have leveraged your methods or technology. Holding such Intellectual Property can dissuade competitors from attempting to litigate for a share of your business – I call this the fear of mutual assured destruction. However, if your company doesn’t have a strong patent portfolio the only thing that stands in the way of patent holders lining up at your door seeking license revenue from you may be your deep pockets and ability to drag on an expensive litigation process beyond their financial means. Intel (under Andy Grove) used to do this with precision to the point where by the time the parties were ready to settle, Intel had already designed around the technology and was rapidly taking over a lion’s share of the market. 

Since most companies can’t operate like Intel, they must do the best they can with the resources they have. In terms of Intellectual Property, this means filing for patents and seeking to acquire other patents relevant to their industry or product focus. Only how does one know which ideas to file or what patents to buy? 

Calculating a Patent’s Value 

Determining the value of a patent or an idea via automation is not totally impossible and certainly does have merit. Only you just need to take the result as one important data point as opposed to some kind of authoritative declaration. The information described in Table 1.0 can be derived from basic information obtained about a patent or a group of patents using a patent data provider. Birds-Eye.Net is happy to recommend a list of patent data providers that can give you access to the raw information you would need to generate what ever you are looking for on your own. To request this list of patent data providers just email us. 

Check out these other Birds-Eye.Net papers/products regarding Patents:

• Automated Patent Evaluation - Automated patent evaluation methods
• Patent Harvesting Starter Kit - Essentials required to start filing patents for your company
• Patent Search and Management Services - A detailed look at your global patent search and management options
• Patent Trolls - Directory of active patent licensees

Can Birds-Eye.Net help you or your Company?
Receive your Birds-Eye.Net articles and white papers hot off the presses by adding our RSS feed to your reader.