Free Patents Online: IP Research and Community

You are here

» Conducting a Patent Search

Conducting a Patent Search

A note on syntax
Boolean Logic
Keyword Searching
Advanced Keyword Syntax and Techniques
    Word Stemming
    Search Term Weighting
    Date Range Searching
Assessing Search Results


1. A note on syntax

One point that we must emphasize is that you must make sure you are familiar with the search language for the database you are using. Unfortunately, none of the different search engines use the exact same search language. If you assume that you can use the same syntax from one patent searching site to another, you are going to have a problem. Often, that problem will be obvious because a syntax error will be reported when you try to run the query. However, you may not get an error! Your query might run just fine, but since the syntax is different, the query will not mean what you intended it to mean. You will get results for a different query than you intended (which almost certainly means you will miss relevant references), and if you don't realize what you did, you may never even know. Learn the syntax of the site you use!

[Back to top  Top]

2. Boolean Logic

Boolean logic (named after the mathematician George Boole) refers to logical relationships between sets. For our purposes, those sets are the sets of results from a search engine.

For instance, a search for "bicycle" might return one set of references, while a search for "wheel" would return a different set of references (some of which probably overlap with the "bicycle" references). But, what if we want a set of results that encompasses the references for both of those terms? We use Boolean commands to tell the search engine to combine the two sets.

The commands used in Boolean logic include "and", "or", and "not" (the USPTO web site uses "ANDNOT" in place of NOT, but we will use NOT here for the purposes of clarity - just realize that you will need to use ANDNOT when actually searching at the USPTO web site). In the example above, if we wanted to all references that contained "bicycle" or "wheel", the search would simply be: bicycle OR wheel.

It would be incorrect to construct the search as: bicycle AND wheel. Why? Use of the "AND" command would only return references that contained both the word "bicycle" and the word "wheel". That is very different than returning all results that contain "bicycle" OR "wheel".

And this query: bicycle NOT wheel, would mean yet something else. It would mean the set of all references to bicycle, with any overlapping references from the set of "wheel" results removed.

We assume that most people are familiar with Boolean operators, and their use in search engines. However, the reason the topic bears discussion is that, while the examples above are quite simple, extremely complex searches can be constructed using Boolean operators. The use of the operators, along with the use of parentheses, if necessary, must be done carefully to make sure that you are constructing the search that you think you are constructing. Here is a more complex example:

bicycle AND wheel NOT (spokes OR seat)

What does that mean? It means the set of all references which contain both bicycle and wheel, but which do not contain either spokes or seat. Note the use of parentheses to indicate grouping. If the query was written without the parentheses, it would mean something different:

bicycle AND wheel NOT spokes OR seat

Without the parenthesis you have requested all references which contain bicycle and wheel, but not spokes, or references which contain seat but not spokes. The first query returns a few thousand results, while the second query returns a few HUNDRED THOUSAND results! So, they are quite different.

The moral of the story: Be careful when constructing your searches. They will get even more complicated when you start adding in date terms, class and subclass terms, and all sorts of other variables. Make sure you asked the search engine what you intended to ask it, and always examine your search results critically so that if any logical errors have occurred, you will detect them.

While technically perhaps not a part of Boolean logic, wildcard characters are important enough, and universal enough, that we will discuss them here. A wildcard character is a special character which means "any character or characters". Wildcards are very helpful in increasing the flexibility of your search queries. Let's examine a hypothetical example:

You are doing a search for a new kind of steel and one of the important points is that it is extremely light compared to other similar materials of the same strength. You have decided that you must use "light" in your keywords, but you realize that there are many forms of the word "light", including "light", "lighter", "lightest", and perhaps "lightweight". You want to be sure that you find any possible form of the word. Since you just learned about the Boolean operator "or", you know you could write such a search like so:

light OR lighter OR lightest OR lightweight

But, that becomes very cumbersome. And what if you forgot one??? Wildcards make this easy. You can write it like so:


The wildcard character says to the search engine "Find me anything that starts with 'light', followed by any other characters (or no other characters)."

A lot simpler, right? The only drawback is that it can be less specific that using all those "or" commands. When you write it all out with "or", you know exactly what it is looking for. When you use wildcards, there might be words that match that you did not think about and are not relevant to your search. In this case, any patents containing "lighting" would be found also. So would any patents containing "lighthouse", or worse yet "lights" - there are probably thousands of patents containing the word "lights". So, maybe this particular use of wildcards isn't a good idea.

That does not mean you should not use wildcards. It just means you need to be careful. If your use of wildcards causes you to accidentally include some common word that is not relevant to your search, you could waste a lot of time. But, if used wisely, they can ease the construction of your queries and make them more comprehensive.

Word stemming can often substitute for wildcards, and will be discussed later.

