Two related issues. One, will a machine translation do? One case found a machine translation insufficient because it contained many untranslated words. Two, what do you do if the Examiner doesn't provide a translation? File a petition.
Details: The MPEP section "Reliance upon Abstracts and Foreign Language Documents in Support of a Rejection" states:
Citation of and reliance upon an abstract without citation of and reliance upon the underlying scientific document is generally inappropriate where both the abstract and the underlying document are prior art. See Ex parte Jones, 62 USPQ2d 1206, 1208 (Bd. Pat. App. & Inter. 2001) (unpublished). ...Why does the MPEP mention "where both the abstract and the underlying document are prior art" ? Abstracts for scientific articles are sometimes published separately than the article itself. In such circumstances, the abstract may be prior art even if the article is not, or vice versa. This wrinkle doesn't come up with patent references, since the abstract is simply part of the patent application.
When both the abstract and the underlying document qualify as prior art, the underlying document should normally be used to support a rejection. In limited circumstances, it may be appropriate for the examiner to make a rejection in a non-final Office action based in whole or in part on the abstract only without relying on the full text document. In such circumstances, the full text document and a translation (if not in English) may be supplied in the next Office action.
So the first thing for you to check is whether or not the abstract is separately published. If not, both are prior art, and the MPEP says the Examiner is required to supply an English language translation of a foreign reference – good news for Applicants. However, a recent BPAI decision interpreting this MPEP section read it differently.
In Ex parte Park, the Applicant argued that the Examiner's reliance on the English language abstract of a reference was improper, citing MPEP 706.02. In a footnote, the Board noted that "contrary to the premise of Appellants’ argument, the Examiner’s reliance on Inada is based exclusively on the English language abstract and not on the underlying non-English language document." By focusing on the language "seeks to rely on that document" in the MPEP provision, the Board seemed to imply that a translation would be required only if the Examiner relied on the underlying document rather than the abstract. Yet in Park both the abstract and the underlying reference were available as prior art – the reference was a Japanese patent publication, not a scientific article.
The dicta in Ex parte Park contradicts two older BPAI decisions from 2005 and 2006. In the first, Ex parte Barraclough, the Board remanded the case back to the Examiner rather than reaching a decision because the Examiner did not provide a translation. As authority for this requirement, the Board referred to MPEP 706.02 and to Ex parte Jones (the same case mentioned in the MPEP). The Board noted that the translation should be already in the record because it functions as a board of review, not of examination. (Note: the remand decision is not available on the BPAI website, but is available through Public PAIR in Appl. No. 10/499,644.)
In the second decision, Ex parte Anelli, the Board criticized the Examiner for waiting until the Examiner's Answer to provide the English language translation, noting that this delay "directly contravenes MPEP 706.02." However, the Board considered this harmless error because they reversed the obviousness rejection for not teaching all the elements.
The Board also touched on the translation issue in another recent decision. A footnote in Ex parte Iekuchi noted that both the Appeal Brief and the Examiner's Answer characterized the rejection as relying on the "abstract only" of a foreign language reference. The Board's failure to address the lack of a translation may be tacit agreement that no translation was required because only the abstract was relied upon – but it's also possible that the Board didn't bring up the issue because the Applicant didn't.
Ex parte Rudd, involved a machine translation of a foreign language reference, where the translation included a number of untranslated words (shown in the text as asterisks). The Board found that many of the specific paragraphs relied on by the Examiner contained untranslatable words as well as grammatical and idiomatic errors, and noted that "major parts of the machine translations are unintelligible to us." The Board remanded back to the Examiner with an order to provide a better translation.
Finally, the recent decision Ex parte Camps held that the requirement for an translation of a foreign language reference is reviewable by petition, not appeal. The Board cited the same Ex parte Jones decision mentioned in the MPEP:
In Jones, our colleagues held that: "... If a translation is not supplied by the examiner, the applicant may wish to consider seeking supervisory relief by way of a petition (37 CFR § 1.181) to have the examiner directed to obtain and supply a translation."
My two cents: One of the underlying rationales in Ex parte Jones was that an abstract for a scientific journals is likely to be in error because these abstracts are written by persons other than the author. This same concern isn't present when the reference in question is a patent application. Nonetheless, in dealing with any reference it's important to have the whole document, because the understanding you get from reading the whole is sometimes different than what you get reading the abstract alone. Therefore, the idea that a translation is not required if the Examiner relies only on the abstract seems fundamentally unfair.
The policy of making this a petitionable matter rather than an appealable one makes sense because requiring the Examiner to take a particular action is a procedural matter. Even so, it's bad news for Applicants because petition practice is slow, unpredictable, and expensive.