- This article was considered for deletion at Wikipedia on November 25 2015. This is a backup of Wikipedia:Cloem. All of its AfDs can be found at Wikipedia:Special:PrefixIndex/Wikipedia:Articles_for_deletion/Cloem, the first at Wikipedia:Wikipedia:Articles_for_deletion/Cloem.
- Wikipedia editors had multiple issues with this page:
Cloem is a French company which applies Natural Language Processing (NLP) technologies to patents ("algorithmic patenting"). Its slogan is "Invent more, better and faster". Cloem's team comprises patent experts and computer linguistics scientists, describing themselves as the "quants of IP". The publicly available roadmap of the company comprises a partial automation of patent application drafting, the detection and injection of trends ("technical patterns") in technology, as well as tools for patent analytics.
Cloem creates, then timestamps and optionally publishes, variants of patent claims, called "Cloems", using Natural Language Processing technologies.
Described by the company's scientists, Cloem technology uses specialized claim parsers, proprietary algorithms to manipulate patents as well as proprietary dictionaries.
A patent claim defines a technical invention. Claims define the boundaries of legal protection. A claim is one unique sentence.
A parser performs the formal analysis of a sentence.
Cloem’s algorithms include the injection of the definition of terms, permutation of the order of method steps, handling of time, transcoding of claim category and creation of intermediate generalizations.
According to Cloem’s approach, one or more words of "base claims" (or "seed claims") are replaced by their synonyms, hyponyms, hyperonyms, holonyms and meronyms. Cloem’s claim drafting model is said to follow industry “best practices” (e.g. USPTO M.P.E.P., EPO Examination Guidelines, Case Law, etc). Cloem implements a proprietary implementation of TRIZ.
Amplifying human creativity, machine can "exhaust" ideas. Even more, if a "technical pattern" (as a combination of new words like "touch screen") can be propagated to the patent corpus. Compared to a word processing software, the proposed algorithmic patenting consists in pushing thesaurus and dictionaries into patent claims. When an inventor provides an invention, he "implicitly" discloses a "class" of inventions and it is rather fair to let the machine "fill-in" the gaps.
A business method patent claim is a software patent claim with a few words being changed ("bank" replaced by "server" and "order" replaced by "message")
The more knowledge silos (the more patent classification classes), the more fees can be collected by the patent system. By contrast the recognition of transpositions of technical patterns (analogy reasoning) would substantially decrease the number of allowable patents.
The man in the street reading a text: people skip words, replace a word by another, misread some others, permute lines, etc. By contrast, the "legal" way to read a text is performed on a "word-by-word" basis (e.g. patent law provisions about support or added subject matter). The automatic creation of variants corresponds to the natural way to read contents ("vibrations" of texts)
New questions for the patent system
The use of dictionaries is not controlled or standardized in patent practices; by contrast, the approach quantifies the use of words. Attorneys routinely manipulate claim language; does the automation of such operations discard generated contents ?
The qualification of "prior art" of Cloems when published texts is an open question: Internet as a source of prior art is a debated topic.
Once published, contents may be pushed to patent offices via "Third Party Observations".
Objective embodiment of the "skilled person"
Patent laws generally rely on the "skilled person" or Phosita to evaluate if a patent claim is obvious or not.
Different tests can be used, including the "problem-solution" approach.
The method according to Cloem's approach reduces this assessment to a quantification of text manipulation.
Enablement and sufficiency of disclosure
Derivate works from an enabling disclosure are probably enabling disclosures. Non-sense variants can be useful.
By publishing variants of a patent claim, adjacent inventions should no longer be patentable. See also here.
If focused to certain technical domains (i.e. using a plurality of patent claims), the associated freedom to operate can be theoretically improved.
Cloem’s approach theoretically can expand and/or densify the public domain (for example when applied to Appropriate technologies) and impede obvious patents.
For example, as alleged by Cloem employees, the approach can break silos of knowledge by mixing-up patent classification classes. If variants are not published, the created variants can be patented (regular marriage between patenting and defensive publishing)
Negative reviews consider Cloem to be an attack to the patent system ("brutalizating the permutation").
Positive reviews consider CLoem as part of the "Legal Tech" movement.
Neutral reviews consider that Cloem follows patent laws as copyleft follows copyright laws.
- "Would you like 10,000 Cloems(tm) with that Patent?" http://patentlyo.com/patent/2014/10/would-cloems-patent.html
- "Technical poems" http://www.bloomberg.com/news/articles/2014-10-03/cloem-pepsico-coca-cola-indyne-intellectual-property
- "Defensive cloems(tm)" http://www.cullens.com.au/news-events/ideas-into-assets-r-blog/defensive-cloems/
- "Computer Generated Patent Claims: Good Thing or Bad Thing?" http://ratiodex.com/computer-generated-patent-claims-good-thing-or-bad-thing/
- "Two new startups that could upend intellectual property laws" http://venturebeat.com/2015/02/16/two-new-startups-that-could-upend-intellectual-property-laws/
- "Algorithmic Patenting" http://yro.slashdot.org/story/15/02/17/2032207/algorithmic-patenting