In through the outfield blog

01 February 2013

The Smart Search format in the Espacenet patent database

For some time the free Espacenet patent database has automatically opened at the Smart Search page, which I find slightly irritating as I normally want the Advanced search format.

I have used the Smart Search for entering patent numbers – this can be in the form 700000 or 70000 for example, or 190520000, and any relevant records will be retrieved. Filing numbers can also be requested. Using country codes such as US or GB is not required but is of course useful if many records are expected (such as lowish numbers). This ability to use numbers without codes used to be available in the Number Search format but this has been incorporated into Smart Search.

Another ability in Smart Search is the ability to use logical operators. This is the name for the ability to say things like “x must be present within 5 words of y”, so that there is a high chance that the words are closely related. Keyword searches would be in the title or the abstract (summary) area.

For example, suppose you are interested in the concept of an umbrella hat. One example is the rather striking drawing given below, by Friedrich Wielandt of Switzerland whose British patent 747154 was published in 1956.

 Rain protective head gear rain hat patent image

If you ask for “umbrella hat” you get 31 hits. If you ask for the words to be no more than 5 words apart you get 284. These would, admittedly, include some “false drops” such as the unwanted phrase “umbrella fabric is arranged between the hat body” which is from a Chinese specification

You can try reducing the 5 word rule to 3 (I personally would never make it less) which in this case would cut the number found to 262, a small loss. Simply using the words in the title or abstract field in the Advanced search, linked by “and”, gives 369 hits.

Results would presumably vary from one subject area to another, and would also depend on the ways words could relate to each other – “light” for example means illumination but also an absence of weight, and there would be odd results in many cases if that word was used. 

Greater precision nearly always means a loss of relevant material, and tools are never a substitute for using your judgment, which in turn depends on what you or a client wants and expects.

Smart Search uses codes to express these operators. The formula used above is:

umbrella prox/distance<5 hat

There is also this looser, but in a way more specific as it ties the words to use within the same sentence, code:

umbrella prox/unit=sentence hat

You could also add a country code such as “and pn=us” if you wanted the search limited to US publications. It is a pity that these codes are hardly intuitive and must be remembered or written down.

It is also a good idea to add truncation symbols: # means one extra character must be present, ? no character or one character (used for most plurals), while * indicates that any number of characters can follow (used, generally, for verbs).

My impression is that most if not all of the other possibilities in Smart Search are already catered for by using Boolean operators such as OR, AND, NOT, or by string searching with words included within quotes to show that they must be found in the same order.

Comments

The comments to this entry are closed.