[Back to top  Top]

3. Keyword Searching

Keyword searching is the heart of a good patent search (although class/subclass "hand" searches have their place also). You can use Boolean logic, date ranges, class and subclass designations, and a variety of other commands to construct just about any search you can imagine.

Assuming that you are familiar with Boolean logic, and the search syntax (a summary of our syntax is available elsewhere), the next thing to do is decide on keywords. Determining the best keywords to use requires some thought, and is very important to getting good search results.

Always remember this: The inventors may not have used the words that come first to your mind. Think of every possible way to describe the invention. Think of words that any patent on the topic absolutely MUST contain, and then words that the patents PROBABLY contain - but they might not. It is crucial to develop a good list, and then implement the searches with appropriate logic. For instance, assume that we are searching for an invention having to do with bicycle tires. The word "bicycle" is almost certainly in any relevant patent. Is the word "tire" in any relevant patent? Probably - but maybe not. Maybe the inventor uses the word "wheel" instead. Or maybe even "rim" (although they are not really the same - do not assume that everyone uses words perfectly accurately).

So, the search that you would want to use might look like this:

bicycle AND (rim OR wheel OR tire)

Understand? If this is not clear to you let us know - we will give you a crash course in search logic and you will be all set.

Once you construct your search, see how many results you get. You will always have a limited amount of time that has been authorized for a search. Most likely it will be 6-12 hours on a patentability search. Given this fact, YOU CANNOT LOOK THROUGH MANY THOUSANDS OF TITLES. There simply is not time. So, if you perform a search and it returns over a couple thousand results, you probably need to rethink your keywords and target your search a bit more carefully. In the above example, the search "bicycle AND (rim OR wheel OR tire)" returned 3000 results, what would you do? I would modify the search like this:

bicycle AND tire

and then see how many results that came back with. Tire is the most important keyword after bicycle. If that cuts it down to somewhere under 1500, then you have an amount that you can probably get through in a reasonable amount of time. If you have more time after you finish reviewing the results of that query, you might try:

bicycle AND wheel ANDNOT tire

Why the "ANDNOT tire"? Because that prevents you from getting any of the same results you got the first time around - it would be a huge waste of time to look through redundant results.

Note: This was a simple example. Often, there will 10 or 20 keywords that you might want to use, and instead of constructing a single search that returns a thousand results, you will use multiple searches, each with different sets of keywords, that each return a few hundred results or less.

[Back to top  Top]

4. Advanced Keyword Syntax and Techniques

a. Quotes

[Back to top  Top]

b. Word Stemming

Word stemming is a language related tool, which determines the root of a word first, and then retrieves all possible variants.

When you search with word stemming ON, documents in your results list would have both the root word, and the variants of the root word. With word stemming OFF, your search is limited to the exact word.

For both quick search and expert search, word stemming is turned ON by default.


Search term Word stemming Matching results would have the word(s)
Compose On (searches for variants of compose) Compose, composes, composed, composing, composable, composition etc
Composition Off (searches for exact word) Composition
Composition On (searches for variants of the root word "compose") Composition, Compositions, Compose, composes, composed, composing, composable etc
Metallic Off (searches for exact word) Metallic
Metallic On (searches for variants of the root word "metal") Metallic, Metal, Metals etc.


[Back to top  Top]

c. Search Term Weighting

Some of your search terms would be more important than others. Search term weighting helps you to tell the search engine which term is more important, and how important it is over the other term.

FreePatentsOnline uses the "^" (caret) operator to facilitate search term weighting. Combine this with relevancy sorting option to display the most accurate results at the top of the list.

The caret - "^" should be followed by a whole number that indicates the relative importance increase. For example, "^2" means that the word is twice as important as a word with no caret, while "^3" means three times as important, etc.

Note: Ensure that there is no space left between the keyword and the weighting operator.

Query Interpretation
DNA^5 or RNA DNA is 5 times more important to the relevancy of documents than RNA. Finds all documents with either DNA or RNA, or both, but all other factors remaining the same, rank documents with DNA higher.


[Back to top  Top]

d. Wildcards

Truncation or Wildcard searching allows you to retrieve documents containing variations of a search term. Wildcard symbols can be used to replace a single letter or a part of a word, or one or more numbers.

FreePatentsOnline use two wildcards: ? to replace exactly one character or number. * to replace zero or multiple characters or numbers.

Query Results list would have
Analy?e Analyse and Analyze
Sul*ur Sulphur and Sulfur
Harbo*r Harbor and harbour


[Back to top  Top]

5. Class/subclass searching

You can find the relevant classes/subclasses, and enter them into the Advanced Search (the search syntax for that would be "ccl/xxx/xxx"). This will return all patents in a given class/subclass. You then scan every title, and read documents of interest in more detail. This is often called a "hand search".

