Table of Contents
The Company
Q: How long have you been in business?
A: Cognition Technologies has been in business since February of 2003, however, the core engineers have been working on this project together for over 20 years through its predecessor companies.
Q: Is Cognition Technologies, Inc. publicly or privately held?
A: The Company is privately held. It has been funded, to-date, by investments from individuals, angel investors and venture capital firms.
Q: Where is Cognition Technologies, Inc. located?
A: 6133 Bristol Parkway, Suite 350, Culver City, CA 90230; (310) 641-7200.
The Technology
Q: When applied to search technologies, does Cognition's Semantic NLPTM have to do intensive linguistic analysis for each query or is that analysis done ahead of time during indexing? If it is done ahead of time, explain how it works?
A: All linguistic analysis is pre-processed. Conceptual interpretations are saved at indexing time, which means that the computationally intensive reasoning does not have to take place during a search.
Q: How does the distributed processing work across multiple machines?
A: The indexer automatically distributes processing on as many servers as the user selects at indexing time.
Q: What is the underlying technology for Cognition's Semantic Natural Language Processing (NLP)?
A: Cognition's Semantic NLP uses patented linguistically-based technology that interprets the meaning of documents and queries. It mediates text processing with concepts rather than word patterns. Concepts enable Cognition's technology to find paraphrases of a concept. & The text "Did they adopt the bill?"; is known by Cognition to relate to information about "the approval of Proposition A", because "adopt" in the text means "to approve", and "bill" in the text means "a proposed law."
Q: How is your technology architected for OEM Applicationegration? C-DLLs, COM, Web Services, other?
A: Cognition has a C++ API with DLL and an ActiveX/com API for both indexing and search. There are Java, Python, Perl, PHP API, and Ruby APIs.
Q: Does your technology store metadata and document content or is it a text reading engine only?
A: It stores metadata, and permits document review and categorization.
Q: Does your technology update its index 'on the fly' so (legal) review information can be immediately searched?
A: The review tool used within the Cognition for Legal domain does in fact update the index on the fly.
Q: Is your search technology ODBC or ADO.NET compliant? If not, what Interface would an OEM use to integrate your search engine?
A: No. Cognition has a C++ API with DLL and an ActiveX/com API for both indexing and text processing. There are Java, Python, Perl, PHP API, and Ruby APIs.
Q: Does your technology offer Search assists? (Thesaurus, Index browser, Clusters, Term Maps?)
A: All of these are built in to the Search capabilities of Cognition's Semantic NLP. The technology uses large semantic databases covering all of the lowercase terms and hundreds of thousands of upper case terms within the English language. Cognition's technology first determines meanings of search terms, and then searches on those meanings, regardless of how they are worded in the target document base. The user may alter the query term meaning selection if desired, but that is rarely necessary. As a result, Cognition takes a query like "fatal fumes", and finds "poisonous gas" and "death from inhalation of steam", etc. Also, it uses a taxonomy to find particular instances of more general concepts. If you search on "vehicle accident", it will find material about "steamship", "bus" and "car."
Q: Does your technology support foreign character sets?
A: Yes -- anything in Unicode.
The Product
Q: Can I add new language to the dictionary on my own?
A: You may add categories of nouns yourself in a special file. You can add new nouns in classes, such as a list of kinds of video recorders, a list of kinds of motorcycles, or a list of kinds of roses. If you have more complicated special vocabulary, it may be introduced into Cognition's Semantic NLP dictionary as a service provided by Cognition (for a fee).
Q: How do you deal with the constantly changing English Language?
A: We constantly add new language or neologisms. The non-technical language gains about 300 new words per year (Cognition already covers nearly all of the common English language).
Q: Does Cognition's Semantic NLP work with other languages?
A: For non-English, it works as a pattern-matcher, with no conceptual reasoning. If the text being processed is "citron" (a fruit) in Spanish documents, it finds that string, but not kinds of "citron" such as "limon" and "naranja."
Q: What happens if Cognition does not understand the concept of the text being processed?
A: Cognition always guesses a meaning. If it is wrong, the user has the option of changing the meaning selection in the "View Concepts" window.
Q: When applied to Search, how long does it take Cognition to perform search queries?
A: Typically less than one second, comparable to commonly used Search engines.
Q: When querying, can I do Boolean searches with Cognition and how does it work?
A: You form Booleans with AND, OR, WITH, NOT WITH, and AND NOT. The elements between the Boolean operators are treated as conceptual queries. If you ask: "(strike in baseball) AND NOT (walk in baseball)" - it will search on "strike", meaning a state of the game baseball, rather than meaning "hit" or "labor walkout" and will exclude the concept of "walk" meaning a state of the game of baseball rather than "walk" meaning "to put one foot in front of the other."
Q: What is "Linguistic Search?"
A: Linguistic Search is a linguistically-based technique that interprets the meaning of queries, and extracts concepts from them. Then it searches for those concepts in a pre-indexed document base. For example, if you ask "Did they adopt the bill?," it will search on the concept "adopt" which can also be expressed as "approve". Thus it will find documents that mention "approve a bill" or "approval of bill", or any of the very many ways that concept can be expressed. This paraphrasing capability enables Cognition to exhaustively find nearly all documents relevant to a query. Notice that concept searching distinguishes between "adopt" meaning "approve", and "adopt" meaning "bring into one's family". Thus it avoids irrelevant retrievals of documents using "adopt" in that other meaning. Another capability is reasoning in the Cognition taxonomy. If you ask about "vehicle", Cognition retrieves documents about all kinds of vehicles from carts to jet airplanes and space ships.
Q: For enterprise applications, what file formats can Cognition read and index?
A: HTML, Word, Word Perfect, RTF, PDF, XML, PST, plain text and many more.
Q: For enterprise Search, do you support structured databases?
A: Yes
Q: Can the customer manipulate meaning interpretation?
A: Yes. The user can edit a file to force particular interpretations of words. In the legal domain, a user could force any instance of the word "matter" to mean a legal matter. Also, the user can force certain words which can either be names or common words, such as Bush and Tenet, to be interpreted as names or not.
Q: Does your technology display hit highlighting in the metadata?
A: Yes.
Q: Does your technology display hit highlighting in the document?
A: Yes, both at the sentence and word level.
Q: Does your technology allow for large documents to be streamed to the viewer?
A: Yes.