[tweetmeme source=”Intellogist” only_single=false]
I wanted to quickly check in from Baltimore where I’m attending the Patent Information Users Group annual conference (by the way, you can follow Intellogist on Twitter to see my live tweets from the conference!). I saw a presentation about a new feature of WIPO’s PATENTSCOPE ® tool today that will absolutely blow your mind. This is the best cross lingual search feature I have seen so far, and it’s available for free, which should really make you sit up and take notice!
Up until now some of you may have been familiar with the cross-lingual search on QPAT. However, the WIPO search tool has some unique features that differentiate it from the QPAT interface. The PATENTSCOPE ® tool uses a number of cross-lingual dictionaries built in house by WIPO to generate keyword queries in other languages. It does this with the aid of a technology discipline filter to help ensure that the meaning of the desired synonym is clear to the system – for example, so that it will not misinterpret a search on bats (the flying mammal) as a search on baseball bats. It also generates a basic set of main IPC classes to be included in the search string; this is a clever way to ensure that the search results are related to the appropriate general technology area.
Rather than taking my word for how this feature can generate totally cool keyword search strings, try it for yourself! First go to the National and PCT Collections Search form of the WIPO PATENTSCOPE ® interface, and choose “Cross Lingual Expansion” from the drop-down menu that appears when you hover your cursor over “search.” Then input your search query in English, French, German, Spanish, or Japanese, choose the level of control you’d like to have over the query (I recommend choosing “Supervised” Expansion Mode just to see how this thing works on a detailed level), and choose your comfort level between “precision” and “recall.”
(Most searches are a trade-off between precision and recall – precision refers to the relevance level of the results one gets, while recall refers to the percentage of relevant results one is able to pick up with a search string. High precision searches will usually be lower on recall, because targeted search strings can miss some useful outliers.)
After you input your query, the system may ask you to confirm that it is truly in the language you selected. This is because with shorter queries the system may not have enough information to recognize the query language correctly and it needs a little help.
Next, the system will show you a list of general technology areas. Choose the areas that are most closely related to your search topic – this will help the program select the most appropriate cross-lingual synonyms from its dictionaries.
In the next screen, you will be shown a list of possible synonyms to your search query. Choose all that apply for each term, and choose “translate selected terms.” Voila! The system will have generated your search query in English using the synonyms you selected, and then will translate those synonyms (using its proprietary dictionaries in combination with a technology filter) into French, German, Spanish, and Japanese. Furthermore, the system will have picked a list of general IPC main classes to add to the search string as a filter. From this screen, you can have further control over the use of stemming, the acceptable distance between words in a phrase, and the field of search (title, abstract, claims, etc.)
As you can imagine, the end result of this very involved process is an absolutely massive search query containing dozens of terms, and which can be executed in conjunction with complex computational tasks like stemming and proximity operators. It remains to be seen whether the PATENTSCOPE ® system suffers any performance issues from this new tool, but right now, they say go for it!
I hope you all go right over there and try this search tool, because I think it will impress you. At the very least, it’s a free way to start expanding your search queries to include some non-English collections (because don’t forget, these queries can be pasted anywhere – you don’t have to only run them in PATENTSCOPE ® ). And one final note – if you can’t think of a query to test this out on, the demonstration performed today was on the simple query “sound isolation,” which illustrated the tool’s capabilities very well – as test demonstration queries usually do :). Happy searching!
This post was contributed by Intellogist team member Kristin Whitman.