To search an entire class, use a wildcard for the subclass (ccl/xxx/*). This would allow you to see all patents under a particular class, but would be impractical for a class that contains a large number of patents.

This seems a lot simpler than keyword searching, right? You do not have to try and guess what words are in relevant patents. You do not have to try and figure out many alternative ways of describing the same device. But, there is a catch. Just like you can never be sure what words an inventor used, you can never be certain in what class/subclass a relevant patent will be found. Often there are multiple possibilities because of classes overlapping, and sometimes you may even find outright mistakes, where a patent is filed in a class in which is just should not be. A good keyword search will find a relevant patent, no matter which class or subclass it is in. You will NEVER find an erroneously classified patent via a class search.

Alas, some people like class/subclass searches (and they do have their place, when trying to be as thorough as possible, to augment keywords searches, but they are not a substitute for keyword searching). This is probably a result of old habits dying hard: a few years ago there was no such thing as keyword search and so hand searches were the norm.

How do you know what the most relevant classes are? The most common way to figure this out is to do a keyword search first, and use the classes/subclasses listed on documents found during that search. You can also keyword search the class/subclass definitions, or ask an examiner (covered previously).

[Back to top  Top]

6. Date Range Searching

Let's pretend we want to do a Validity search on patent, 1,234,567, filed on August 7, 1994, that claims "a lighter, more aerodynamic design for a bicycle rim by using an oval cross-sectional shape". We have decided that the keywords we want to search on initially are "bicycle" and "rim". What is the syntax for the search we want to construct? Remember, nothing filed after August 7, 1994 counts, because we have to prove the present inventor WAS NOT THE FIRST to invent such a thing, so all relevant prior art must pre-date his application. The search is thus:

APD/1/21/1790->8/17/1994 AND bicycle AND rim

That means, "All patents applied for between January 21, 1790, and August 17, 1994, that contain the words bicycle and rim". Why 1790? That is the first year that the U.S. patent office existed. You can use 1790 always, just to be safe, or, in a high-tech field, you can use a more recent year if you are sure it pre-dates the development of the technology. "APD" is the abbreviation for Application Date here and at the USPTO. Again, know the syntax of the search engine you are using!

[Back to top  Top]

7. Assessing Search Results

Now that you have constructed your search, you find that there are hundreds or thousands of results. How do you know what is useful and what to ignore? To answer that question, we need a bit of background on patent law (DO NOT CONSTRUE THIS, OR ANYTHING ELSE ON THIS SITE, AS LEGAL ADVICE. SEEK APPROPRIATE COUNSEL).

The Test for Obviousness

When a patent examiner reviews an invention, the most difficult task he is faced with is deciding if the invention is "unobvious". As an aid to addressing this issue, the following theoretical question is posed: "Given an inventor with average skill in the appropriate arts, average creativity, and a complete knowledge of the prior art, would he have come up with this invention?" If the answer is yes, then the invention may be judged an obvious extension of existing technology, and therefore not patentable.

Note that the use of the phrase "a complete knowledge of the prior art" means just that - complete. The hypothetical inventor is assumed to have knowledge of each and every piece of prior art known to man that is relevant to the invention under examination.

So, the criterion for "unobviousness" is not simply knowing facts that someone else did not know. It is more a question of inspiration; of exceeding the average level of creativity possessed by the hypothetical inventor.

Isn't this quite a subjective criterion? Absolutely, but that is how it is done. The objective is to prohibit people from obtaining patents on trivial modifications of existing inventions. For instance, if you had invented a new kind of bicycle, would it be fair for someone else to be granted a patent for a bicycle that was identical to yours, except that they had chosen to specify that it was made of a different material, or perhaps had a frame whose angles differed by tiny amounts (and they had no data to prove that these tiny changes improved the functioning of the bicycle in any substantial way)?

No, that would not be fair. And they would never get a patent due to such trivial changes, because their "invention" would be judged an obvious extension of yours.

The Doctrine of Equivalents

The concept of the Doctrine of Equivalents goes hand-in-hand with the test for obviousness. The Doctrine of Equivalents is a mechanism which broadens the scope of claims to include subject matter that is not explicitly claimed, but which would be considered equivalent by an expert in the field.

For instance, let's go back to our hypothetical bicycle patent. We had stated our claims as:

Claim 1: A lighter, more aerodynamic design for a bicycle rim by using an oval cross-section.
Claim 2: The rim of Claim 1 where the cross-sectional piece is formed from aluminum.

Suppose those were our only two claims. Now suppose that a competing bicycle manufacturer made a rim almost identical to ours, but they made the rim out of another high-strength, lightweight allow, such as titanium, or magnesium.

We sue them, alleging that such a trivial modification should not permitted. Our argument is that it should be obvious to anyone familiar with bicycle rims and the materials out of which they are made, that making the same rim, but out of a different alloy, is still an infringement of our patent. They have done nothing novel. In fact, the different alloy they used is even well-known to those in the bicycle industry, so the modification required no creativity at all.

