AN OVERBROAD PATENT ON using word stems to translate unknown words - This application from Microsoft seeks to patent the idea of...Machine translation of unknown words by finding known word stems and adding inflections! 10 minutes of your time can help narrow US patent applications before they become patents. Follow @askpatents on twitter to help.

QUESTION - Have you seen anything that was published before 1/31/2012 that discusses:

  • Using word stems to guess at translations of "out of vocabulary words" in machine translation

If so, please submit evidence of prior art as an answer to this question. We welcome multiple answers from the same individual.

EXTRA CREDIT - the applicant is claiming preprocessing the morphological analysis by looking for transliterations, misspellings, substituting colloquial expressions if recognized.

TITLE: Resolving out-of-vocabulary words with word stem analysis

Summary: [Translated from Legalese into English] A method for translating "out-of-vocabulary" words which involves looking for word stems, translating the word stem and inflecting the word stem in the foreign language

  • Publication Number: US20130197896 A1
  • Application Number: US 13/362,595
  • Assignee: Microsoft Corporation
  • Prior Art Date: Seeking prior Art predating 1/31/2012
  • Open for Challenge at USPTO: Open through 1/28/2014
  • Link to Google Prior Art Search - "Find Prior Art"

Claim 1 requires each and every step below:

A method under control of one or more processors specifically configured with executable instructions, the method comprising:

  1. receiving, by a server, a request from a computing device to translate one or more words from a source language to a target language;

  2. determining that a phrase table excludes a particular word of the one or more words but includes a remainder of the one or more words;

  3. determining whether a morphological analysis of the particular word identifies a stem;

  4. in response to determining that the morphological analysis of the particular word identifies the stem, translating the stem from the source language to the target language to create a translated stem;

  5. inflecting the translated stem based on the morphological analysis of the particular word to create a translated word corresponding to the particular word;

  6. translating the remainder of the one or more words from the source language to the target language to create a translated remainder;

  7. creating a translation of the one or more words based on the translated word and the translated remainder; and

  8. sending the translation to the computing device.

In English this means:

A method of machine translation, comprising:

  1. Receiving request to translate one or more words to target language

  2. Determining that the words are "out-of-vocabulary" (i.e. not in the lookup table)

  3. Determining that the words contain word stems

  4. Translating the word stems into the target language

  5. Inflecting the word stems in the target language

  6. Translating the remaining words

  7. Sending the translated words to a computer

Good prior art would be evidence of a system that did each and every one of these steps prior to 1/31/2012

You're probably aware of ten pieces of art that meet this criteria already... separately, the applicant is claiming looking for transliterations, misspellings, substituting colloquial expressions if recognized

"Using word-stem analysis to translate unknown words from the Applicant"

What is good prior art? Please see our FAQ.

Want to help? Please vote or comment on submissions below. We welcome you to post your own request for prior art on other questionable US Patent Applications.


"Towards the Use of Word Stems and Suffixes for Statistical Machine Translation" (2004) discusses this technique.

2 "Determining that the words are 'out-of-vocabulary'" seems trivial, but they reference this:

the system is able to produce correct or approximatively correct translations even if the full form has not been seen in the training corpus

3 "Determining that the words contain word stems" seems covered by:

We introduce two types of transformations to the verbs...

The full form of the verb is replaced with its base form


Therefore we reduce the words of this language [Serbian] into stems. We split the word into stem and suffix and then drop the suffix.


Gregg Shorthand built words using stems and inflections from the 1930s at least through the 1990s. 'Brief Forms' and extrapolated inflections were common. Several good articles exist, including this:


Other older shorthand systems including Pitman and Duployan had similar structures. Machine stenography is still common in courts of law, and uses 'chords' played on a small keyboard to represent word stems and inflections. Most of these systems are patented, and there may be at least one open source machine shorthand system in use.

Microsoft's attempt to patent such a system appears to be ridiculous.


This article from 1994 covers PC-KIMMO version 2, a then-new computer program designed to perform "morphological analysis" on words in natural languages.

Examples are given in the article for English, but it is strongly hinted that the analysis could easily be adapted for other languages, and they mention several languages (including Finnish) which motivated the need for this style of language parser. They specifically note that version 2 adds the ability to determine the part of speech that a word belongs to (noun, verb, adjective...)

It would then be a straightforward and obvious step to map the word-stems and inflections to their equivalents in another language, reverse the analysis there, and thus produce a transliteration.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.