As we’ve proved in past posts, machine translations aren’t always accurate, especially for translations of patent documents. In a previous post, Landon IP’s Director of Translation Services Sonja Olson summed up the problem with machine translations: “machine translation will get the gist of the document, but it will lose the nuance. Machine translation can plug words together, but it can’t understand the sentence as a whole.” Machine translations can be useful for browsing documents for relevance during the search process, but professional human translations should be used for filing or legal purposes.
Today, we’ll take a look at two machine translation tools that can be used for initial prior art search purposes: 2lingual for non-patent literature (NPL) and Patent Translate for patent documents on Espacenet. Both tools utilize the machine translation functions of Google Translate.
After the jump, we’ll look at how 2lingual and Patent Translate can help you expand your prior art search and evaluate non-English patent documents for relevance!
A search service similar to Sobotong is 2lingual.com. What I like more is that the results are displayed in two columns which allows a quick overview of the results in both languages.
Like Sobotong, 2lingual is a bilingual search interface that allows users to search for non-patent literature in two languages simultaneously. Unlike Sobotong, 2lingual provides a side-by-side search interface, where the search results for both languages appear in adjacent columns. 2lingual was created by H.K. Tang, and the platform can be used to conduct a Google search or a Bing search (under two different URLs). According to the “About” section of the website, the Google search version of the site includes the following features:
- Real-time search suggestions
- Query translation option
- Side-by-side dual language search results
The Bing search option includes the previously mentioned features, plus:
- Spelling corrections
- Cached pages
- Related search links
For both the Google and Bing search options, 37 languages are currently supported.
2lingual also hosts two voice-search tools, which allow the users to search in 66 different spoken languages (voice-to-search tools require a Speech-to-Text capable browser):
On the 2lingual homepage, the user can enter a search term and select two different languages from drop-down menus below the search form. The search results appear in two columns below the form, with results in the first column in the first selected language and results in the second column in the second selected language. The search term for the second column has automatically been translated into the second selected language, but the user can choose to de-activate the automatic query translation.
Results appear in lists of 10 results to a page, with the site title, a snippet with search terms in red font, the URL of the site, and a thumbnail image (if available) appearing for each search result. The user can navigate to one page of search results in one column while staying on another page of results in the second column. The user selects the result title to open the site in the same window.
About Patent Translate
The service uses Google’s Translate technology and enables translation from and to English for French, German, Spanish, Italian, Portuguese and Swedish, covering approximately 90% of all patents issued in Europe. By the end of 2014, the service will also be able to translate patents from and into all 28 languages of the EPO member states, as well as Chinese, Japanese, Korean and Russian.
Using Patent Translate
- Find a patent document in Espacenet.
- Choose the part you want to translate – abstract, description or claims.
- Select the target language, and click the patent translate button.
- Once you see the translation, you can mouse-over to see the original text sentence by sentence.
When Should Machine Translation Tools be Used?
2lingual and Patent Translate are both examples of machine translation tools that can be very useful during the first stage of a prior art search. 2lingual can expand the scope of a non-patent literature search to include results in multiple languages (although a professional patent search by a multilingual translator is a much more effective method for identifying non-English prior art). Patent Translate can be used to evaluate the relevancy of patent documents found while searching Espacenet. However, as the disclaimer above the Patent Translate translation warns: “This automatic translation cannot guarantee full intelligibility, completeness and accuracy.”
The machine translation of patent documents and other prior art by services like Google Translate offer translations that are often disjointed and sometimes incoherent. Machine translations are a useful way to determine the relevance of a document to a search, but any official or legal actions should only be conducted with human-translated document.
Do you have any suggestions for useful machine translation tools? Tell us about them in the comment section!
This post was contributed by Joelle Mornini. The Intellogist blog is provided for free by Intellogist’s parent company Landon IP, a major provider of patent searches, trademark searches, technical translations, and information retrieval services.