Their counter-argument is this: "You had an opportunity to state in your claims exactly what your patent was to legally cover. If you had desired to do so, you could have included magnesium and titanium along with aluminum in your Claim 2. Or, you could have rephrased your Claim 2 to state that the rim was formed from aluminum or any other high-strength, lightweight alloy. You didn't. You should not be allowed to claim that our modification infringes on your patent when it was your own lack of forethought or bad writing that caused you to omit the proper language which would have made it clear that you intended to claim more materials than just aluminum".

Who is right? There is no clear-cut answer to that question. Both sides have legitimate points. In fact, this murky issue is so important that the Supreme Court has addressed it at least twice in recent history, once in 1997 and once in 2002, both times affirming the use of the Doctrine of Equivalents. Of course, like the test for obviousness, just what is "equivalent" and what is not, is a judgment call that must be made on a case-by-case basis.

The main difference between the test for obviousness and the Doctrine of Equivalents is in when they come into play. The test for obviousness comes up during the process of examining a new patent application. The Doctrine of Equivalents is not used for patent applications. Rather, it is applied to the claims of an existing patent by the court system when deciding whether someone is infringing.

So what does all this have to do with patent searching? Quite a bit. The point behind all this is simply this: When conducting a search, you do not need to find prior art that EXACTLY mimics the invention in the disclosure. If you find the exact invention in the prior art, it is said that the prior art "anticipates" the invention. Anticipation is great when it happens, but most of the time you are going to find prior art that is close, but not an exact match. And that's fine - because of the obviousness test, and the Doctrine of Equivalents, extremely similar prior art may serve just as well as anticipation.

Enough Background! What is relevant and what is not???

Now that you understand how patent examiners and judges determine whether one invention is unallowably close to another invention, how do you, the patent searcher, decide what prior art to cite when conducting a patent search?

Just like the application of the obviousness test or the Doctrine of Equivalents is a judgment call, so is what prior art to cite. However, you have a much easier time of it than a patent examiner or a judge. Why? Because rather than trying to be fair, you are trying to be thorough. When an examiner or a judge is in doubt, they still have to come to a decision, no matter how tough it is. You do not, because you follow this simple rule: If in doubt, cite it. It is much worse to leave out a relevant reference than it is to waste a bit of time by citing a reference that may later be determined to not be relevant.

i. Inventions with Few Features

Note that most disclosures only have a few points on which you will be searching. For instance, the disclosure might describe a bicycle, but the search would not be on an entire bicycle, but rather on a specific piece of a bicycle which was novel, such as an improved rim or an improved pedal. The disclosure should make it clear what aspects of the device are the subject of the search.

In the normal case of a disclosure with no more than a handful of important points on which to search, anything that parallels an important feature in the disclosure is relevant. The prior art that you cite does not need to parallel every aspect of the invention; it may share only one important feature. So, if you believe that you have found prior art that anticipates an aspect of the disclosure, or if you believe that a judge or examiner would consider it an obvious modification of the disclosure, cite it.

Note that it is better to find a single piece of prior art that addresses all facets of the disclosure, but that is not always possible, so you cite several patents that, together, build up a case for each point in the disclosure being anticipated or obvious.

ii. Inventions with Many Features

You may want to search on an invention that has numerous important aspects. This can occur with a complex invention with many supposedly-novel aspects, or in patents that seek to patent a multi-step process.

So what happens where the invention is a 10-step process, and not all of these steps are necessarily novel (though some may be), but the combination of these particular steps in that particular order is novel? Do you cite every piece of prior art that mentions any of the steps in any context? Probably not - it's a judgment call.

iii.Where Can Prior Art Come From?

The simple answer is: anywhere! For instance, when searching patents for prior art, it does not matter if you find relevant information in the Background, in the Abstract, in the Summary of the Invention, in the Detailed Description, in the Figures, or in the Claims. If relevant information is present, you can use it.

From the standpoint of efficiency, you will usually have a pretty good idea if anything worthwhile is going to be found in a given patent by reading the Abstract, and then the Summary of the Invention. If those sections look promising, but you haven't found exactly what you are looking for, use your browser to search for specific words in the patent - often the quote you want is buried in the Detailed Description and there is no point in reading a 50 page patent to find that one gem when you can use the "find" function on your web browser.

Similarly, when looking for prior art in non-patent literature, the information can come from anywhere (subject to the date requirements of certain types of searches). A web site is fine. A technical journal is fine. A book is fine. Even product literature or promotional material is fine. Anything that serves to document, in reasonable detail, that the concepts you are searching for already exist (or already existed before a given date when dealing with Validity searches) should be cited.

[Back to top  Top]